Reflections on English Word-Formation

We are all familiar with coming across a new word, whether it has just been
invented or whether we have just not met it before. How do we invent new
words? How do we understand words that we have never heard before? What
are the limits on the kinds of words we produce? How have linguists and
grammarians dealt with the phenomenon of creating new words, and how
justified are their ways of viewing such words? In this concise and compelling
book, Professor Bauer, one of the world’s best-known morphologists, looks
back over fifty years of his work, seeking out overlooked patterns in word-
formation, and offering new solutions to recurrent problems. Each chapter
deals with a different morphological problem, meaning that the book can
either be read from start to finish, or alternatively used as a concise reference
work on the key issues and problems in the field.

laurie bauer is Emeritus Professor of Linguistics, Victoria University of

Wellington, New Zealand. He is the author of many books on linguistics,
including The Oxford Reference Guide to English Morphology (with Lieber
and Plag), which won the LSA’s Leonard Bloomfield Prize. He won the Royal
Society of New Zealand’s Humanities medal in 2017.
Reflections on English
Word-Formation

Laurie Bauer
Victoria University of Wellington
Shaftesbury Road, Cambridge CB2 8EA, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre,
New Delhi – 110025, India
103 Penang Road, #05–06/07, Visioncrest Commercial, Singapore 238467

Cambridge University Press is part of Cambridge University Press & Assessment,

a department of the University of Cambridge.
We share the University’s mission to contribute to society through the pursuit of
education, learning and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781009559973
DOI: 10.1017/9781009559935
© Laurie Bauer 2025
This publication is in copyright. Subject to statutory exception and to the provisions
of relevant collective licensing agreements, no reproduction of any part may take
place without the written permission of Cambridge University Press & Assessment.
When citing this work, please include a reference to the DOI 10.1017/9781009559935
First published 2025
Ronnie Li/500px/Getty Images
A catalogue record for this publication is available from the British Library
Library of Congress Cataloging-in-Publication Data
Names: Bauer, Laurie, 1949– author.
Title: Reflections on English word-formation / Laurie Bauer, Victoria University
of Wellington.
Description: Cambridge, United Kingdom ; New York, NY : Cambridge
University Press, 2025. | Includes bibliographical references and index.
Identifiers: LCCN 2025000454 (print) | LCCN 2025000455 (ebook) | ISBN
9781009559973 (hardback) | ISBN 9781009559966 (paperback) | ISBN
9781009559935 (ebook)
Subjects: LCSH: English language – Word formation.
Classification: LCC PE1175 .B286 2025 (print) | LCC PE1175 (ebook) |
DDC 425/.92–dc23/eng/20250520
LC record available at https://lccn.loc.gov/2025000454
LC ebook record available at https://lccn.loc.gov/2025000455
ISBN 978-1-009-55997-3 Hardback
ISBN 978-1-009-55996-6 Paperback
Cambridge University Press & Assessment has no responsibility for the persistence
or accuracy of URLs for external or third-party internet websites referred to in this
publication and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
For EU product safety concerns, contact us at Calle de José Abascal, 56, 1°, 28003
Madrid, Spain, or email [email protected].
 Contents

Acknowledgements page vii

Conventions and Abbreviations viii

1 Introduction 1

Part I Basic Questions 11

2 Reflections on the Background to the Study of Word-Formation 13
3 Reflections on Why We Need Word-Formation 35
4 Reflections on the Recognition of Novelty in Words 39
5 Reflections on Blocking and Competition 43
6 Reflections on Potential and Norm 50
7 Reflections on Definition by Stipulation and on Word-Classes 56
8 Reflections on Analogical Word-Formation 64
9 Reflections on the Nature of the Lexeme 69

Part II Semantic Questions 75

10 Reflections on How Words Bear Meaning, and What This
Implies for Complex Words 77
11 Reflections on Tautology and Redundancy 90

Part III Syntactic Questions 95

12 Reflections on Recursion 97
13 Reflections on Problems with Heads in Word-Formation 102
14 Reflections on Coordination in Word-Formation 110

v
vi Contents

Part IV Interfaces 117

15 Reflections on the Interface between Word-Formation
and Phonology: Morphophonemics 119
16 Reflections on the Interface between Word-Formation
and Syntax 131
17 Reflections on the Interface between Word-Formation
and Phonetics 138
18 Reflections on the Interface between Word-Formation
and Orthography 144
19 Reflections on the Interface between Word-Formation
and Borrowing 148

Part V Patterns of Word-Formation in English 153

20 Reflections on the Limits of Conversion 155
21 Reflections on Back-Formation 164
22 Reflections on Coordinative Compounds 168
23 Reflections on the Irregularity of Prepositions 174
24 Reflections on Reduplication 179

Part VI Historical Questions 185

25 Reflections on Dead Morphology 187
26 Reflections on Compounds in English and in Wider Germanic 193

Part VII Questions Involving Inflection 207

27 Reflections on Inflection inside Word-Formation 209
28 Reflections on Canonical Form 218
29 Reflections on the Spread of Regular Inflection to Simple
and Derived Forms 225
30 Conclusion 232

Index of Topics 238

Index of English Word-Forming Elements 241
 Acknowledgements

I should like to thank the team at Cambridge University Press for their
enthusiasm for the project and their professional approach to this (and, in my
experience, every) book project. They were instrumental, among other things,
in finding referees for the work when it was first proposed, who provided
extremely thoughtful and helpful feedback. I should also like to thank Liza
Tarasova and Natalia Beliaeva for reading and commenting on an earlier draft,
and providing invaluable criticism, insights and suggestions.

vii
 Conventions and Abbreviations

Conventions and Abbreviations

‘ ... ’ round (a) meanings/glosses, (b) technical terms being discussed
“ ... ” round direct quotations
· [decimal point] in glosses, boundary between morphs not indicated by the spelling
(also used in the corresponding gloss)
[. . .] to indicate morphosyntactic bracketing; to indicate a phonetic
transcription
* to indicate an impossible formation; to show the position of
a missing letter
/ ... / to enclose phonemic transcriptions
. [full stop/ to divide words in glosses
period]
~ replaces the meaning of the base in a gloss, or allows for a blank to
be filled in by the reader
italics for cited words, phrases, sentences, affixes not in displays
small capitals (where needed) (a) for lexemes and (b) in glosses

Abbreviations in Glosses
1, 2, 3 1st, 2nd, 3rd person
abl ablative
acc accusative
adv adverb
dimin diminutive
gen genitive
le linking element
nmlz nominalization
nom nominative
N, V, A noun, verb, adjective
pl plural
poss possessive
sg singular

viii
1 Introduction

1.1 What Is Word-Formation?

‘Word-formation’ is the standard label for the way in which words are built up
from smaller recurrent formal elements. These smaller formal elements are
generally presumed to be linked in a fairly direct way to the meaning of the
word that is formed from them. Word-formation is widely assumed to be made
up of at least two distinct types: compounding or composition, the way in which
compounds like molehill and threadbare are constructed from smaller words;
and derivation, the way in which derivatives like unfriendly and discovery are
constructed from prefixes, suffixes and a word which is their base.
Given the well-known ambiguity of the term ‘word’ in modern linguistics, it
might be better if word-formation were termed ‘lexeme-formation’. However,
the term ‘word-formation’ was established before the term ‘lexeme’ became
normal, so ‘word-formation’ is the usual term. But the term ‘word-formation’ is
also sometimes used to include inflectional morphology, and so to deal with the
construction of word-forms such as covered and elephants; this means that
precisely what is included under word-formation is not necessarily fixed (see
Chapter 2). The standard notation is that covered (in italics) is a word-form
belonging to the paradigm of the lexeme cover (in small capital letters), but
this depends on several assumptions being agreed to (see Section 2.4).
Some people prefer the term ‘lexical morphology’ to ‘word-formation’ (e.g.
Coates 1987). This seems to imply that there is nothing in word-formation that
is not part of morphology. The difficulty with this is that there is not necessarily
agreement on what the term ‘morphology’ encompasses, either. While every-
one agrees that compounding, inflection and derivation are part of morphology,
it is not necessarily true that everyone agrees that the formation of words by
conversion (e.g. the link between the verb to cuddle and the noun a cuddle) is
part of morphology, and neither is necessarily true that everyone agrees that the
formation of initialisms such as MIT from Massachusetts Institute of
Technology (note that the initialism is not MIOT) is part of morphology. Yet
both of these might well be included as part of word-formation.

1
2 1 Introduction

We can already see that word-formation is plagued by terminological prob-

lems, and some of these will be expressly addressed in this book (see e.g.
Chapters 9, 20 and 22). Terminological problems cannot be solved by fiat. But
if they are made explicit, authors can position themselves in regard to the major
points of contention involved.
Word-formation can be viewed from at least two different viewpoints. It can
be viewed as dealing with the analysis of known words, or it can be viewed as
dealing with the ways in which new words are created by speakers. Since the
1970s – possibly before – the predominant approach has been the latter, at least
overtly. However, most handbooks provide many examples of the former, and
the two viewpoints are not necessarily kept strictly apart. Both approaches will
be considered here in different chapters. The latter perspective gives rise to the
question of productivity, whether and to what extent individual morphological
processes (including compounding and the use of prefixes and suffixes) are
used to create previously unknown words. Productivity has proved to be a very
difficult area of study (see Bauer 2001). Despite arguments made, for instance
by Bauer (1983), productivity is often seen as being one of the ways in which
the creation of new words differs from the creation of new sentences. But just
how variable productivity works, how it can be measured (if at all) and how it
may be constrained are still questions that can be debated, and some of these
questions will be raised below (e.g. in Chapter 5).

1.2 Is English Word-Formation Different from Word-Formation

in Other Languages?
The short answer to the question raised in the heading here is ‘yes’. All languages
differ in the details of how their word-formation works, what semantic categories
are marked in the word-formation (Japanese and Swahili mark causatives overtly
in their morphology, while English tends not to), what kinds of formal means are
used in the creation of words (English tends not to use infixation or reduplicating
prefixes while many other languages do), and what kinds of pattern are found
frequently and what kinds are found rarely. But that is not the reason for having
a book about English word-formation. In principle, I could have written about
word-formation in a different language, or across languages, and either of those
topics might have made a useful contribution. But English is not only a familiar
language for many linguists for whom it is not a first language, it is a language for
which a great deal of data is available, in the form of dictionaries (most notably
the Oxford English Dictionary), corpora illustrating usage, wordlists, coverage of
word-formation specifically and easily available examples of real usage by
language users. It is also the language for which I have the best intuitions
(however dangerous intuitions on usage may be). For all these reasons, focusing
on English provides materials which allow me to do what I want to do in this
 1 Introduction 3

book, namely examine the ways in which word-formation can be described and
the problems associated with them. This also means, I hope, that many of the
questions that are raised here with specific reference to English will actually be of
relevance to other languages and allow for discussions on the ways in which
languages can differ. Sometimes, for clarity, comparisons are made with other
languages (as has already been done above, see, in particular, Chapter 26), but the
focus is on the way in which these shed light on what happens in English.

1.3 The Historical Context

Any book is written in a context that is constantly changing. In particular that
context involves the way in which scholars understand the topic that they are
dealing with, involving the theories they believe in, the terminology they use
and the elements that they perceive as being relevant. The historical context,
however, is not necessarily uniform across a field such as linguistics. Ideas
which are no longer viewed as current in one area of linguistics may still be
considered standard in another. This is inevitable in any subject in which the
theories that were dominant when one generation of scholars was trained are no
longer dominant when new scholars are trained one or two generations later. In
the case of word-formation, the perceived importance of morphology in
a grammar has changed markedly over just a few generations, going from
being a central aspect of grammar to a marginalized area of study and back to
a central topic of focus within linguistics (Coates 1987). There is little point in
going through the changes in underlying philosophy which have led to this
position, but to consider the different structures and fundamental notions that
have been left behind might clarify some of the ways in which word-formation
has been dealt with.
The twentieth century provides a textbook example of the way in which
linguistic entities can come into and go out of fashion, when we consider the
notion of the morpheme. The morpheme seems to have been named in the late
nineteenth century, although the notion had been available before then. There
were two views of the morpheme, a European one and a North American one, but
they merged on the North American pattern. A morpheme is a meaningful
element of a word. If we consider the word unfriendly, we can see that the
word is made up of three elements. The core of the word is friend, friend is turned
into an adjective by the addition of the suffix -ly, and friendly is made negative by
the addition of the prefix un-. Each of friend, -ly and un- either is a morpheme or,
in slightly later analyses, represents a morpheme. If we follow the latter view,
a morpheme is an abstract unit (not a form which can be heard or written),
realized by one or more forms. If we follow the European tradition (following
Saussure 1916), the morpheme is a minimal sign: it has a form and it has
a meaning, but contains nothing smaller with the same qualities. If we follow
4 1 Introduction

the North American tradition (following Bloomfield 1935), a morpheme is

a minimal unit of grammatical analysis; again, it contains nothing with the
same qualities. In both cases, the notion of the morpheme was modelled on the
notion of the phoneme. In the European tradition, it might have bound variants, in
the North American tradition it can be viewed as having allomorphs. The bound
variants or allomorphs are variant forms which are restricted to particular
environments. Consider the words embark and entrain meaning, respectively,
‘to put onto a ship’ and ‘to put onto a train’. Each is made up of two forms, em-
and bark, en- and train. The difference between em- and en- has nothing to do
with the meaning, but the bilabial nasal appears before a bilabial consonant,
while the alveolar nasal (at least in the written form) is the default case, occurring
where there is no need for em-. Because of their distribution, the two do not
contrast and they also share a meaning, and so can be seen as variants of the same
abstract item. They are distinct morphs representing the same morpheme.
(Actually, I have simplified a little, because we find enmesh more frequently
than we find emmesh, because /nm/ is a permissible medial consonant sequence –
e.g. in enmity – while /nb/ would not be possible in *enbark.) By the 1950s, the
morpheme was a standard element of grammatical description, and worked
brilliantly for a great deal of English word-formation (and morphology more
generally in many languages), exploiting the very strong notion of allomorphy,
very much in the structuralist tradition.
Scholars of the time knew that there were problems with this picture, but felt
that the notion was valuable enough for it to be worthwhile working round the
difficulties these caused. However, more and more problems were found, and
as scholars wanted to write explicit grammars in the innovative Chomskyan
tradition that became dominant in the 1960s, the problems became viewed as
insuperable. This happened first in the description of inflectional systems,
although similar examples from derivational morphology could be used to
make the same points. The problems were raised and elaborated in works
such as the hugely influential Matthews (1972) and Anderson (1992), and
can be found summarized in works such as Anderson (2015) and Bauer
(2016, 2019). The result was that by the 1990s, the morpheme as it had been
presented by the structuralists was no longer considered to be a tenable theor-
etical notion, at least within theoretical morphological studies, though psycho-
linguists continued to work with the notion.
Although this is not the place to rehearse all the arguments against the
morpheme, we can say briefly that we find instances where a single morph is
associated with more than just one meaning, instances where a single meaning
is associated with several morphs; we find instances where we have
a recognizable morph, but the meaning usually associated with it is not associ-
ated with a particular word in which it occurs; we find morphs which do not
appear to carry any meaning at all and meanings which do not appear to have
 1 Introduction 5

any morph for them to be associated with; we have instances where a meaning
appears to be associated with a process rather than with a form, and many
instances where it is not clear how far we can stretch the notion of allomorphy.
This cumulation of problems leads to difficulties in applying the notion of
morpheme to many languages, but then the question becomes what can we
replace the morpheme with.
In some instances, particularly in cases of complex inflectional systems, it
seems to be possible to build up some morphs by the application of phono-
logical rules (Matthews 1972). Meanings can be associated with whole words
or sequences of words rather than with individual morphs (Matthews 1972,
Booij 2010). In other cases, we might observe formal patterns which do not
directly correlate with particular meanings, but which nevertheless tend to be
meaning-bearing (Aronoff 1994) or, if not, at least important in morphological
patterning and sometimes more diachronically stable than the individual words
that instantiate them. Note in all this that while morphemes may not be used,
morphs – the formal side of morphemes – tend to persist, and what is lost is the
direct link between individual forms and individual meanings.
Ironically, against such a background, the study of word-formation in
English, with its concentration of prefixes and suffixes, seems stuck on the
idea of the morpheme. There is still a view that words are made up of formal
elements, to each of which can be attributed a meaning or a function. This looks
remarkably like analysis into morphemes except that there is recognition that
not all morphs have meaning (the -t- in dramatist is either a meaningless
extender or is part of an allomorph of drama which occurs before certain
suffixes), that not all morphs have a consistent meaning (the -er in dish-
washer can denote either a person or an instrument, although some less specific
meaning may allow for a single gloss to cover both), we can have meaning
spread over multiple morphs (as in enliven where prefix and suffix together
provide the causative meaning) and more than one meaning in a single morph
(as in song, which can be seen as containing the meaning of ‘sing’ as well as the
meaning of ‘noun’) and we can have meanings not associated with any morph
(as with the difference between the verb to whisk and the noun a whisk). That is,
those who study word-formation seem more willing than those who study
inflection to work round the problems associated with the morpheme rather
than simply discarding the notion. Discussions which seem to assume
a morpheme-like unit will also be found in this book.
Developments in semantic theory have also had a strong influence on the
study of word-formation. The notion of prototype (Rosch 1973, Taylor 2003)
has influenced the way in which we view not only the meanings of words, but
also the way in which we envisage categories. For instance, we can now think
of the suffix -er as having a prototypical meaning centring on the notion of
agency (discoverer), but fading off into instruments (sharpener), experiencers
6 1 Introduction

(lover), locations (diner) and even patients (keeper) rather than as having a set
of fundamentally unrelated meanings. Where categories are concerned, we can
think of a class such as compounds having more central members and less
central members, rather than being defined by a fixed set of criteria or tests. If
girl Friday is a compound, it is not as central a compound as flower-girl,
because girl Friday is left-headed (see further Chapter 13). An alternative is
to see a category such as compound as a canonical category, again with more
and less central members of the category (Corbett 2010). Within Cognitive
Linguistics, the importance of figures of speech has been seen as central for the
semantics of word-formation because so many words involve figures of speech.
For instance, a whirlybird ‘helicopter’ is not literally a bird, but resembles
a bird in that it can fly, a jailbird ‘prisoner’ is not literally a bird, but is like
a budgerigar in that he or she is kept in a cage, and a thunderbird is not a bird
because it is a mythical creature, once believed to cause thunder. Recognizing
the figurative expressions for what they are makes it easier to explain the often
complex meanings of words created by word-formation, as well as the mean-
ings of simplex words.
Corresponding to these various developments, there have been many differ-
ent theoretical approaches to morphological study. In the early 1960s, within
Chomskyan generative grammar, morphology was seen as part of syntax, in
line with the view of morphemes that was still current at the time. Lees (1960)
provides an illustration of this general approach. Later (most obviously in
Chomsky and Halle 1968, but much discussed before that date), some of the
workings of morphology were subsumed in phonological rules, in a movement
that eventually led to Lexical Phonology and Morphology. Viewing morph-
ology as syntactic persists in Distributed Morphology (Halle and Marantz 1994
and a large amount of more recent work). This contrasts with a movement
towards seeing morphology as an independent area of linguistics, not just part
of other larger fields, a view which is made overt in the title of Aronoff (1994),
but is also to be seen in a lot of works where the notion of morphological
paradigm is given a central position, both in inflection and in derivation. This
point of view is found in various theoretical positions starting with Bybee
(1985) and is increasingly important today.
While this brief outline does not cover all the theoretical positions that have
been taken over the last century or so (for more detailed coverage, see Stewart
2015), readers can expect to find individual scholars taking positions which call
on several of these approaches in slightly different ways, and giving focus to
different aspects of the ways in which word-formation can be dealt with. There
is no consensus, but there is much to be learned from studies which take
different perspectives on the topic.
 1 Introduction 7

1.4 Why Study Word-Formation in Particular?

We could ask why we study anything at all, from how earthworms mate to why
the universe is expanding, and I imagine that the answer would be much the
same: because it’s fascinating and we are curious about how the universe
works. One of the topics within the range that we could wonder about is why
we should study language. Again, its fascination could be part of the answer,
but we might have some extra reasons (just as there might be extra reasons for
the study of anything). In the case of language, we could argue that one of the
strangest things about human beings is not only that they communicate by
means of language, but that it is hard to stop them talking! While other species
undoubtedly communicate, the human communication system is far more
complex than any other we understand, and it is definitional of human beings.
We might be classified as Homo loquens. Studying language is therefore one
way to try to understand part of what it is that makes us human. More than that,
most of us speak more than one language, and those people who do not are at
least aware that although language may be a part of what defines humankind, it
is not always the same language, and that various languages may well be totally
incomprehensible to people who have at least one language to help them
communicate with others. Most of us, however many or how few languages
we speak, will also be aware that if we cannot use language to communicate,
our ability to communicate at all drops drastically. Such observations raise
many questions, including: what do languages share? Are all languages really
just dialects of Human? Or are they so radically different that knowing one is
little help in learning another? Is it always possible to translate between
languages? Can we learn Dog and can dogs understand Human (see
Anderson 2004)?
Once we have decided to study language, we still have a huge problem:
languages have so many properties that it is hard to know where to start
studying them. Do we treat languages as organized sound (gesture in the case
of sign languages), as words chained together, as a means of transferring
meaning by having different ways of expressing a huge number of meanings,
as a way of organizing human interaction in such a way as to promote
a coherent society, as a way of reflecting the societal structures within which
we operate, as some kind of code that keeps changing? Language has all those
aspects. Reducing our focus to word-formation has some benefits as a way into
this morass of complex interactions: it concerns words, which speakers of
European languages at least think they have some understanding of; it deals
not only with forms but also with meanings; although we can invent new words,
most messages are made up of familiar ones, while most sentences (at least in
academic discussions) are not at all familiar; and last but not least, words are
fun. Many a comedian has asked questions like if a vegetarian eats vegetables,
8 1 Introduction

what does a humanitarian do, or if sink, sank and sunk are all related words, are
think, thank and thunk related in the same way? Language is endlessly fascin-
ating, and word-formation, quite apart from being interesting in its own right,
makes a good entry point into the wider field of study.

1.5 This Book

Although this book is about word-formation, specifically about word-
formation in English, it is not a textbook in the sense that it does not attempt
to provide a systematic and thorough discussion of the topic. The title
Reflections is intended to signify that it covers topics that I have found to be
of particular interest in the fifty years I have been thinking and writing about the
area, but also that the aim is to provide new insights or points of discussion, not
just to provide a summary of the state of the art. The topics covered are loosely
collected into thematic groupings, but are quasi-independent: the chapters can
be read in isolation, although places where the individual chapters are linked
are indicated, as they already have been above. Some matters which may
appear minor are covered, some which may seem more important are not.
But while topics which are vital to various theories about the way word-
formation might work are covered, explanation of individual theories are in
general not covered: theories come and go, but how word-formation is actually
used lies at the heart of the theories.

Challenge
Find examples from English derivational morphology where the mor-
pheme does not function in the way that is expected, as set out in this
chapter. How important are these examples in the general scope of word-
formation?

References
Anderson, Stephen R. (1992). A-morphous Morphology. Cambridge: Cambridge
University Press.
(2004). Dr Dolittle’s Delusion: Animals and the Uniqueness of Human
Communication. New Haven, CT: Yale University Press.
(2015). The morpheme: Its nature and use. In Matthew Baermann (ed.), The Oxford
Handbook of Inflection. Oxford: Oxford University Press, 11–33.
Aronoff, Mark. (1994). Morphology by Itself. Cambridge, MA: MIT Press.
Bauer, Laurie. (1983). English Word-Formation. Cambridge: Cambridge University
Press.
(2001). Morphological Productivity. Cambridge: Cambridge University Press.
 1 Introduction 9

(2016). Classical morphemics: Assumptions, extensions, and alternatives. In

Andrew Hippisley & Gregory Stump (eds.), The Cambridge Handbook of
Morphology. Cambridge: Cambridge University Press, 331–55.
(2019). Rethinking Morphology. Edinburgh: Edinburgh University Press.
Bloomfield, Leonard. (1935). Language. London: Allen & Unwin.
Booij, Geert. (2010). Construction Morphology. Oxford: Oxford University Press.
Bybee, Joan. (1985). Morphology. Amsterdam: Benjamins.
Chomsky, Noam & Morris Halle. (1968). The Sound Pattern of English. New York:
Harper & Row.
Coates, Richard. (1987). Lexical morphology. In John Lyons, Richard Coates,
Margaret Deuchar & Gerald Gazdar (eds.), New Horizons in Linguistics 2.
London: Pelican, 103–21.
Corbett, Greville G. (2010). Canonical derivational morphology. Word Structure 3,
141–55.
Halle, Morris & Alex Marantz. (1994). Distributed Morphology and the pieces of
inflection. In Kenneth Hale & Samuel J. Keyser (eds.), The View from Building
20. Cambridge, MA: MIT Press, 111–76.
Lees, Robert B. (1960). The Grammar of English Nominalizations. Bloomington:
Indiana University Press.
Matthews, P.H. (1972). Inflectional Morphology. Cambridge: Cambridge University
Press.
Rosch, Eleanor H. (1973). On the internal structure of perceptual and semantic categor-
ies. In T.E. Moore (ed.), Cognitive Development and the Acquisition of Language,
New York: Academic Press, 111–44.
Saussure, Ferdinand de. (1916). Cours de linguistique générale. Paris: Payot.
Stewart, Thomas W. (2015). Contemporary Morphological Theories. Edinburgh:
Edinburgh University Press.
Taylor, John. (2003). Linguistic Categorization. 3rd ed. Oxford: Oxford University
Press.
Part I

Basic Questions
2 Reflections on the Background to the Study
of Word-Formation

2.1 Introduction
To study word-formation is to investigate the way in which words are formed.
But that does not allow us to say precisely what is included in word-formation.
Although the word cat is made up of the phonemes /k/, /æ/ and /t/, we would
probably not call this word-formation. So a fundamental question about word-
formation is what it covers. Although there is a great deal of dispute as to how
best to deal with word-formation in a grammar, there is probably rather less
dispute as to what the problems facing an attempt to do so are. Many of these
problems will be dealt with, in one way or another, later in the book. Here, the
aim is simply to raise the issues.

2.2 Two Traditions

Word-formation has been studied in two main traditions, reflecting the way in
which the study of linguistic structure as a whole has been discussed. There is
the tradition of considering the patterns, exemplifying them, considering limits
on them and what they mean. In relatively recent linguistic terminology, we
might call such approaches ‘structuralist’, although many such discussions pre-
date the era which we now think of as the time of structuralist linguistics,
starting with Saussure. (We might alternatively use the term ‘taxonomic’, if this
term had not been debased in the linguistics of the early 1960s.) Then there is
a tradition of viewing the patterns of word-formation as the output of a series of
formal rules. Such approaches go back to the precursors of Panini, and continue
today in many formal models including the A-morphous Morphology of
Anderson (1992) and Distributed Morphology (Halle and Marantz 1993).
The fact that one of these examples focuses on the phonological make-up of
words and the other sees words as being made up of elements that are linked by
syntactic rules is not really relevant to this fundamental division. In principle,
works in the structuralist and rule-based traditions can be translated into the
other, though the focus is rather different. The formal mechanisms involved in
the second of these approaches may vary considerably, and in doing so may

13
14 Part I Basic Questions

capture different generalizations, but all such approaches share the view that
word-formation is underlyingly regular and productive in that new words (new
lexemes) can be created by means of the already known rules.
There is a third, more modern, tradition for studying language structures,
though it does not translate easily into the other two. It is the tradition of
Cognitive Linguistics, with various branches including Construction Grammar,
Usage-Based Linguistics and Exemplar Grammar (for a useful introduction,
see Bybee 2023). What these share is that they see the patterns of language
emerging from the cognitive structures of the human brain and the way in
which language is used by its speakers, and they see meaning as the central
driver of linguistic structure. Where this influences the study of word-
formation, it brings the focus onto networks of words and partial resemblances
between words. For many working in this tradition, the use of nominalizations
might be more important than the way in a which a particular affix forms
nominalizations. This includes the importance of paradigms in various ways.
Those who study constructions look at the ways in which sequences of elem-
ents are built up, and how their meanings may be idiosyncratic to the construc-
tion rather than derived entirely from the meanings of the smaller elements in
the sequence.
Whichever of these traditions the individual scholar may wish to pursue,
there are three questions which need to be answered before the description is
undertaken: what is formed in word-formation? Is word-formation a coherent
object of study? And what are the limits of word-formation? None of these is as
simple as it might appear.

2.3 What Is a Word?

It is well known that we have no universal definition of what a word is (Bauer
2000, Dixon and Aikhenvald 2002, Hippisley 2015, Wray 2015). Neither do we
have a definition of a word which will help us where English is concerned. Most
authorities end up using a modified orthographic criterion: a word is surrounded
by spaces or punctuation marks in a written text. Quite apart from the places
where this does not give a reasonable outcome, it is back to front. As a general
principle, we would expect that the place where we write the spaces should
derive from the nature of the spoken language, not vice versa. Such a criterion
gives a good practical guide, but one that started from the spoken language would
be preferable. This is made more difficult, however, by the fact that phonological
words and morphosyntactic words do not necessarily coincide – as well as by the
fact that definitions for both are inconsistent in the literature (Aikhenvald et al.
2020). For example, for some scholars, prefixes are not part of the same
phonological word as their bases, for others, some words like brokenness consist
of two phonological words because of the geminate /nn/ which can occur only
 2 Background to the Study of Word-Formation 15

over morphological boundaries. From a semantic point of view, on the up and up

(‘continually improving’) and kick the bucket (‘die’) seem to be single words,
although the spelling system marks them as being made up of multiple words.
And from an orthographic point of view, the fact that we can write wordforma-
tion, word-formation or word formation indicates that spaces are not necessarily
reliably indicative of status. Where nouns and verbs are concerned, an inflec-
tional marker (where there is one) typically shows the end of the word, but there
are huge numbers of words where this does not apply, and care must be taken
with -ing and -ed (or its irregular congeners) because it is not clear whether
participles count as inflectional. There are also problems with items such as
ladies-in-waiting where the inflection does not occur on the right-hand edge of
the item, and so leaves it unclear as to whether this should be treated as a word or
not. All of this provides a huge theoretical problem, which is rarely seen as
a practical problem. That is because word-formation is defined in its own terms,
without reference to the word as such.

2.4 Inflections versus Derivation

There is a long tradition of word-formation studies which sees word-formation as
embracing derivation and compounding (and possibly some less regular minor
patterns) but excluding inflection. Standard descriptive grammars of Indo-
European languages, when they mention word-formation, tend to make such
assumptions, and the notion is entrenched in the title of the classic text by
Marchand (1969) (The Categories and Types of Present-Day English Word-
Formation), which does not include inflectional morphology. There is a more
recent tradition, in which word-formation involves all patterns (or all rules leading
to patterns) of elements smaller than the word, including inflection (Baker and
Hengeveld 2012: 23, Lieber 2022: 106). The difference hangs on the researcher’s
view of the split morphology hypothesis (Perlmutter 1988, Booij 1996). The split
morphology hypothesis holds that inflection is distinct from (is split off from)
derivation, and those who believe in split morphology see inflection as categorially
different from derivation; those who reject split morphology do so on the basis that
inflection and derivation have more in common than they have distinguishing
them, so that they should be treated the same way in a grammar.
Arguments can be provided on both sides of the issue, and have important
theoretical implications, but there does not seem to be any absolutely decisive
factor which can be used to determine the answer. Some of the arguments are
sketched below.
Against split morphology:
• Both inflectional and derivational morphology make use of the same set of
formal processes (though not necessarily in any individual language – all
16 Part I Basic Questions

prefixes in English are derivational), processes such as prefixation, suffix-

ation, apophony, reduplication.
• Although there are many criteria which have been presented for distinguish-
ing between inflection and derivation (see, notably, Plank 1994), they are not
always easy to apply, they do not always agree, and the borderline between
them is often fuzzy – there is dispute in the literature as to whether the -ly that
forms English adverbs is inflectional or derivational; the result is that there
may be dispute as to whether individual processes in a given language are
inflectional or derivational, and it may not be clear how to distinguish
inflection from derivation in a given language, if it is possible at all (see
Bauer and Bauer 2012 on Māori, for instance).
• There appear to be languages with morphological structures that do not
distinguish inflection and derivation in any clear way.
• Many apparently inflectional markers can have uses which appear to be
derivational, including plural markers in a language like English with bellow
versus bellows, glass versus glasses (‘spectacles’), look versus looks, and
also including participial forms.
For split morphology:
• Without split morphology, there can be no notion of the lexeme, which is defined
in terms of inflection, and which is a widely accepted notion in morphological
studies as well as in lexicography (where the term ‘lemma’ is often used).
• All the criteria that distinguish between inflection and derivation lead us to
believe that there is at least a canonical difference between them, because
although they do not always agree, they do align in many instances.
• Although subsidiarily, all the arguments which lead to the establishment of
the lexeme also lead to the adoption of the split morphology hypothesis.

2.5 The Limits of Word-Formation

The next question concerns the limits of word-formation. If derivation and
compounding are at the core of word-formation, how far does it extend? On the
one hand we have phonaesthemes (Firth [1930] 1965), which are pieces of
phonology whose influence in the meaning of a word is diffuse, such as the /ɡl/
at the beginning of glum and gloom (see Section 17.1), at another extreme we
have phrasal verbs like look up in look up the answer, which in English are
usually treated as being at the intersection of syntax and idiomaticity.
Somewhere between these extremes we have back-formation, clipping, con-
version, reduplication, internal modification, acronyms, initialisms and pos-
sibly neoclassical compounds and closely related forms. Even within
compounding, we may argue as to whether, for example, Paris–Rome in the
Paris–Rome flight is or is not a compound (see Chapter 22). If we do not accept
 2 Background to the Study of Word-Formation 17

the split morphology hypothesis, we also have to consider inflection. And there
are vast numbers of multiple word expressions (MWEs) such as up to date, by
and large, black and blue, which in some ways are more like lexical items than
like freely constructed syntactic phrases, whatever their etymological origins
might be. Bauer (2019a) presents a view of word-formation in which all these
things might be seen as relevant, but that view is intentionally provocative, and
many, perhaps most, scholars would want to be more conservative in what
word-formation deals with. Note that there is no claim here than any one
answer is ‘right’ and others are ‘wrong’, merely that there are many possibil-
ities, and that what is included may influence the form of the theory that is used
to discuss such matters.
Another way in which scholars of word-formation have to determine what is
and is not relevant to their study is in deciding what the limits are to
a synchronic description of word-formation. Bauer (1983) (see also Bauer
2001) spends a lot of space discussing lexicalization and productivity: prod-
uctivity is involved in whether and to what extent a morphological pattern can
be extended to create new forms, and words which are established in the
community but which have features (forms, meanings, etc.) which cannot be
attributed to the word-formation pattern are lexicalized to some extent. (I have
formulated this statement here in terms of patterns, but in Bauer 2001 it is
formulated in terms of rules.) To use some well-discussed examples, lady
(derived historically from elements meaning ‘loaf kneader’), Arabic (with
stress on the initial syllable despite stress normally falling on the syllable
preceding the suffix -ic), blackmail (where mail used to mean ‘coin’ but no
longer does), bishopric (where the suffix -ric does not occur except when
attached to bishop), depth (where the suffix -th is not used any more to create
new words and where the vowel of the base is not predictable) are all lexical-
ized in one way or another. The question is: which of these (if any) should be
included in the synchronic description of English word-formation (or, to
formulate in perhaps a more challenging way, should the synchronic rules of
English word-formation allow the generation of any of these forms)? If such
words are not part of the productive patterns of English, which is the implica-
tion of the challenge just issued, how should they be treated in a description of
English and what, if any, part do such words have to play in the description of
current English word-formation?
A certain amount of terminology will help with the discussion. Bauer (1983)
introduces the term ‘institutionalization’. A word is institutionalized when it is
recognized in the speech community even though it is in principle in line with
productive patterns and could be created by them. The word hunter is institu-
tionalized as a ‘person who kills wild animals for food’, because the term is
familiar in the community and could be created from hunt + er as a pattern or as
a rule. The same is true of hunter meaning ‘a horse used in hunting’. However,
18 Part I Basic Questions

hunter is lexicalized in the sense ‘pocket watch whose face is protected by a lid’
because in the current state of the language we cannot relate that use to the
elements hunt + er. Bauer (1983) also uses the term ‘established’ for words
which are either lexicalized or institutionalized.
A word is analysable to the extent that the language specialist (perhaps the
speaker of the language, though that is less clear) can see the elements which
make it up and see that they might be relevant to the interpretation of the word.
Blackmail is analysable in this sense: we may not know what mail is supposed
to mean in blackmail, but we are not in doubt as to it being a relevant element in
the make-up of the word.
A word is transparent to the extent that the linguist or speaker can see what
the elements in the word are and what they mean, and thus how they contribute
to the overall meaning of the word. Warmth is analysable and transparent,
health might be analysable, but is not transparent, at least not for most speakers.
This is despite the fact that -th is not productive, and thus health is not
institutionalized. The borderline between analysable and transparent is fuzzy
because individual speakers/analysts may differ on the extent to which, for
instance, filth or dearth belong in either of the classes. The opposite of
‘transparent’ is ‘opaque’, and both terms are gradable in this usage.
Most textbook analyses of the patterns in English word-formation are based
on transparent words. For example, Bauer et al. (2013: 196) provide
a discussion of nominalizing -al (arrival, denial, rehearsal) despite saying
that it is “minimally productive” and despite citing only one word which might
suggest that it is actually productive and despite the fact that other authors (e.g.
Dixon 2014: 345) claim that it is not productive. It produces (or used to
produce) words that are transparent, however. Quirk et al. (1972: 1001) discuss
the verbalizing suffix -en (deaden, ripen, widen), where the form and meaning
are transparent, despite the fact that -en is probably no longer productive (Plag
1999: 218), the most recent widely accepted formation being the decidedly
jokey embiggen from The Simpsons. Examples could be found in many other
sources. It is not clear whether this is ever an overt policy in such books, but it
makes sense in terms of students who wish to analyse words they come across:
if a word is not analysable, it will not be clear that it is meaningful to look for
elements in the word (think of lady); if it is not transparent (e.g. tootle listed by
Marchand 1969), finding elements will not help the analysis much. The place
where handbooks might look at analysable but not transparent words is where
the relevant pattern is transparent in some instances and not in others. For
example, the suffix -th might be discussed because of words like warmth and
despite words like dearth.
But if this solution is a good practical one, it is less clear that it is a sound
theoretical one. Parallels in syntax are hard to come by, but we would not write
a rule of current English syntax which allows a preposition to be coordinated
 2 Background to the Study of Word-Formation 19

with an adjective (in and pretty) just because we have the expression by and
large (historically this had a different analysis but it is difficult to justify it in
current usage: by and large is lexicalized). If something is lexicalized, it can no
longer be used in the creation of new forms (for one reason or another –
a change of meaning in an element, a change in the patterns that the grammar
accepts, and so on). Theoretically, a stronger position is that the current rules or
patterns are only those which are still productive. In this position, there is no
rule of -th suffixation (and thus no -th suffix) because it is not productive.
Similarly, we can ignore the pattern illustrated by spoilsport and dreadnought
because that particular compound pattern is no longer part of English
morphology.
If we accept this position, it brings with it some immediate benefits. The most
obvious one is that the scope of English word-formation shrinks considerably. We
no longer need to worry about whether -sion, -tion, -ion are variants of the same
underlying element (allomorphs), and if so what conditions their distribution,
because the only productive nominalization marker from the set is -ation and the
others are lexicalized. By the same token, we do not have to worry about the
morphophonological rule which gives us the alternation found in collide and
collision, because if the word collision is lexicalized we cannot generate it in the
current state of the language, and, therefore, we do not have to worry about its
phonology if that phonological form arises purely in connection with some
lexicalized morphological process.
Another benefit is that such a principle draws a line between synchronic
morphology and etymology: morphology deals with what is possible in the
current state of the language, while etymology deals with the history of words
and how they came to have their various meanings. Etymological information
may have little to do with the current meaning of the word, as when the English
word nice has a Latin etymon meaning ‘ignorant’. On the other hand, the
etymology may make clear precisely how the current meaning of an extant
word, but one no longer governed by current morphological patterns or rules,
came to be. For example, etymologically speaking, we can recognize an old -er
suffix in words like chatter and stammer which indicates repeated action.
At the same time, there are some costs. The most obvious of these can be
seen in the status of a word like collision. The straightforward way to deal with
such words would be simply to list them in the lexicon. This would then seem to
imply that collide and collision are not linked, and that, in fact, there is
enormous redundancy in the lexicon because the meaning embedded in related
words has to be specified more than once. If we want to link them, we have to
have recourse to some separate mechanism, something like the via rules of
Natural Generative Phonology (Hooper 1976) or the redundancy rules of
Jackendoff (1975), which is not the same as the normal rules of morphology
and phonology. To some extent, we may be able to shrug this off; we already
20 Part I Basic Questions

have forms in the lexicon with redundant semantics – forms like dog/canine,
horse/equine, tree/arboreal (canine must repeat the semantic information
associated with dog, even though there is no formal link). But not all scholars
will be willing to accept such a large increase in memorized material at the
expense of so much computed material, even if we know that speakers often
have dual access to apparently morphologically complex words, both by
whole-word access and through computation of meaning from form
(Stemberger and MacWhinney 1986, Vannest et al. 2005).
For the sake of the argument, let us assume that we accept the position that
only productive processes can be part of morphology. However much it may
sound like a reasonable hypothesis, it strikes a number of practical problems.
Where does the morphologist get data on the productivity of processes?
Although the Oxford English Dictionary (OED) makes some attempt at cover-
ing all of English, no other English dictionary gets close. Even the OED has
problems. One of these is that it does not usually list words for which it does not
have evidence of repeated use in the community. While a large number of low-
frequency forms created with a particular morphological process is usually
seen as primary evidence of the productivity of that process (Bauer 2001: 150),
the OED’s methodology inevitably hides some of the relevant evidence.
For this reason, as well as others, many linguists use corpora, preferably
large ones, as sources of data. Corpora allow rare examples of relevant forma-
tions to be found, but bring with them their own problems. One problem is the
nature of the examples uncovered. Of course, they tend to be a random sample
of possible formations, but it is frequently unclear whether the words formed
are genuine usages of word-formation as language use, or whether they illus-
trate deliberate jokes on the part of the writer – puns, deliberate references to
other words or expressions, indicators, in fictional texts, of a character’s ignor-
ance or status. In principle, we might think, this is unlikely to be a problem: the
value of corpus data is to find patterns, including patterns that the researcher
was not aware of, rather than to find rare examples of unlikely formations. To
a large extent, this is true, but in cases of marginal productivity, whether
a particular morphological process is used at all may be important.
We can consider a genuine example. Bauer et al. (2013) use large corpora to
find data on which to base their description. When it comes to negative prefixes,
they find surprising data. The overuse of un-, as the most productive negative
prefix, is expected, as we can already find it in widespread usages such as
untypical (partly replacing an earlier atypical) or undeterminable for indetermin-
able. But the discovery of many instances where in- appears to replace un- is
unexpected: forms such as inbearable, inbelievable, inintelligent, incertainty.
Even less expected are cases where the orthographic allomorphy of in- is not
observed, with examples like inbalance, inperfections, inprecision, inprescrib-
able (Bauer et al. 2013: 361). Here the corpus examples seem to indicate
 2 Background to the Study of Word-Formation 21

productivity of the in- prefix, where previous authors had claimed it to be

non-productive, but they also appear to show the breakdown of the expected
allomorphy. The breakdown of the allomorphy could be because English
often has a morphophonemic writing system, and one place where this
shows in the writing of <unkind> rather than *<ungkind> (although [ŋ]
was an allophone of /n/ at the time when the spelling was established); we
also usually find <enmesh> rather than <emmesh>, which would follow the
expected form if a following bilabial automatically gives rise to [m]. Note,
moreover, that what was found in the corpus was not a single form, but
a series of forms, sufficient to look like evidence of something new.
However, one factor casts doubt on all of this: on a typewriter keyboard
<m> is adjacent to <n> and <u> is adjacent to <i> – this suggests that any of
these forms could easily be typographical errors caused by the nature of the
technology. Since we cannot know, the status of the corpus data is thrown
into doubt. This is an extreme example, but does indicate that we have to be
careful with apparently clear corpus examples.
In a wider sense, whatever our source of data, it may not be clear just what
indicates productivity. Consider the following examples.
I have not got the true egocentricity of the true artist in blonde-ishment. (Jane
Duncan. 1959. My Friend Muriel. London: Macmillan, p. 131)
“they can wipe up any spilth”. Spilth was a word coined by Russell, or so they
believed. How the language had managed without it, they couldn’t imagine.
(Rebecca Tope. 2018. The Stavely Suspect. London: Allison and Busby, p. 206)
the apparent safety of the now ubiquitous rideshare was just that – an apparency.
(Elizabeth Breck. 2020. Anonymous. New York: Crooked Lane, p. 21)
the place seemed poorer than he’d expected, since state capitals were usually
stuffed with well-paid bureaucrats. Maybe, he thought, Delaware didn’t well-pay
its bureaucrats. (John Sandford. 2020. Masked Prey. New York: Putnam, p. 183)

I would suggest that none of these is evidence of productivity. Duncan’s

blonde-ishment might be a Scotticism, but rather it seems to have been formed
on blandishment, as a joke, and -ment is usually added to verbs rather than
adjectives. Spilth, perhaps based on filth, does fit the pattern of -th formations,
but this is a very isolated example of the use of the suffix. Apparency fits all the
patterns, and is listed in Marr (2008) as archaic. Bauer et al. (2013: 197) list
some unfamiliar examples of -cy (most of them are either established or have
final -cy instead of the expected -ce, where the two alternate without motiv-
ation), but there is little evidence of productivity rather than rarity. Well-pay as
a back-formation from well-paid is not a typical pattern of back-formation,
although other verbs are created by back-formation from bases including a past
participle.
22 Part I Basic Questions

Cannon (1987) provides an analysis of the items in a set of dictionaries of

new words that appeared in the 1970s and 1980s, looking at the macro-patterns
of innovation that are registered in such works. Unfortunately, this particular
genre of dictionary seems to have gone into decline since then, so that equiva-
lent studies based on later data are not possible, although the OED can be
searched for words which are first attested in or after a particular year. Despite
many criticisms that can be made about Cannon’s interpretation of his sources,
his results are of interest.
Given what has been said here about word-formation, we might expect
Cannon to have found most of the new words are compounds, then instances
of prefixation and suffixation with established affixes, then a mixed set of
clippings, blends, back-formations and neoclassical compounds, perhaps with
a handful of instances of reduplication.
But this is not what Cannon finds, as is shown by his summary (Cannon
1987: 279). He does find that compounds are the most frequent type of
formation: 29.5 per cent of the words investigated, most of which are nouns.
Prefixations and suffixations provide only 21.6 per cent of the formations.
Given that number, the proportion of shortenings (including abbreviations,
acronyms, back-formations and blends) at 18 per cent seems much higher
than might be expected, 14.4 per cent of new meanings for existing forms
(very often figurative extensions, such as lifer extended from a prisoner to
someone serving in the army) also seems high and so does 7.5 per cent of
borrowings – a category that word-formation specialists ignore, though it is
familiar to lexicographers and lexicologists. In comparison, only 4.1 per cent of
the words were instances of conversion, which seems low. Even if the figures
cannot be completely trusted (Spaceship Earth is cited as an abbreviation of the
title of a book Operating Manual for Spaceship Earth), the proportions given
by Cannon raise questions about the way in which vocabulary is increased, and
suggest that word-formation is less important than morphologists tend to think.
Another way to approach the same question of the extent to which
word-formation is important in English, is to consider the most frequently
encountered words of English. The first 10,000 most frequent words in the
British National Corpus (Davies 2004–) were considered. The first thou-
sand, the third thousand, the fifth thousand, the seventh thousand and the
ninth thousand were analysed. The number of clearly prefixed words, the
number of clearly suffixed words and the number of compounds were
counted (if more than one formation-type was present, the external one in
the word was counted). Also counted, but in a separate category, were the
number of words that might be considered morphologically simplex, the
number of words that are morphologically complex, but in French (gov-
ernment), Latin (introduce) or Greek (epiglottis), and the number of direct
loans from other languages which might be considered morphologically
 2 Background to the Study of Word-Formation 23

simplex in English, whatever their origins. Instances of abbreviations and

inflected forms were noted, but are not reported here, because there were
relatively few. These numbers must be taken with a large pinch of salt,
since borderline cases abound and a different count could come up with
different answers: for example, words like output were considered to be
compounds, not instances of prefixation; words like preponderance were
considered to be instances of suffixation rather than learned morphology.
Nevertheless, they should give a first approximation to a view of the
sources of new words. First, the number of examples of English word-
formation are considered. Although the numbers in each cell are presented
as number of attestations, adding a decimal point allows them to be read
as percentages since each is a count out of 1,000. The numbers involved
are of interest, as well as the way they change depending on the frequency
of the words in use.

1−1,000 2,001−3,000 4,001−5,000 6,001−7,000 8,001−9,000 Total/5,000

Compounds 9 25 38 56 55 183
Prefixes 32 31 63 63 93 282
Suffixes 70 138 116 154 110 588
Total 111 194 217 273 258 1,053

Overall, around 25 per cent of words are morphologically complex, with more
in the less frequent words. This contrasts with the number of words which are
simplex, either because they are monomorphic or because they are loans which
cannot be analysed in English.

1−1,000 2,001−3,000 4,001−5,000 6,001−7,000 8,001−9,000 Tota/5,000

Simplex 661 551 432 338 273 2,255
Loans 10 47 97 137 180 471
Total 671 606 529 475 453 2,726

This leaves the words that are complex in one of the major donor languages for
English. Although most of these come directly from the donor languages, some
have been formed in later usage within English (consider telephone which was
not a word needed by the Greeks!).

1−1,000 2,001−3,000 4,001−5,000 6,001−7,000 8,001−9,000 Total/5,000

Learned morphology 182 218 211 201 251 1,063

This means that learned morphology is just as important in numerical terms for
speakers and learners of English as native morphology. The learned morph-
ology is not necessarily as informative as the English morphology, for example,
24 Part I Basic Questions

induce ‘lead in’ will take a lot more explanation than inlay will require, and to
cover all learned morphology will require information from at least three
languages (Greek, Latin and French – for example, antagonize, educate and
department, respectively), some languages being used more in some domains
than in others. Nevertheless, the importance of this aspect of English word-
formation should not be underestimated.
We can also ask what formations count as word-formation. For instance,
does expletive insertion as in absobloodylutely and kangafuckingroo count as
word-formation? It has been treated as such by, for example, Aronoff (1976)
and Bauer (1983), and it has the potential to throw light on the way in which
words can be constructed, but as a case of what is more widely known as
marginal morphology (Dressler 2000) or expressive morphology (Zwicky
and Pullum 1987) its status as a part of word-formation is in question. In fact,
despite these labels, even its status as morphology might be questioned: is
kangafuckingroo simply a syntactic alternative to fucking kangaroo, or is it
genuinely morphological? The spelling as it has been used here might suggest
a single word (and therefore morphology), but it is often hyphenated, just like
this-person-is-a-jerk attitude (Bauer et al. 2013) which does not necessarily
imply that person is not a word in that expression. In a different way, we
might query the status of initialisms such as CID ‘Criminal Investigation
Department’, FBI ‘Federal Bureau of Investigation’, imho ‘in my humble
opinion’, lol ‘laugh out loud’, SAS ‘Special Air Service’, YHA ‘Youth Hostels
Association’. Each of these represents some kind of syntactic string, though
users may not always be aware of what the individual letters stand for. They
are word-like in that they are fixed expressions for a given content, sentence-
like in that their meaning can be unpacked into a syntactic unit, and like
neither in that they are spelled out as individual letters. It may also be relevant
to consider whether an initialism is a word when it is synonymous with
a syntactic phrase.
We can also ask about constructions which may span or cross the divide
between word-formation and syntax. The obvious relevant formations here are
phrasal verbs. In English, these are typically treated as syntactic constructions,
while in German apparently comparable constructions are treated as morpho-
logical constructions. In both cases the direct object can interrupt the verb +
particle (something that seems typical of syntactic constructions):
I call my girlfriend up every day.
Ich rufe meine Freundin jeden Tag an [‘I call my girlfriend every day up’]
In German, though, the infinitive form of the verb is anrufen (a single word)
while in English the infinitive is call up (a sequence of two orthographic
words). This seems to make the distinction; but if we do not think of words
 2 Background to the Study of Word-Formation 25

as orthographic units, we might argue for a parallel treatment between the

English and the German constructions.

2.6 Are There Rules of Word-Formation?

The question of whether there are rules of word-formation can be answered
either way if we define our terms appropriately. For example, if we ask whether
gutcat is a compound of English, we might answer that it is not, because it is
a rule of English compounds that the superordinate terms must appear on the
right, so the word must be catgut (either that, or the word must mean something
else). In saying so, we assume that we are dealing with a rule, and so rules exist.
On the other hand, if we ask whether Congolese is a good word of English, we
might answer that although it is used, it does not seem to follow any rule of
English because the <l> is inexplicable. At this point we must assume that the
creation of words is not (or is not always) constrained by rules.
At a more general level, we might ask whether word-formation operates in
a rule-governed way, just like syntax. (The “just like syntax” part of this might
be cheating a little, since it is not entirely clear that syntactic rules always work
in the way that we expect linguistic rules to operate.) This is a much bigger
question, and correspondingly more difficult to answer. It will be discussed at
various points in this book. But it assumes that grammatical structures are
created by a series of rules which are always available and cannot be circum-
scribed by the presence of individual words, sounds or word-parts, and that
there is freedom to insert words into the grammatical structures thus created.
That is, if we have a sentence like
I have to see a man about a dog.
it must also be possible to have a grammatically equally good sentence of the
form
My uncle’s father has to mutter a tower about this shilling.

Trained linguists may have no problem with this (though naive speakers
probably will), but when it comes to parallel discussions about the availability
of words, intuitions (even among linguists) differ, and it is harder to reach
a consensus. It is not clear whether we can discard the metaphor of a rule in
word-formation, or whether we have to modify our understanding of what
a rule means in word-formation in some principled way. If we want to replace
the notion of a rule, then we have to replace it with something, and we do not
have anything that is as widely understood in the linguistic community or as
theoretically well developed as the rule.
26 Part I Basic Questions

2.7 Are the Elements of Words Meaningful?

At first sight, the question of whether word elements are meaningful has a self-
evident answer. If we take a word like cats, we can split it into two elements, of
which the first means ‘cat’ and the second means ‘more than one’, and cats
means ‘more than one cat’. If we take a word like bookshelf, it means ‘a shelf
for books’ and the meanings of book and shelf are involved in that overall
meaning – though if we are careful not to cheat, we might ask why it is not
books-shelf since it is for more than one book, and how we know it is for books
and not, say, made out of books, like metal shelf.
But those who have seen the TV series The Wire may recall that a lake trout is
a sea-fish (not a fresh-water fish) and is not a kind of trout. As one of the
characters in the series comments: “No lake, no trout”. Such examples are
commoner than we might think.
buttercup is not a hyponym of cup; foxglove is not a hyponym of glove;
glowworm is not a hyponym of worm; hedge sparrow is not a hyponym of
sparrow; sea wolf is not a hyponym of wolf; starfish is not a hyponym of fish
butterfly has little if anything to do with butter; catgut does not come from a cat;
a cherry birch has nothing to do with cherries; dogwood does not appear to have
anything to do with dogs; a wheatear (‘bird sp.’) has nothing to do with wheat
(its name seems to be a corruption of an earlier form of ‘white arse’)
a howler ‘very funny joke’ does not howl; a reefer ‘marijuana cigarette’ (no
longer current) does not reef; a sleeper on a railway track does not sleep
“What I don’t get”, she was saying, “is where they get off calling this Long
Island Iced Tea. There must be half a dozen different kinds of booze in it, but
is there any tea at all?”
“You’re asking the wrong person.”
“No tea”, she decided. (Lawrence Block. 2000. Hit List. London: Orion, p. 138)
I collected Martin’s car for the slow slog to Freddie’s place, wondering for the
ten thousandth time why they call it ‘rush hour’ when you can’t move faster
than a crawl. (Sarah Paretsky. 2013. Critical Mass. New York: Putnam, p. 278)
“He deadheaded to Rome for two days and is now back in London’
‘Deadheaded?” Gershwin asked.
“Taking a flight but not working it.” (J.A. Kerley. 2014. The Memory Killer.
London: HarperCollins, p. 35)
And now I come to think of it, I haven’t a dashed clue what an eave is either.
And how do you drop them? (Alexis Hall. 2022. A Lady for a Duke. London:
Piatkus, p. 213, with reference to to eavesdrop)
The first set of words above are typically classed as exocentric compounds,
but bahuvrihi compounds like redhead (‘one who has a red head – by virtue of
having red hair’) have not been included in this list. That is, there are several
 2 Background to the Study of Word-Formation 27

ways in which compounds might not denote a hyponym of their head element,
which often makes it look as though the elements of words are not meaningful.
Another source of apparently meaningless elements in English word-
formation is unique morphs. Unique morphs are morphs that occur in only
one place in the linguistic system. Vim used to be one such, occurring only in
vim and vigour, though modern usage allows it rather wider usage, kith is
another, appearing only in kith and kin.
They cheered and drank, knocking it back with vim. (Karen Swan. 2020. The
Hidden Beach. London: Macmillan, p. 107)

When unique morphs are elements in words it is typically because what was
once a recognizable element has, due to phonetic change or dialectal variation,
become unrecognizable, and has lost its meaning as an independent unit.
A classic example is cran in cranberry. Cran is etymologically a variant of
crane (the bird), though the meaning link is now lost, despite the closely related
spelling. Many linguists call unique morphs ‘cranberry morphs’ because of this
example, although other berry names might have provided better examples:
bilberry and whortleberry (dialectal names for the same plant and its fruit) now
contain unique morphs, and the link between whortleberry and hurtle and hurt
is completely lost. Whort, like rasp in raspberry, was once the name of the fruit,
without any berry element. Other examples are provided by cobweb, fenugreek,
lukewarm and dishevelled. The cob element in cobweb, also found in attercop
(now familiar from the fiction of J.R.R. Tolkien), is related to German Kopf
‘head’. With the berry names in particular, what the unique morph does is
distinguish one kind of berry from another. In this regard, such elements are no
different from the goose in gooseberry (which has nothing to do with geese)
and the locust in locustberry (which does not appear to have anything to do with
locusts): they are simply names which allow contrasting types to be named.
A cranberry, on this reading, is just something we can call a berry which is not
a strawberry, raspberry, gooseberry and so on. Finally, note the prefix in
midwife, which is not related etymologically to mid from middle, but to the
German preposition mit, Danish and Swedish med, ‘with’. The problem here is
common enough that some mechanism or set of mechanisms is required to
account for such instances, even if they are relevant only in a minority of words
overall.

2.8 Many of the Fundamental Principles of Word-Formation

Do Not Function
Bauer et al. (2013: 635–9) draw attention to the fact that many of the principles
which have been suggested as providing overall constraints in word-formation
do not work as they are supposed to, or at least, do not work all the time. In
28 Part I Basic Questions

some cases, it is a matter of just what the constraint is supposed to do, or

a matter of how the constraint (or its field of application) is to be defined, in
other cases, the proposed rule appears not to hold true. Some of these will be
discussed in later chapters, such as blocking (see Chapter 5) and the right-hand
head rule (see Chapter 13). Some can be discussed here briefly.
Botha (1984: 137) proposes the no-phrase constraint, phrased as “Lexical rules
do not apply to syntactic phrases to form morphologically complex words”. It
has been pointed out by many that the name of this constraint appears to disprove
the constraint, as do many phrasal compounds such as There is a sort of Oh-what-
a-wicked-world-this-is-and-how-I-wish-I-could-do-something-to-make-it-better-
and-nobler expression about Montmorency (Jerome K. Jerome, Three Men in
a Boat, 1889). However, if such expressions are taken not to involve lexical rules
or not to create morphologically complex words it might be that the constraint
holds. The constraint certainly needs to be clarified.
Lieber (1981: 173) provides the repeated morph constraint, phrased as “No
word formation process . . . can apply to its own output”. This blocks sequences
of identical affixes, such as we might find in doer-upperer, mini-mini-dress, re-
reread, sub-subcontractor. We would also need to ask whether having
a compound as an element within another compound counts as a process
applying to its own output. At the same time, the constraint is not without
merit, since the exceptions are rare, and are usually pragmatically predictable.
Again, greater clarification is required.
Aronoff (1976: 48) provides the unitary base hypothesis, namely that “A
W[ord] F[ormation] R[ule] will never operate on either this or that” the bases it
can operate on are “always unique”. Presumably a unique base type could be,
for instance, a base which is either a noun or an adjective, since noun and
adjective share the feature [+N] in Chomsky’s (1970) classification of word
classes, and it could be, for example, a noun denoting a human being. Just what
is included and what is excluded as a possible unitary description may not be
entirely clear. However, the prefix un- in English can attach to verbs (undo), to
adjectives (unkind) and to nouns (unperson), and this would seem to be
excluded.
Corresponding, in some ways, to the unitary base hypothesis, we also have the
unitary output hypothesis (Scalise 1984: 137) which “does not allow a particular
phonological form to be considered a single affix if it produces outputs with
different category labels or different semantics”. That is, the suffix -er that is used
in smaller cannot be the same suffix as the -er that produces defender because
the -er in smaller creates adjectives, the -er that produces defender creates nouns,
and the meanings in the two cases are different. As far as this goes, it seems very
reasonable: we have to say we have two homophonous affixes -er1 and -er2,
which attach to different bases and have different meanings. But now consider
the cases with un- cited above. They do not produce categories, they just fail to
 2 Background to the Study of Word-Formation 29

change categories, but the un- in undo creates a verb with a reversative meaning,
the un- in unkind carries a negative meaning and the un- in unperson creates
a meaning which denies the suitability of the label used as a base (Bauer et al.
2013: 364–6). Are these differences sufficient for the constraint to be called into
force and for us to say that we must have three different prefixes? And if not
(which is the usual position taken by scholars) why not? Or, if we consider the
suffix -er in defender, that in diner (‘place where one can eat’) or that in scratcher
(‘type of lottery ticket’), do the different meanings of ‘agent’, ‘location’ and
‘patient’ mean that we should say that we have three suffixes here (and if not,
why not)? Finally consider the English prefix a- in words such as alike and
askew. Most of these words (though not all) can be adjectives (The twins are very
alike, Her hat was askew) or adverbs (Treat all your children alike, The picture
hung askew). Does that mean that we should say that we have a1- creating
adjectives and a2- creating adverbs or not? All these examples show that matters
are not necessarily simple.
There may also be other criteria which can be useful in determining whether
homophonous forms are a single affix or not (Bauer 2003: 146–52): whether they
show the same allomorphy, whether they are both/all available, whether they
potentiate the same subsequent affixation, and so on.
Aronoff (1976: 21) also introduces the word-based hypothesis, requiring
that “A new word is formed by applying a regular rule to a single already
existing word”. Again, this may be a definition of regular word-formation
processes, or it could be intended as a constraint on how word-formation
operates. If the latter, it probably fails, for instance with blends which, at
the time Aronoff was writing, were considered to be irredeemably irregu-
lar but have now been found to have a great deal of regularity in their
formation (see e.g. Lappe 2007). If it is a definition, it excludes some
things which others might wish to include under word-formation (see
Bauer 1980).
Beard (e.g. 1995) proposes a separation constraint, that is that the phono-
logical part of word-formation (e.g. affixation) should be kept separate from the
semantic part of word-formation. Such a position allows generalizations over
the semantics of nominalizations, for example, without having to restate them
separately for -ation, -ure, -ment and so on. I suspect that this has not gained
greater adherence because there is no generally accepted principle to allow its
implementation. However, Bauer (2019b) distinguishes between formal para-
digms and functional paradigms, which allows a similar consideration from
a different point of view.
The surprise with all these proposed principles/constraints/hypotheses is not
that there are so many of them, nor that they require further specification or
even replacement. The surprise is that there is not more discussion of them in
the literature, at least defending them as default expectations in word-
30 Part I Basic Questions

formation, and considering how they relate to what the linguist is supposed to
do with those parts of lexical creativity that they exclude.

2.9 Is Word-Formation Necessarily Morphological?

The general view of morphology is that it divides into three major areas:
compounding, derivation and inflection. These are the three main ways of
creating new words (in one sense or another), and morphology is distin-
guished from syntax because it deals with the internal structure of words
rather than with the ordering of words in sentences. Of these three branches,
word-formation is seen as a cover term for compounding and derivation,
while inflection is seen as more syntactic, in the sense that it marks the
function of words within the sentence. All of this, of course, can be ques-
tioned (see Section 2.4 on the division between inflection and derivation, for
instance). One of the reasons why it might be questioned is that there are
borderline issues between the three areas mentioned here. Sometimes the
borderline issues are concerned with diachronic change. For instance, the
derivational suffix -dom in kingdom stems historically from a word corres-
ponding to modern doom so that a compound has become a derivative, and the
same is true of the suffix -ric in bishopric, which is etymologically related to
the German word Reich ‘empire’. The other way round, the increasing
modern use of ism as a word in its own right implies that words like eclecti-
cism, usually seen as derivatives, could now be seen as compounds. For the
borderline between inflection and derivation, there is dispute in the literature
as to whether the -ly that creates adverbs like thinly, wisely is inflectional or
derivational (or, indeed, whether the question has been misleadingly
formulated).
More generally, though, there are many things which might count as new
words which are not created by processes that are normally thought of as
morphological. These include acronyms like scuba (from Self-Contained
Underwater Breathing Apparatus) and laser (from Light Amplification by
Stimulated Emission of Radiation), blends like spork (from spoon + fork) and
Chunnel (from Channel + tunnel) – and for some authorities these two
examples illustrate different phenomena – and clipping as in phone (from
telephone), photo (from photograph) and flu (from influenza). Bauer (2019a)
presents an argument for seeing all these types, and others, as being part of
word-formation, but such a position is very controversial.
We also find things which might be considered words or might be considered
pieces of syntax. If artist in residence is a piece of syntax (even that is an open
question), then can lady-in-waiting be a word, or must it also be a syntactic
construction, and what about Middleton-in-Teesdale? Such questions extend to
 2 Background to the Study of Word-Formation 31

a much wider range of multiword expressions from How are you doing? and So
it goes to Rain cats and dogs.
Finally, we should consider figures of speech. When people started using the
form crown to mean not a head ornament but the government, a new way of
speaking was inaugurated. We might argue as to whether the two meanings of
crown are a matter of polysemy or homonymy, but crown can now be used with
verbs like decide as well as with expressions such as is made of. Is this word-
formation? The most general interpretation is that it is not, but lexicographers
list new meanings alongside new forms when looking at lexical innovation.
Such questions can be resolved by stipulation, but it would be preferable to
have some motivated position for including or excluding the various types and
circumscribing the area of word-formation (see Chapter 7).

2.10 Unpredictability
Wherever we look in word-formation, it seems that there is a great deal of
unpredictability. Rules work well in describing some of the formations we
wish to term instances of word-formation, but not all. In some cases, the
meaning of elements is directly relevant to the meaning of the formed word,
but again not in all. Some constraints seem to work well much of the time, but
not all of the time. A dishwasher can be a person or a machine, but a car
washer does not seem to be institutionalized as either. Even if we are familiar
with individual letters from a word being used to make a new word (if it is,
then, a new word), the expression on the QT (from on the quiet and meaning
‘secretly’) is an unusual type of formation. At every stage, we have to ask
whether the unexpected is just the way that word-formation works or whether
it means that certain formations cannot really be considered as a part of word-
formation; we have to ask whether the unpredictable formation should be
ignored or taken to be part of the range of constructions that word-formation
deals with; we have to ask whether our ways of building such material into
a grammar have to be flexible enough to allow for the diversity we find. The
more we examine the detail, the more the apparently odd turns out to be more
predictable that we thought (see Lappe 2007 for an excellent case in point). So
there may be reasons for attempting to extend our field of study beyond the
blandly regular. But at some point we seem to hit the irredeemably unpredict-
able, and we have to decide what to do with that when we meet it.

2.11 Conclusion
What counts as word-formation is controversial, how word-formation is to be
described is controversial, whether word-formation exists in a different module
of the grammar from syntax is controversial, and how data is to be collected to
32 Part I Basic Questions

prove productivity is controversial. The inevitable outcome is that anything

that is done to describe word-formation will be controversial in some way.
Most of the questions that have been raised here are matters of choice for the
researcher, which should ideally be clarified (justified to the extent that is
possible) as a preliminary to work in the field. The only place where I think
that a decision is really central to the topic, is that I believe that the distinction
between word-formation and etymology should be drawn on the boundaries of
productivity, even though such a position brings with it serious problems of
determining what is productive, particularly in borderline cases.

Challenge
Make a case either supporting the division made here between synchronic
morphology and etymology or rejecting it. If you reject the difference in the
way that it is sketched here, how do morphology and etymology fit into your
view of how we should model word-formation? Do you have an alternative
way of delimiting synchronic morphology (or, if not, what kinds of argument
might be used to support such a position)?

References
Aikhenvald, Alexandra Y., R.M.W. Dixon & Nathan M. White. (2020). The essence of
‘word’. In Alexandra Y. Aikhenvald, R.M.W. Dixon & Nathan M. White (eds.),
Phonological Word and Grammatical Word. Oxford: Oxford University Press, 1–24.
Anderson, Stephen R. (1992). A-morphous Morphology. Cambridge: Cambridge
University Press.
Aronoff, Mark. (1976). Word Formation in Generative Grammar. Cambridge, MA:
MIT Press.
Baker, Anne & Kees Hengeveld. (2012). Linguistics. Malden, MA: Wiley.
Bauer, Laurie. (1980). In the beginning was the word. Te Reo 23, 73–80.
(1983). English Word-Formation. Cambridge: Cambridge University Press.
(2000). Word. In G. Booij, C. Lehmann & J. Mugdan (eds.), Morphology: An
International Handbook of Inflection and Word-Formation. Berlin: de Gruyter,
247–57.
(2001). Morphological Productivity. Cambridge: Cambridge University Press.
(2003). Introducing Linguistic Morphology. 2nd ed. Edinburgh: Edinburgh University
Press.
(2019a). Rethinking Morphology. Edinburgh: Edinburgh University Press.
(2019b). Notions of paradigm and their value in word-formation. Word Structure 12,
153–75.
Bauer, Laurie & Winifred Bauer. (2012). The inflection–derivation divide in Māori and
its implications. Te Reo 55, 3–24.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
 2 Background to the Study of Word-Formation 33

Beard, Robert. (1995). Lexeme-Morpheme Base Morphology. Albany: State University

of New York Press.
Booij, Geert. (1996). Inherent versus contextual morphology inflection and the split
morphology hypothesis. Yearbook of Morphology 1995, 1–16.
Botha, Rudolf P. (1984). Morphological Mechanisms. Oxford: Pergamon.
Bybee, Joan. (2023). What is usage-based linguistics? In Manuel Díaz-Campos &
Sonia Balasch (eds.), The Handbook of Usage-Based Linguistics. Hoboken, NJ:
Wiley, 9–29.
Cannon, Garland. (1987). Historical Change and English Word-Formation. New York:
Lang.
Davies, Mark. (2004–). British National Corpus (from Oxford University Press).
Available online at www.english-corpora.org/bnc/.
Dixon, R.M.W. (2014). Making New Words. Oxford: Oxford University Press.
Dixon, R.M.W. & Alexandra Y. Aikhenvald. (2002). Word: A typological framework.
In Robert M.W. Dixon & Alexandra Y. Aikhenvald (eds.), Word: A Cross-
Linguistic Typology. Cambridge: Cambridge University Press, 1–40.
Dressler, Wolfgang U. (2000). Extragrammatical vs. marginal morphology. In
Ursula Doleschal & Anna M. Thornton (eds.), Extragrammatical and Marginal
Morphology. Munich: Lincom, 1–10.
Firth, J.R. ([1930] 1965). Speech. In The Tongues of Men and Speech. Oxford: Oxford
University Press, 139–211.
Halle, Morris & Alec Marantz. (1993). Distributed morphology and the pieces of
inflection. In Kenneth Hale & Samuel J. Keyser (eds.), View from Building 20,
Cambridge, MA: MIT Press, 111–76.
Hippisley, Andrew. (2015). The word as a universal category. In John R. Taylor
(ed.), The Oxford Handbook of the Word. Oxford: Oxford University Press,
246–69.
Hooper, Joan B. (1976). An Introduction to Natural Generative Phonology. New York:
Academic Press.
Jackendoff, Ray. (1975). Morphological and semantic regularities in the lexicon.
Language 51, 639–71.
Lappe, Sabine. (2007). English Prosodic Morphology. Dordrecht: Springer.
Lieber, Rochelle. (1981). On the Organization of the Lexicon. Bloomington: Indiana
University Linguistics Club.
(2022). Introducing Morphology. 3rd ed. Cambridge: Cambridge University Press.
Marchand, Hans. (1969). The Categories and Types of Present-Day English Word-
Formation. 2nd ed. Munich: Beck.
Marr, Vivian (ed.). (2008). The Chambers Dictionary. 11th ed. Edinburgh: Chambers
Harrap.
OED. Oxford English Dictionary [online]. oed.com.
Perlmutter, David M. (1988). The split morphology hypothesis: Evidence from
Yiddish. In Michael Hammond & Michael Noonan (eds.), Theoretical
Morphology: Approaches in Modern Linguistics. San Diego: Academic,
79–100.
Plag, Ingo. (1999). Morphological Productivity. Berlin: Mouton de Gruyter.
Plank, Frans. (1994). Inflection and derivation. In Ron E. Asher (ed.), The Encyclopedia
of Language and Linguistics. Oxford: Pergamon, 1671–8.
34 Part I Basic Questions

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. (1972).
A Grammar of Contemporary English. London: Longman.
Scalise, Sergio. (1984). Generative Morphology. Dordrecht: Foris.
Stemberger, Joseph & Brian MacWhinney. (1986). Frequency and the lexical storage of
regularly inflected forms. Memory & Cognition 14, 17–26.
Vannest, Jennifer, Thad A. Polk & Richard L. Lewis. (2005). Dual-route processing of
complex words: New fMRI evidence from derivational suffixation. Cognitive,
Affective, & Behavioral Neuroscience 5, 67–76.
Wray, Alison. (2015). Why are we so sure we know what a word is? In John R. Taylor
(ed.), The Oxford Handbook of the Word. Oxford: Oxford University Press,
725–50.
Zwicky, Arnold M. & Geoffrey K. Pullum. (1987). Plain morphology and expressive
morphology. Proceedings of the Thirteenth Annual Meeting of the Berkeley
Linguistics Society, 330–4.
3 Reflections on Why We Need
Word-Formation

As far as I know, there is no language that does not have word-formation. Since that
implies that there is no language without words, and that there is no language
which does not have specific ways of constructing words as opposed to sentences,
this might seem odd. Why would all languages have words if they could use
morphemes (or minimal signs, or standard formatives – whatever terminology the
reader is easy with) instead? Why complicate matters? It is sometimes suggested
that isolating languages do not have words (Hockett 1944: 255, cited in Dixon and
Aikhenvald 2002: 3, argues that “there are no words in Chinese”, although such
a claim was controversial at the time and is probably even more controversial
today – Dixon and Aikhenvald 2002: 32–3). Note, though, that even in a language
like Chinese (Mandarin) we can have a sequence of meaningful elements like
huā mù ‘flower tree’ which means ‘vegetation’, something that we would probably
call a ‘word’. In other languages, things which have some of the features we think
of as belonging to words look rather more like syntactic structures to those who are
mostly familiar with Indo-European languages, as in the following example, where
the decimal point divides the meaningful elements of the whole from each other.
West Greenlandic
puu·ssa·qar·ti·nngil·ara
bag·future·have·causative·negative·1sg.3sg.indicative
‘I have no bag for it’
Even more problematic is that we have no good definition of a word, so that it
is hard to determine precisely how we recognize words when we meet them
(Bauer 2000, Dixon and Aikhenvald 2002, Hippisley 2015).
In some places, it looks as though we might be able to function quite well
without word-formation.
you can’t arrest them on cop-ly intuition? (Robert B. Parker. 2003. Stone
Cold. London: Murray, p. 229)
Judges hate to issue them on cop intuition. (same volume, p. 240)

35
36 Part I Basic Questions

In the first example, cop-ly is a case of word-formation (it creates an

adjective from a noun using a suffix). In the second example, we find the
noun used unchanged. In this example, we might argue that we have, despite
that, used word-formation, because cop intuition is a new compound, but we
could equally argue that cop is just being used in attributive position as
a syntactic structure, and no word-formation has taken place. Furthermore,
had Parker written that something could not happen on a cop’s intuition, the
meaning would be the same, but there would be no question that we were
dealing with syntax. The point, though, is not to argue about which solution is
the best one in terms of style, communicative effectiveness or grammar, but to
point out that even if we think that the purpose of word-formation is to do one
or both of causing an old word to be seen as belonging to a new word-class
(transposition) or to expand our vocabulary, we could invent a syntax which
would do that without having to make appeal to word-formation as a separate
part of the grammar. A similar conclusion is suggested by the next example.
A utilitarian concrete-and-glass building. (Matthew Palmer. 2015. Secrets of
State. New York: Putnam, p. 40)

Concrete-and-glass is a syntactic structure, while concrete-glass could have

been used as a clear piece of word-formation, or even conglass.
When we start looking, there are plenty of places where we have an appar-
ently syntactic alternative to a process of word-formation. This is particularly
noticeable with compounds, where there are several syntactic constructions
which, in some cases, can cover the same content as a compound. This is most
clearly the case when the two expressions exist and are synonymous or nearly
synonymous or are closely parallel. We find examples where an adjective is
used to modify a noun or where a genitive is used with a noun. Neoclassical
formations can also be used in similar places, but then the compound looks to
be the more syntactic of the two patterns.

Adjective
atom bomb atomic bomb
car mechanic automotive engineer
cow pox bovine encephalitis
dog-tooth (only in metaphorical canine tooth
contexts)
home office domestic economy
Genitive
cowlick cow’s lick
lady-smock lady’s smock
milk thistle lady’s thistle
Neoclassical
compound
dictionary-making lexicography
life sciences biology
woman-hating misogyny
 3 Why We Need Word-Formation 37

The fundamental principle underlying all of this is one that is usually known
as Zipf’s law. Zipf actually has several laws, and the one in question here is the
law of abbreviation (Zipf [1949] 1965: 38), formulated as “the length of a word
tends to bear an inverse relationship to its relative frequency”, that is long
words tend to be rare, short words tend to be common. If we look away from the
apparent circularity of Zipf’s use of the term ‘word’ here, and say something
like ‘meaning-bearing construct’, we can see that words are likely to be more
frequent than syntactic paraphrases, and are thus better suited to being rela-
tively short, if they are used in a repetitive manner. Of course, not all words are
likely to be used in a repetitive manner, and then length per se does not matter.
William James. He was the groundsman, handyman, if-there’s-any-sort-of-
difficulty-ask-William-and-he’ll-fix-it-for-you person about the place.
(Laurence Meynell. 1978. Papersnake. London: Macmillan, p. 10)
Words can be made shorter than typical syntactic phrases in several ways:
• restrict the amount of modification that is possible;
• compress the syntax by omitting redundant material;
• have processes for making words create (in most cases) shorter strings than
those which create syntactic structures.
Typically, established compounds do not allow independent modification of the
elements. That is, if we take blackbird to be a compound, we cannot have *rather
blackbird or *black omnivorous bird if we are to retain the meaning of ‘blackbird’.
For many, this is part of the definition of a compound. This comes down to the
difference between naming and describing, where word-formation provides
a name for an entity, action or quality and syntax describes it. Words, of course,
can enter into syntactic descriptions, so that an unusually speckled blackbird is
a grammatically appropriate form which does not break the spirit of the restriction.
We find compression in the omission of grammatical words such as articles
and prepositions, for example in the difference between a book-cover (word-
formation) and the cover of a book (syntax). The use of figurative expressions can
also provide compression, since, for example, lollipop man omits a lot of detail in
comparison with man helping children cross the road who has a stick with a road
signal on it that looks like a lollipop. Similarly, so-called synthetic (or verbal-
nexus, or secondary) compounds like bus-driver omit prepositions and articles in
comparison with syntactic paraphrases such as the driver of the bus.
The use of affixes and not just words not only makes the output shorter, but
also binds the elements together more tightly because affixes cannot typically
stand alone. We also find that a process such as clipping is available to make
longer words shorter when they become more frequently used, as in flu for
influenza, phone for telephone, telly for television. Acronyms perform similar
function for longer names (e.g. for names of governmental bodies).
38 Part I Basic Questions

All of this is a matter of tendencies rather than a matter of absolutes. Green

belt looks, superficially at least, like a piece of syntax, even though it is the
name of an entity rather than or as much as a description of it. At the same time,
it does not allow modification of the adjective if the meaning is to be retained:
greener belt or rather green belt are possible if describing an article of clothing,
but not if talking about the green belt round a town. Find the lady looks like
syntax, but is the name for a card game, beloved of mountebanks, in which
people bet on which of three inverted playing cards is the queen.
Words recur more often than phrases do, so words need to be, other things
being equal, shorter than phrases. Processes used for making words can be used
to make long words, just as syntactic constructions can be adopted as labels, so
there is not a hard-and-fast dividing line. Nevertheless, word-formation pro-
cesses are typically aimed at making more compact expressions than syntactic
processes. Because of the requirements placed on words, those processes are
often different from those used in syntax, and because of the ubiquity of words
or word-like constructions, word-formation patterns are common, even if they
are not the only way of forming new words.

Challenge
Genitives have multiple functions in English, as well as in other languages. The
genitive in the women’s magazine seems to function more like the modifier in
a compound and less like a possessive marker than the genitive in the women’s
experience. Can you find criteria for distinguishing between the two? Do you
think the two differ in syntactic structure? Do you think that the type in the
women’s magazine represents a type of compound? Why (not)? Are there other
functions of the genitive? Does the genitive function the same way in any other
language with which you are familiar?

References
Bauer, Laurie. (2000). Word. In G. Booij, C. Lehmann & J. Mugdan (eds.),
Morphology: An International Handbook of Inflection and Word-Formation.
Berlin: de Gruyter, 247–57.
Dixon, Robert M.W. & Alexandra Y. Aikhenvald. (2002). Word: A typological frame-
work. In Robert M.W. Dixon & Alexandra Y. Aikhenvald (eds.), Word: A Cross-
Linguistic Typology. Cambridge: Cambridge University Press, 1–40.
Hippisley, Andrew. (2015). The word as a universal category. In John R. Taylor (ed.),
The Oxford Handbook of the Word. Oxford: Oxford University Press, 246–69.
Hockett, Charles F. (1944). Review of Linguistic Interludes and Morphology: The
Descriptive Analysis of Words by E.A. Nida. Language 20, 252–5.
Zipf, George K. ([1949] 1965). The Psycho-Biology of Language. Cambridge MA: MIT
Press.
4 Reflections on the Recognition of Novelty
in Words

Van Santen (1992: 63–74) points out that if we believe that morphological
processes can be productive, we must be able to recognize existing words in
order to see the productivity. Following Corbin (1987), many authorities
distinguish between two meanings of productivity: a process is available if it
can be used, and profitable to the extent that it is actually used. Van Santen
(1992: 72; my translation) says that “Productivity is manifested in the space
between the existing and the impossible”. This would seem to imply that
speakers can recognize individual words which they know and are aware of
words which are new in their experience.
Although there is clearly some truth to this – as will be shown just below –
it seems unlikely that it is true across the board. The places where it seems to
fail are those where the relevant process is particularly productive (profit-
able). Consider the suffix -ing. Although I know that I have heard the form
revolting (both as a part of a finite verb group and as an adjective), and I know
that because I have heard jokes which play on the possible ambiguity of
expressions such as Sire, the peasants are revolting, I would be in doubt as to
whether or not I had heard the word misreporting, although I feel rather more
secure in saying that misreporting as an adjective is not familiar. Consortable
is listed in Lehnert (1971), but not in Marr (2008), even though Marr (2008)
lists consort as a transitive verb, which means that consortable ought to be
possible. It does not seem familiar, but I cannot be sure that I have not met it
previously. Lehnert (1971) also lists composable, disposable, opposable,
proposable, supposable, transposable. Of these, I recognize disposable and
opposable, I find composable and proposable to be unfamiliar, and I am in
doubt as to the status of supposable and transposable. These are personal
reactions; others will react differently. The point is, though, that individuals
may not be able to draw a firm line between the item-familiar (Meys 1975)
and the new. This was also the experience I had when asking people questions
about the words discussed in Section 15.6. People found whether or not they
were familiar with a given word a reasonable question to be asked, and in
most cases had no problem in providing an answer. In some cases, though, it
proved difficult to decide.

39
40 Part I Basic Questions

The situation is so normal that in occurs frequently in fiction. There, it is

not the familiarity to which attention is drawn, but the instances where the
word mentioned is not familiar. There are a number of stock reactions, and
a number of stock ways of posing the question as to the word’s status.
Examples are given below.
She . . . “commonized” her accent, adopting the singsong tones of
Birmingham. (M.C. Beaton. 2004. Deadly Dance. New York: St Martin’s
Minotaur, p. 157)
and Devereux making fun of his juniority. Was there such word? If not, there
ought to be. (Carola Dunn. 2007. The Bloody Tower. New York: St Martin’s
Minotaur, p. 32)
“imagining your consternation”, Leo said.
“I am indeed consterned”, I said, not caring that it wasn’t a word. (Jack
Frederickson. 2013. The Dead Caller from Chicago. New York: Minotaur, p. 305)
They’re probably wondering where I’m going so they can roadblock me up
ahead. (Is roadblock a verb? It should be.) (James Patterson and David Ellis.
2013. Mistress. New York: Little and Brown, p. 377)
“He was a bit Gothy himself.”
“I’m not sure that Gothy is a word, actually.” (Stephen Leather. 2014.
Lastnight. London: Hodder & Stoughton, p. 205)
“We have been hard at work unraveling the unravel-able.”
“That’s not a word.” (Clive Custler with Grant Blackwood. 2010. Lost
Empire. London: Michael Joseph, p. 88)
Was he being politically incorrect by giving her a pass on the basis of her
height? Was he being . . . well, he wasn’t sure the word existed, but was he
being a sizeist? A heightist? (Lawrence Block. 2013. Hit Me. New York:
Mulholland, p. 32)
“He was a great believer in free spiritism among young people”,
“Free spiritism?” Campion snapped. “That’s not even a word, let alone
a philosophy.” (Mike Ripley and Margery Allingham. 2014. Mr Campion’s
Farewell. London: Severn House large print, p. 229)
“‘Selective patriation’? What in God’s name does that mean?” The prime
minister looked from one to the other until his gaze settled on Derek Farmer.
“Is it even a word?”
“If it wasn’t a word before, it is now, Geoff”, said Farmer. (Andy McNab.
2015. State of Emergency. London: Bantam, p. 68)
“He must have behaved in a most unvicarish fashion.”
“Is ‘unvicarish’ a word?”
“I suppose not”, Susan said. (Eloisa James. 2017. Seven Minutes in Heaven.
New York: Avon, p. 16)
 4 Recognition of Novelty in Words 41

I came by my Francophilism, to invent a word, honestly. (Harlan Coben.

2017. Don’t Let Go. New York: Dutton, p. 51)
“You look like a person who relishes solitude”, I said. “I pride myself on my
ability to spot a fellow Churchyardian. Forgive me if I’m wrong.”
“There’s no such word as ‘Churchyardian’”, Collier said.
“There is now”, I told him. “I’ve just made it up.” (Alan Bradley. 2019. The
Golden Tresses of the Dead. London: Orion, p. 159)
It does not matter that some of the coinages here are probably intended to be
jocular, the important thing here is that they are thought worthy of comment.
Some words are simply accepted, some are accepted but acknowledged to be
new and of these some are seen as fulfilling a legitimate need. The coinage is
seen as an invention, not a proper or real word, in some cases as invalid because
new, or as perfectly good word once it has been coined. What these examples
have in common – though there are many that do not show this feature – is that
they are overtly recognized as being innovative.
The everyday experience of such instances is the claim that something is not
a word in Scrabble or other word-games. The proof of something being a word,
in this sense, is that it is found in the dictionary that is being used for reference.
Given that dictionaries differ wildly in the number of words they list, this claim
is more a claim about the familiarity of a word than it is about its actual
existence in the language (whatever that means). For example, elect(or)ess is
listed in Marr (2008) but not in Thompson (1995). It does indicate that speakers
are aware of words which they expect to find (or not to find) in a dictionary, and
thus what is known and what is new.
More often than not, however, new words in texts pass unremarked upon and
unglossed (Renouf and Bauer 2001). We cannot assume that this means they are
not noticed, but it may mean that they are intended as jokes or that their meaning is
considered to be so self-explanatory in context that no special comment is required.
In short, van Santen’s observation is supported by the data, although there is
a fuzzy area where speakers are insecure, and that insecurity seems to be
precisely with words coined by the most productive processes. This might be
seen as a confirmation of Aronoff’s (e.g. Anshen and Aronoff 1997) position
that the most productive uses of the most productive morphological processes
are not listed. But even that requires some moderation. Productive processes
can give rise to item-familiar words (we know words like driver, killer, lover
formed by the very productive affix -er), but because of the productivity of such
processes, we cannot store all the possible outputs, and we do not know
whether they are familiar or not.
42 Part I Basic Questions

Challenge
Choose half-a-dozen words which are at least four syllables long at random from
a dictionary or vocabulary list. Check the relative frequencies of the words.
Think of a way to check whether people find these words familiar or not (you
might want to check what they know about each word – collocations, meanings,
domain of usage). How far does frequency correlate with familiarity? What other
factors are relevant? If your words are morphologically complex, how far do
your consultants deduce meanings of unfamiliar words from the meanings of
bases and affixes?

References
Anshen, Frank & Mark Aronoff. (1997). Morphology in real time. Yearbook of
Morphology 1996, 9–12.
Corbin, Danielle. (1987). Morphologie dérivationelle et structuration du lexique, 2 vols.
Tübingen: Niemeyer.
Lehnert, Martin. (1971). Reverse Dictionary of Present-Day English. Leipzig: VEB.
Marr, Vivian (ed.). (2008). The Chambers Dictionary. 11th ed. Edinburgh: Chambers
Harrap.
Meys, W.J. (1975). Compound Adjectives in English and the Ideal Speaker-Listener.
Amsterdam: North Holland.
Renouf, Antoinette & Laurie Bauer. (2001). Contextual clues to word-meaning.
International Journal of Corpus Linguistics 5, 231–58.
Santen, Ariane J. van. (1992). Produktiviteit in Taal en Taalgebruik. Unpublished
doctoral dissertation, University of Leiden.
Thompson, Della (ed.). (1995). The Concise Oxford Dictionary. 9th ed. Oxford: Oxford
University Press.
5 Reflections on Blocking and Competition

It makes such good sense that speakers of languages should abhor absolute
synonyms that it almost seems churlish to question the notion that a new
complex word should never mean precisely the same as an existing word.
Given that transmission is an automatically available (see Chapter 4 for the
term) nominalization from transmit, if we have transmittal it must have some
specialized meaning that speakers wish to distinguish from the sense of trans-
mission. On some occasions, the difference will be a matter of style or colloca-
tion or regional dialect rather than strictly a matter of sense, but the general idea
that absolute synonyms are not found seems to be a well-established principle.
Aronoff (1976) terms this ‘blocking’, and the term has become established,
even though Aronoff (2023: 52) admits that it is not a particularly good term for
the phenomenon, which does have other names, such as ‘pre-emption by
synonymy’ (Clark and Clark 1979) and ‘the rule of the occupied slot’ (Regel
der besetzten Stelle, Burgschmidt 1977: 43). Aronoff (1976: 43) defines block-
ing as the
non-occurrence of one form due to the simple existence of another
and this seems to imply that the synonym need not arise from the same base (as
in the transmit case), but can be more general so that, to use a widely cited
example (Bolinger 1975: 109), the existence of thief blocks the use of stealer
with the same meaning.
We can distinguish several types of blocking, which have sometimes been
treated separately in the literature.
1 Inflectional blocking: the presence of an irregular inflectional form blocks
the application of a regular inflectional form to the same base. This has been
argued to be an instance of the Elsewhere Condition, and alternatively has
been argued to be a result of Level Ordering. Some scholars seem to retain
the term ‘blocking’ only for derivational morphology, thus excluding such
cases, which might then be seen to be instances of suppletion.
2 Individual blocking or token blocking (Rainer 1988): the presence of one lexical
item blocks the coinage or institutionalization of one other in a non-systematic

43
44 Part I Basic Questions

manner. The thief example above might illustrate this, or the lack of a derived
verb from full because of the existence of fill. Aronoff (2023: 52) talks about
lexical blocking here, which he sees as the classic case of blocking.
3 Domain blocking or type blocking (Rainer 1988): the coinage of some new
lexeme is blocked by an overall constraint on the usage of a particular
morphological process. For example, the creation of verbs in -en on the
basis of adjectives is blocked where the adjective base does not end in an
obstruent, so that *bluen, *greenen, *souren are not possible (the process is
probably not available any longer, which might then be seen as the absolute
blocking of any more forms on this particular pattern).
In Chapter 29 several examples are given of blocking failing to apply to
inflectional forms. The question there is whether the places where blocking
fails can be systematized in some way, but as far as English is concerned there
does not seem to be any absolute rule – or if there is, it is yet to be spelled out
fully. More generally, the phenomenon is known as ‘overabundance’
(Thornton, e.g. 2011), and is much more widespread in a range of languages
than was realized in the 1980s. What we can say here is that inflectional
blocking does not work as well as was once believed.
We can also find cases where individual blocking fails. A simple example is
orientate, which has virtually replaced an earlier orient (in either case,
a verbalization of the noun orient). In this case there is an intermediate step:
orientation could be the nominalization of the verb orient or of the verb orientate
and orientate probably arose through back-formation from orientation.
A separate case that seems to have worked the other way round is compute,
replacing an earlier computate. Again, the nominalization computation is
ambiguous, but this time, the shorter verb appears to have triumphed. A verb in
-ate may also explain why exacerbation is now used in place of an earlier
exacerbescence. Regardless does not appear to have prevented the rise of
irregardless. Disfranchise has been replaced by disenfranchise. As with many
pairs of nominalizations from the same base, commission and commitment are
usually used with different meanings, although both have a range of meanings.
But dictionary definitions (e.g. Marr 2008) give both of them the meaning ‘act of
committing’, which suggests that there is at least overlap between them.
Blocking would seem to presuppose that there should not be overlapping.
The coexistence of approval and approbation is probably due to the different
style levels of the two words, approbation being far more formal, and this leads
to different collocations, so that approval rating is found but not approbation
rating. At an earlier stage, approbation and proof competed for the same place,
but are now semantically different.
There are numerous words which do not share a base and yet appear to share
at least some senses. Among others, we can think of abbreviate and shorten,
aggravate and exacerbate, surrender and relinquish. Others are relatively easy
 5 Blocking and Competition 45

to find in any thesaurus. Some of these differ in style level, and they often have
different collocations. Yet you can both surrender and relinquish territory to an
enemy. Environmental conditions can exacerbate or aggravate an illness, and
someone’s intervention can aggravate or exacerbate a situation.
Although Stratton (2023) does not focus on the existence of synonymy
per se, by considering changing patterns of synonymy for words meaning
‘man’ in the history of English, he illustrates a societal need for synonymy
and shows that synonyms can differ by text-type, as well as gaining differing
shades of meaning, and in some cases disappearing, over longer periods.
Bauer et al. (2013: 636) comment that many synonymous derived forms are
found in corpora. Some of their examples could well be the result of memory
lapses under the pressure of seeking words, which gives the appearance of
leading to new coinages which, however, are often not institutionalized (omit-
ment, for instance, in the place of the lexicalized omission), but others like
educationist and educationalist are listed in dictionaries as synonyms, and the
difference may be more a matter of fashion than of differing sense. This seems
to contradict the view from Bauer (2003: 81) that it is institutionalization that is
blocked rather than coinage.
We also find instances where affixes in competition appear to vary in the
speech of the community or of individuals without affecting the meanings.
Forms ending in -ance/-ence and -ancy/-ency and forms ending in -ic and -ical
provide multiple examples (even if there are examples where the endings are
used to create distinct meanings). Residence and residency may mean different
things, but complacence and complacency, luxuriance and luxuriancy, persist-
ence and persistency are synonymous pairs. Economic and economical may
differ in meaning, but episodic and episodical, logarithmic and logarithmical,
philologic and philological, strategic and strategical can all be synonymous
(Marr 2008). In many other cases (for both patterns) only one of the pair is in
general use or is in general use among professionals in the relevant area. For
instance, most linguists at the present time use phonetic, phonological, mor-
phological, semantic and syntactic, but non-linguists may vary in their usage.
Some authorities (e.g. Di Sciullo and Williams 1987) suggest that intelli-
genter is blocked by more intelligent, thus allowing syntactic constructions to
block morphological ones. There are various problems here, even if we agree
that intelligenter is unlikely to be used. The first is that we can argue that more
intelligent is periphrastic morphology rather than syntax (if there is a genuine
distinction to be drawn here), so that it is just morphology blocking morph-
ology. The second is that trisyllabic adjectives (and longer) do occur with -er
from time to time, although apart from curiouser and forms like unhappier they
do not usually become established, so that it is difficult to say that intelligenter
is actually not possible; we also occasionally find disyllabic adjectives which
usually reject -er being used with it.
46 Part I Basic Questions

You’re already rich and famous . . . and you’re going to be richer and
famouser. (Lawrence Block. 2003. Small Town. London: Orion, p. 335)
It’s getting mysteriouser. (Jeffery Deaver. 2003. The Vanished Man. London:
Hodder & Stoughton, p. 46)
I’ve been up for three hours, nervouser than a nun at a penguin shoot. (John
Sandford. 2017. Golden Prey. New York: Putnam, p. 19)
“A substantial sum?”
“The substantialler the better.” (Grace Burrowes. 2017. Too Scot to Handle.
New York: Forever, p. 156)
Finally, some of the people who claim that intelligenter is blocked also claim
that more smart is blocked by smarter, where the blocking goes in the other
direction. Actually, most monosyllabic adjectives can occur with more (for
some that is the only option: consider right). At least one website presents the
argument that smarter and more smart are not synonymous (https://allthediffer
ences.com/difference-between-more-smart-and-smarter/) and avoids the issue
in that way. But we can also find more smart as a straightforward comparative,
in a decade from now, the human capital will have grown enormously, with the same
amount of people, just much more smart than they were before. (www.accenture.com/
nl-en/blogs/insights/the-smart-workforce-amplifying-human-capital-by-making-the-s
martest-people-even-smarter accessed 8 July 2023)
The whole subject is fraught; blocking sometimes seems to work, but
sometimes it does not. When it does not, some of the reasons for its failure
are easy to explain, but, again, they are not always easy to explain. Some claims
about blocking are wrong, some of them may be considered to be tendencies
rather than fixed patterns of behaviour, but without more detailed analysis, we
probably have to conclude that blocking does not automatically apply. Plag
(2003: 63–8) suggests that more frequent forms are more likely to block
competitors than less frequent forms. While this makes sense, some of the
examples above suggest that even that rule does not always work. Certainly,
such a rule could explain how the form orient should end up being blocked by
a new form orientate after the event once orientate has become common, but it
cannot explain why the older, and at the time more frequent, form should not
block the innovative form. If this solution is to be turned into a theoretical
principle, a more subtle picture of just what is going on needs to be worked out.
Where different words with potentially the same meaning are created from
the same base, we talk of competition. Blocking, other things being equal,
should then lead to the prevention of competition. To the extent that this is not
what happens (as with -ence/-ancy examples cited above) it is another sign
that blocking has failed. Another example is the coexistence of both rigidify
and rigidize or rigidity and rigidness (Marr 2008). Even then, we have to ask
 5 Blocking and Competition 47

what it means for either of two (or more) processes to coexist. Do both forms
have to be institutionalized? Do both processes have to be equally usable in
every instance? Do the two have to be completely synonymous, or is it
sufficient for them to have overlapping meanings? Do they have to have the
same connotations, so are meaningless and meaning-free in competition if the
latter implies that it is a good thing to have no meaning? Are childish and
childlike equally usable if the former implies that resembling a child is bad
because of immaturity, while childlike implies that it is good because of the
implied innocence? And do the same implications have to hold in every case,
so that summerlike and summerish have to differ in the same way (at some
level) as childlike and childish?
Some of these questions are answerable in principle, others may not be. If we
take it that the meaning associated with words becomes more specific as the
words are used more in the community (see Section 10.4), then we can say that
meaning that is accrued through this process is not part of the competition but is
developed after the competition has applied (or has failed to apply). On the
other hand, one of the ways that competition can be valuable to the community
(as we have seen with several examples including transmission and transmit-
tal) is that it allows distinctive meanings to be shown by different forms, even
when the same base seems to be central to creating a relevant word. This might
have to be considered a different use of competition. The difference is one
between competition where two processes can apply, and which is used is
random and does not lead to a semantic distinction, and competition where
a competing form is used precisely because a semantic difference is required.
However, the situation that holds with -ic and -ical, with -ence and -ency
seems to give rise to a paradox. We can see that the two or more suffixes come
into English from borrowings which are motivated either in the donor language
or by coming from a different route from a common etymon, and we can thus
see that this gives rise to some words in English with one alternative and some
with another. But the reduplication of effort has to be redundant, so why is one
pattern not then removed in favour of the more frequent pattern, removing the
redundancy, as seems to be happening with the loss of orient, albeit slowly?
The answer would seem to be that the individual words become item-familiar
in English – at least once they reach an appropriate level of frequency or
familiarity – and that individual frequency maintains the form against challen-
gers from other patterns. But that gives precedence to the individual word over
the pattern. Yet the whole notion of productivity gives precedence to the pattern
over the individual word. This seems to mean that we can account for token
blocking if we consider single words, but type blocking only if we consider
productivity, which involves patterns (or rules). If that is true, either token
blocking and type blocking are two entirely distinct phenomena, or the two
approaches conflict with each other. Token-blocking, in principle, cannot occur
48 Part I Basic Questions

with productivity, because there is no pattern or rule which allows the blocked
words to be coined. Type blocking cannot occur if precedence is given to the
individual words, because each case has to be considered sui generis and
cannot be explained by an overall rule or pattern. However we choose to
proceed, we should avoid giving conflicting phenomena the same label.
We can get a different view of competition if we look at matters from an
onomasiological point of view (Štekauer 2005, Grzega 2009) where the speaker
seeks a word to fill a gap in their vocabulary rather than seeking a new word which
is compatible with but extends the rules or patterns. In this view, presumably,
almost anything can be in competition with anything else. English has multiple
expressions to mean ‘no longer alive’, including at rest (a prepositional phrase),
deceased (a past participle), stone dead (a compound), dead as a doornail (an
idiomatic phrase) and dozens of others, including many figurative usages, cir-
cumlocutions, fixed syntactic phrases and so on. Even though these are not all of
an equivalent style level, they make the point that this particular slot in our
vocabularies can be filled by many forms. Not only is there choice in the pattern
of word-formation that is used (when one is used), there is choice in the particular
lexeme that is used to head the chosen expression or from which to derive the
expression. Notably, there are alternatives to the use of word-formation, and if we
look at lists of near synonyms in a thesaurus, we might consider that word-
formation is a minor way of filling such gaps.
I have left more questions unanswered here than I have solved. The notion of
blocking seems to apply to moderate the productivity of patterns which are in
competition. A more subtle analysis might conclude that the two are not
related, depending on how we define competition and the boundaries of
competition, and on just how we see blocking as working. It seems clear that
a more restrictive view of blocking is required if it is to be really useful for
scholars of word-formation.

Challenge
Find several verbs which have more than one nominalization listed in diction-
aries. Find as many examples of the use of each nominalization as you can. Do
the uses of the nominalizations of every verb overlap, or are the meanings
distinct for each nominalization? Do your findings confirm the conclusions
found in dictionaries or not?

References
Aronoff, Mark. (1976). Word Formation in Generative Grammar. Cambridge, MA:
MIT Press.
(2023). Three ways of looking at morphological rivalry. Word Structure 16, 49–62.
 5 Blocking and Competition 49

Bauer, Laurie. (2003). Introducing Linguistic Morphology. 2nd ed. Edinburgh:

Edinburgh University Press.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Bolinger, Dwight. (1975). Aspects of Language. 2nd ed. New York: Harper, Brace,
Jovanovich.
Burgschmidt, Ernst. (1977). Strukturierung, Norm und Produktivität in der
Wortbildung. In Herbert E. Brekle & Dieter Kastovsky (eds.), Perspektiven der
Wortbildungsforschungen. Bonn: Bouvier, 39–47.
Clark, Eve V. & Herbert H. Clark. (1979). When nouns surface as verbs. Language 55,
767–811.
Di Sciullo, Anna Maria & Edwin Williams. (1987). On the Definition of Word.
Cambridge, MA: MIT Press.
Grzega, Joachim. (2009). Compounding from an onomasiological perspective. In
Rochelle Lieber & Pavol Štekauer (eds.), The Oxford Handbook of
Compounding. Oxford: Oxford University Press, 217–32.
Marr, Vivian (ed.). (2008). The Chambers Dictionary. 11th ed. Edinburgh: Chambers
Harrap.
Plag, Ingo. (2003). Word-Formation in English. Cambridge: Cambridge University
Press.
Rainer, Franz. (1988). Towards a theory of blocking: The case of Italian and German
quality nouns. Yearbook of Morphology 1988, 155–85.
Štekauer, Pavol. (2005). Onomasiological approach to word-formation. In Pavol Štekauer
& Rochelle Lieber (eds.), Handbook of Word-Formation. Dordrecht: Springer,
207–32.
Stratton, James M. (2023). Where did wer go? Lexical variation and change in third
person male adult noun referents in Old and Middle English. Language Variation
and Change 35: 199–221.
Thornton, Anna M. (2011). Overabundance (multiple forms realizing the same cell):
A non-canonical phenomenon in Italian verb morphology. In Martin Maiden, John
Charles Smith, Maria Goldbach & Marc-Olivier Hinzelin (eds.), Morphological
Autonomy: Perspectives from Romance Inflectional Morphology. Oxford: Oxford
University Press, 358–81.
6 Reflections on Potential and Norm

Although morphologists have spent a lot of time discussing actual and potential
words, lexicalization and productivity, there is one gap that has not been dealt
with well, namely the notion of norm. The idea of the norm was introduced by
Coseriu (1952) for what people actually say as opposed to all the things that it
would be grammatical for them to say. A similar idea, though without the
terminology, is discussed by Pawley and Syder (1983) in a rather different
context.
To illustrate the point, consider how you might say goodbye to a friend to
whom you are speaking English. You have many choices, depending on who
you are talking to, how friendly you are, where in the world you are, how old
you are and your friend is, whether you are serious or joking and so on. Such
possible expressions include things like goodbye, good day, so long, farewell,
laters, see you (at Christmas), ciao and sayonara. But although it would be
grammatical to say until we see each other again, you would not say it. The
equivalent of until we see each other again would be said in German, Italian or
Russian, but in English it is not part of the norm (Bauer 2022).
When we discuss word-formation we often discuss the patterns in terms of
their productivity (see Section 2.5). We know that we can create compounds
of the form N + N, so we might expect that we can add the noun muscle to the
noun dystrophy to give muscle dystrophy. While that might be a potential
word of English (if we assume that blocking will not prevent it – see
Chapter 5), it is not what we say, because we use an adjective to give muscular
dystrophy. On the other hand, if we know that we can use muscle (and because
of that, the adjective muscular) to mean ‘strong, although without literal
muscles’, we might expect to be able to say muscular car, when the expres-
sion that is part of the norm is muscle car. In a similar vein, alongside burial
we might expect *marrial (where we find marriage), alongside laughter we
might expect *coughter (where we find cough), and alongside adventure we
might expect *inventure (rather than the invention that we find). What this
means is that, besides knowing what is grammatical in a language, we have to
know what forms among the grammatical possibilities are actually used, just

50
 6 Potential and Norm 51

because that is the way we say it. It is not clear how we built that into a model
of how word-formation works.
A fairly standard picture of how word-formation has been assumed to work for
many years now runs as follows. We identify a gap in our vocabulary. To fill it,
we identify a word-formation process which can fill the gap (it is not clear how
that is done). The process must be productive in order to be able to fill a gap, and
the word which we are attempting to coin to fill the gap is then a potential word.
This implies that we must have a base or bases for creating a suitable word as
well as a suitable word-formation process. Again, it is not clear how that potential
base is chosen, since it may, for example, involve some figurative use of
language. If the potential word is not blocked (to the extent that blocking
works, see Chapter 5, and to the extent that blocking can work at the coining
stage rather than at the stage of institutionalization – Bauer 2003: 81), the
potential word is then taken as an actual word. That actual word may or may
not become established in the community for expressing the content that we
could previously not express in a word. It may, but need not, later become
lexicalized, or just vanish from use.
As is suggested here, there are gaps in our understanding of the process
which, whatever model of word-formation we may be using, tend to focus on
the formal application of the process and the shape of the output rather than on
the process of choosing between alternatives. But the focus here is how we can
move from a potential word to an actual word and from there to something
which is part of the norm, which are psycho-social questions rather than formal
questions.
We can begin with potential words. If we accept the notion that there exists
a set of bases for creating words and a set of rules/processes/patterns by which
words are formed, and furthermore we accept that all that is required for a word
to be formed is a suitable base and a process that can apply to that base, then the
notion that there are some words which are possible but are not observed is
inevitable. Indeed, for Aronoff (1976), it is the job of a generative morphology
to determine just what the set of all possible words is, whether they are attested
or not. Rainer (2012) divides those words into potential words and virtual
words, those which are prevented from surfacing as actual words because of
blocking. We have seen, though (Chapter 5), that sometimes blocking appears
to fail, so that we cannot be sure just which words fit into which category if we
accept that distinction. It seems safer to ignore the distinction.
At the same time, experience suggests that not all words which fit extant
patterns can be formed. Some words are deemed aesthetically impossible,
which may or may not prevent their coinage. Adams (1973: 2) quotes such an
opinion of the word aviation from 1909 which clearly neither prevented its
formation nor its institutionalization; on the other hand, ignoration (a nom-
inalization from ignore) is listed in the OED, but is not in usage for reasons
52 Part I Basic Questions

which are obscure but which might include aesthetic ones. Although we have
verbs from summer and winter meaning ‘to spend the summer/winter’ as in
They summered in a cottage at the seaside, there is no equivalent verb to
autumn. But this could be a potential word, although it is not clear whether the
verb in They Apriled in a cottage by the seaside is even a potential word. Was
the attested mouthfulness ever a potential word before it was produced? Or
sniggeruity (although this is clearly a joke)? Or charismability, whose mean-
ing is not clear?
I wished I’d brought my apples. All I needed was to chew on something with
mouthfulness. (Ann Prospero. 2000. Almost Night. London: Penguin, p. 143)
Rincewind was pretty sure horses couldn’t snigger, but this one radiated an air
of sniggeruity. (Terry Pratchett. 1998. The Lost Continent. New York:
Harpertorch, p. 199)
Kyla . . . whose number of Vanity Fair covers alone testifies to her charism-
ability. (Richard E. Grant. 1998. By Design. London: Picador, p. 92)

We appear to have several problems here: we cannot tell whether something

is or is not a potential word, some words are coined which were not potential
words, we have words that have parallels and yet are not coined and we cannot
tell whether they were never potential words (and why) or whether they remain
potential words, and with some potential words that are coined we sometimes
seem to have a meaning attached to them which is not predicted to occur.
One of the most frequent sources of words which do not appear to be
potential words created by processes of word-formation is the apparently
random deletion of phonological material from a much longer expression
which is often used in the names of chemicals and the like. Barnhart at al.
(1990) give many examples of the phenomenon, including lysostaphin (from
LYSO-dissolution+STAPHylococci+IN), pronethalol (from PROpyl+amiNE
+meTHyl+naphthALeme+methanOL) and ras (from RAt+viruS). The main
constraint at work here seems to be having a pronounceable output, perhaps
with some material from as many of the major elements of the original as
possible.
Part of the difficulty here is the question of what makes a particular
expression a potential word. If we assume something like a word-formation
rule, this seems to imply that anything which the set of rules permits to be an
output, and which is not already a known word, is a potential word.
Furthermore, it seems to imply that anything which is not permitted by the
rules is not a potential word. When we have a large enough corpus to consider,
however, we appear to find things that are not permitted by the rules, but
which nevertheless appear. If all such cases fail to become established, there
is little problem, although there may be a philosophical problem of how
 6 Potential and Norm 53

things that are not licensed can be generated in a model that uses rules. If this
is true, then mouthfulness is a potential word (which, in fact becomes estab-
lished, as in the following example).
It could be described as mouthfulness [. . . which] includes not only assisting
in enhancing the intensity of the five basic tastes, but also enhancing the edge
or peripheral flavours of the base flavours (www.foodnavigator.com/News/
Promotional-Features/Angeotide-delivers-a-superior-mouthfulness-that-enh
ances-taste accessed 11 August 2023)
Sniggeruity, in contrast, is not attested elsewhere in an internet search.
Examples like autumn (verb) and charismability cited above can be found
with an internet search, but are very rare. Yet inbearable, found by Bauer et al.
(2013: 361), and surely not generally accepted because of the allomorph of in-
before <b>, is found in an internet search, but usually, though not always,
corrected to unbearable.
The problem with all this is that even if we prefer analogy (see Chapter 8)
or paradigmatic structure to rules (or an equivalent), the same problems arise.
The notion of potential word seems to be rather more slippery than we might
expect.
When it comes to the notion of norm, things are even more difficult. There
is a link with lexicalization, in that the more a word is used, the more likely it
is to become lexicalized, and the more likely it is to be part of the norm, but
some words seem to become norms in their specialist areas quite quickly, if
for no other reason than that there is no alternative name available. There is
a distinction, though, in that lexicalization is a matter of more or less, while
being part of the norm seems to be a matter of yes or no, although the size of
the community that recognizes a norm may vary considerably, and the norm
may change over time. A recent expression which seems to have become part
of the norm in Scotland (I am not aware of it being used more widely, but it
may be) is the use of Scooby for ‘clue’. The phrase comes from rhyming
slang, with the full version being Scooby-Doo (a cartoon character), a name
which rhymes with clue.
we don’t have a Scooby where she’s staying. (Val McDermid. 2014. The
Skeleton Road. London: Little, Brown, p. 145)
It may be the case that attested words are automatically assumed to be part
of the norm, but that can change very quickly. In New Zealand, the
Australasian word stoush ‘disagreement, fight’ is no longer recognized by
young speakers, although it is still used in newspapers as a useful way of
saving space. A New Zealand cartoon had a dejected-looking man saying, “I
was going to shoot through, but nobody knew what it meant any more” (shoot
through means ‘to move on, to leave’). These examples indicate that the norm
54 Part I Basic Questions

is partly (perhaps largely) a matter of frequent usage, part of the ebb and flow
of words as any language changes. This means that it is a separate phenom-
enon from others typically used in word-formation, but perhaps necessary for
the diachronic study of word-formation, just as for the diachronic study of all
vocabulary usage.
Certainly, we should not underestimate the importance of norm on the
complex words that we are likely to meet. There is a certain amount of variation
between speakers, or between dialect areas and so on, but for me at least,
cooker is an instrument and cook is a person, while catcher is a person and
catch (on a door) is an instrument. If they happened to be the other way round,
we must assume that the grammar would be just the same as it now is. I say
accountancy but inheritance and not inheritancy. Although there has been
variation in the past, most linguists today use syntactic but phonological.
This may not be a matter of word-formation, but what the norm happens to
be can influence the productivity of different patterns, and at that point word-
formation is involved.

Challenge
Make a list of words in -ce or -cy which you (as an individual or as a group)
use, including those where you use both (with the same or different mean-
ings). Then check in a dictionary to find which member(s) of the pair are
listed. Is there any reason for the discrepancies? If you look in the Oxford
English Dictionary, do the two show different periods of use? Do rhyming
bases tend to show the same pattern? Are there adjectives in -ant that have no
corresponding -ance or -ancy form? If you had to create a corresponding
noun, what would it be? Can you tell why? If you prefer, you can try the same
exercise with -ic and -ical. If you try both, do you get parallel results in the
two cases?

References
Adams, Valerie. (1973). An Introduction to Modern English Word-Formation. London:
Longman.
Aronoff, Mark. (1976). Word Formation in Generative Grammar. Cambridge, MA:
MIT Press.
Barnhart, Robert K., Sol Steinmetz & Clarence L. Barnhart. (1990). Third Barnhart
Dictionary of New English. New York: H.W. Wilson.
Bauer, Laurie. (2003). Introducing Linguistic Morphology. 2nd ed. Edinburgh:
Edinburgh University Press.
(2022). What you must say, what you can say and what you do not say. In
Andreea S. Calude & Laurie Bauer (eds.), Mysteries of English Grammar.
New York: Routledge, 11–20.
 6 Potential and Norm 55

Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Coseriu, Eugenio. (1952). Sistema, norm y habla. Montevideo: Facultad de
Humanidades y Ciencias, Instituto de Filología, Departamento de Lingüistica.
OED. The Oxford English Dictionary [online]. oed.com
Pawley, Andrew & Frances Hodgetts Syder. (1983). Two puzzles for linguistic theory:
Nativelike selection and nativelike fluency. In Jack C. Richards & Richard W. Schmidt
(eds.), Language and Communication. London: Routledge, 191–225.
Rainer, Franz. (2012). Morphological metaphysics: Virtual, potential, and actual words.
Word Structure 5, 165–82.
7 Reflections on Definition by Stipulation
and on Word-Classes

In an ideal world, our classes of entities would be easily classified, unambiguous

and non-intersecting. We do not live in an ideal world. For instance, consider the
class of mammals. If we look round the farmyard in Europe, we might conclude
that mammals are creatures which give birth to live young, which feed their
young on their milk, the young sucking on a nipple, and which are covered in
hair. We exclude ducks and hens because they have feathers, not hair and because
their young are not born live but are hatched from eggs, we exclude fish because
they have scales and not hair, and so on. But if we move from Europe to
Australia, we meet monotremes (the platypus and the echidna) which have
hair, but which lay eggs, which feed their young from their milk, but which do
not have nipples. Our criteria no longer match up, and we have to create new
categories, defined in different ways. In this case, the reanalysis is possible if we
divide mammals into placental mammals and monotremes. The mammals of the
European farmyard may be prototypical mammals (in the sense of Rosch 1973
and Taylor 2003), but the criteria we used to define them are not, in the light of
further knowledge, accurate in defining mammals as a natural kind.
If we look at citrus fruit, we may recognize an orange and a lemon, say, and may
be familiar with other terms such as grapefruit, kumquat, lime, mandarin, tanger-
ine, tangelo and ugli fruit, but we may not have any criteria that we can success-
fully use to distinguish between these entities. Colour is a poor guide because
lemons and oranges may be green before they are ripe. Size is awkward because
although a mandarin may usually be smaller than an orange, it is possible to have
small oranges and large mandarins. Shape is difficult to classify. If we are telling
a small child what the difference is, we may resort to definition by ostentation: this
fruit in my hand is a mandarin. Even the labels we use may be misleading: in
French, a lime may be called a citron vert ‘green lemon’, a wild lime is not a lime,
and neither is a Spanish lime; a grapefruit has nothing to do with a grape.
In the case of mammals and citrus fruit, genetics can be used to keep track of
the various subspecies. In linguistics, though, we have no such fallback.
A linguistic category such as sentence has no genetic definition, it has to be
defined, typically in terms of its form and its function within a linguistic

56
 7 Definition by Stipulation and Word-Classes 57

system. Functional and formal criteria do not always align. This has implica-
tions for the terminology of linguistics and the way the terminology is used.
Consider the definition of the linguistic term affix. Bauer (2004) defines an
affix as
A type of obligatorily bound morph . . . attached to a base of a particular word-
class.
This assumes that we know what a morph is, what a base is and what a word-
class is, but those terms (or at least, two of them) are defined in their turn in
Bauer (2004) and a definition of word-class can be found in other publications.
Now consider the morph -ish in greenish. Its looks as though it may fit the
definition of an affix, although we would have to know rather more to be sure:
how the “particular word-class” is defined for -ish, for example. Now consider
the following example.
“He must have been pleased to move back here.”
Daisy wrinkles her nose. “Ish . . .”. (Elly Griffiths. 2016. The Woman in Blue.
London: Quercus, p. 242)
The question is whether -ish is an affix in this example. It has no base and is
apparently not obligatorily bound. We now have to re-evaluate. If -ish is not an
affix in this example, it is presumably a base. Is it a base in this example but not
in greenish, or is it a base in both instances, in which case is greenish really
a compound? If -ish is an affix in greenish, but not in the cited example, what is
it? Some authorities like to create a new category of affixoid (e.g. Kastovsky
2009, Ralli 2020) for items which are intermediate between lexemes and
affixes, which sort of fixes the problem: -ish is an affixoid and not an affix at
all. Bauer’s definition of affix just needs to be modified. But there is an
alternative solution. We simply state that -ish is an affix wherever it arises. It
is simply being quoted as a form in the cited example. If we do that, -ish
becomes an affix not because it fits with a regular pattern of form and/or
function which might define it as such, but because we have stipulated that
this is the case.
Stipulation might appear in this characterization to be a very bad way to
conduct a scientific description (it is, in any case, deliberately set out here to be
seen as such). But it is more common in linguistics than might be thought.
Consider another common definition, this time from Bauer 2017: 3 – the
definition in Bauer 2004 is different):
compounds . . . are often defined as words whose elements are words.
Now consider the expression Many hands make light work. Is this
a compound? It is certainly an expression whose elements are words, but is it
a word whose elements are words (for further discussion, see Chapter 9 and
58 Part I Basic Questions

Section 16.3)? It is fixed in the same way that word is, it does not allow
elements within it to be swapped out with synonyms (or with antonyms). It
tends to be learned as a single unit. It is very like a word. But does being very
like a word make it a word? I would expect people to deny that Many hands
make light work is a word, and would expect them to claim that it is a syntactic
structure – perhaps a sentence. But then what about man about town? This
appears to have a syntactic structure, too, similar to book about linguistics (not
a sentence, but nevertheless syntactic), but this time it is easier to find people
who are willing to see this as a word, along with superficially similar construc-
tions such as man-of-war, lady-in-waiting ormother-in-law. Are these, then,
words, and if so is man-about-town (and the other examples) a compound,
a word whose elements are words? And if it is not, why is it not? Here it seems
that stipulation may be the only way we can resolve the dilemma. We either see
these things as compounds or we see them as syntactic structures, but those
two solutions are generally considered incompatible (see Bauer 2025, for
discussion).
As a final example, consider the notion that endocentric compounds in
English are hyponyms of their right-hand element. Street corner denotes
a kind of corner, not a kind of street. Bauer et al. (2013: 434) exclude some
constructions which are not hyponyms of the right-hand element from the set of
compounds, thus raising the possibility that right-headedness is part of the
definition of a compound (though Bauer et al. do not advocate this position). If
we were to take that position, we would be stipulating right-headedness as
a requirement for a compound. We would therefore exclude from the set of
compounds items such as girl Friday, or endgame (girl Friday denotes a type of
girl, not a type of Friday; endgame denotes a kind of end, not a kind of game).
We would then have to determine what such expressions are and what their
grammatical structure is. And yet endgame is a word whose elements are
words, so the stipulation would have the effect of circumscribing the set of
compounds, presumably on the grounds that in most cases it is true that right-
headedness is a prototypical feature of compounds.
As with the other examples considered here, creating new types has the
inevitable effect of increasing the number of borderline cases, and thus making
it harder to distinguish consistently between types and increasing the require-
ment for criteria to help make decisions. If our criteria are not hard and fast but
are prototypes, then we simply increase the uncertainty at borderlines, because
we have to determine how close to a prototype something has to come before it
is accepted as being part of a relevant category, and that requires some kind of
measurement of approximation (a measurement we do not have). Alternatively,
we have to allow some freedom to be insecure about whether things fit into one
category or another. While this has some appeal, in effect it simply throws
things back on intuition and does away with the need for scientific criteria at all.
 7 Definition by Stipulation and Word-Classes 59

One of the places where all this becomes relevant is with word-classes.
Word-classes are notoriously difficult to set up and to define. All discussions of
word-classes have to deal with the problems this gives rise to (see e.g. Crystal
2004, Hollmann 2020), but the picture of word-classes that seems to be
dominant in studies of word-formation is simplistic and gives rise to a range
of questions.
Most handbooks of word-formation talk about processes (typically, processes
of affixation) which create nouns, verbs, adjectives, rather than about processes
which create, say, human nouns, telic verbs and gradable adjectives. Similarly,
affixes are, in general terms, said to derive adjectives from nouns or verbs from
adjectives rather than adjectives from human nouns or verbs from predicative
adjectives. There are, of course, exceptions, such as comments on the adjectival
suffix -ly in English being productive only on human nouns (which may not be
accurate, because it could probably be used on words denoting intelligent aliens
or dwarfs/elves/fairies/gods, etc.; not only spectatorly but also demonly are
attested – Bauer et al. 2013: 306). The practice is not necessarily harmful, but
it does make certain presuppositions about the way in which word-classes in
English work. It assumes that a word like noun, for example, denotes the highest
level of abstraction for that category, and all other types of noun can be safely
subsumed under that single label (just as monotreme can be safely subsumed
under mammal, even if the denotata differ in important ways). Monotreme and
placental mammal may be incompatible with each other, just as abstract noun
and concrete noun may be incompatible terms, but they fit neatly into a scale of
hyponymy. It must be noted, though, that a given noun can be a count noun,
a concrete noun and a human noun at the same time, so the incompatibility does
not always hold. Second, it assumes that there are a very limited number of
relevant categories, that they are all incompatible with each other, and that we
know what they are. That interjections are rarely mentioned in works on word-
formation implies that they are not relevant to the topic, that adverbs and
prepositions are sometimes mentioned implies that word cannot be both at the
same time (though see Chapter 23 for some discussion). Another way of looking
at this is to say that noun implies not adjective, and so on. There may also be an
implication that a set of nouns and verbs or nouns and adjectives and so on is not
a word-class (though note the remark by Brugmann 1891: II: 93 that “Speaking
generally, no sharp distinction between substantive and adjective can be drawn in
the Indo-Germanic [i.e. Indo-European] languages”). Most of this is probably
controversial, possibly problematic, and not least for the study of word-
formation. Moreover, most of these problems are well known.
We can begin with the word-classes of bases. We should recall here
Aronoff’s (1976: 48) unitary base hypothesis, namely that “A W[ord]
F[ormation] R[ule] will never operate on either this or that”, WFRs always
operate on a unique base-type.
60 Part I Basic Questions

Now consider the suffix -er. It attaches most obviously to verbs, giving
forms like attacker, boaster, clinger, dancer, owner, womanizer and hun-
dreds of others. But -er is also found attached to nouns, for example, in
falconer, islander, lifer, peasouper (Bauer et al. 2013: 217–18). Having
made this observation, how should we interpret it? One interpretation is that
we are not dealing with the same -er in the two instances. The suffix -er1
attaches to verbs, while -er2 attaches to nouns, and, presumably, we have
a number of other affixes attaching to adverbs (outsider) and numbers
(tenner). This saves the unitary base hypothesis, but probably goes against
an intuition that the same affix is involved in all of these. While intuitions
can be wrong, any semantic differences between -er1 and -er2 could be
argued to be the result of the word-class of the base, rather than the result of
a different affix. An alternative analysis is that the class of nouns and verbs
together form a natural class which can function as the requisite unique
base-type. The difficulty here is that for most authorities, from the classical
grammarians to Chomsky (1970), nouns and verbs are maximally distinct
word-classes, and cannot easily be seen as forming a superclass. Similarly,
while we can find counter- attached to adjectives (counter-intuitive), to
nouns (counterexample) or to verbs (counteract), we cannot simply say
that counter- can attach to the set of words, because counter- does not
attach to prepositions. While, in principle, it might be possible to see any
superset of word-classes as forming separate word-classes, this seems to go
against the fundamental spirit of the suggested hypothesis. A third possibil-
ity – though probably not a realistic one – would be if the nouns and verb
used in the bases for -er nominalizations had some feature or features in
common which could be seen as providing the underlying uniqueness. I am
not aware of any such suggestion.
Classes of base and classes of output are also important in dealing with
derivation and with conversion. I shall concentrate on conversion here, though
many similar points could be made with reference to overt derivation. One of
the major determinants of conversion, according to most authorities, is that
conversion involves a change of word-class. That is a word which belongs to
one word-class loses the features of that word-class and takes on the features of
another without there being any change in form. For example, the noun position
ceases to be a noun and becomes a verb in a sentence like They positioned
themselves to compete in Europe. To know whether this condition is met, we
have to be able to tell whether the input (in this instance, the noun) and the
output (in this instance the verb) belong to different word-classes, which
implies that we know what the word-classes are. In the case of position, the
answer is taken as clear-cut, but there are other instances where that might not
be the case. Two examples will be considered here, though others are
potentially relevant.
 7 Definition by Stipulation and Word-Classes 61

Consider a name like Leigh. In a sentence such as Leigh walked into the
room, this is clearly acting as a name, denoting a particular individual, known
to both speaker and listener. However, in a sentence like There are three Leighs
in our class, it is less clear. Leigh here denotes a person who is called Leigh, and
the fact that several can be co-present in a particular space indicates that the
word no longer has unique reference. In such an instance, a word like Leigh is in
the same paradigm as a word like linguist, and as such appears to be acting as
a common noun (the use of a capital letter in English should not be taken as
significant in this regard; the English use of capitals is often indiscriminate).
The question is, if Leigh has shifted from being a name to being a common
noun, has conversion occurred? The general answer to this question is ‘no’.
Although I have carefully called Leigh a ‘name’ in one of its functions, it is also
often termed a ‘proper noun’, and if we believe that label we must say that
Leigh has shifted from being one kind of noun to being another kind of noun,
and that is not a change of word-class. But at this stage we are simply playing
with terminology. If I want to term Leigh a ‘name’ and not a ‘proper noun’, can
I then claim that conversion is involved here? That is, is what determines
a word-class a random choice of label, or is there some inherent content
which determines the matter. We could build an argument either way: names
like Leigh do not (in English) form part of determiner phrases (we cannot say
the Leigh or this Leigh without shifting to a common noun – although we might
be less sure about my/our/your Leigh); alternatively, part of the definition of
a noun is that it can act as a semantically crucial part of the subject of a verb, and
Leigh walked into the room shows that Leigh fulfils this function. I do not want
to solve this conundrum, merely make the point that what the word-classes
are – and consequently, what conversion is – may be a matter of interpretation
and argumentation rather than something which is automatically obvious.
The second example I want to deal with here is genuinely unresolved. It is
the matter of how to deal with participial forms like interesting and building
(and also like reserved, but I shall ignore that type simply to save space). On the
one hand, there is a verb to interest, with a form interesting (as in I hope that
their proposal is interesting their potential backers), on the other, interesting is
typically an adjective (as in That’s a very interesting observation), and building
is typically a noun (both as in The building is four storeys high and Their
building a block of flats on our fence-line looks like maliciousness). Do we,
then, have conversion between verb and adjective (interesting) or between verb
and noun (building)? One problem is that the -ing in the verbal interesting is
usually taken to be inflectional, and inflectional forms are not usually said to
occur inside conversion (at least not in English, though other languages have
some such examples). So, what are the possibilities here? We can defy the usual
assumptions, and see this as conversion. We can deny that it is conversion,
because interesting and building are verb forms in all their occurrences. We can
62 Part I Basic Questions

claim that it is conversion, but conversion between a verb and a participle,

making the assumption that a participle is a different word-class from a verb
(this assumption was made by classical grammarians, on the basis that parti-
ciples show tense and case, while verbs show only tense and nouns only case).
We can assume conversion from some other starting point (perhaps all -ing
forms are fundamentally nominal or fundamentally adjectival). We can deny
that it is conversion, because conversion does not start from an inflected form.
There may be other solutions, too. The important point, of course, is that unless
we know what our assumptions about the set of word-classes we are dealing
with are, we cannot determine whether this is conversion or not. It would be
nice to be able to say that our assumptions are well founded and properly
justified. But too often they are simply accepted without being questioned, and
we end up with the notion that interesting and building are verbs because the
forms interesting and building are sometimes word-forms in a verbal paradigm.
These are not the only problems in word-formation that depend on an analysis
of word-classes, but the same general point remains true: without a better view of
what a word-class is, and how word-classes work, a view which allows us to
argue about the way in which different word-classes relate to each other, we do
not have a way of arguing a case on these issues, and we are left with no
alternative but to stipulate a solution. A stipulated answer may be useful, but
the answers would feel rather more robust if we had a better way of arriving at
them. It may be that word-formation allows for a way into solving the dilemma.

Challenge
Can you argue a case in favour of names either belonging to the same word-
class as nouns or being in a separate word-class from nouns?

References
Aronoff, Mark. (1976). Word Formation in Generative Grammar. Cambridge, MA:
MIT Press.
Bauer, Laurie. (2004). A Glossary of Morphology. Edinburgh: Edinburgh University
Press.
(2017). Compounds and Compounding. Cambridge: Cambridge University Press.
(2025). Fixity or why English may not have compounds. In Sara Matriciano-
Mayerhofer, Johannes Schnitzer & Elisabeth Peters (eds.) (2025). Patterns,
Variants and Change: Through the Prism of morphology. Studies in Honour of
Franz Rainer. Strasbourg: Éditions de Linguistique et de Philologie.
Bauer, Laurie. Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Brugmann, Karl. (1891). A Comparative Grammar of the Indo-Germanic Languages.
Trans. Robert Seymour Conway and W.H.D. Rouse. New York: Westermann.
 7 Definition by Stipulation and Word-Classes 63

Chomsky, Noam. (1970). Remarks on nominalization. In Roderick A. Jacobs &

Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar.
Waltham, MA: Ginn, 184–221.
Crystal, David. ([1967] 2004). English word-classes. In Bas Aarts, David Denison,
Evelien Keizer & Gergana Popova (eds.), Fuzzy Grammar: A Reader. Oxford:
Oxford University Press, 191–211 [reprinted from Lingua 17, 24–56].
Hollmann, Willem B. (2020). Word classes. In Bas Aarts, Jill Bowie & Gergana Popova
(eds.), The Oxford Handbook of English Grammar. Oxford; Oxford University
press, 281–300.
Kastovsky, Dieter. (2009). Astronaut, astrology, astrophysics: About combining forms.
classical compounds, and affixoids. In R.W. McConchie, Alpo Honkapohja &
Jukka Tyrkkö (eds.), Selected Proceedings of the 2008 Symposium on New
Approaches in English Historical Lexis. Somerville, MA: Cascadilla, 1–13.
Ralli, Angela. (2020). Affixoids. In Lívia Körtvélyessy & Pavol Štekauer (eds.),
Complex Words: Advances in Morphology. Cambridge: Cambridge University
Press, 217–37.
Rosch, Eleanor H. (1973). On the internal structure of perceptual and semantic categor-
ies. In T.E. Moore (ed.), Cognitive Development and the Acquisition of Language,
New York: Academic Press, 111–44.
Taylor, John. (2003). Linguistic Categorization. 3rd ed. Oxford: Oxford University
Press.
8 Reflections on Analogical Word-Formation

While there is a great deal of word-formation that can be described in terms of

rules applying to bases and affixes, there are also cases of words being formed
on the basis of single parallel form. Although the term ‘analogy’ has something
of a bad press in linguistics because it has been used to cover so many things
from sound patterns to syntactic changes, and because it seems to be difficult to
formulate in terms of a theory of what it permits and what it does not permit, it
seems impossible to ignore the notion of analogy in these cases. I use the term
here without any theoretical assumptions, and certainly with no presupposi-
tions as to how an analogy is to be formulated or annotated, which I see as being
irrelevant for the discussion of the examples discussed here.
The label ‘creativity’ is sometimes used to denote innovation which is not rule-
based, though it is sometimes used more restrictively (e.g. to denote the creation of
morphologically simple forms (Bauer 2001: 63)). We might, therefore, see ana-
logical formation as being creative rather than productive. The first example to be
dealt with here seems to suggest that such a proposal is not always appropriate.
This first example to be considered, the use of -er and -ee to create contrast-
ing roles in relation to some, usually verbal, base, is interesting because there
seem to be perfectly good patterns, which can be expressed as rules, for the use
of both of these suffixes. Nevertheless, corresponding nouns in -er and -ee
seem to arise adjacent to each other in texts, and often with parallel unusual
features. This happens so frequently that direct comparison between the two
words (or sometimes between the base and the suffixed word) must be
a relevant factor in the coinage. This feature of such words has been com-
mented on in the literature on the suffix -ee. Examples are provided below, and
some remarks on the examples follow them.
The bellower was Harmon Crundall – and the belowee the mysterious Mrs
Smith. (Joan Hess. 1986. The Murder at the Murder at the Mimosa Inn.
New York: St Martin’s, p. 41)
The toaster is hoping the toastees will have no troubles but little babies.
(Laurence Sanders. 1989. Stolen Blessings. Wallington, Surrey: Severn
House, p. 78)

64
 8 Analogical Word-Formation 65

Candor was dangerous to both candorer and candoree. (Dave Duncan. 1996.
Present Tense. New York: Avon, p. 139)
“Some guys are just born to have the shit kicked out of them.”
“And stomped”, said the other. “Like the world is divided into stompers and
stompees and he’s a stompee.” (Stephen Dobyns. 1998. Saratoga Strongbox.
New York: Viking, p. 46)
Sluggers don’t much like getting their ass kicked by the designated sluggee.
(Robert B. Parker 1998. Sudden Mischief. New York: Berkley, p. 63)
Lady Bella was the seducee, never the seducer, and would never overtly flirt.
(Skye Kathleen Moody. 1998. Wildcrafters. New York: St Martin’s Press,
p. 100)
It was bad enough to consider being the dumper. To be the dumpee was
terrifying. (Wes Craven. 1999. Fountain Society. New York: Simon &
Schuster, p. 18)
I thought I was the sneaker when I was really the sneakee. (Clive Cussler.
2000. Blue Gold. London: Simon & Schuster, p. 38)
We’re lover and lovee. (Robert Littell. 2013. A Nasty Piece of Work.
New York: St Martin’s, p. 184)
“So you think he has rebounded quickly from his loss?”
“I do. I think he’s a fast rebounder. And his reboundee is named Susan Baird.”
(David Rosenfelt. 2014. Hounded. New York: Minotaur, p. 172)

Note that in many cases, not even the -er form is established in the commu-
nity: a sneaker is usually a shoe rather than a person, a dumper is usually
a truck, stomper is not established, a toaster is usually a kitchen implement, and
candorer is possible for -er suffixation on a nominal base, but is not item-
familiar. With lover, the word is established with the right meaning, but the
person in the reciprocal relationship is usually also a lover. This seems to imply
that it is not simply a matter of the -ee form copying the -er form (although the
greater productivity of -er may mean that such a pattern is frequent), there is
some mutual support in the coinages. It also appears, from the examples
provided above, that these paired formations are often very consciously cre-
ated, which some scholars, following Schultink (1961), consider to mean that
the coinage cannot be a matter of productivity. Some of the formations are also
clearly intended to be jocular. While this may be perfectly compatible with the
conscious nature of the formations, it does not in itself mean that they are not
perfectly acceptable words.
The other examples considered here are far less systematic. In the first
example, it is not entirely clear that the substitution is in English.
66 Part I Basic Questions

In Jim Beam-o veritas. (Jonathan Nasaw. 2003. Fear Itself. New York: Simon
& Schuster, p. 102)
On one level this is a Latin proverb, with a loanword embedded in it. But, of
course, the proverb is used in English, even if cited in Latin, and Jim Beam (a
brand of whiskey) was not available in the Latin period. But whether it is
a relevant example or not, it is similar to others in that the base of an affixed
word is substituted for a familiar one. Other examples follow.
If the magic was in the ear of the behearer . . . then Lucy seemed ready to settle for
that. (Gavin Lyall. 1993. Spy’s Honour. London: Hodder & Stoughton, p. 310)
“If he can philander”, Susie said in an airy way, “so can I. He womanises,
I man-ise.” (Susan Moody 1994. The Italian Garden. London: Hodder &
Stoughton, p. 69)
The man is a known modelizer. (Wes Craven. 1999. Fountain Society.
New York: Simon & Schuster, p. 42)
and Devereux making fun of his juniority. Was there such a word? (Carola
Dunn. 2007. The Bloody Tower. New York: St Martin’s Press, p. 32)
The Hitler Youth turn into the Hitler Oldth. (Andy McDermott. 2014.
Kingdom of Darkness. London: Headline, p. 257)
The comments on conscious, and sometimes jocular, formations that applied
to the examples with -ee above also apply here. Sometimes here the model on
which the new word is built is not overt: with man-ize we know that it is based
on womanize because it is there in the text, but modelize, based on the same
original, is presented without guidance for the reader, who has to deduce the
pattern. The same is true with juniority, based on seniority. With the case of
behearer, the reader has to make bigger steps to fill in the meaning, particularly
since the original is not beseer, but to be in the ~ of the be~er provides a more
extensive pattern for the reader to use.
The same kind of pattern is found when new words are substituted in
otherwise familiar compounds.
If she’s dead, is that a corpsenapping? And is that a crime? (Stuart Woods.
1991. New York Dead. London: HarperCollins, p. 241)
More correctly, if less grammatically, what is produced by the home musician
is an ‘alongsideput’. (Stephen Davies 2003. Themes in the Philosophy of
Music. Oxford: Oxford University Press, p. 101)
case that met military specifications for being ruggedized, meaning it was
vapor-proof and dust-proof and everything-proof. (Patricia Cornwell. 1998.
Point of Origin. London: Warner, p. 196)
 8 Analogical Word-Formation 67

She’s just a journeyman. Journeywoman. Is that a word? (Renee Patrick.

2016. Design for Dying. New York: Doherty, p. 89)
Vapor-proof may well be a productive use of the element proof, but everything-
proof seems to be a wild generalization based on the previous two examples. It
might be argued to be a productive use of a productive pattern, but the close
parallels with the other examples suggest otherwise.
The next example is very much a one-off, based on a case of univerbation that
probably would not be counted as word-formation in the usual understanding.
You wouldn’t get stars staying there, only wannabes and usetabes. (Barry
Norman. 1998. Death on Sunset. London: Orion, p. 93)

My last examples show reanalysis of the affix, so that what was originally
part of the base is reassigned to the affix. Such cases are of interest because they
show speakers following an overall gestalt of what a word must look like rather
than following a series of patterns or rules which focus on and manipulate bases
and affixes.
They [cows] had personalities – or cow . . . cowanalities – or whatever you
want to call it. (Radio New Zealand, Saturday Morning, 4 November 2006)
“You a socialist?”
“I’m a nothingalist.” (Brian Freeman. 2014. Season of Fear. London:
Quercus, p. 74)
Jingle bells / Batman smells / Robin ran away, / The Batmobile / Has lost its
wheels / Now it’s a Batmosleigh. (children’s rhyme)
The types of example that have been explored here may not be exhaustive,
and just how much of word-formation operates on such a basis is not clear. It
might be argued that most blends, neoclassical formations and derivation
function on the basis of such templates rather than on the basis of minimal
meaningful elements. This is the kind of model that has been espoused by
Bybee (e.g. Bybee 1985) for some time. It is hard to imagine what might be
viewed as appropriate evidence to support such a hypothesis as opposed to, say,
a rule-based hypothesis, but the alternative, that words such as the ones
discussed in this chapter and the regular patterns that are generally discussed
in the handbooks should be created by totally separate mechanisms also seems
to be missing something.

Challenge
The suffix -ee has been discussed by several linguists, and many examples can be
found, as well as overt commentary, in the relevant literature. On the basis of such
examples or on the basis of examples you yourself can find (but not in dictionaries,
68 Part I Basic Questions

where the original patterns of usage are not usually given), collect a sample of
about twenty words (more if you are working as a team) containing the suffix and
the ways in which they are used in early (perhaps unique) attestations. Is formation
of -ee words regularly a matter of analogy? Or in how many instances do you
think that the words are formed independently of corresponding -er words or
a corresponding base? How can you tell? Does this influence your view of the way
in which words are formed in English? Why (not)?

References
Bauer, Laurie. (2001). Morphological Productivity. Cambridge: Cambridge University
Press.
Bybee, Joan. (1985). Morphology. Amsterdam: Benjamins.
Schultink, Henk. (1961). Productiviteit als morfologisch fenomeen. Forum der Letteren
2, 110–25.
9 Reflections on the Nature of the Lexeme

Since word-formation is usually considered to be about the formation of words in

the sense of lexemes, and since compounds are words, in the sense of lexemes,
whose elements are also words (lexemes), and since we have topics discussed in
word-formation texts which might not be considered to result in lexemes (for
instance, are initialisms like UN ‘United Nations’ lexemes?), and the same might
apply to larger units which might be thought of a syntactic (are phrasal verbs
lexemes, is the expression kick the bucket ‘die’ a lexeme?) the definition of the
lexeme is clearly a matter of interest in word-formation. It is often simply
assumed, or given a superficial definition which does not cover all that the analyst
needs to know. A definition of the lexeme as a word in the sense that cat and cats
represent the same word answers only some of the potential problems.
It is perhaps worth noting that the lexeme is a modern interpretation of the
grammar of the classical languages, Greek and Latin, in particular under the
influence of Lyons and Matthews (see references below). In those languages,
words as they occur in sentences have certain ‘accidental’ properties which are
not essential to central notion of the word (Lyons 1968: 198). This explains the
older term ‘accidence’ for what we now more usually call ‘inflection’ (the
spelling is now general, though inflexion used to be the standard British
spelling). It is accidental that in the Latin sentence Brutus Caesarem occidit
‘Brutus killed Caesar’ that Caesarem is marked as being in the accusative case,
because it is not essential to the nature of Caesar that it should be.
Lyons rephrases this a couple of years later when he says
[A lexeme is] a unit which is manifest in one ‘form’ or another in sentences,
but which is itself distinct from all its forms. (Lyons 1970: 21)

That is, the lexeme is the word with all its inflections stripped away (or,
probably equivalently, encompassing all of its inflected forms); this makes it an
abstract form, rather than one that a listener can hear, or a speaker produce. This
also implies that it is a term defined for the purposes of analysis, and, as such, it
is open to varying definitions. It is perfectly possible for the lexeme to be
defined in different ways by different authorities.

69
70 Part I Basic Questions

What is not clear in Lyons’s definition given above, is whether lexemes must
always be represented in inflected forms or whether an uninflected item can be
classed as a lexeme. Similar problems arise with parallel definitions from other
linguists:
Ein Lexem kann nur definiert werden als eine Menge von grammatischen Wörten, die
denselben Stamm enthalten.
[A lexeme can only be defined as a set of grammatical words which share a stem.]
(Bergenholtz and Mugdan 1979: 117, my translation)

the lexeme is:

1. A complete sign on a particular linguistic level, namely the lexicon;
2. A class of variants, namely word-forms;
3. An abstract unit of the language system. (Lipka 2002: 89)
[A lexeme is a] word seen as an abstract grammatical entity, represented concretely by
one or more different inflected forms according to the grammatical context. (Carstairs-
McCarthy 2002: 144)

A lexeme is the abstract unit that stands for the set of inflectional forms. (Booij 2015: 158)
Bergenholtz and Mugdan (1979: 118) clarify their position by saying that
with and not are lexemes despite having only one form, Lipka (2002) implies
that such words are lexemes by annotating them in small capitals. However,
Lyons (1977: 452–3) avoids calling the a lexeme, but implies that this might be
a lexeme on the grounds that “[this and these] might be said to be forms of
this . . . while the definite article is invariably the” (I have adjusted the notation
for the lexeme in this quotation).
We seem, therefore, to have two distinct meanings for the term ‘lexeme’. It is
not clear whether Matthews’s definitions of the lexeme as a “lexical word”
(Matthews 1972: 161) or “the fundamental element in the lexicon of
a language” (Matthews 1991: 26) fit with Lyons’s usage or not, but I suspect
not. There is a distinct French tradition where there is a clear distinction, drawn
in rather different terms. For Fradin (2013: 102), as translated by Boyé (2018:
20), the lexeme has the following features:
It is an abstract unit to which word-forms are related; this unit captures the variations
across word-forms.
It possesses a phonological representation which gives it prosodic autonomy.
Its meaning is stable and unique.
It belongs to a category and can have argument structure.

It belongs to an open-ended set and can serve as output and input of derivational
morphology.
 9 The Nature of the Lexeme 71

In this, it contrasts with the ‘grammeme’, which includes words such as

prepositions, determiners and conjunctions (whether this corresponds exactly
to the English term ‘grammatical word’ in the sense used by Bauer et al. 2013:
10, where it contrasts with ‘lexical word’, is not necessarily clear). In an earlier
French tradition (Martinet 1967: 16), the term ‘morpheme’ would have been
used rather than grammeme: in either case, there is a mismatch between the
French terms and what appear to be the corresponding English terms.
At this point, we have three definitions of the lexeme, which overlap in that,
for example, Latin puella ‘girl’ is a lexeme which encompasses word-forms
such as puellam (‘girl.acc.sg’), puellae (‘girl.nom.pl’). A word like incurably
(created by word-formation but having only one form) is probably a lexeme for
all, words like the and with, are not lexemes for all, but are for others, and this
may be described as a lexeme if we believe it inflects.
This still leaves many questions unanswered. The most important of these is
whether English -er in owner and -ation in civilization are lexemes, on the
grounds that they are listed in the lexicon and so are listemes in the terminology
of Di Sciullo and Williams (1987). Inflectional affixes are clearly not lexemes,
at least in modern word-based theories, since they are built up by phonological
rules to create word-forms from lexemic bases without creating compounds.
Derivational affixes may not be either, if we believe that the distinction
between a derivative like smoker and a compound like smokehouse is that the
compound is a lexeme which contains two lexemes, and the derivative is
a lexeme built from the appropriate stem of a single different lexeme.
Derivational suffixes are also not (or are not generally) phonologically autono-
mous and cannot be the output of derivational morphology (to use Fradin’s
terms). On the other hand, derivational affixes may have stable semantics (if
not always unique semantics), are abstract units which may vary in form, and
probably have to be listed, rather than built up by phonological rules in the way
that inflectional affixes are in Matthews (1972) and other works with a similar
philosophy. The weight of the evidence is thus that they are not lexemes, but
simply calling them ‘elements’ (Matthews 1970: 112) or ‘forms’ (Lyons 1977:
452) or ‘formatives’ does not help solve the problem of whether they are to be
classed with other types of grammatically relevant chunks or whether they are
a class unto themselves. In the French tradition, they can probably be classed
with the grammemes, which might provide insights. The question here is not
directly linked to the question of what a lexeme is, but it arises from consider-
ing that wider puzzle.
The next question is whether a lexeme, not including compound lexemes,
can contain multiple lexemes as constituents. For Lyons (1977: 23) this is
possible, and items such as kick the bucket ‘die’ are ‘phrasal lexemes’.
Although they do not use the term, Dixon and Aikhenvald (2002: 7) approve
of the general notion. Again, it is not clear whether for Lyons a phrasal lexeme
72 Part I Basic Questions

has to inflect, so that kick the bucket can be a phrasal lexeme because we can
have kicked the bucket while on the take ‘dishonestly accepting bribes’ does not
inflect and therefore is not a phrasal lexeme. The lack of clarity thus becomes
more widespread. Di Sciullo and Williams (1987: 3) link such phrases to the
notion of listedness (i.e. whether the speaker/listener has to memorize the
words or not) and they call listed items ‘listemes’. Listemes seem to include
words as a subtype; another sub-type comprises syntactic structures “The listed
syntactic objects are the idioms” (Di Sciullo and Williams 1987: 5). Carstairs-
McCarthy (2002: 144) appears to take a similar view, but calls the listed items
‘lexical items’, a term which I prefer. Whether the lexical items encompass
more than just the idioms may be controversial: I would take the view that cool
as a cucumber and To be or not to be are also lexical items, despite not being
idiomatic, on the basis that they are memorized and thus listed. Strictly
speaking, the view where syntactic lexical items, particularly at least some
idioms, are lexemes and the view where lexemes are a special type of lexical
item indicates another potential difference of definition, although the difference
is not particularly meaningful.
The conclusion here is that although there is wide agreement about the
notion of lexeme in its most central or canonical uses (e.g. when we say that
am is a form of the lexeme be – or, more explicitly, a word-form belonging to
the inflectional paradigm of the lexeme be), once we move away from that
point of agreement we find a great deal of disagreement and thus potential
misunderstanding. It is easy enough to define a series of terms so that most of
the points of disagreement are made unambiguous, but it would be nice if
morphologists could agree on what that series of definitions comprises. It might
be useful to start from the notion that although lexemes are listed, not every-
thing that is listed is a lexeme (though borderlines might be hard to determine);
a further step, that there is a level of analysis where sentences are seen as strings
of lexemes, would require rather more consensus-building. Even further, we
might be able to agree that anything which inflects is a lexeme (e.g. mothers-in-
law), even if not all lexemes inflect (e.g. with). Such adjustments to the
definition of lexeme, though relatively minor, would help clarify the nature
of a central notion in modern morphology. If these proposals were accepted, the
French tradition would still be separate from the anglophone tradition.

Challenge
Can you find any benefits either to using the term ‘lexeme’ only for words than
can inflect or to using the term ‘lexeme’ for words which do not inflect as well?
Would those benefits also be relevant to all lexical items? Does this help you
decide whether lexemes are a subtype of lexical item or whether lexical items
are a subclass of lexeme? Do you consider names to be lexemes? Does this
 9 The Nature of the Lexeme 73

affect your preferences for the definition of the lexeme? Do you think that the
nature of the lexeme changes from one language to another, depending on how
inflection works in the particular language?

References
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Bergenholtz, Henning & Joachim Mugdan. (1979). Einführung in die Morphologie.
Stuttgart: Kohlhammer.
Booij, Geert E. (2015). The structure of words. In John R. Taylor (ed.), The Oxford
Handbook of the Word. Oxford: Oxford University Press, 157–74.
Boyé, Gilles. (2018). Lexemes, categories and paradigms. In Olivier Bonami,
Gilles Boyé, Georgette Dal, Hélène Giraudo & Fiammetta Namer (eds.), The
Lexeme in Descriptive and Theoretical Morphology. Berlin: Language Science
Press, 19–41.
Carstairs-McCarthy, Andrew. (2002). An Introduction to English Morphology. Edinburgh:
Edinburgh University Press.
Di Sciullo, Anna Maria & Edwin Williams. (1987). On the Definition of Word. Cambridge,
MA: MIT Press.
Dixon, R.M.W. & Alexandra Y. Aikhenvald. (2002). Word: A typological framework.
In R.M.W. Dixon & Alexandra Y. Aikhenvald (eds.), Word: A Cross-Linguistic
Typology. Cambridge: Cambridge University Press, 1–41.
Fradin, Bernard. (2013). Nouvelles approches en morphologie. Paris: PUF.
Lipka, Leonhard. (2002). English Lexicology. Tübingen: Narr.
Lyons, John. (1968). Introduction to Theoretical Linguistics. Cambridge: Cambridge
University Press.
(1970). Introduction. In John Lyons (ed.), New Horizons in Linguistics. Harmondsworth:
Pelican, 7–28.
(1977). Semantics. 2 vols. Cambridge: Cambridge University Press.
Martinet, André. (1967). Éléments de linguistique générale. Paris: Colin.
Matthews, P.H. (1970). Recent developments in morphology. In John Lyons (ed.), New
Horizons in Linguistics. Harmondsworth: Pelican, 96–114.
(1972). Inflectional Morphology. Cambridge: Cambridge University Press.
(1991). Morphology. 2nd ed. Cambridge: Cambridge University Press.
Part II

Semantic Questions
10 Reflections on How Words Bear Meaning,
and What This Implies for Complex Words

10.1 Introduction
Perhaps because of the notion from Chomsky (1965) that what linguists are
trying to explain and model is the usage of the ideal speaker-listener, and
perhaps because of the availability of standard dictionaries, linguists – and, it
must be said, non-linguists, too – operate on the principle that words have
definite and knowable meanings, which can be fully described. Furthermore,
we tend to operate on the principle that what we use when we produce instances
of language is these fully explicit meanings. The first point that I want to make
here is that for a large proportion of the words we think we know, perhaps all of
them, this is not strictly true, which seems to imply that we may not actually
communicate as much as we think when we use words. Second, I want to look
at the implication for this on the semantics of complex words. This includes
looking at the way meaning in complex words is deduced in a word-based
model of morphological structure.

10.2 Some Examples from Personal Experience

In this section, I provide some examples from my own recent experience to
illustrate some of the types of information that must underlie what we know
about words. Because I am citing my own experience, the information is
personal, but not only do I believe that the scenarios will be familiar to others,
if not the specific examples involved, I believe that to some extent all experi-
ence of vocabulary items is personal, and that the overlap between personal
experiences provides a superordinate level of experience which we might call
vocabulary of English (or any other relevant language). My own reactions are,
of course, skewed to some extent in that I am a linguist with an interest in
vocabulary, etymology and the like, and so have available some information
which not everyone will share, but I do not believe this matters in the larger
scale of things.
My first example is the word voe. I am aware that there is a word voe, which
I could legitimately use in a game of Scrabble. I don’t know what it means at all.

77
78 Part II Semantic Questions

I have a form to which I can attach no meaning. Note that this is not a claim that
the meaning of voe is undiscoverable: if it were, then, presumably, we would
not want to say that voe was a word, like smofle which, as far as I am aware, is
not listed anywhere as a word of English.
At the other end of the scale, I was recently looking for a word whose form
escaped me (temporarily). I knew that it meant writing about saints, and knew
that the implication behind the word was that in such writing only the good
things about the person are reported. I knew that it was a Greek compound.
I wanted to use the word figuratively for a description of a real person (and not
a saint) who had been described in glowing terms in something I had read. The
word, of course, was hagiography, but for a few minutes I had a meaning, but
no form. An equally familiar experience is illustrated in a passage from
Wodehouse (2012). Wodehouse writes:
I knew something was going to happen. You know that pre-what-d’you-call-it you get
sometimes? Well, I got it then.
Wodehouse’s narrator knows the meaning of the word and something of the
form (premonition starts with pre-, although, interestingly, this is in the spelling
and not in the pronunciation), but no more.
In the song ‘Me and Bobby McGee’ (Kristofferson and Foster c.1969), the
singer says that he “took [his] harpoon out of [his] dirty red bandanna”. I have
never (to my knowledge) heard the word harpoon with this meaning anywhere
else. In the following line the singer says he “was blowin’ sad”, which suggests
a wind instrument, and if it could be wrapped in a bandanna, I assume it means
what I would call a mouth organ or harmonica – though other instruments
would be possible. I have to deduce this from the context, just as I have to
deduce, later in the song, that Salinas is a town in America that I happen not to
be familiar with (it is in California, near Monterey). Although I could legitim-
ately assume that harpoon is a malapropism or a dialect word – either American
English or regionally within the USA – I actually assume that it is a slang word,
and the justification for this assumption might be of some interest, though it
would be difficult to deduce. At no point do I assume that harpoon must refer to
an instrument for killing whales. This illustrates firstly that we can build up
a meaning from very few clues and that it is easy to reject a meaning attached to
a form when it doesn’t make sense, and thus assume some kind of homonymy.
That there might be homonymy does not appear to cause problems with
understanding.
In the same song, the singer says that they “finally sang up every song that
driver knew”. Many dictionaries do not list sing up at all, presumably because
of its low frequency, and those that do, like Courtney (1983), list it only as it
parallels speak up. I am not aware of having come across sing up in the sense
used in this song anywhere else. But it makes a paradigm with forms like drink
 10 The Meaning of Words 79

up, eat up, use up, all of which are listed by Courtney (1983). Again, we have
a novel form, but one which is part of an existing paradigm, which makes it
easy to understand, and allows productive use of a construction which is not
compositional.
Compare this with a word that is well established in my vocabulary, and
where I have quite a lot of information: diamond. For the word diamond,
I might have all the following information.
• Diamonds are hard.
• Diamonds are made of carbon.
• Diamonds are mined.
• Diamonds are used for jewellery.
• Diamonds also have industrial uses.
• Diamonds come in several colours (including blue, yellow and pink).
• Diamonds are precious because expensive.
• Diamonds can be smuggled.
• Diamonds reflect light.
• Diamonds are cut to maximize this feature.
• According to the commercial slogan, diamonds are forever.
• According to a cynical song, diamonds are a girl’s best friend.
• Diamonds are used for engagement rings.
• The shape ‘◊’ is supposed to represent a diamond, and is called a diamond.
• This shape is also used for a suit in cards.
• Diamonds in cards is a red suit.
• The nine of diamonds is supposed to be the curse of Scotland.
• Fake diamonds can be made of zircon.
• Diamond is a term used in baseball (but I’m not sure what it implies).
• Diamonds are traded in London and Antwerp.
• The size of diamonds is measured in carats.
• The best diamonds are said to be of the first water.
• Diamond collocates with mine, bracelet, necklace, ring, cutting, smuggling.
If we say She wore her diamonds, we probably mean a necklace, and perhaps
more.
This set of statements about the word diamond cannot be exhaustive, but
provides at least a starting point. Although I have, for practical reasons,
expressed this information in linguistic statements, some of the information
may be held in my brain in experiential ways, for example, the hardness of
diamonds. Note, though, that despite all this knowledge, I cannot necessarily
recognize a diamond – a zircon or even glass could probably fool me with ease.
I do not know how big a diamond would have to be to be a one-carat diamond.
I am not sure of the rank of diamonds relative to other suits in the card-game
bridge. There is a lot of information here, but the fact that I cannot recognize
80 Part II Semantic Questions

something that is called a diamond in the same way that I can recognize
something called a book is of interest.
The problem here is an old one. Locke was worried about how we are to
define gold, given that most of us do not have the technical knowledge to
distinguish gold from other things. We can call something gold not because we
have sufficient information to know without any doubt that it is gold, but
because we are willing to believe some authority that tells us it is gold.
New Zealand children throughout the country have a consistent story about
the daddy longlegs spider (Pholcus phalangioides). The story is that this is the
most poisonous spider in the world, but fortunately its jaws are too weak (or too
small) to allow it to bite humans. This is almost entirely false (the creature is
a spider; there are other creatures called daddy longlegs in other places in the
world, not all of them are spiders). The creature is not particularly poisonous
and can – and does – bite humans (Sirvid n.d.). The point of this example is that
speakers can have totally erroneous beliefs about the world (and about the
words which we use to help conceptualize the world), and can use such errors to
define words (‘the most poisonous spider in the world’). The definitions that
individuals use do not have to be based on scientific truth to be able to function
appropriately in the world.
Weinreich ([1962] 1980: 298) reports an experiment with dictionary mean-
ings he tried with graduate students. He took a number of near-synonymous
adjectives, including gloomy, morose, saturnine, sulky and sullen, and provided
the students with their dictionary definitions, asking them to match each
definition with a word. “The results”, he reports, “were poor”. I repeated the
same experiment with undergraduate students and a different dictionary some
years ago, and the results were rather more promising. The value of dictionary
definitions is not the point here, though. The question is: if speakers know these
words and do not see them as synonyms in a particular text, what kinds of
information might distinguish them? Some of the answer is in collocations:
gloomy can collocate with house, room, picture (especially when used figura-
tively), outlook, as well as with words denoting people. All of these words can
collocate with a noun denoting a person or a person’s face. They differ in style:
saturnine is literary, morose is rather formal, while gloomy and sulky are
neither. Sulky gives me the impression of not getting what you want, possibly
being rather immature, and with the possibility that the emotion is short-lived.
Sullen, on the other hand, seems rather longer-lived. Sulky suggests a pout,
while the others do not. There are several points to make about this example.
First, but perhaps least importantly, is that writing really good definitions of
words is extremely difficult – even if Weinreich’s experiment may have gone
beyond what we normally require of such a definition. Second, that some of the
information we use to distinguish between words is not the type of information
 10 The Meaning of Words 81

which usually gets into definitions. Third, just what we know about individual
words goes beyond the material provided by a typical dictionary definition.
Everyone familiar with Lear’s (1871) poem ‘The Owl and the Pussy-Cat’ will
also be familiar with the term runcible spoon. Runcible, evidently, collocates
with spoon. But what does it mean? Since Lear invented runcible as a nonsense
word (and used it several times, although most famously in this poem), we might
justifiably argue that it has no meaning, although it does have a collocation. Not
only does this show that words can exist in the community but have no meaning,
it also shows that words whose meaning we do not know can have firm colloca-
tions. At a later date, hypostatization stepped in: if we have a word, there must be
a thing. Runcible spoon came to mean what we now call a spork. But Lear also
had a runcible cat, so if runcible spoon means ‘spork’ it is not clear how the
meaning of runcible can be equally used of a cat. If this example of knowing
a collocate without knowing the real meaning seems too far-fetched, consider
aneroid. This clearly collocates with barometer, but until I looked it up in
a dictionary, I didn’t know that aneroid means ‘not using liquid’.
There are many words for which I know a superordinate term, but not what
differentiates this particular word from other hyponyms of the superordinate.
For instance, I know that sarsaparilla, sassafras and mint julep are drinks, but
otherwise I know virtually nothing about them – not even whether they are
alcoholic or not. I know that beguine, galliard and lambada are dances, but
little more. I know that barouche, brougham and carriole are types of horse-
drawn vehicle but would not recognize any of them specifically. I know that
bézique, faro and euchre are card games, but do not know how to play any of
them. Such imperfect knowledge of the meanings of words must be very
common, and indicates that speakers do not have the perfect knowledge of
their vocabulary that is presupposed by most semantic theories. Despite the fact
that I know only the superordinate of each of these, I assume that they are not
synonyms. In a similar way, although I know the word hurtle, I am surprised to
find it defined in a dictionary (Marr 2008) as “to move rapidly with a clattering
sound” – I only know that it means ‘move rapidly’ and sound does not figure in
the way I understand the word. In this particular instance, it may be that the
dictionary is incorrect or out of date. The following examples from the British
National Corpus (BNC) (Davies 2004–) do not appear to imply noise at all,
even if the majority of examples in the BNC might be argued to involve sound.

Araminta’s callous mouth and Agnes Diggory’s sympathetic eyes hurtled

through her mind.
Clutching a Union Jack, three-year-old Louis hurtled into the arms of his dad.
A stone hurtled through the air and the dogs scattered.
the situation in East Germany could hurtle out of control.
Now a blue-grey merlin, the smallest of the falcons, hurtles across the heath.
82 Part II Semantic Questions

But whatever the majority of speakers think that hurtle means, and independent
of the accuracy of the dictionary definition, the examples collected here indi-
cate that we do not have a full knowledge of words we can (in some sense)
understand and possibly use.
I know the word mustelid from media discussions of predation against birds,
and know that it denotes a family of carnivorous animals, of which the weasel
is, for me, the prototypical member. Apart from weasels and stoats, I do not
know what animals are members of the family. That is, I know the superordin-
ate term, but not much about its hyponyms. Until I looked this up, I thought that
ferrets were probably members of the family (and they are), had not thought of
martens or polecats at all, and was surprised to find that badgers are mustelids.
Here I know (in some sense) the superordinate term, but do not know the full
range of its hyponyms, and so have a restricted impression of the range of
meaning mustelid might carry.
As a young child (probably under the age of 7), I was relatively regularly told
that I was a spuniachie-looking thing (/spjuːniæxi/). I was told this by my
mother, whose parents were Scottish, and by my aunt, her sister. I deduced that
the word meant ‘skinny, weak and feeble’. I assumed that this was just another
Scottish word, but when I lived in Scotland later in my life, I could find nobody
who recognized the word. I now assume that spuniachie was a family word, not
known in any wider community. In the British Woman’s Own for
14 November 1981, a woman from near Bristol reported that in their family
puddycovers was used for ‘gloves’, and poggolies for ‘slippers’. Words need
not have a very wide distribution at all, and even family words may have quite
a short lifespan. Speakers may or may not be aware of the restricted nature of
such words. Words may belong to very restricted communities, and even then
be transient.
What these various examples show is that our individual knowledge of the
words which we recognize and use is spotty, and incomplete, occasionally even
false. Furthermore, what we think words mean may or may not coincide with
what we can learn from a dictionary definition, and in any case includes more
than can typically be found in a dictionary definition. This is not to say that
a dictionary definition is not useful (although it might mean that, even in law
courts, rather less infallibility should be attributed to such definitions than is
often the case), just that it is not quite what is held in the head of a speaker,
especially because – though not only because – it is fully specified.
Although the examples used here are my own, the general observation is not
new. Clark (2023) talks of such instances as being those where speakers and
listeners gradually acquire parts of the meanings of individual words. She
argues that our understanding of individual words only has to be good enough
to allow for communication on one particular occasion, but that speakers
typically need more information about the words used than do listeners.
 10 The Meaning of Words 83

10.3 Some Implications

The information that is given about diamond (Section 10.2) indicates that much
of the information we have in our heads, both about the nature of the world and
the words which we use to describe that nature, is encyclopaedic information.
Although this might be a controversial claim, it is made advisedly, and
a justification for the position is given in Bauer (2005), where it is pointed
out that, in some cases, information which would uniquely identify a particular
real-world entity is not the most useful information for a real speaker, and there
may be multiple pieces of information of this kind. This standpoint is also
accepted within Distributed Morphology (see Siddiqi 2019: 149 and the com-
ment in Spencer 2019: 210). I shall not try to justify this position any further
here.
Many of the examples given above also hold the implication that what
speakers know about individual words can change. I could come to learn
what a galliard is, either in terms of what it looks like when danced, or in
terms of the music to which it is danced. I could check precisely what voe
means. I have learned that a badger is a mustelid. Readers could have learned
from what was said above that the nine of diamonds is called ‘the curse of
Scotland’ (the reason for this name is disputed).
The example of sullen, sulky, etc. indicates that what we know about words is
more than a clear-cut definition will provide: we also know about collocates,
about what used to be called ‘connotations’ and about the stylistic implications
of using one word over another. In other words, we must have more information
about individual words than is usually provided in a dictionary definition
(though some dictionaries do attempt to provide some of this information).

10.4 Applying This to Complex Words

If all this is true of morphologically simple words, we must assume that it will
also be true of morphologically complex words. The difference is that morpho-
logically complex words tend to make some information explicit in a way that
need not be true of simple words. For example, words beginning with un- are
usually negative (usually because this does not apply to words which beginning
in under-, for example), words ending with -ation are usually nouns, words
ending in -ee that has been attached to a verb usually denote people, and so on.
Sometimes, we do not need much more information than this. If we know what
happy means, we can work out a great deal about unhappy; if we know what
license means as a verb, we can predict much about licensee, although not that
the word is often used of someone licensed to sell alcohol.
Herein lies the problem for deducing the meaning of a complex word from its
morphological form. The morphology provides a framework within which we
84 Part II Semantic Questions

can seek a meaning but does not fully explicate the meaning. The examples
given earlier show that this is not unexpected.
Morphologically complex words differ from simple ones in that some
information is provided in the word. In compounds, the given information is
in the meanings we, as individuals, already hold about the elements. In wind-
mill that information may be more useful than in penknife, because we no
longer use penknives to cut quills for writing. Nonetheless, the information is
given, we just have to interpret it. One part of interpreting it – and this applies
equally to derivatives – is knowing which homonym or polyseme is selected for
the meaning in the complex word. To understand canonical, we have to
understand canon refers to a standard criterion as opposed to a body of
clergymen. To understand draftee we have to recognize that draft does not
have any of the meanings that appear in bank draft, draftsman (draughtsman),
or drafty (draughty). To understand draughtsman and draught horse we have to
recognize different meanings of draught. This loss of possibilities available for
the base word is a well-known principle of the semantics of new complex
words (Bauer and Valera 2015).
Some affixes carry meaning in much the same way that lexemes do. This is
typically the case of prefixes in English-coined words such as post-, pre-, re-, some
of which might be considered to be prepositions/adverbs, as in byproduct, over-
bridge. Some suffixes are just as explicit, though typically with a more specific
meaning: suffixes such as -fold, -scape, -teen. Other affixes may not carry specific
meaning in the same way, but may simply mark the word-class of the output. Such
instances include -al (creating adjectives), -ar, -ion, -ity, -ize, -ness, -ous. Parental
tells us little more than that the word is an adjective connected with the word
parent, stupidity is a noun connected with being stupid. Some affixes do both,
affixes such as -ee, -ess, -hood, -less. The elements of neoclassical formations,
when they attach to English words, tend to have a density of information that is
similar to that of lexemes, but as they become more used as English forms, they
become less specific in their meanings: -itis in Mondayitis is not as specific as -itis
in appendicitis, where it still carries its original meaning of ‘inflammation’.
With affixes, as with lexemes, we have to deal with homonymy or polysemy.
Adverbial -s in afterwards, downstairs, indoors has a homonymous affix in
Babs (from Barbara), Debs (from Debora), Wimblers (from Wimbledon);
nominal -al in arrival has a homonymous affix in adjectival -al in parental.
The different readings of -er in driver (agent or instrument), believer (experi-
encer), diner (agent or location), jumper (agent or clothing) are usually treated
as polysemy (although the distinction is not particularly clear in this example).
There are two claims from Cognitive Linguistics about the semantics of such
affixes (Panther and Thornburg 2002, Basilio 2009). The first is that affixed
words have a metonymical relation to their base, the second is that the readings
of polysemous affixes are in a metonymical relationship with each other. The
 10 The Meaning of Words 85

two are clearly related. To the extent that either of these holds, and there is
a great deal of evidence to support them, we may not need to specify the
distinctions in linguistic structure, because they are automatically covered by
cognitive processes. Nevertheless, in both the homonymy case and the poly-
semy case, the users of the words have to be able to determine whether the affix
is appropriate and how it is being used. Some authorities (Falkum and Vicente
2015, Pustejovsky 2017) suggest that such meanings are underspecified; an
alternative view is that, from a linguistic standpoint, they are overspecified
when they become fixed in particular usages. That is, the linguistic structure is
inherently vague, and it is usage that determines just how individual words are
likely to be used.
If the meanings of complex words, at least in the sense that they are
determined by a grammar, are inherently vague, then it is up to the user to
decide what is a plausible reading of the word, in context, provided that it is
within a permitted shell of variation. In many cases, we do not even have to
know what the precise relationship is, as long as we know that there is one. It is
well known that the precise relationship holding between police and dog in
police dog may not be clear (is the dog associated with the police, is the dog
trained by the police, is the dog working for the police, is the dog owned by the
police and so on?). The grammar says that there is a relationship (in some sense,
a defining one), and that it is likely that a police dog is a type of dog, and that
may be all we know, or we may fill in something more precise for ourselves.
This is similar to knowing that a lambada is a dance, but not knowing any more.
Sometimes we may know that whatever the relationship is, it is the same as in
another, already known word. We understand goulash communism on the basis
of spaghetti western, and whatever the relationship between the elements is, we
recognize it as being the same in both cases. That is, we see the analogy
between the forms and use it to interpret the less familiar word (see Chapter 8).
Equally, we may recognize that elements do not have a common meaning.
The -age that occurs in postage does not have the same reading as the one that
occurs in wastage, and it may not be clear whether either of those shares
a reading with frontage. Linguistically, what we can recognize here is a noun
related to the meaning of its base. The details depend on usage, and are
gradually built up in the individual, as the word becomes more familiar, and
in the community as more individuals agree on the frame in which the words
are used.
The implication here is that words in the community, just like words in the
individual, start off with minimal specification, and gradually become more
fully semantically specified. Any theory of the lexicon must allow for this
process, and allow for the fact that different individual speakers will have
different bits of information about each word. This seems to imply that, in
the community, words may function with a range of partly defined meanings,
86 Part II Semantic Questions

not necessarily all compatible, which may develop into new meanings for the
word and cause old meanings to vanish. In the first few months of 2023, I heard
the word exasperate used by people in news broadcasts and radio interviews to
mean ‘to make worse’: This will only exasperate the situation. I have also heard
a professional broadcaster say that because of traffic jams there were a lot of
exacerbated motorists trying to get to work. This is not the first time in the
history of English that a word for ‘make worse’ has come to be used to mean
‘annoy’; the same trajectory can be found earlier for aggravate. At the moment,
there is apparently some confusion in the community about what both exasper-
ate and exacerbate mean. We cannot yet tell what the outcome will be. In the
meantime, the community must have dual meanings for both words, even if
these have not yet got into dictionaries.
The take-away message from all of this is that linguists should not expect the
fully specified dictionary meaning to be what individual members of the speech
community know, and should not expect that fully specified meaning to be
deducible from the form of a new complex word. Rather, the form provides
only a minimal amount of information about the meaning of a word; most of the
interesting detail is provided by knowledge of the real world and knowledge of
the way the word is used – something that can only arise when the learner has
heard the word used many times in context. It remains true that the speaker
usually needs a rather more elaborated meaning of the word than the listener
does, but, of course, that cannot be guaranteed in cases like daddy longlegs.

10.5 Derivation in Word-Based Morphology

Word-based morphology (including some forms of word-and-paradigm or
a-morphous morphology) has largely been developed as a way of dealing
with inflectional morphology. In word-based models, the word is the funda-
mental unit of analysis, and the meaning of the elements within the word is
derived from the meaning of the word as whole. Either the meanings of the
elements are abstracted from the words using parallel words to deduce this, or
the material added to stems is constructed through a series of rules which create
phonological form rather than sequences of morphological elements (Blevins
2006). In either instance, the meanings attached to the grammatical material
(i.e. not to the stem) are given by the grammatical system of the language
concerned. If the language has inflectional case, the number of cases and their
fundamental nature is supplied by the grammar, with a limited number of
potential contrasts; if the language has inflectional tense, the way of dividing
time into a limited number of options is presented by the way the language is
used. Learners of the language who are not linguists may not name the various
possibilities, but the system is there in the language to which they are exposed.
The function associated with the -us in Latin lupus ‘wolf’ is also associated
 10 The Meaning of Words 87

with the function assigned to -a in puella ‘girl’, but with a different set of stems.
The meanings provided by the language fill empty slots that the language
requires to supplement the stem. That is, the meanings (which, as linguists,
we might gloss as ‘nominative singular’ or ‘3rd person of the present indica-
tive’ and so on) are added to the meaning of the stem without much that is
unpredictable except in idiomatic structures such as the plural of brain being
used to mean ‘intelligence’.
The meanings associated with the stems are not limited in parallel ways, as
we have seen. The learner’s understanding of the stem increases with exposure
to the stem, in the way discussed earlier in this section. Even the meaning
associated with transpositional derivation is not simply additive because, for
instance, postage has a meaning of ‘cost’ which is associated directly neither
with the stem nor with the nominalization (not even with having
a nominalization with -age, which in other words such as marriage does not
carry that meaning). In other words, the way meaning arises in derivational
morphology and inflectional morphology is different enough to suggest that
derivational morphology cannot be treated in the same way as inflectional
morphology in a word-based model. We cannot look at a word like carriage
and deduce which part of the meaning arises from which part of the form or
present a series of rules that will produce the form required to mean ‘a section
of a train used for transporting passengers’. This is part of the overspecification
of derived forms in comparison with what can be calculated on the basis of its
formal make-up. The overspecification is clearly word-based: ‘used for trans-
porting passengers’ is not an association with either carry or with -age.
Consider, then, a word like roaster. The meaning of the stem, roast, is given
(though not necessarily fully known), in the same way that it would be in
roasted. The rest of the meaning is not limited by the linguistic system, and
some of it is not carried by the linguistic system: whether a roaster is a person or
a utensil seems to be something that falls into that category, along with the fact
that a roaster can be something destined to be roasted or a very hot day. These
various meanings of roaster have in common that they all indicate an entity
connected with roasting. That is the most that can be attributed to the suffix. But
how does that meaning arise? If the meaning is associated purely with the word
roaster, then any affix would be suitable, just enough to indicate that derivation
occurs, and possibly to indicate the word-class of the output. That is not how
English (or any other language with derivation) works. Attaching the meaning
to the suffix -er, on the other hand, would be tantamount to having morphemes,
which word-based morphology does not want. So we have an option, following
Beard’s work, of saying that we have a nominalization, but the form of that
nominalization is, in some way, idiosyncratic and not necessarily directly
linked to the form used. This seems plausible, but would need some elaboration
on how it would work.
88 Part II Semantic Questions

A better alternative for deducing the meaning of a derivative – though not

a sufficient one – is to start from the root and find the largest section of the word
for which a meaning is familiar in one’s personal lexicon. The person hearing
the word then has to add such meaning as is definitely provided by the
remaining affixes (sometimes no more than word-class), then deduce the
meaning of the word in context. Finally, relevant inflectional meaning has to
be added (or, in English, where a form like inform may be an infinitive, an
imperative, or a present-tense form for most persons of the verb, the inflectional
meaning must be deduced from context). In cases where a derivative shows
a meaning of the base that is not available to the base in isolation (Bauer and
Valera 2015), such a process will not work, and further deduction or calculation
will be required. That this process demands a distinction between inflection and
derivation (and thus demands a lexeme) is not a problem in a theory that is
based on the lexeme.

Challenge
Find some unfamiliar derived lexemes (this book contains many examples).
What can be deduced about their meaning on the basis of their form? What
parts of their meaning cannot be deduced? Contrast this with some very
infrequent inflected forms of morphologically simple lexemes (word-forms
like greyest, strived). What parts of the meaning can be deduced from the
form in such cases? If you assume that the full meaning is associated with
the word, how do you relate phonological form with the meaning
expressed? If you assume that the meaning associated with the form is
supported by the form, what elements of the meaning can you find support
for? What do you conclude about what the form of a word tells you about
its meaning?

References
Basilio, Margarida Maria de Paula. (2009). The role of metonymy in word formation:
Brazilian Portuguese agent noun constructions. In Klaus-Uwe Panther, Linda
L. Thornburg & Antonio Barcelona (eds.), Metonymy and Metaphor in
Grammar. Amsterdam: Benjamins, 99–144.
Bauer, Laurie. (2005). The illusory distinction between lexical and encyclopedic infor-
mation. In Henrik Gottlieb, Jens Erik Mogensen & Arne Zettersten (eds.),
Symposium on Lexicography XI. Tübingen: Niemeyer, 111–15.
Bauer, Laurie & Salvador Valera. (2015). Sense inheritance in English word-formation.
In Laurie Bauer, Lívia Körtvélyessy & Pavol Štekauer (eds.), Semantics of
Complex Words. Cham: Springer, 67–84.
Blevins, James P. (2006). Word-based morphology. Journal of Linguistics 42, 531–73.
Chomsky, Noam. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
 10 The Meaning of Words 89

Clark, Eve V. 2023. A gradualist view of word-meaning in language acquisition and

language use. Journal of Linguistics 59, 737–62.
Courtney, Rosemary. (1983). Longman Dictionary of Phrasal Verbs. Harlow: Longman.
Davies, Mark. (2004–) BYU–BNC [based on the British National Corpus from Oxford
University Press]. Available online at http://corpus.byu.edu/bnc/.
Falkum, Ingrid Lossius & Augustin Vicente. (2015). Polysemy: Current perspectives
and approaches. Lingua 157, 1–16.
Kristofferson, Kris & Fred Foster. (c.1969). ‘Me and Bobby McGee’. Sung on the album
Kristofferson (1970), released by Monument [a version of the song sung by Janis
Joplin differs in some relevant respects from the version cited here].
Lear, Edward. (1871). Nonsense Songs, Stories, Botany, and Alphabets. Boston:
Osgood.
Marr, Vivian (ed.). (2008). The Chambers Dictionary. 11th ed. Edinburgh: Chambers
Harrap.
Panther, Klaus-Uwe & Linda L. Thornburg (2002). The roles of metaphor and meton-
ymy in English -er nominals. In René Dirven & Ralf Pörings (eds.), Metaphor and
Metonymy in Comparison and Contrast. Berlin: De Gruyter Mouton, 279–322.
Pustejovsky, James. (2017). The semantics of lexical underspecification. Folia
Linguistica 32, 323–47.
Siddiqi, Daniel. (2019). Distributed Morphology. In Jenny Audring & Francesca Masini
(eds.), The Oxford Handbook of Morphological Theory. Oxford: Oxford University
Press, 143–65.
Sirvid, Philip. (n.d.). Daddy long-legs spider (https://collections.tepapa.govt.nz/topic/
9428 accessed 31 August 2021).
Spencer, Andrew. (2019). Manufacturing consent over Distributed Morphology. Word
Structure 12, 208–59.
Weinreich, Uriel. ([1962] 1980). On Semantics. Edited by William Labov & Beatrice
S. Weinreich. Philadelphia: University of Pennsylvania Press [originally published
in International Journal of American Linguistics 28(2), pt. 4 (1962), 25–43].
Wodehouse, P.G. ([1919] 2012). Doing Clarence a bit of good. In My Man Jeeves.
Project Gutenberg EBook #8164.
11 Reflections on Tautology and Redundancy

11.1 Introduction
Anyone who remembers the political situation in the 1970s will recall the
Strategic Arms Limitation Talks between the USA and the USSR. Strategic
Arms Limitation Talks being too long for the headline writers, an acronym was
used instead of the full expression. But the talks were not simply referred to as
SALT, but rather as SALT talks. Since the T in SALT stands for talks, this was, in
effect, Strategic Arms Limitation Talks talks. This looks like a prima facie case
of tautology, saying the same thing twice, classically viewed as a stylistic fault.
However, one of the words that is sometimes seen as a near-synonym of
tautology is redundancy: if you say the same thing twice, one of those expres-
sions is unnecessary or redundant. The difference is that while tautology is
traditionally seen as a bad thing, redundancy may be a good thing.
All natural language contains redundancies. If we say these things, we have
made both words plural, though it would be less redundant to mark the plurality
only once. If we say pill and then bill, the two are distinguished not only by the
fact that the /b/ may be slightly voiced, but by the fact that the /p/ is aspirated
and uttered with greater force. Logically we should not need three ways to
make the distinction. If we say I’m coming the am and the -ing go together to
make up the continuous aspect, though the -ing alone would convey the
message. Legal language is full of expressions like goods and chattels, which
really mean ‘goods and goods’ but are intended to make sure everything
possible is covered by the phrase. We know that the Arabic spelling system
does not mark vowels, and *v*n *n *ngl*sh w* c*n *nd*rst*nd * wr*tt*n t*xt
w*th**t v*w*l l*tt*rs (it is a lot harder if the position is not marked or if only
the vowels letters are left instead of the consonants). All this means that when
we hear someone speaking to us in a crowded room or another place where
there is a lot of background noise, if we are listening to someone speaking over
a bad phone line, or if we are slightly deaf, we can still understand speech
because so much redundancy is built in. At a different level, orators and
teachers find it useful to repeat key points of a message to make sure it is
grasped by the audience: they may do this in slogans which simply repeat the

90
 11 Tautology and Redundancy 91

same wording, or they may reformulate the message, but repetition, a special
case of redundancy, is not necessarily a bad thing.
Two questions arise: how much duplication does word-formation create?
And is it tautology or redundancy when we find it?

11.2 Acronyms and Initialisms

The kind of apparent tautology illustrated by SALT talks is quite common with
acronyms and initialisms. Some examples are given here.
ABS system, ATM machine, Covid virus/disease, GPS system, HIV virus,
ISBN number, PIN number, RAT test, SAT test, VIN number

In most of these cases, it is the head word of the underlying phrase which is
repeated, though note Covid, which may refer to the virus as well as the disease.
This is not an absolute rule, though. An example which is often cited is Please
RSVP, where the SVP stands for French s’il vous plaît, ‘please’. The trouble
with acronyms and initialisms for the user is that people tend to forget what the
letters stand for, and simply use the new form as a name, without any analysis
of that name. Many people do not know what Covid actually stands for, not
even whether it refers to the disease or to the virus which causes the disease.
Where the acronym or abbreviation represents foreign words, as with RSVP,
KGB, SPQR, the lack of complete understanding is likely to be much greater.
Even internet abbreviations, typically much more recent and more transparent,
can be misunderstood: some people thought that lol meant ‘lots of love’ when it
first came out, rather than ‘laugh out loud’.
If you are not fully aware of what the letters are supposed to stand for, or
if you are not sure if your interlocutor knows what they stand for, then
adding some clarification can be useful, and may not actually be tautology
at all in the mind of the user, because there is no sense of repetition being
involved. Where the acronym or initialism is fully understood, the extra
gloss can be omitted, and we can talk about ABS, ATM, ISBN and so on
with no new head noun.
Although these examples look as though they are tautological, they probably
are not for many users; they may actually be useful for some listeners, and those
speakers for whom they are not necessary can easily omit them.

11.3 Tautological Compounds

Tautological compounds are those where one of the elements appears to
recapitulate information provided in the other element. There is not always
consensus as to what counts as a relevant example, but some probable cases are
given here.
92 Part II Semantic Questions

alleyway, cobblestone, cod fish, collie dog, courtyard, elm tree, flagstone, oak
tree, pathway, pine tree, poodle dog, pussycat, tuna fish, widow woman
Although these all look parallel, they are not all the same. To begin with,
some are in more general use than others. Collie dog seems more usual to me
than poodle dog, in turn more usual than Labrador dog (which I have seen
listed but am not familiar with). Clearly, other speakers may feel differently
about these, but the fact of variation is not in question. Second, some of these
contrast with a different right-hand element: oak tree contrasts with oak wood,
which could be seen as an argument against the use of tree being tautologous.
Elmwood and pinewood are also found. Cod fish and tuna fish have no corres-
ponding contrasting element.
They differ in other ways as well. It seems that any tree name can have the
element tree inserted after it: beech tree, birch tree, fir tree, kauri tree, palm
tree, sycamore tree. It is not clear whether there is analogy with expressions
such as apple tree, cherry tree, pear tree or not (though note we get banana
palm, date palm rather than banana tree, date tree). Not every fish can have
fish inserted after it, though: herring fish, sardine fish, snapper fish, trout fish,
all sound odd to my ears, although whether they are impossible is something
that would require further research. Courtyard seems to be like the phrases
goods and chattels, last will and testament in the sense that court is French
and yard is English, and putting the two together simply covers all the
possibilities.
On the other hand, all the examples given here would mean the same if the
right-hand element were omitted, and the shorter version is probably the more
common. This suggests that the right-hand element is redundant in these
cases. Although this is true of widow woman today, there used to be a term
widowman (‘widower’), so that widow has not always been restricted to
women.
Benczes (2014: 445) argues that compounds of this type are not tautological
because they are
used to dignify and upgrade concepts via the conceptual metaphor more of form is
more of content, whereby a linguistic unit that has a larger form is perceived to carry
more information (that is, more content) than a single-word unit.
While I would not wish to dismiss this notion, it does raise the question of
whether forms like oak tree are single words or word-sequences, and it also
raises the question of whether adding an element like tree can provide informa-
tion on the “dignity” of the word rather than to provide semantic information
about the kind of plants involved. The lack of redundancy could also be due to
other reasons, for example a requirement for emphasis, which might explain why
oak and oak tree can both be used, another factor which Benczes recognizes.
 11 Tautology and Redundancy 93

11.4 Tautological Names

Tautology in names often arises through words from another language being
used in the relevant expressions. Consider, for example, the River Avon in
Stratford-upon-Avon. Avon is a Celtic word meaning ‘river’ (in modern Welsh,
it is spelled afon), so that the River Avon is, technically speaking, just repeating
the same content twice. There are many more examples from all over the world.
Consider, for instance, Māori loans in New Zealand, which give Lake Rotoiti
and Mount Maunganui. Rotoiti means ‘little lake’, Maunganui means ‘big
mountain’, so that the names, if translated, give ‘lake little lake’ and ‘mount
big mount’. The La Brea Tar Pits in Los Angeles, when the Spanish is
translated, give ‘the the tar tar pits’. Penhill, in Yorkshire, contains another
loan from Celtic and means ‘hill hill’.
Since the point of these examples is presumably that speakers do not
understand the non-English label, it is hard to say that these are really tautolo-
gous. Avon (and so on for the other examples) is just a name, and the English
version is the only piece to provide information. This is no different in principle
from what happens in a word like children. Child is the singular, and the plural
used to be childer (still found in some dialects). But when -er stopped being
used to mark plurals, a new plural was needed, and the -en (as in oxen) was
added to childer to make it plural. Dutch still has the plural kinderen from kind
‘child’ where the two plurals are more clearly marked. Other double plurals in
English include bacterias, dices, phenomenas, some of which are not viewed as
standard.

11.5 Conclusion
It seems that such tautology as there appears to be in these formations is not
clearly tautologous for language users, and, indeed, is often not even redundant.
In terms of the norm of English (see Chapter 6), that is just the way things are
done. It may be part of the phenomenon of using longer words for important
things (Janda 2021), but even that is not clear.

Challenge
Look for double plurals in some large corpus. How common are they? Do
they all have the same basic motivation of ensuring that the plurality is
clearly marked? Do fish, sheep and deer get marked for plural? Why (not)?
How do double comparatives and superlatives (more prettier, most pretti-
est) fit into the general pattern? Do we find double marking of derivational
categories?
94 Part II Semantic Questions

References
Benczes, Réka. (2014). Repetitions which are not repetitions: The non-redundant nature
of tautological compounds. English Language and Linguistics 18, 431–47.
Janda, Richard D. (2021). Perturbations, practices, predictions, and postludes in
a bioheuristic historical linguistics. In Richard D. Janda, Brian D. Joseph &
Barbara S. Vance (eds.), The Handbook of Historical Linguistics, vol II.
Hoboken, NJ: Wiley, 523–650.
Part III

Syntactic Questions
12 Reflections on Recursion

It is noted in Bauer (1983: 69) that in a few places in English there appears to be
recursion in suffix sequences. Two patterns are given there, one involving -ation,
-al and -ize, where any one of the suffixes can be the first in a sequence of three, as in
sensationalize
formalization
civilizational
and the other involves the suffixes -ic, -al and -ist in a similar way:
classicalist
nationalistic
egotistical
Both these examples give rise to problems of analysis. Is -ation a single
suffix, the only productive affix containing the Latin -ion formative or is it
a sequence of -ate and -ion? Is -ical an independent suffix, or is it a sequence
of -ic and -al? And is -istic an independent suffix, or is it a sequence of -ist and
-ic? (There is also the possibility that both patterns exist.) Merging of affix
sequences into single affixes has been termed affix-telescoping (Haspelmath
1995: 3–6). Whether telescoping applies or not may not matter in the sense
that any rule-based system is going to have to allow recursion in this set of
affixes, and therefore allow for potential words like sensationalization (noted
by Miller 1981: 114) as well as for words like institutionalizationalize, which
do not occur. Ljung (1970: 13) notes that the longest suffixal string he has
attested contains just four suffixes, but as we have already seen, such numbers
depend on how you count some suffixes.
Sequences of two suffixes are common in English, and are dealt with in
descriptive terms in Bauer et al. (2013: ch. 27) and in theoretical terms in works
in Lexical Morphology and by Fabb (1988). It is worth noting that such sequences
do not generally allow for recursion of the same suffix. Note, however, the
construction illustrated by breaker-upperer, discussed in Chapter 13, the word
perfectionation, where the -at(e) is arguably an independent suffix, so this is not
strictly a straightforward repetition, and fractionation, where the same is true, and
where the first -ion is attached to a bound root, which may be important. Note also

97
98 Part III Syntactic Questions

the rather unusual example below (which might be argued to involve homonymous
affixes rather than repetitions of the same affix).
I’m only kind of Jewish. I’m Jewish-ish. (John Connolly. 2018. The Woman in
the Woods. London: Hodder & Stoughton, p. 69)
Repetitions of the same sound-sequence are allowed, e.g. in adulterer,
ringing, but even such sequences are not common.
Sequences of three suffixes are rare, anyway. Some examples are given below.
Christianizer
dictatorialness
establishmentarian
justifiableness
musicianship
polarization
polarizer
pressurizer
provincialization
Prussianization
sensationalism
truthfulness
verbalizer
The list of types here is not exhaustive, and this list is based on established
words, while nonce words may provide a wider range of types. Yet there is
sufficient data here to suggest some generalizations.
The suffix -ize always permits subsequent suffixation with -ation. Since this
is one of the main sources of -ation suffixation, this is not surprising. The same
seems to be true with -er added to -ize. Again, given the productivity of -er
suffixation on verbs, this is not surprising. Given that -ize can attach to words
with several different suffixes, this means there is a large pool of available
bases to lead to three-suffix sequences. It also suggests that computerizer (with
recursion of -er) should be possible, though it mainly seems to be used as
a trade name. The same is true with containerizer.
The same argument applies to -ness. This suffix attaches very freely to
adjectives, and there are many adjectives that are created by adding a suffix to
a noun, which may already end in a suffix, so a route to three-suffix sequences is
assured. Some adjectives of Latinate origin (or suffixes in such words) prefer -ity
to -ness, but -ness is always available as an alternative – not always
a synonymous alternative – even, for instance, with realness alongside reality.
Dixon (2014: 390–1) cites some examples with four derivational suffixes in
a sequence, but this seems to be seen as something plausible rather than
something attested. Nationalistically is attested, if most often in dictionaries,
 12 Recursion 99

but with -istic and -ical in the word, counting the affixes seems fraught. Dixon
even suggests subsidizationalistic as having five, but this is not attested (and
might be four if -istic is seen as a single affix).
This gives rise to a hypothesis about possible sequences of suffixes. The
hypothesis is as follows:
In any sequence of two or three derivational suffixes at the end of a word, if the
word is not item-familiar with all the observed suffixes, just one suffix has been
added to a word which is item-familiar.
Bauer et al. (2013) list a few apparent exceptions (such as foolageness and
witchessery), but these turn out not be exceptional, because foolage and
witchess are in use, even if they are not particularly common. If the hypothesis
holds up, then recursion of suffixes is not a genuine phenomenon. Where we
think we find it, it is the result of an analysable suffix being added to a (more or
less) unanalysable word, and any recursion is the result of the same affix being
found in established words and available for productive use. The hypothesis
does not mention sequences of four suffixes, because it assumes affix-
telescoping. If any such really exist, I would guess that the same rule holds.
Dealing with prefixes is slightly more difficult in that it is not necessarily clear
whether a given form is a prefix, an initial combining form from a neoclassical
formation or a preposition forming the first element of a compound. Bauer et al.
(2013) are quite liberal in what they accept in their lists. Even then, their examples
fit with the proposed hypothesis. The jocular psychosociopseudohistorian (cited by
Bauer 1983: 68) goes against the general pattern (although these are combining
forms rather than real prefixes), but this may be part of its deliberate effect. Bauer
et al. (2013: 498) give only sequences of two prefixes, though they also find sub-
sub-sub-contractor (2013: 499), and Dixon (2014: 392–4) also finds a limit of two.
If this holds, it has important implications. First, although compounds may
show different bracketings, derivations must always show left-branching for
prefixes and right-branching for suffixes: [(prefix) [(prefix) [(prefix) [[[[root]]]]
(suffix)] (suffix)] (suffix)]. Only when prefixes and suffixes interact do genuine
ordering problems arise (Bauer et al. 2013: 501). Second, the most marginal
affixes are likely to be the most productive ones, the innermost (those nearest
the root) the least productive ones. Inflection has not been mentioned here, but
if an inflectional affix is added – possibly leading to a sequence of four
suffixes – it is (nearly) always on the righthand edge of the word, and inflection
is, by definition, extremely productive.
Note that the repetition of prefixes is possible, as long as this makes
pragmatic sense: meta-meta-rule is grammatical (Dixon 2014: 393 cites others
that sound totally plausible, Bauer et al. 2013: 499–500 give some attested
examples), and Bauer et al. also cite examples of synonymous or nearly
100 Part III Syntactic Questions

synonymous prefixes being chained: micro-minivan and mega-superstar. The

hypothesis given above for suffixes seems to apply here for prefixes.
It is unlikely that any of the mega-mega-rich will be invited to that particular
blow-out. (Ben Elton. 1989. Stark. London: Sphere, p. 13)
Stein (1977) points out that we prefer to say things like re-write again rather
than re-re-write, not unfair rather than un-unfair and rather shortish rather than
shortishish, which means that the need for the repetition of affixes is not great.
At the same time, such limits as there are must probably be considered to be
pragmatic limits rather than being determined by the structure of the grammar.
Recursion in compounds is well attested at the level of the word-class: that is
long compounds made up of sequences of nouns are regularly found, for
example in headlines, although they rarely become established. Even here,
the length of such compounds seems to be more limited in English than in other
Germanic languages. Hansen (1938: 113) comments that in Danish “news-
papers have been seen to start competitions where competitors had to create the
longest word” (my translation). Quite apart from finding a definition of a word
in English which would support such an enterprise, the result in practical terms
would probably be quite limited: six elements is very long for English.
Louis and I sat eating chocolate-chip-cookie-dough ice-cream. (John
Connolly. 2000. Dark Hollow. London: Hodder & Stoughton, p. 233)
Recursion of the same lexeme in a compound is rare, but not impossible,
most obviously in repeated-element compounds:
“Mine is green.”
“Green green? Or more like an olive green?” . . .
“The salesman called it olive”, he said. (Lawrence Block. 2000. Hit List.
London: Orion, p. 118)
Think me up a way to earn some blunt, would you? Not earn-earn it, but come
into it proper-like. (Grace Burrowes, 2017. Too Scot to Handle. New York:
Forever, pp. 236–7)
This is just a storm-storm. But it’ll be a whopper. (A.J. Finn. 2018. The
Woman in the Window. New York: Morrow, p. 137)

These repeated-element compounds or ‘identical constituent compounds’

(Hohenhaus 2004) usually give rise to the meaning ‘a real ~, a proper ~,
a prototypical ~’, but as Hohenhaus shows, not inevitably. They might be con-
sidered another type of tautologous construction (see Chapter 11), but the meaning
suggests that they are not tautological. They do seem to show recursion, though.
Since a syntactic construction can be used to modify a head noun, and this is
usually considered to form a compound (Bauer et al. 2013: 456–7), syntactic
recursion can also be relevant in compound formation.
 12 Recursion 101

And there was straight extortion, too, genuine strong-arm, give-me-the-

money-or-I’ll-blow-your-brains-out scenarios. (Stephen Solomita. 1989.
Force of Nature. New York: Putnam, p. 24)

Recursion is found in compounds where ‘recursion’ simply means that nouns

can be strung together, and in derivation in the same sense that affixes can be
strung together. Repetition of the same affix (or the same word in compounds)
is rare, and occurs in a few relatively predictable contexts. This means that
recursion of a specific element is rarer than might appear at first glance.

Challenge
What evidence can you find that might resolve the status of -ation, -ical and
-istic in English? Do these suffixes always have the same status, or can the same
sequence realize two distinct patterns of underlying elements? Does it make
any difference if you consider only productive uses (e.g. if you consider only
words found in newspapers or online that are not listed in some reasonably
extensive dictionary)?

References
Bauer, Laurie. (1983). English Word-Formation. Cambridge: Cambridge University
Press.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Dixon, R.M.W. (2014). Making New Words. Oxford: Oxford University Press.
Fabb, Nigel. (1988). English suffixation is constrained only by selectional restrictions.
Natural Language and Linguistic Theory 6, 527–39.
Hansen, Aage. (1938). Indledning til nydansk grammatik. Århus: Aarhus Universitetsforlag.
Haspelmath, Martin. (1995). The growth of affixes in morphological reanalysis.
Yearbook of Morphology 1994, 1–29.
Hohenhaus, Peter. (2004). Identical constituent compounding – A corpus-based study.
Folia Linguistica 38, 297–331.
Ljung, Magnus. 1970. English Denominal Adjectives. Lund: Acta Universitatis
Gothoburgensis.
Miller, George A. (1981). Semantic relations among words. In Morris Halle,
Joan Bresnan & George A. Miller (eds.), Linguistic Theory and Psychological
Reality. Cambridge, MA: MIT Press, 60–118.
Stein, Gabriele. (1977). The place of word-formation in linguistic description.
In Herbert E. Brekle & Dieter Kastovsky (eds.), Perspektiven der
Wortbildungsforschung. Bonn: Bouvier, 219–35.
13 Reflections on Problems with Heads
in Word-Formation

Williams’s (1981) right-hand-head rule, proposed to be general rule of

morphological structure, continues to be very influential insofar as it relates
to English. Although the right-hand-head rule, which states that the rightmost
element (morpheme) in a word is the head of that word, has since been subject
to a great deal of criticism (see e.g. Bauer 1990, 2019: 61–5), where compounds
are concerned it has held up remarkably well, and may even be considered part
of the definition of a compound in English.
The head of a compound is usually defined as being the element in the
compound from which the compound-as-a-whole inherits its word-class (noun,
adjective, verb), from which, other things being equal, it receives its inflectional
behaviour and which is the superordinate term (sometimes called a ‘hyperonym’)
of the compound. Snowgoose is a noun because goose is a noun, its plural is
snowgeese because the plural of goose is geese, and snowgoose is hyponym of
goose, and so goose is the head in snowgoose. Although there are many major
classes of exception to these general principles (some of them dealt with else-
where in this volume, see e.g. Chapter 29) these principles are generally taken to
hold for endocentric noun and adjective compounds in English, and if verbal
compounds are excluded, it is because some authorities still see verbal com-
pounds as being rather marginal in English.
One problem with what has just been said is the inclusion of the word
endocentric. Since endocentric means ‘having an internal centre’, where centre
is another term for head, this risks becoming circular. Endocentric contrasts
with exocentric, and an exocentric construction is one that does not have an
internal head, and when applied to compounds (Bloomfield 1935, although the
term appears to go back to Aleksandrow 1880 – see Carr 1939: xxvii), means
that there is an external element which fulfils the functions of a head. That
external element, however, is invisible. To take a standard example, consider
the compound redcap. Redcap is not a hyponym of cap, and though cap is
a noun and redcap is also a noun, redcap is not a noun by virtue of this fact, but
by virtue of the fact that it denotes an entity. Just what entity it denotes depends
to a large extent on the dialect you speak: it could be a railway porter (US
English), it could be a military policeman (British English), it could be a goblin
(Scottish English) or a bird. But like the names for other things of the same

102
 13 Problems with Heads in Word-Formation 103

type, redcap is a noun. There is nothing in any of this to suggest that its plural
should be anything but regular, so we gain no extra information from the fact
that it is. The missing head must be something that bears a meaning such as
‘person’ for the first two readings, a meaning such as ‘creature’ for the ‘goblin’
reading, and ‘bird’ for the last reading. Just what else this presumed head
contains is obscure. Because of the right-hand-head rule, we assume that it has
an underlying position to the right of redcap, and is either deleted on the way to
the surface, or is an element which never has any form, however this is to be
achieved in a grammatical model. We do have a problem in that big top ‘circus
tent’, presumably also an exocentric compound since it is neither a big nor
a top, has no obvious meaning to attribute to the head. If the head means ‘tent’,
then we must account for the fact (as we do with other examples) that big top
tent is a rather odd expression. And with redhead ‘person with red hair’
(another exocentric compound of similar structure) we must account for the
fact that redhead person is not grammatical; we would need to have red-headed
person (Bauer 2022).
Such factors might lead us to seek a better solution for these exocentric
compounds, and fortunately one is easily available in the literature (Bauer
2016, 2017b and references there). The proposal is that just as we can talk
about the crown and mean ‘the monarch, the person closely associated with the
crown because of wearing it’, using a figure of speech (here metonymy), so we
can talk about a redcap and mean ‘the entity that wears/possesses/has a red
cap’. Metonymy, or more specifically synecdoche, is involved. If redcap means
what it does by virtue of various figurative readings, we do not need an empty
head element, just as we do not need an empty element as the head in crown
‘monarch, government’. The right-hand-head rule must apply to the compound
as it appears (so the head is cap), but the hyponymy rule is not relevant because
the figurative interpretation makes it irrelevant, just as crown has to be referred
to as it when it means the headwear, but as he or she when it means ‘monarch’.
Not only does this avoid having a word like redcap whose meaning is explained
both by it being a figurative reading and by it being an exocentric compound, it
also allows us to extend the right-hand-head rule for English compounds in
a satisfactory way. Moreover, it may be enough to extend the notion of
endocentricity so far that we can get rid of the notion of exocentricity com-
pletely (for further arguments on this, see Bauer 2016).
That is not the end of the story, though. There is another set of compounds,
coordinative compounds, where this solution seems less successful.
Coordinative compounds are those that are made up of two (or more) elements,
each of which is on an equal footing in interpreting the compound, just as we
find in syntax when words or other constructions are coordinated. Since just
where the borderline of coordinative compounds runs in English is awkward
(Bauer 2023, ch. 22) just a handful of relevant examples will be cited here.
104 Part III Syntactic Questions

They include Alsace-Lorraine, HarperCollins, dinner-dance, historical-

philosophical, bitter-sweet, cough-laugh, drink-drive. These have problems for
the notion of head. For example, while Alsace-Lorraine is a hyponym neither of
Alsace nor of Lorraine, dinner-dance is a hyponym of both dinner and dance.
Perhaps because of the existence of these two types, some scholars see coordina-
tive compounds as having no head, and some see them as having two heads,
although it is not clear just what is implied by any construction having two heads,
just as it is not necessarily clear what is implied by a construction having no head
(an invisible external head is only one possible interpretation of this – for
extensive discussion, see Matthews 2007). In syntax, a more modern position
than that of Bloomfield (1935) is that all structures are endocentric, and so all
have a head. We might expect the most helpful criterion to be just what the
compound-as-a-whole inherits from the individual elements. English provides us
with very little information on this, though French and Spanish, for example,
tend to get the gender of the compound from the left-hand element: French taxi-
camionnette ‘taxi van’ gets its masculine gender from taxi (Arnaud 2004: 337),
though we might not be sure whether taxi van or its French counterpart are
genuinely coordinative compounds. English compound verbs, like drink-drive,
might be expected to provide similar information, but not many of them have
irregular past tenses, and with drink-drive, the lexeme occurs almost exclusively
in the infinitive (often appearing in attributive position) or in the -ing form. In any
case, in English any inflection has to occur on the right-hand edge of the word,
whether we think this is the head or not: we could never have a form like
*drinking-drive, unless drink-drive was thought to be a syntactic sequence of
two independent words. Most dictionaries do not list a past tense for joy-ride, but
Collins (2006) gives joy-rode. Another potential example, test-drive, is listed in
dictionaries as having the past tense test-drove, but it is not clear whether it
should be interpreted as ‘to test and to drive simultaneously’ (a coordinative
reading) or ‘to drive by way of a test’ (a subordinative reading); if it is to be
glossed as ‘to test by means of driving’ then not only is it subordinative, but left-
headed in terms of the hyponymy criterion.
One piece of information that English does allow us to consider is which
element speakers feel to be the most important one in such constructions.
Unfortunately, the place where we have information is in a type not listed
above, and often (misleadingly) called appositional compounds, like singer-
songwriter, actor-director and possibly café-restaurant in which the same
entity is named by two facets of its nature. The type is very common, not
only in English, but in other European languages, too, but whether these
examples are compounds or not may be controversial, an alternative analysis
seeing them as instances of syntactic coordination without an overt coordinat-
ing element. If we accept their compound status, though, then the compound is
a hyponym of each of its elements, as we have seen with other examples above.
 13 Problems with Heads in Word-Formation 105

to bask in the painter-writer-musician glamour of the place. (Lawrence Block.

1999. The Burglar in the Rye. Harpenden, Herts.: No Exit, p. 9)
Stephen wasn’t happy with only being a doctor, however. He was a doctor-
screenwriter-producer. (Randall Hicks. 2005. The Baby Game. San Diego:
Wordslinger, p. 32)
That’s a question you cannot ask a lawyer. Or even teacher-lawyers. Or even
builder-teacher-lawyers. (Nury Vittachi. 2008. Mr Wong Goes West. Crows
Nest, NSW: Allen & Unwin, p. 81)
Bauer (2010: 39) cites the example below, which seems to indicate that the
speaker sees the left-hand element as being the more important in such a word.
“I am a lawyer-musician, not a musician-lawyer”, he says. “My calling is
the law.”
A similar implication is provided by the following.
My best friend’s a demon witch. Or it’s probably a witch demon, because
there’s more of the witch. (Nora Roberts. 2022. The Choice. New York:
St Martin’s Press, p. 161)

This is surprising, because if only one element is the head, we would expect
it to be the right-hand element in English, following the right-hand-head rule.
This is obviously less clear than it might seem.
A similar problem is raised by some coordinative blends. The difference
between a tigon and a liger is whether the lion is the sire or the dam: sire first,
dam second. The same is true of the difference between zorse and hebra, but does
not appear to apply to labradoodle, for example, even though poobrador is also
attested. In those instances where the distinction is upheld, however, the two
elements coordinated in the blend are not of equivalent status. Here it is not
clear to me which is perceived as the more important: the dam, whose status as
a parent is beyond doubt, or the sire, in line with our culture’s traditional focus on
the male. Another case in point is Japlish (and also other parallel words, though
not always to the same extent) which can be a variety of Japanese with a lot of
English words, or a variety of English with strong Japanese influence (Bauer
2017a: 161). Because the evidence comes from very few examples, perhaps the
best we can say in all these cases is that in coordinative word-formation, the
position of the head seems to vary, and not be consistently on the right. If we see
these words as having two heads, we have to account for differences of interpret-
ation, which might be harder than allowing left- or right-headed constructions.
There is also a very small number of apparent compounds which are left-
headed. The two most frequently cited examples are endgame (which is a kind
of end rather than a kind of game) and man/girl Friday, which is not a hyponym
of Friday. We can dismiss these as not being compounds, but then we want to
know what their structure is. They could be abbreviated forms of end of the
106 Part III Syntactic Questions

game and man/girl named Friday (possibly man/girl of Friday), in which case
they might be seen as compressed forms of structures such as maid-of-honour,
but if that is the case we might expect to find more of them, and for the
derivation to be clearer.
Another set of compounds also provides difficulties. These are compounds
like woman doctor. Compounds of this type look as though they are coordina-
tive compounds, because they can be glossed as ‘a woman and a doctor’ (see
Chapter 22), and some authorities treat them in this way (e.g. Fabb 1998).
Rather more scholars, though, right back to the Sanskrit grammarians, see such
forms as being subordinative, or more specifically, attributive or ascriptive,
with the first element being a gender marker. English has many compounds of
this form, including boy scout, girlfriend, gentleman-farmer, hen pheasant,
manservant, nanny goat or, with a different format, she-wolf.
However, while we have expressions where the gender-marker comes
first, as in the examples already cited and in baby rabbit, buck rabbit, bull-
calf, bull elephant, cock lobster, cock robin, dog-fox, nanny goat, tom cat;
we also find expressions where the gender-marker (sometimes an age-
marker) comes second, even though its function seems to be entirely paral-
lel: Arab mare, Clydesdale stallion, Hereford bull, lion cub, turkey hen,
woodcock. The order of the elements is largely predictable: a gender marker
goes in front of a species name, but after the name of a breed. But this is not
the whole story, since cub goes second (except in cub reporter, cub scout),
and hen can be either first or second in position. The question of headedness
here is important. Storch (1886 as cited in Carr 1939: xxviii) sees the second
type here as being left-headed. If we want to say that all these are right-
headed, then we have to assume that the gender is sometimes more important
than the animal, while on other occasions the animal is more important than
the gender, but it is not clear why this difference should be so predictable. If
we want to see a difference in headedness, we need to explain why the right-
hand-head rule should fail to apply in such a minor class.
A different minor-class is discussed by Bauer and Renouf (2001). This class
is made up of compound adjectives whose second element is only: fruit-only
(conserve), oestrogen-only (pill). As Bauer and Renouf point out, oestrogen-
only is a hyponym of oestrogen, so that on semantic criteria it looks as though
the left-hand element is the head. They also point out that if this is just a matter
of a syntactic expression being used attributively, we would expect only
oestrogen, which is the more common ordering in syntactic usage. If oestrogen-
only is not a compound, its left-headedness might not be a problem, but in that
case, we need an alternative analysis for it, and none seems to be forthcoming.
Similar problems arise with words such as sugar-free, but the free here may be
an affix rather than a compound element.
 13 Problems with Heads in Word-Formation 107

A different kind of problem is raised when the suffix -er is added to some
complex verbs. To think about this, we need a little background. Carstairs-
McCarthy (2002) suggests that words cannot inflect internally, so that because
we can have maids-of-honour but not *maid-of-honours, maid-of-honour must
be a phrase, but because we can have mother-in-laws (although mothers-in-law
is also quite normal), mother-in-law can be seen as a single word. This notion
does not seem to be particularly controversial, although it suggests that inflec-
tional markers go on the right-hand edge of a word, rather than on the head of
a word, which is often claimed (e.g. Zwicky 1985). We would expect, on the
basis of this precedent, to find derivational affixes on the right-hand edge of
a word, and mostly that is the case. There is one construction where this does
not work, though.
I always thought of him as a bit of a hanger-oner, if you know what I mean.
(Hazel Holt. 1997. Mrs Mallory and the Only Good Lawyer. New York:
Dutton, p. 125)
Doris, headwaitress/receptionist at the George and Dragon, was one of the
best finder-outers in the village. (Hamilton Crane. 1997. Bonjour,
Miss Seeton. New York: Berkley, p. 27)
A real fixer-upper [viz. a house in decay]. (Donna Andrews. 2003. Crouching
Buzzard, Leaping Loon. New York: St Martin’s Minotaur, p. 121)
You and I are the hander-outers. (Kerry Greenwood. 2004. Earthly Delights.
Crow’s Nest, NSW: Allen & Unwin, p. 106)
The netherworld of stardom-hanger-on-er. (Randal Hicks. 2005. The Baby
Game. San Diego: Wordslinger, p. 19)
I wasn’t sure that she herself wasn’t one of the holder-outers. (Susan Moody.
2006. Quick off the Mark. Sutton, Surrey: Severn House, p. 178)
She never gave up, which was odd, because she considered herself
a champion giver-upper. (Kate Kessler. 2016. It Takes One. New York:
Redhook, p. 4)

If, for example, give up is a phrase (which is how these verbs are often treated
in English), then we would expect giver-up. If give up is a word (which is how
similar verbs are treated in German, for instance), we would expect give-upper.
What we get in these cases is a blend of the two. The -er is added to the head of
the phrase (or, possibly to the verb, which provides the same result), and then is
added on the right-hand edge of the word, thus illustrating perfectly the half-
way status of constructs like give up. This happens, though, at the expense of an
apparent reduplication of the suffix, which is not normal behaviour. We would
not expect to find a person who is looked up (either visited, or searched for) to
be termed a lookee-uppee, though it is not clear what we could call them.
Neither would we expect, even if we accept the construction type sketched
108 Part III Syntactic Questions

above, to find, as we do in some varieties of English, a giver-upperer, where the

third -er seems completely redundant.
The Breaker Upperers is a New Zealand romantic-comedy film (Wikipedia)
The classroom mixer-upper-er (www.educationalvantage.com/EDVAN/prod
uct/PR309 accessed 11 July 2023)
improvised ping pong ball picker-upperer (www.reddit.com/r/mildlyinterest
ing/comments/ab5ufc/this_improvised_ping_pong_ball_pickerupperer_
made/ accessed 11 July 2023)

Since the idea of affixes being added to the right-hand edge of words in
English has been raised in this section, it is worth adding that occasionally we
find examples which seem to support such a conclusion. Two such examples
are given below.
Then he called a week later and said never mind, so I never minded. (Barbara
Parker. 2000. Suspicion of Malice. New York: Dutton, p. 65)
make suring they work (Radio New Zealand, 7 p.m., 20 March 2012)
Such examples may not be part of the normal standard grammar of English,
but they arise, and they contradict the common assertion that inflections are
added to the head of a word (see Bauer 2017a: 29–32 for more discussion).
Overall, the right-hand-head rule, which starts out looking so persuasive,
leads us into places where it creates problems and fails to give clear-cut
answers. It may still operate as a canonical rule – and, indeed, we have
seen some evidence that its application may actually be wider than was
originally thought. We still need some way of dealing with the problem
areas, though.

Challenge
Can you find any examples of English verbal coordinate compounds which
must be glossed as coordinate and cannot be glossed as being subordinate? Ask
the same question with regard to nouns and adjectives, though these two types
are slightly easier to deal with. Do you think there are any coordinate com-
pounds in English? Justify your position. Can you find any examples of words
that look like compounds but are clearly left-headed? How can you be sure that
they are both compounds and left-headed?

References
Aleksandrow, Aleksander. (1880). Litauische Studien, I. Nominalzusammensetzungen.
Doctoral dissertation from the University of Dorpat (now Tartu, Estonia).
 13 Problems with Heads in Word-Formation 109

Arnaud, Pierre J.L. (2004). Problématique du nom composé. In Pierre J.L. Arnaud (ed.),
Le nom composé: données sur seize langues. Lyon: Presses Universitaires de Lyon,
329–53.
Bauer, Laurie. (1990). Be-heading the word. Journal of Linguistics 26, 1–31.
(2010). Co-compounds in Germanic. Journal of Germanic Linguistics 22, 201–18.
(2016). Re-evaluating exocentricity in word-formation. In Daniel Siddiqi &
Heidi Harley (eds.), Morphological Metatheory. Amsterdam: Benjamins, 461–77.
(2017a). Compounds and Compounding. Cambridge: Cambridge University Press.
(2017b). Metonymy and the semantics of word-formation. In Nikos Koutsoukos,
Jenny Audring & Francesca Masini (eds.), Morphological Variation: Synchrony
and Diachrony. Proceedings of the Mediterranean Morphology Meetings vol. 11,
1–13. Available at http://mmm.lis.upatras.gr/index.php/mmm/issue/view/352
(2019). Rethinking Morphology. Edinburgh: Edinburgh University Press.
(2022). Exocentricity yet again: A response to Nóbrega and Panagiotidis. Word
Structure 15, 138–47.
(2023). Coordinative compounds, including dvandva. In Peter Ackema, Sabrina
Bendjaballah, Eulàlia Bonet and Antonio Fábregas (eds.), The Wiley Blackwell
Companion to Morphology. Hoboken, NJ: Wiley.
Bauer, Laurie & Antoinette Renouf. (2001). A corpus-based study of compounding in
English. Journal of English Linguistics 29: 101–23.
Bloomfield, Leonard. (1935). Language. London: Allen & Unwin.
Carr, Charles T. (1939). Nominal Compounds in Germanic. London: Oxford University
Press.
Carstairs-McCarthy, Andrew. (2002). An Introduction to English Morphology. Edinburgh:
Edinburgh University Press.
Collins (2006). Collins English Dictionary. 8th ed. Glasgow: HarperCollins.
Fabb, Nigel. (1998). Compounding. In Andrew Spencer & Arnold M. Zwicky (eds.),
The Handbook of Morphology. Oxford: Blackwell, 66–83.
Matthews, P.H. (2007). Syntactic Relations: A Critical Survey. Cambridge: Cambridge
University Press.
Storch, Theodor. (1886). Angelsächsische Nominalkomposita. Strasbourg: Trübner.
Williams, Edwin. (1981). On the notions ‘lexically related’ and ‘head of a word’.
Linguistic Inquiry 12, 245–74.
Zwicky, Arnold M. (1985). Heads. Journal of Linguistics 21, 1–29.
14 Reflections on Coordination
in Word-Formation

In phrases, elements of the same type can coordinate with each other quite freely.
Just what ‘of the same type’ means may require some explanation, since the type
may be a matter of word-class or of function or both: Cars and buses are allowed
in the city centre is fine because cars and buses are both nouns and are both
subjects of the verb, The cover of the book is red and white is fine because red and
white are both adjectives and both subject complements, All the vehicles going
into the city centre are red and cars is odd, not only because the message seems
unlikely – although if we had said are red cars it would be equally unlikely but
less grammatically weird – but because we are trying to coordinate an adjective
and a noun. Come here and sit down is fine because we are coordinating two verb
phrases, but Come and relaxing is odd because although come and relax are both
verbs, they are not sufficiently similar for the coordination to work.
Things are not always so easy to explain. He wore a green and yellow shirt
follows the rules, and is perfectly acceptable, the once and future king (a phrase
from Sir Thomas Mallory, describing King Arthur) is also fine although once is
an adverb and future is (superficially, at any rate) a noun, but both act as modifiers
dealing with time. Blue Danish porcelain is fine, but blue and Danish porcelain
is not (if it means the same thing) because we are not coordinating like with like.
But now consider the following. She was sure she had seen blackbirds and
bluebirds in the tree. We coordinate two nouns both of which are the direct object of
see, and all is well. She was sure she has seen black and bluebirds in the tree. We are
apparently coordinating two adjectives both of which modify the same noun, but
now there is something wrong. Yet She was sure she had seen yellow and green
leaves on the tree is fine again. And it is perhaps not altogether clear whether She
was sure she had seen red and grey squirrels in England is like black and bluebirds
or like yellow and green leaves. Red and grey in this sentence are both classifying
adjectives (red squirrel and grey squirrel are both types of squirrel while yellow and
green leaves are descriptive adjectives but do not create classes of leaf). One
solution to the problem is that modifiers to a noun head can be coordinated
(provided they are suitably alike) in a syntactic construction (a phrase), but not in
a compound (Payne and Huddleston 2002: 448–9). Moreover, the same is true of
the head nouns in such a construction: we can have black cars and buses (syntactic

110
 14 Coordination in Word-Formation 111

construction) but not blackberries and birds (at least not if the black also modifies
the birds). We can find many examples which seem to support such an explanation:
we do not get *They are building a new motor and railway, *Tear and raindrops
were running down her face, *I’ve got a tooth and headache. But it turns out that it
is difficult to be sure, and that speakers do not agree. Payne and Huddleston classify
ice-cream as a compound using this criterion, claiming that you cannot have ice-
lollies and creams, but Bauer (2014) cites examples such as the following:
Living on the broken dreams of ice lollies and creams (www.melodramatic
.com/node/70347?page=1 accessed 12 January 2011)
Far too many ice-lollies and creams had been consumed but we were all happy
little campers (http://yacf.co.uk/forum/index.php?topic=33253.120 accessed
16 January 2011)
These nine months were filled with dripping ice lollies and creams, spilt soft
drinks and lost maltesers (http://keeptrackkyle.blogspot.com/2006/07/tidy
ing-up.html accessed 16 January 2011)

Payne and Huddleston (2002: 450) admit that the line between compounds and
phrases is blurred, but see the fundamental distinction as vital and the failure of
the coordination criterion in some places as a price worth paying. We could also
claim that ice-cream is a compound for some speakers and a phrase for others.
If we look at what happens in German, we find that the coordination is rather
less problematic in what are clearly compounds (some examples from Wiese
1996: 70–1, Fleischer and Bartz 2007: 92). Dutch seems to resemble German, and
the Dutch facts are discussed by Booij (1985), but are omitted here to save space.

German example Gloss Comment

Hals- und Beinbruch neck and leg.breakage equivalent to English
break a leg
Mast- und Schotbruch mast and sheet.breakage the seafaring equivalent
Frühlings- und spring and autumn days
Herbsttage
Hol- und Bringedienst fetch and carry.service
Ostersonntag oder Easter Sunday or Monday
-montag

The phenomenon has a specific name in German, apparently from Grimm,

‘decomposition’ (Erben 1975: 30). But German scholars look at the phenomenon
from a different angle. Since Hals- und Beinbruch (and equivalently for the other
examples) must come from Halsbruch und Beinbruch (‘neck.breakage and leg.
breakage’), they see this not as coordination but as deletion of a repeated element.
The claim is not that the German tradition necessarily transfers to English; that is
a dangerous assumption to make – see Chapter 26. What is interesting is that there is
a completely different way of looking at the situation, which might make it easier to
understand.
112 Part III Syntactic Questions

One of the benefits of the German solution is that it generalizes beyond

compounds to other types of word-formation, specifically affixation. In both
English and German, it is sometimes possible to coordinate prefixes. The English
examples below come from Di Sciullo and Williams (1987: 105–6) and Bauer
et al. (2013: 434), the German examples come from Wiese (1996: 70).
pre- and post-test methodology
pro- or anti-feminist
mini or micro engines
super and supra national
Über- oder unterbau ‘super- or substructure’
pro- und antiamerikanisch ‘pro- and anti-American’
In both English and German, we can find coordinated bases using the same
suffix, but this is not generally available.
child- and home-less
business and family-wise
sugar- and fat-free
mütter- und väterlich mother- and father.ly
Brüder- oder little brother or sister (where the suffix -chen is
Schwesterchen translated as ‘little’)
*Komponist- und female composer and teacher (the suffix -in means
Lehrerin ‘female’)
*damp- and stoutish

Booij (1985) and Wiese (1996) point out that where affixes can take part in
this process in Dutch and German, the deleted element must be a phonological
word. This is compatible with the view expressed by Di Sciullo and Williams
(1987: 105), who say of prefixes that are allowed to stand “these prefixes have
achieved a tentative status as a kind of free form”. This tentative status is
a status as a phonological word. Wiese (1996: 70) points out that in German
there is some variability in what speakers allow in such deletion, and that not
only is further research required, but that it is clear that the deletion is not
confined to places where coordination is involved. Since English has not been
considered from this point of view, and since examples of the phenomenon are
rare in English, the comments hold to an even greater extent.
If the phonological form is relevant in English, too, this explains why we can
have pre- and post-war but not, say, *dis- and re-colour, why sugar- and fat-
free is fine, but *damp- and stoutish is not. But there are other things going on.
The first is that we cannot leave a bound root standing alone or delete it.
*re- and insist
*persist and vade, *superpose and sede (note that super is a phonological
word)
 14 Coordination in Word-Formation 113

The exception is if the bound root is a neoclassical element used as a prefix to

an English word.
socio- and psycho-linguistics
We do not appear to be able to do this if there is an entire neoclassical
formation involved or if the neoclassical element is a suffix.
*neur(o)- and psychology
*music- and icon-ology
In English, suffixes must not only be phonological words, but apparently
have to have the form of words, so, as pointed out by Di Sciullo and Williams,
we cannot have
*educate and rehabilitation

(meaning ‘education and rehabilitation’) any more than we can have

*compact and completeness
where -ness is not a phonological word. Child- and home-less (given above)
seems to imply a full vowel in less, since /ləs/ would not be a phonological word.
There are many possible patterns for coordination in a 3-term compound,
and often at least two interpretations of the patterns. We ignore constructions
such as art galleries and museums, head, shoulders, knees and toes, lake view
and balcony, where the coordination involves two or more independent phrases
with no interaction. Some of the examples here are taken from the British
National Corpus (Davies 2004–). Impossible combinations must, of course, be
made up, and the reader may disagree with the asterisks.

Internally bracketed items act Externally bracketed item treats the

Bracketing independently internal items as a unit
1 [[A B] and C] data collection and analysis
fruit trees and bushes
gas exploration and production
?ice-creams and lollies
sunrise and set
2a [A [B and C]] city shops and sights copper pots and pans
Christmas cards and trees
2b anti-Christian and Jewish anti-rules and regulations
3a [A and [B C]] air- and ground-search
*black and bluebirds
black and whiteheads
earth and moonquakes
?fire and candle-light
?foot or basketball
114 Part III Syntactic Questions

*ginger and shortbread

*guide and cookbooks
horse or camelhair
in- and out-flow
landlord and tenant act
sesame or rape oil
*tooth and headache
*swim and sportswear
road and rail crossings
water- and weatherproof
3b pre- and post-operative
pro- and anti-British
4a [[A and B] C] health and safety standards life and death decisions
oak and ash trees cease and desist order
food and drug administration
husband and wife team
salt and pepper moustache
steak and kidney pie
4b salt and sugar free

The rows labelled (b) show examples with affixes, the other rows show
compounds or potential compounds. Since one of the claims about such
examples is that they distinguish between compounds and phrases, I do not
want to make strong claims about the accuracy of those claims: readers must
decide whether it is always the case that the acceptable sequences are phrases,
and the unacceptable ones are compounds. They must also determine to what
extent blank cells can be filled with acceptable or unacceptable examples. On
the basis of the examples given above, it seems that the unacceptable examples
represent a very small section of the possible patterns (given that even
unacceptable examples have to make sense for the test to be a fair one: sesame
and motor oil might not be acceptable, but that is because a coordination of the
two seems unlikely; foot and handball may be better than foot and basketball
because of the parallel between hand and foot).
As a final point, consider the prefixes which seem to be able to coordinate.
We have seen that there may be a restriction to prefixes which are phonological
words, so that a-, be-, en-, for-, in- ‘negative’ and so on may be excluded by this
proviso. But the pairs that seem perfectly possible include
mini- and maxi-
mini- and micro-
mono- and poly-
pre- and post-
pro- and anti-
super- and supra-
 14 Coordination in Word-Formation 115

Those that do not appear to be possible (or are rare), include

ante- and post- Ante- and Postpartum (M. Hoedemaker, D. Prange
and Y. Gundelach. 2009. Body condition change
ante- and postpartum, health and reproductive
performance in German Holstein cows. Reprod
Dom Anim 44, 167−73)
endo- and exo- endo-skeleton and exo-skeleton seems to be the
more common usage
sub- and super- subscript and superscript is used rather than sub-
and super-script
pseudo- and quasi- pseudo- and quasi-isometrics (www.sportsmith
.co/articles/pseudo-isometrics/ accessed
28 February 2024)
Not only are there very few, the majority of these are antonymic (only
pseudo- and quasi- and mini- and micro- might be considered quasi-
synonymic). Examples such as up- and down- may involve compounds rather
than prefixes, it is often difficult to tell. The restriction to prefixes which are
opposites excludes most prefix pairs, and, for example, rules out all negative
prefixes from occurring in this construction.
The whole question of coordination within compounds and derivatives is
more complex than is generally acknowledged. Not only do intuitions differ
about the data, but the number of factors involved makes it difficult to work out
precisely what is going on. Parallels with languages like Dutch and German do
not appear to work completely, and Smith (2000) claims that the situation in
German has been oversimplified anyway. One of the factors he feels needs to be
included is parallel forms of modification. Part of the problem for the linguist
trying to describe this area is that relevant data is hard to find, with negative
data impossible to find. However, since coordination is possible within deriva-
tives, albeit rare in English, it seems unlikely that it can be used to define
a compound.

Challenge
The examples presented above as instances of different bracketing patterns
are incomplete and subject to challenge, since different people might not
agree on what is or is not possible in English, and will certainly disagree as to
what is a compound in English. Do you agree with the asterisks that are
given? How far can you extend the data set, especially with reference to the
blank cells? If you wanted to test the various hypotheses on coordination, or
look for further restrictions on coordination, how would you find relevant
data? In particular, to what extent can relevant data be found via automated
searches?
116 Part III Syntactic Questions

References
Bauer, Laurie. (2014). Grammaticality, acceptability, possible words and large corpora.
Morphology 24, 83–103.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Booij, Geert E. (1985). Coordination reduction in complex words: A case for prosodic
phonology. In Harry van der Hulst & Norval Smith (eds.), Advances in Non-Linear
Phonology. Dordrecht: Foris, 143–60.
Davies, Mark. (2004–). British National Corpus (from Oxford University Press).
Available online at www.english-corpora.org/bnc/
Di Sciullo, Anna Maria & Edwin Williams. (1987). On the Definition of Word.
Cambridge, MA: MIT Press.
Erben, Johannes. (1975). Einführung in die deutsche Wortbildungslehre. Berlin:
Schmidt.
Fleischer, Wolfgang & Irmhild Barz. (2007). Wortbildung der deutschen Gegenwarts-
sprache. 3. unveränderte Auflage. Tübingen: Niemeyer.
Payne, John & Rodney Huddleston. (2002). Nouns and noun phrases. In
Rodney Huddleston and Geoffrey K. Pullum (eds.), The Cambridge Grammar of
the English Language. Cambridge: Cambridge University Press, 323–524.
Smith, George. (2000). Word remnants and coordination. In Rolf Thieroff,
Matthias Tamrat, Nanna Fuhrhop & Oliver Teuber (eds.), Deutsche Grammatik
in Theorie und Praxis. Tübingen: Niemeyer, 57–68.
Wiese, Richard. (1996). The Phonology of German. Oxford: Oxford University Press.
Part IV

Interfaces
15 Reflections on the Interface between
Word-Formation and Phonology
Morphophonemics

15.1 Introduction
It is well-known that various morphophonemic alternations arise under
word-formation in English, and these were of particular importance in the
theory of Lexical Phonology and Morphology (sometimes called Level
Ordered Morphology though ‘stratum’ is preferred to ‘level’ here, see
Allen 1978, Kiparsky 1982, Mohanan 1986, McMahon 1994, Bauer
2003a: 166–95). While the simplistic versions of Lexical Phonology that
introduced the theory in the 1970s and 1980s have been shown to be
incorrect in their predictions in several ways, the morphophonemic alter-
nations themselves have rarely been challenged. For example, the phon-
emic alternation found in morphologically defined environments such as
adhere (with /r/) and adhesive (with /s/ or /z/, depending on the speaker) is
recognized as a bona fide phonological phenomenon. In this chapter, five
case studies are considered briefly, suggesting that, if lexicalization is
taken seriously, the range of morphophonemic patterns is much smaller
than is generally allowed for, but also that some of the alternations may
not be related to phonology at all.

15.2 Sequences of Stop Consonants

In the middle of words, English allows sequences of non-identical stop
consonants (including nasals as stop consonants) in which the second
consonant is voiced, with various constraints. We find the patterns illus-
trated below.
ɡm agma, paradigmatic
ɡn lignite, signify
mb amber, umbrella
mn amnesia, amnesty
nd conduce, endure, handful, pander
ŋɡ bungalow, finger, linger

119
120 Part IV Interfaces

Such clusters are not found word-finally (except for /nd/), and neither are
they found in some other environments, specified in terms of structural features
of the relevant words. In standard morphophonological descriptions of English,
we can thus find statements (often formulated as a rule) that
Stop + voiced stop clusters not including /d/ are simplified by deletion of the
plosive element (or if there is no plosive, one of the nasals) at the end of a word
and in some other definable environments.
We will return to the other definable environments, but they include before
the suffix -ing, so that singing (in standard British English, though not in all
varieties) is /sɪŋɪŋ/ and not */sɪŋɡɪŋɡ/. This looks like a straightforward mor-
phophonemic rule, where the common meaning is reflected in the spelling,
which can be considered morphophonemic in this regard, and whose limited
use makes it easy to get oversight of the process.
With the aid of a reverse dictionary, ninety-eight words of English ending in
the relevant written sequences of consonants (<nd> was not included) were
discovered. Then Marr (2008) was used to find related words in which these
sequences were followed by a vowel, except that a following <le> was permit-
ted and sometimes /r/. This option was chosen as the only places where it is
possible to have the sequences of consonants pronounced. Instances where
extra words included a case which was already listed (e.g. designable and
designability) were not counted twice. Following -ing and comparative -er,
superlative -est were ignored because they are usually seen as inflectional.
When those cases where no relevant words were found in Marr (2008) were
removed, sixty-two base words remained, with 130 related words where the
consonant sequence was followed by a relevant segment. Pronunciations, when
not given in Marr (2008), were also checked in other reference works, though
not all were listed (e.g. womby related to womb, a Shakespearian usage).
The first thing to notice is that the rule as stated above is shown to be
incorrect. Contrary to the predictions of the rule, some words are listed with
two-consonant pronunciations in standard reference works. Iamb and dithy-
ramb can both be found listed with the consonant sequence pronounced at the
end of the word. Limn is often given as having /mn/ before -ing. Although iamb
and dithyramb are both borrowed from Greek, other Greek words on the list,
such as paradigm, do not share this possibility, so that etymology does not
appear to be a decisive factor here.
We have to ask what general result we might expect. It could be that
Germanic affixes behave differently from learned ones, with the proviso that
occasionally the etymology does not entirely reflect current linguistic pattern-
ing. In light of the large literature on Lexical Phonology and Morphology or
Stratal Morphology, it could be that individual affixes, independent of their
origin, show their own consistent behaviour with regard to the sequence
 15 Word-Formation and Phonology 121

simplification. The least expected result would be for individual affixes to

behave in an unpredictable manner.
The behaviour of the adjective-forming -y (e.g. clingy) and the noun-
forming -y (e.g. thingy) both behave as expected. The rule applies before
both on every occasion, so that there is only one consonant pronounced.
The suffix -ity (n = 3, e.g. benignity) behaves consistently in the other way:
the rule never applies before it, so that both consonants are pronounced.
Between them, these two examples are consistent with either the etymo-
logical hypothesis or the Stratal hypothesis.
The suffix -le (if we can count it as an English suffix) has a two-consonant
pronunciation before it, and the suffix -ism has a single consonant pronunci-
ation before it, but neither has many examples (tangle and dongle are possibly
not real derivatives, twangle is rare; the only relevant example of -ism in my
data set is foreignism). But to the extent that they can be accepted, they suggest
that the strict etymological hypothesis is wrong.
More interesting are those suffixes which behave inconsistently.
-er usually has only one consonant before it, but limner can have two.
-ic usually has both consonants pronounced before it, but tombic can have no /b/.
-ish usually has only one consonant before it, but youngish can have two.
-ist usually has two consonants pronounced before it, but columnist can have
no /n/.
-ous would normally be expected to have two consonants before it, as in
fungous (although that has truncation), but wrongous may have no /ɡ/.
These might be explained as errors – although that is a dangerous line to take,
given that these pronunciations are widespread enough to make it into reference
works – or, more obviously, as being individually lexicalized cases. They could
be treated as coming from a different dialect, but there is no evidence of
consistent behaviour. But if they are lexicalized, we have reason for surprise
because most of these words are rare, and we would expect rare words to follow
the rules if there are rules.
We also find a lot of words where what follows our presumed consonant
sequences can scarcely be interpreted as an English affix. Bombard, clangour,
gangrel, hymnody, phlegmon, rhombos, rhombus illustrate some of the range of
forms found. In these instances, as might be expected, the words are treated as
if they do not have affixes, and both consonants are pronounced. Note, how-
ever, columel ‘a small column’, where even the orthography does not include
an <n>, and clangour which some people – like me – pronounce with no /ɡ/. All
of these must count as lexicalized forms because they cannot be produced in the
current state of the language, and that makes the morphophonemics irrelevant,
since their forms must be listed.
122 Part IV Interfaces

This example shows the value of a concept of lexicalization and taking the
current patterns of word-formation as being the productive patterns and not
all those forms which happen to be analysable. However, that is not the end of
the story.
Many people don’t like the word wronger. They will say that if you want to
compare wrong, you have to say more wrong: Both of these things are wrong,
but the first is more wrong than the second. Similarly, they will prefer more
right to righter, something that descriptive grammars support.
However, even speakers that say they don’t like the word wronger will
pronounce it /rɒŋə/. The pronunciation /rɒŋɡə/ is clearly wronger than the
pronunciation /rɒŋə/. This is odd, because if we consider the other words
which seem to belong to the same pattern, long and strong, they have a /ɡ/ in
the pronunciation when they are in the comparative form: /lɒŋɡə/, /strɒŋɡə/.
Younger works the same way, and the superlatives follow suit. We would
expect the high-frequency words to influence the form of the low-frequency
word, and for all of these adjectives to behave the same way, but they do not.
On the other hand, the comparative of numb is number /nʌmə/ (not homoph-
onous with number ‘numeral’). We can again appeal to lexicalization. But for
adherents of Lexical Phonology, there is at least a puzzle here, when inflection
behaves like Stratum 1 affixation instead of like Stratum 2 affixation, and where
inflection, often defined by its regularity, behaves just as irregularly as deriv-
ation does.

15.3 Velar Softening

English spelling has a rule, with some minor exceptions, that <c> before <e>,
<i> or <y> is pronounced /s/, but before <a>, <o>, <u>, a consonant or a word-
boundary is pronounced /k/. The rule applies where no morphological inter-
vention occurs, and is illustrated by words like fence, cite, cycle, call, colour,
cucumber, crash or tic. However, when this rule applies over morphological
boundaries it has the effect of changing a word-final /k/ to a word-medial /s/
before an affix boundary. In some instances, the following <i> acts as
a diacritic, further changing the /s/ to a /ʃ/, as in magician, special and so on.
This morphophonemic rule has been given the name of ‘velar softening’
(Chomsky and Halle 1968), which can be used here.
We should note that velar softening is not really a phonological rule; it is an
orthographic rule, based on Romance traditions of spelling. Not all final /k/
sounds are affected, only those which are spelled <c> (never those spelled <ck>
or <k> or <ch>). Words like monarchist (monarchic, monarchism, etc.) do not
show velar softening, adhocism (perhaps not widely used) shows no softening,
retaining /k/, as does arced, and neither do words like zincify (also spelled
<zinkify, zinckify>, zinciferous and other related forms. In fact, this is a rule
 15 Word-Formation and Phonology 123

which applies only if the input ends in -ic, and probably (even if not strictly
a case of English word-formation, but a borrowing), the suffix -ic. But beyond
that, velar softening does not apply before all affixes beginning with the
relevant vowel letters. If we were to invent electricish or academicy they
would have /k/ at the end of electric and academic rather than /s/ (see rheumat-
icky in Marr 2008). So the rule of velar softening (or this version of the rule,
based on the letter <c>), to the extent that it is a phonological rule at all, has to
be specified as follows:
/k/ in the suffix -ic is replaced by /s/ before the affixes -ian, -ity, -ist, -ism, -ize, -ify.
It is not entirely clear whether these sequences are productive or not. Bauer
et al. (2013) cite examples containing these sequences, but very few of their
examples are clearly new. There is also some doubt as to whether the rule of
velar softening is available to speakers at all: when Ohala (1974) asked
consultants to produce unfamiliar (because invented) words ending in such
sequences, velar softening was not consistently applied, and even when it was
applied, it is not clear whether it was applied as a morphophonemic rule or as
a rule of orthography. The suffix -ic, though, “shows some productivity”
(Bauer et al. 2013: 303), so we might expect the affix sequences to be
available, if not common, and velar softening thus also to show some
productivity.
This example shows just how difficult it can be to determine whether a given
pattern is lexicalized or not, whether it is a morphological pattern or
a morphophonological pattern. But it is also a warning not to assume too easily
that observable patterns over familiar vocabulary are necessarily productive.

15.4 Degemination
Degemination is the simplification of a sequence of two adjacent identical
phonemes to a single phoneme. There are no instances of geminate consonants
and no geminate vowels internal to a morph in English (Bauer 2003b). For
example, a word like missel has only a single /s/ in the pronunciation, and the
<ss> is simply an indication that the preceding vowel is short. Over boundaries
between an affix and a base, however, gemination is found. For example
misspell has geminate /ss/
posttonic has geminate /tt/
unnamed has geminate /nn/
embalmment has geminate /mm/
palely has geminate /ll/
Upton et al. (2001) indicate that British and American English may differ in
some of these words, with fewer geminates in American English, but I am using
124 Part IV Interfaces

British pronunciations as the basis for my discussion. In all of these instances,

a consonant on the margin of an affix is identical to a consonant on the abutting
margin of the base. Since both consonants can be motivated from the form of
the relevant element in other contexts, the geminate is expected, and what is not
necessarily expected is degemination in such positions, that is the deletion of
one of the identical consonants. Degemination can be heard, for example, in
fully (whose elements are full and -ly, but English orthography does not allow
a spelling with three identical consonants). The question, therefore, is when
does degemination occur.
Retention of the geminate seems to be the general pattern with prefixes,
which are often more word-like than suffixes. This seems to be true with both
consonants and vowels, though geminate vowels are extremely rare, perhaps
restricted to co-own(er(ship)), although there seems to be no principled reasons
why others should not exist. However, there are exceptions. Some examples
showing degemination, not always to the exclusion of retained gemination, are
given below. Most of the words cited also have morphologically related words,
which are relevant (e.g. connote, dissatisfy).
connotation
dissatisfaction, dissemble, dissever, dissimilarity, dissoluble, dissymmetry
granddad, granddaughter
illegal, illegible, illegitimate, illiberal, illicit, illiterate, illogical
immaterial, immature, immeasurable, immemorial, immiscible, immobile
innocuous, innumerable, innumerate, innutritious
From the point of view of a Stratal theory of morphology, this looks as though it
may be easily explicable. The prefix grand- (if it is a prefix and not a compound
element) loses its final /d/ before any consonant (grandchild, grandmother) and all
the other examples illustrate Stratum I prefixes. However, things are not quite that
simple. Not all Stratum I prefixes allow degemination. Sub- (which must be on
Stratum I because it is stressed in subplot and subsequent) does not permit
degemination in subbreed, subbranch. Also, more importantly, dis- shows variable
degemination, so that degemination is not automatically related to Stratum I. There
are ways round these problems in various models of Stratal Morphology, but they
weaken the predictions of the theories.
Suffixes are harder to deal with. Although there are relatively many suffixes in
English, very few provide relevant data. These main ones are -ly (adverbial and
adjectival), and, potentially, -ness. With -ly, a geminate is typically retained after
a stressed syllable, but degemination occurs after an unstressed syllable; with
-ness, a geminate is typically retained (but far less in American English). Some
examples with -ly from three pronunciation dictionaries are presented below.
 15 Word-Formation and Phonology 125

Jones et al. (2003) Upton et al. (2001) Wells (1990)

Word +Gemin −Gemin +Gemin −Gemin +Gemin −Gemin

dully ✓ ✓ ✓ ✓ ✓
foully ✓ ✓ ✓
really ✓ ✓ ✓
shrilly ✓ ✓ ✓ ✓
severally ✓ ✓ ✓
wholly ✓ ✓ ✓ ✓

Degemination seems to be a result of high frequency, but with some varieties

more susceptible to it than others, but there is not absolute agreement as to
when it occurs, even in the same variety.
Here, then, we have a morphophonemic process whose application is not
entirely predictable, which is influenced by frequency and by the individual affixes
(possibly the individual lexemes) involved. That means that it is quite difficult to
explain just what the process is, and even more difficult to predict the outcome. The
phonological process can be easily described, but not when it will apply. Something
which looks simple, is actually more complex than appears at first sight.

15.5 Alternations before -ian

There is one apparently morphophonemic rule which is clearly productive,
even though it affects relatively few forms. It is brought about by suffixation of
adjective-forming -ian, usually considered a variant of the -an suffix or having
an -i- extender (Bauer et al. 2013: 181). Since the use of the extender is
unpredictable in general, but is usual in this construction (note Elizabethan as
an exception), -ian is treated here as a separate suffix.
The suffix -ian is attached to proper nouns to make adjectives. Typically, the
vowel immediately before the suffix is stressed and long (Christian provides an
exception, probably because of the two consonants following the stressed
vowel, as with Egyptian). Some examples are given below. The suffix is
transcribed as disyllabic /iən/ here (as in Wells 1990, Jones et al. 2003),
which accounts for the main stress pattern on the antepenult, though this is
sometimes reduced to monosyllabic /jən/.

Amazon/Amazonian /æməzəʊniən/
Aristotle/Aristotelian /ærɪstətiːliən/
Babylon/Babylonian /bæbɪləʊniən/
Bacon/Baconian /beɪkəʊniən/
Caesar/Caesarian /siːzeəriən/
126 Part IV Interfaces

Devon/Devonian /dɪvəʊniən/
Handel/Handelian /hændiːliən/
Johnson/Johnsonian /ʤɒnsəʊniən/ (and other names ending in -son)
Lilliput/Lilliputian /lɪlɪpjuːʃən/
Venus /Venusian /vɪnjuːziən/
Venice/Venetian /vɪniːʃən/
I have omitted a lot of examples where the facts of the case may be obscured
by other factors, for example where the base word ends in <ia> (Patagonia/
Patagonian) or some phonological material is deleted (Olympus/Olympian) or
some unpredictable consonant change has occurred (Troy/Trojan < earlier
Troian), for example, but that should not affect the general pattern. Because
the pattern is productive, the list cannot be finite.
If we look at the above examples in terms of the alternating vowel sounds,
we find the following patterns:
ə alternates with əʊ
ə alternates with iː
ɒ alternates with əʊ
ə alternates with eə
ʌ alternates with juː
ə alternates with juː
ɪ alternates with iː

If we take the point of view that /ə/ is the neutralization in unstressed

syllables of (nearly) all the other vowels, then we might have to replace
the /ə/s in the above with the relevant underlying vowel to see a pattern. This
gives problems with many of the words listed such as Amazon, Aristotle,
Bacon, Caesar where, precisely in the base forms illustrated here, there is no
independent evidence as to what the underlying vowel might be. In words
ending in -son we might presume that we have an underlying /ʌ/. We then have
/ʌ/ and /ɒ/ alternating with /əʊ/ in apparently the same environment, and /əʊ/
and /juː/ both alternating with /ʌ/ in apparently the same environment.
However, if instead of looking for alternating vowel sounds, we look at the
vowel letters, things become clearer. What we have is the long vowel in the
environment before -ian which corresponds to the short vowel that is written in
the base word. The only trouble with this solution is that it goes against all that
we usually assume about the nature of orthography: specifically, the usual
assumption is that orthography derives from the spoken language, not vice
versa. So this is a surprising, and perhaps disturbing, finding.
It is, however, supported by some further examples. Words which end in <f>
and <w> have an -ian form which contains /v/ (written <v>) in the correspond-
ing position. There are few relevant forms, but the patterns appear to be
productive to the extent that this is a relevant consideration.
 15 Word-Formation and Phonology 127

Aronoff/Aronovian /ærənəʊviən/
Harrow/Harrovian /hærəʊviən/
Marlowe/Marlovian /mɑːləʊviən/
Shaw/Shavian /ʃeɪviən/
Skiddaw/Skiddavian /skɪdeɪviən/
Snow/Snovian /snəʊviən/
Since the /f/ and the /w/ do not share phonetic qualities, and the <w> is not
even pronounced as a consonant in these examples, and the stressed vowel in
Shavian can best be explained in orthographic terms, such examples seem to
strengthen the case that the basis of this word-formation pattern is orthographic
rather than phonological.

15.6 Participial -ed before Other Suffixes

The allomorphy of past tense and past participial -ed is familiar from many
textbooks. The variant /ɪd/ is heard following /t/ or /d/, otherwise the pronun-
ciation is determined by the voicing of the segment immediately preceding
the suffix. We find the variants illustrated in wanted /wɒntɪd/, watched /wɒʧt/,
loved /lʌvd/. In a small number of words, this patterning breaks down when
the participle is used adjectivally. We say /blesɪd/ rather than /blest/ for
blessed (or we used to in attributive position), we say /lɜːnɪd/ rather than
/lɜːnd/ for learned (learnt does not seem to be used attributively of a person)
and /bɪlʌvɪd/ rather than /bɪlʌvd/ for beloved. We could add naked and sacred,
although these are no longer felt to be participles, and naked probably never
was one, from the etymological point of view. This seems to be part of
retaining old participial forms in adjectival use that is also illustrated by
drunken, new-mown, shaven and the like. We could treat these forms as
belonging to different lexemes from the base verb (as long as our model
allows for lexemes). This might help explain why we can have an aged man
(/eɪʤɪd/) but would have to have an aged piece of beef (/eɪʤd/) (Quirk et al.
1972: 246).
Participles acting as adjectives can, when it makes sense, also be turned into
manner adverbials by the addition of -ly and into nouns by the addition of -ness.
When the verb ends in /t/ or /d/ these have no option but to take the /ɪd/
allomorph of -ed: admittedly, animatedly, belatedly, contentedness, disap-
pointedly, excitedly, excitedness, and so on. Equally, those adjectives where
the <ed> is pronounced /ɪd/ even when it would not be the regular form, keep
that pronunciation when there is a subsequent derivation: blessedly,
learnedness.
But where the /ɪd/ allomorph is not automatically required by the phonology
of the adjacent consonants or the phonological make-up of the adjective, things
128 Part IV Interfaces

are less clear. Fixedly and markedness have the /ɪd/ pronunciation, even though
their phonology does not seem to require it, determinedly and tiredness do not.
Some specification of just what the relevant -ed suffix is does little to clarify
matters. The -ed that has been discussed so far is added to verbs. There is
another, homophonous -ed which is added to nouns, and these nouns may
become elements in compounds, so that we find bearded, red-handed (with /ɪd/
because of the preceding /d/), good-humoured, good-natured (which have /d/
rather than /ɪd/). It is not automatically obvious that the two types of -ed will
function in the same way (though we might expect them to, if they belong to the
same morphome – Aronoff 1994). Fowler (1965 sv -edly) treats them all as the
same, but he also treats those that have /ɪd/ as a function of the /t/ or /d/
preceding the /ɪd/ in the same way as those where the /ɪd/ is less obviously
justified. I shall ignore the denominal -ed, though I think it behaves in much the
same way as the deverbal one.
We also need to ask whether a prefix on the adjective makes any difference:
is reserved any different from unreserved, ordered any different from dis-
ordered? The main difference does not appear to be phonological, but perhaps
a matter of norm (see Chapter 6): the prefixed forms are more regularly used as
adjectives. It should also be noted that many dictionaries do not give pronunci-
ations for relevant forms – either because they are assumed to be obvious, or to
save space, or both – and in some cases do not list them at all. Intuition is not the
best data to use in instances like this, but may be what is available.
Jespersen ([1942] 1961: 29) thinks he has the answer. The /ɪd/, he suggests,
“forms a connecting link (syllable) between stem and ending, but it is not
required when the stem ends in an unstressed syllable”. Certainly, many of the
instances where the /ɪd/ arises are found where the verb has final stress:
allegedly, ashamedly/-ness, assuredly, cussedly/-ness, markedly/-ness, sup-
posedly. Equally, there are instances where the verb does not have final stress,
and the /ɪd/ is missing: bewilderedly, determinedly, hurriedly, impoverishedly.
But Jespersen himself does not know which class ashamedly fits into, and Jones
et al. (2003) give preparedly /prɪpeədli/ but preparedness /prɪpeərɪdnɪs/. Also
tiredly and tiredness have no /ɪd/ although they are on a monosyllabic (and thus
stressed) base.
Fowler (1965) has a different solution. Although his terminology is not the
same as mine, he would say that the pronunciation with /ɪd/ is lexicalized: we
use the /ɪd/ pronunciation only when we know that the word requires one (from
previous experience of the individual word). For Burchfield (1996), completely
rewriting Fowler’s article, this cannot be true because some unfamiliar words
nevertheless get given the /ɪd/ pronunciation; in other words, the rule is
productive.
I would support Burchfield in this view, but there are problems. What is the
productive rule (Burchfield admits that he doesn’t know)? Do all speakers share
 15 Word-Formation and Phonology 129

the same view of what the rule is? Reservedly (listed by Jones with /ɪd/) could
get that pronunciation because of unreservedly, or the two could independently
be following the same rule. Burchfield cites admiredly, depressedly, harassedly,
labouredly, scatteredly and veiledly from the OED (none mentioned by Jones
et al. 2003). From Wikipedia (https://en.wiktionary.org/wiki/Category:Englis
h_terms_suffixed_with_-ly accessed 7 June 2023) we can add abashedly,
allowedly, authorizedly, barbedly, bedraggledly, belabouredly, bemusedly,
blurredly and dozens of others. Burchfield says that for him, some of this list
would have /ɪd/, but doesn’t tell us which ones. Even a small-scale survey
suggests that not everyone agrees on all of these, so there is the possibility that
there is an ongoing change affecting these cases, which makes the rule even
harder to discern. The change is probably a loss of the /ɪd/ pronunciation, which
is, in any case, a remnant of a much older pronunciation of such forms. But if
there are unfamiliar forms which speakers pronounce with the /ɪd/ pronunci-
ation, then the problem remains. Again, a small-scale survey suggests that
speakers use both pronunciations in words they claim not to know, and that final
stress in the base is not a determining factor for which pronunciation is
produced.
The bigger question that this example raises is whether we can talk of rules in
a situation where the productivity of a given pattern is fading. What we seem to
find is conflicting usages based on principles which are not clear, varying from
speaker to speaker and even within the usage of a single speaker. This means
that it is not even straightforward to talk in terms of lexicalization of individual
examples, although that may be part of a solution.

15.7 Conclusion
Morphophonemic rules often apply in unpredictable ways, and some of them
are not even morphophonemic, because the variation is based on the orthog-
raphy rather than on phonemic alternations. This leads to a grammar which is
hard to write or hard to apply. Just how the speaker knows when a given rule
applies is not clear, but it seems to be based on experience of the relevant
formatives or lexemes, rather than on formulating a rule.

Challenge
There is alternation between the phonemes /s/ and /ʃ/, and correspondingly
between /z/ and /ʒ/, in a number of related words in English, such as substance
and substantial, confuse and confusion, space and spacious. Find a representa-
tive set of related pairs using dictionaries, reverse dictionaries, phonology
books, pronunciation guides and any other sources available to you. Can you
always tell what the relevant pairs are? How would you want to treat such pairs
130 Part IV Interfaces

in a linguistic description of present-day English? What benefits would you

gain from your way of treating such pairs as opposed to other potential ways of
dealing with them? What might be the major disadvantages of your proposed
way of dealing with the pairs?

References
Allen, Margaret R. (1978). Morphological Investigations. PhD dissertation, University
of Connecticut.
Aronoff, Mark. (1994). Morphology by Itself. Cambridge, MA: MIT Press.
Bauer, Laurie (2003a). Introducing Linguistic Morphology. 2nd ed. Edinburgh:
Edinburgh University Press.
(2003b). The phonotactics of some English morphology. In Henrik Galberg Jacobsen,
Dorthe Bleses, Thomas O. Madsen & Pia Thomsen (eds.), Take Danish – for
Instance: Linguistic Studies in Honour of Hans Basbøll. Odense: University
Press of Southern Denmark, 1–8.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Burchfield, R.W. (1996). The New Fowler’s Modern English Usage. 3rd ed. Oxford:
Oxford University Press.
Chomsky, Noam & Morris Halle. (1968). The Sound Pattern of English. New York:
Harper & Row.
Fowler, H.W. (1965). Modern English Usage. Second edition revised by Sir
Ernest Gowers. Oxford: Oxford University Press.
Jespersen, Otto. ([1942] 1961). A Modern English Grammar on Historical Principles.
Part VI: Morphology. London: Allen & Unwin/Copenhagen: Munksgaard.
Jones, Daniel, Peter Roach, James Hartman & Jane Setter. (2003). Cambridge English
Pronouncing Dictionary. 16th ed. Cambridge: Cambridge University Press.
Kiparsky, Paul. (1982). Lexical morphology and phonology. In-Seok Yang (ed.),
Linguistics in the Morning Calm. Seoul: Hanshin; 3–91.
McMahon, A[pril] M.S. (1994). Lexical phonology and morphology. In R.E. Asher
(ed.), The Encyclopedia of Language and Linguistics Vol. 4. Oxford: Pergamon,
2155–60.
Marr, Vivian (ed.). (2008). The Chambers Dictionary. 11th ed. Edinburgh: Chambers
Harrap.
Mohanan, Karuvannur Puthanveetil. (1986). The Theory of Lexical Phonology.
Dordrecht: Reidel.
OED. The Oxford English Dictionary [online]. oed.com
Ohala, John. (1974). Experimental historical phonology. In John Anderson &
Charles Jones (eds.), Historical Linguistics II. Amsterdam: North Holland,
353–87.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. (1972).
A Grammar of Contemporary English. London: Longman.
Upton, Clive, William A. Kretzschmar Jr & Rafal Konopka. (2001). The Oxford
Dictionary of Pronunciation for Current English. Oxford: Oxford University Press.
Wells, J[ohn] C. (1990). Longman Pronunciation Dictionary. Harlow: Longman.
16 Reflections on the Interface between
Word-Formation and Syntax

Syntactic structures arise in word-formation as apparent bases, and although

some scholars have suggested strict limitations on such usage, it appears to be
extremely free in modern usage. Examples of relevant formations will be given
here, and two questions will be dealt with: what is the output of the formation?
And how is its use limited in word-formation (if at all)?
There are two, apparently distinct, patterns of such usage: syntactic struc-
tures arise as complex modifiers in what are usually taken to be compounds,
and syntactic constructions are found as the bases in affixation. The two will be
dealt with individually before generalizations are sought.

16.1 Syntactic Bases in Affixation

Some examples of the phenomena are listed here, ordered by the suffix used,
with an adjectival example first.
trying to sound completely don’t-care-ish. (Judy Astley. 2013. In the Summer
Time. London: Bantam, p. 62)
Very much a tooth-for-toother was our Tris. (Susan Moody. 2016. Quick off
the Mark. Sutton, Surrey: Severn House, p. 169)
“Free spiritism?” Campion snapped. “That’s not even a word, let alone
a philosophy.” (Mike Ripley and Margery Allingham. 2014. Mr Campion’s
Farewell. London: Severn House Large Print, p. 229)
born of the boundless can-doism of the immediate post-war. (Harry Stein.
1995. The Magic Bullet. New York: Delacorte, p. 14)
He’s such an I’m-right-ist. (attested in conversation, 1996)
Managing-a-tight-budget-itis. (Judith Cutler. 1999. Dying by Degrees.
London: Headline, p. 170)
uppityness and snotty-nosed finer-than-thou-ness. (Trevanian, 1998. Incident
at Twenty-Mile. New York: St Martin’s, p. 201)
the that’s-the-way-it-is-ness (Louisa Luna. 2020. The Janes. Melbourne: The
Texts company, p. 62)

131
132 Part IV Interfaces

his familiar, blank expression, his irritating look of not-really-there-ness.

(Paula Hawkins. 2017. Into the Water. London: Doubleday, p. 150)
Rankin watched it go, saying several un-bright-spring-day things under his
breath. (Gavin Lyall. 1999. Honourable Intentions. London: Hodder &
Stoughton, p. 82)
Lattice windows give it a story-book appearance, while inside it’s even still
more once-upon-a-timeyfied. (Jerome K. Jerome. [1889] 1968. Three Men in
a Boat. Harmondsworth: Penguin, p. 173)

Most of the examples cited above, as well as most of those cited by Bauer
et al. (2013) suggest that the phrase used in the base must be a familiar one, or if
not particularly familiar, a direct quotation, and some scholars have suggested
this as a restriction. It seems likely that this is an epiphenomenon of needing the
word to be of a reasonable length, since most such examples use a phrase made
up of two or three words. Despite this, there are examples which show that the
unfamiliar is also possible. Bauer et al. (2013) cite the following:
does nothingness, metal-lumpish, post-Connery-as-Bond, pre-Redford-and-
Hoffman, senior-skier-hood
One factor that does appear to be relevant is the productivity of the affix. It
seems as if it is the most productive affixes which are found in such
constructions.

16.2 Phrases as Modifiers

This situation is very common, particularly, though not exclusively, when the
phrase modifies a noun. Some examples are given below.
William James. He was the groundsmen, handyman, if-there’s-any-sort-of-
difficulty-ask-William-and-he’ll-fix-it-for-you person about the place.
(Laurence Meynell. 1978. Papersnake. London: Macmillan, p. 10)
Natalia wasn’t sure – not committed-for-the-rest-of-her-life sure – that’s what
she did want. (Brian Freemantle. 1996. Charlie’s Chance. Bath: Chivers
Large Print, p. 279)
She was dressed all in black – not a scruffy, Camden, it-doesn’t-show-the-
stains and-hides-my-fat-black, [hyphenation as in the original] but a designer-
label it-allows-me-to-express-my-elegant-simplicity sort of black. (Lauren
Henderson. 1996. Too Many Blondes. London: Hodder & Stoughton, p. 151)
that infuriating I-know-I’ve-been-bad-but-I’d-probably-do-it -again grin.
(Barbara Seranella. 1999. No Offence Intended. London: Hale, p. 13)
The theory has been dubbed the “Did He Fall (on purpose)? Theory”. (Robert
Rankin. 2003. The Hollow Chocolate Bunnies of the Apocalypse. London:
Gollancz, p. 56)
 16 Word-Formation and Syntax 133

our fear-of-terrorist-atrocity society. (Dick Francis. 2006. Under Orders.

London: Joseph, p. 87)
Eric’s Aunt I-told-you-she-was-wrong-for-you Lena. (Patricia Smiley. 2005.
Cover Your Assets. New York: Mysterious Press, p. 65)
As in scream-out-loud, best-sex-I’ve-ever-had good? (Maggie Sefton. 2013.
Poisoned Politics. Detroit: Thorndike, p. 264)

Assuming that the examples here represent the same phenomenon as the use of
syntax in the base of derivatives, which seems reasonable, though it may not be
open to proof, the requirement of familiarity or direct citation seems to be
completely disproved. While the example from Smiley might be a rather lengthy
instance of direct quotation, the examples from Meynell and Henderson seem
unlikely to be quotations or familiar expressions.
The major question these examples raise is the nature of the output. Is scream-
out-loud good (to use one of the shortest examples here) a compound, or is
scream-out-loud simply a word (which may or may not clearly belong to a word-
class) which is inserted into a syntactic structure (the two need not be incompat-
ible). One piece of evidence, of dubious reliability, is the spelling. Except in the
case flagged above, which may be considered a typographical error, the usual
spelling convention for these constructions is for the syntactic construction to be
hyphenated (presumably to indicate its unity), but for that construction not to be
linked by a hyphen to the head of the phrase. This seems to imply that the
syntactic construction is considered to be a word, but that the head is not part of
the same word. In English, this could be a reflection of the stress, since such
expressions are stressed on the head of the phrase. But it is notable that parallel
compounds in other Germanic languages bind the head together with the syntac-
tic phrase into a compound. Consider the examples below (Bauer 1978: 186)
which illustrate the difference.
Swedish
hon hade komochtagmigomdukanminen på sig
She had come-and-catch-me-if-you-can-look.the on her
‘She was wearing her come-and-get-me-if-you can-look’
Danish
hvorfor-skal-man-op-om-morgenen-stemme
Why-must-one-up-in-morning.the-voice
‘Why-do-I-have-to-get-up-in-the-morning voice’
This contrasts with what happens in French, where compounds are not
signalled by orthography in the same way, and the syntactic construction is
not linked to the head of the entire expression.
134 Part IV Interfaces

French
son côté m’as-tu-vu
3SG.POSS side me-have-you-seen
‘his/her have-you-seen-me side’

16.3 Discussion
The first point to note is that syntactic phrases can act as words even if they are
not involved in further word-formation, although it is not obvious whether this
makes them lexemes (see Chapter 9). The main difference between these word-
like objects and the syntactic constructions we have been looking at is that
many of those cited above are not frequent enough to become item-familiar,
and that those items which do become item-familiar can last long enough for
their grammar to become outdated, as with forget-me-not (as opposed to
*don’t-forget-me), apparently borrowed into several languages from Old
French. Established examples are nonetheless rather rare.
attorney general, by and large, farewell, go-between, man-at-arms (from
earlier man-of-arms), wannabe

New such expressions are, however, possible.

What was he? Her ex? Her erstwhile lover? Her lover-in-abeyance?
(Val McDermid. 2006. The Grave Tattoo. London: HarperCollins, p. 5)
These examples raise the question of whether the use of hyphens is
significant or not. It might well be that any multi-word expression (MWE),
hyphenated or not, is a relevant example. This might include any or all of the
examples below.
all things considered, as warm as toast, bed and breakfast, bite the bullet, an
eye for an eye, a piece of cake [= ‘easy’], put up with, rich pickings, under the
weather
In fact, it might not even be necessary that multiple orthographic words are
involved. Perhaps deceased in the example below is a syntactic expression
which has been used as a word with the genitive marking part of the structure of
the surrounding sentence. Alternative analyses are, of course, possible, too.
I’m handing you a batch of email correspondence between my client and
various members of the deceaseds’ families. (M.R. Hall. 2012. The Flight.
London: Mantle, p. 237)
If this is the case, then we have a mode of word-formation which is simply
a matter of syntax. Formal models of word-formation would probably need to
 16 Word-Formation and Syntax 135

build in a loop from the syntax back to the morphology to allow for this, but
easier ways of dealing with the descriptive problem might be possible. If the
relevant new words are simply syntactic constituents rank-shifted to be words,
then these words can act just like other words in being able to take affixes and to
be used as modifiers to rather more canonical words. Syntactic rules might have
to be expanded slightly to allow this new kind of word to act in this way with
adjectives and verbs (which is why the comment was made above that these
new words might not belong to a word-class), but that would not mean a major
disruption to syntactic patterns.
However, this argument does not hold up easily. It fails to hold up because
most of these postulated words cannot act as syntactic heads. In this, they differ
from the MWEs mentioned above, which can act as syntactic heads. However
conjoined nouns are dealt with in the syntax, bed and breakfast can be dealt
with in the same way; bite the bullet acts syntactically just like any verb phrase,
and like any other verb phrase, for example, confers sentencehood on the
construction of which it is a part: I bit the bullet and presented my apologies.
On the other hand, I told you she was wrong for you cannot be a noun, verb,
adjective or adverb – it can only be a sentence unless it is used in attributive
position. We can find, though, occasional exceptions.
not in this atmosphere of “we’ve-all-come-out-here-to-enjoy-ourselves-
let’s-get-on-with-it”. (Agatha Christie. 1964. A Caribbean Mystery. Leicester:
Ulverscroft (large print), p. 81)
indulge your carnal appetite in exchange for a little looking-the-other-way.
(Gerald Hammond. 2004. Dead Letters. London: Allison & Busby, p. 87)
The chief was a has-been, big-city detective, (Sandra Brown. 2005. Chill
Factor. Waterville, Maine: Thorndike large print edition, p. 10)
I don’t need a tagalong. (Iris Johansen and Roy Johansen. 2014. Sight Unseen.
New York: St Martin’s, p. 53)

Even these examples might be considered not to contain a new word (despite
the hyphens). Alternatively, we can modify the constraint (as do Bauer et al.
2013: 490) and limit it to ruling out phrases as heads in compounds.
Furthermore, it appears that such expressions need not be syntactic
constituents. Bauer et al. (2013: 457) comment that “determiner phrases
seem not to be allowed”. As stated, this must be wrong in the light of
examples such as no-phrase constraint (Botha 1984: 137) or no-frills
airline, but it contains some truth since *her the-fear-of-terrorist-atrocity
society is not possible. Just what the real constraint might be is not easy to
see, though, since Bauer et al. (2013: 457) cite this-person-is-a-jerk atti-
tude, which could easily be preceded by an article or a possessive pronoun.
It may only be a(n) and the which are ruled out in initial position in the
136 Part IV Interfaces

phrasal element, but even that seems dubious in the light of the following
examples from COCA (Davies 2008–).
The His Dark Materials books by Philip Pullman
poems have appeared in the His Rib and Got Poetry anthologies as well as in
various literary journals
the North American leg of his the Thrill of It All Tour
he’d published his The Principles of Psychology three years earlier
Orson Welles (1915–1985) adapted War of the Worlds for his The Mercury
Theatre on the Air series
His The New York Times Best Seller

Bauer et al. (2013: 457, see also Bauer 1978) also comment that the
syntax need not be a constituent, citing thumbs-up sign, and later (2013:
488) they also give as an example tortoise-and-hare syndrome, where
ordinary syntax would probably demand articles (the syndrome recalling
the tortoise and the hare).
However we interpret this or try to model it, we have to know just what
structures are included and what are not. Is this, for instance, the same structure
as gives us compounds before suffixes, as in the following examples?
asshole-ishness explains a lot of things these days. (Dallas Murphy. 1992.
Lush Life. New York: Pocket Books, p. 158)
He . . . could be described as either muscular or couch-potato-esque. (Sarah
Andrews. 2002. Fault Line. New York: St Martin’s, p. 52)
It was all very Girl-Guide-ish. (Vivienne Plumb. 2003. Secret City. Auckland:
Cape Catley, p. 178)
It’s too science-fiction-y. (Nury Vittachi. 2008. Mr Wong Goes West. Crows
Nest, NSW: Allen & Unwin, p. 183)
Do the following examples count as instances of the same type, or are they
(or some of them) of a different type?
black-robed attorneys, go-fast stripes, no-go area, red-light district, three-
syllable word
Plag (2003: 221–2) deals with some of these, but it has to be acknowledged
that the borderline between things that might be compounds and things that are
not, or between words and syntactic structures, is very fuzzy in this area. In the
case of three-syllable word, for instance, there is no plural marking on syllable,
which is typical of what happens in compounds, yet three-syllable French
word, where the expression is interrupted by another adjective, makes the
whole thing look more like a piece of syntax than like a word.
 16 Word-Formation and Syntax 137

Just what is going on here remains obscure, but it does look as though the use
of syntactic phrases inside complex expressions is more a matter of word-
formation than simply a matter of syntax, and subject to slightly more restrict-
ive rules than plain syntax.

Challenge
The most constrained set of syntactic expressions seems to be those that can
occur as the base in further derivation. Find further examples (Bauer et al. 2013
give some found in COCA, for instance), and see what the constraints are. Are
the constraints related to those on derivation on adverbial bases discussed in
Chapter 23?

References
Bauer, Laurie. (1978). The Grammar of Nominal Compounding. Odense: Odense
University Press.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Botha, Rudolf P. (1984). Morphological Mechanisms. Oxford: Pergamon.
Davies, Mark. (2008–). The Corpus of Contemporary American English (COCA).
Available online at www.english-corpora.org/coca/
Plag, Ingo. (2003). Word-Formation in English. Cambridge: Cambridge University
Press.
17 Reflections on the Interface between
Word-Formation and Phonetics

Sound symbolism, the use of the sounds of language to represent or reflect the
world in different ways, is a topic that has fascinated linguists and philosophers
for millennia (see Hinton et al. 1994, Feist 2013, Benczes 2019 among hun-
dreds of others). Most of the phenomena that are discussed under this title,
though, do not have anything directly to do with word-formation. For example,
the use of onomatopoeia in representing animal sounds (cuckoo, oink, miaow)
but also in far less direct reflections of natural sound (flutter, grate, twitter) have
become ordinary words, with form made up of phonemes, whose form varies
with language change: would squeak feel like such a good representation of an
un-oiled gate or a mouse in its Middle English form of [skwɛːk]? They are also
extremely culture-bound: pigs may say oink in Greek as well as in English, but
they do not say that in Swedish or Dutch. But if this is word-formation, it is
a word-formation of a very different kind from that which gives us inexpres-
sibility. Similarly, it appears that many languages have a close front vowel in
words for things that are close and an opener and/or back vowel for things that
are far away (as in English here and there, near and far). Lengthening the
stressed vowel in enormous may make it clear to one’s interlocutor that the
entity described is particularly big. But none of these things are matters of
word-formation. Since this book is about word-formation, the relevant factors
here are those where word-formation seems to be related to the sounds used,
and there are far fewer such instances. Two will be of particular interest here:
phonaesthemes and diminutive marking, and rhyme will be briefly considered.

17.1 Phonaesthemes
One of the obvious places where word-formation may interface with phonetics
is in phonaesthemes. Phonaesthemes are sequences of sounds that, while not
being morphemes, nevertheless carry some, typically vague, meaning relevant
for the word in which they appear. For example, the sequence /ʌmp/ is often
thought to provide some meaning referring to a dull sound, as in bump
(although there are not particularly many relevant words) or something awk-
ward, as in frump and slump. Similarly, the sequence /sl/ is often thought to

138
 17 Word-Formation and Phonetics 139

indicate slipperiness to some extent, as in slide, slime, slip, slither, slope (see
Adams 2001: 126–7). Bergen (2004) finds that such patterns are used by
speakers to link words together, and in that sense are real to speakers of
English, and Bauer (2019) suggests that there is some point in seeing them as
part of word-formation, though under the heading of ‘resonances’ rather than as
‘morphemes’.
One phonaestheme often cited for English is the one with the sequence /ɡl/.
The oddity of this sequence is that it is associated with two distinct meanings,
one of them to do with light, often faint or reflected light (glimmer, glisten),
while the other meaning is to do with dullness, depressedness or lack of
transparency (gloaming, gloop, glum). To see the value of phonaesthemes,
words beginning with /ɡl/ in English were investigated.
Marr (2008) provides a list of 129 words beginning with /ɡl/, with many
derivatives which are ignored here. This list was then compared with a frequency
list from COCA (Davies 2008–), and only words from the most frequent 10,000
words of English were retained. This left a list of just thirty words, which were
likely to be familiar to all competent speakers of English (the list could have been
extended beyond the 10,000-word mark, but any cut-off point would be arbi-
trary). Of these just four reflected the dullness meaning, twelve could be thought
to be related to the light meaning, and the final fourteen were not obviously
related to either (assignment to these semantic classes is often personal, so that
other observers might find different values, but the abundance of the ‘neither’
category is striking). The fact that the majority of words with initial /ɡl/ belonged
to neither pattern seems to be a problem for the notion of phonaesthemes as
meaningful and useful elements in word-formation.
Yang’s (2016) Tolerance Principle seems relevant here. Yang sets out
parameters for how many exceptions there can be to a given pattern for it to
be perceived as a ‘rule’, that is, something that can be used productively in the
creation of new forms. Yang (2016: 67) sets up quite specific numerical
expectations, and although we do not need here to examine the mathematics
behind his predictions, he proposes that for a body of forms of size twenty, the
maximum number of allowed exceptions is seven, if a rule is to be productive,
and for fifty examples, the maximum number of exceptions is thirteen.
Fourteen exceptions with twelve forms complying with one of our patterns is
far too many, and the other pattern is even less compliant. This makes it sound
as though the /ɡl/ phonaesthemes cannot be rules and cannot be productive, if
Yang is right. This seems to contradict Bergen’s (2004) position.
Such a conclusion, though, may be overly hasty. Bergen says that speakers use
phonaesthemes to link established words, not that phonaesthemes are productive,
and although Abelin (1999) talks in terms of productivity, he too looks at
perceived links with constructed words. If phonaesthemes are not productive,
they do not all fall within the purview of Yang’s Tolerance Principle. We thus
140 Part IV Interfaces

have a compromise solution available here, which offends neither Abelin (1999)
and Bergen (2004) nor Yang: phonaesthemes are perceived in words, where they
are discovered over sets of semantically related words that happen to contain
common sound sequences, but cannot be used to create words; they are a feature
of word-analysis rather than a feature of word-formation. This would explain
why phonaesthemes can be ambiguous (as stressed by Feist 2013). If this is
a tenable position, its implications are yet to be discovered.

17.2 Diminutives
One place where phonetics is often thought to have a place in word-formation is
with diminutives (and, possibly the same thing, with hypocoristics or pet names).
Jespersen (1922) is one scholar who finds, cross-linguistically, a close front
vowel in words meaning ‘small’ or ‘a short period of time’. Diminutive markers
are supposed to follow this general trend (Jespersen 1922: 402). Bauer (1996)
examines this claim in a cross-linguistic sample of thirty languages that have
overt markers for both diminutives and augmentatives, and finds that as
a linguistic universal, there is no overall trend in this direction. But he points
out that there are some languages in which the proposed pattern does seem to
exist. The question then becomes whether English is one of those languages.
English does not have many diminutive suffixes, and they are not used as
widely as diminutive suffixes in Dutch, German, Italian or Spanish.
Nevertheless, there are some relevant forms.

Suffix Examples Sound Comment

-er: fresher, footer, rugger, soccer ə Also with additional -s
-ette: kitchenette, roomette, statuette e Also with other meanings
-ie: auntie, ciggie (‘cigarette’), doggie, i Also spelled with <y>
goalie
-kin: catkin, lambkin, pannikin ɪ Very rare
-let: booklet, piglet, streamlet ə
-ling: foundling, weakling; duckling, gosling, ɪ
spiderling
-o: ambo, drongo, lesbo, medico əʊ More common in Australia
-s: ducks, pops, preggers s~z Alone, often in vocatives or
hypocoristics
-zza Bazza (from Barry), Mazza (from ə used only in hypocoristics
Amanda)

If we just look at this set of suffixes, it is hard to see that close front vowels
feature particularly significantly in diminutive markers in English. There are,
though, a significant number of other ways of marking the small in English,
 17 Word-Formation and Phonetics 141

including clipping, and the forms hypo-, mini-, micro-, nano- and possibly -ola
(as in aureola) and -een (as in poteen, an Irish loan) (see Bauer et al. 2013). It
could be argued that to the extent these show relevant vowel sounds, they are
vowels sounds from other languages, and not particularly relevant. Alternatively,
they may be borrowed precisely for their phonetic form. English does have other
ways of marking diminutivization, though, illustrated in
beddy-byes, drinky-poo, kitten, owlet, toothy-peg
This is a mixed set, and not everyone will consider them relevant. But to the
extent that they are relevant, they do not add greatly to the notion that close
front vowels occur in English diminutives. The one factor that does point in that
direction is that of all the diminutive suffixes in English -ie is by far the most
productive, overwhelming all other models except possibly clipping (which is
iconic in a different way). In terms of token-frequency, then, close front vowels
are the rule.

17.3 Rhyme in Word-Formation

There is an argument to be made that rhyme is phonetic rather than phono-
logical. How far this argument is accepted will depend on just what is con-
sidered to be phonology, and just what is considered to be phonetics.
The argument in favour of viewing rhyme as phonetic is that phonologically
distinct items can nevertheless be perceived as good rhymes. For example,
sword, cored and awed all rhyme in Standard Southern British English, despite
the fact that the /d/ in cored and awed alternates with /t/ and /ɪd/ on different
bases while the /d/ in sword does not, and despite the fact that the <r> in cored
represents an alternation between /r/ and Ø, while there is no /r/ pronunciation
of the <r> in sword.
The argument against seeing rhyme as phonetic is that detailed phonetic
analysis shows that freeze and frees are phonetically distinguishable (see Plag
et al. 2017, Seyfarth et al. 2018). If we see these as being homophones (and
therefore as rhyming) we must be operating at the phonemic level (which is
phonological) rather than at the phonetic level.
For present purposes, it is not particularly important which of these positions
we want to take – all that is at stake is where such material might be dealt with
in this book. Treating it here is not necessarily unmotivated.
The first question we need to ask about rhyme is what its function is outside
of its poetic usages, where it has functions such as making the poetry’s structure
more predictable and thus easier to remember and also more enjoyable for the
listener. There is a huge literature on the subject, and I do not wish to get
embroiled in it here. It seems likely that rhyme outside poetry draws attention to
the rhyming words, perhaps linking them together in some way. Just what that
142 Part IV Interfaces

attention involves is not necessarily simple. Bauer and Huddleston (2002:

1666) talk of a “trivialising effect” in compounds whose elements rhyme
such as copshop and gang-bang. But any trivializing may be the result first,
of the ludic element involved in rhyme and second of drawing specific attention
to the words in a way which contrasts with the serious nature of the real-world
referents involved. Items such as walkie-talkie, which may be based on
reduplication rather than compounding, show the attraction of rhyme for the
user, and expressions such as hoity-toity have only the rhyming form and no
semantic content, to justify them. Rhymes are clearly fun, and because rhyme is
relatively rare in ordinary usage, it is surprising. Note the standard reaction
when a person produces a rhyme without having planned it: I’m a poet and
I don’t know it.
Alliteration and assonance have similar effects, but are even less clearly
phonetic rather than phonological.

Challenge
How many words can you find that end in <ug>, like bug, slug, smug? Does
the /ʌɡ/ have a meaning in some of them? If so, what is the meaning, and in how
many of them is it found? Is there any external motivation for the sound to
represent that particular meaning? Do speakers agree? Try the same exercise
with initial /fl/ or with final /ʌf/. Do we need to distinguish between different
types of phonaestheme, and if so, how?

References
Abelin, Åsa. (1999). Studies in sound symbolism. PhD dissertation, Göteborg
Universitet. Acta Universitatis Gothoburgensis 17.
Adams, Valerie. (2001). Complex Words in English. Harlow: Longman.
Bauer, Laurie. (1996). No phonetic iconicity in evaluative morphology. Studia
Linguistica 50, 189–206.
(2019). Rethinking Morphology. Edinburgh: Edinburgh University Press.
Bauer, Laurie & Rodney Huddleston. (2002). Lexical word-formation. In Rodney
Huddleston & Geoffrey K. Pullum, The Cambridge Grammar of the English
Language. Cambridge: Cambridge University Press, 1621–1721.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Benczes, Réka. (2019). Rhyme over Reason: Phonological Motivation in English.
Cambridge: Cambridge University Press
Bergen, Benjamin K. (2004). The psychological reality of phonaesthemes. Language
80, 290–311.
Davies, Mark. (2008–). The Corpus of Contemporary American English (COCA).
Available online at www.english-corpora.org/coca/
Feist, Jim. (2013). ‘Sound symbolism’ in English. Journal of Pragmatics 45, 104–18.
 17 Word-Formation and Phonetics 143

Hinton, Leanne, Johanna Nichols & John J. Ohala (eds.). (1994). Sound Symbolism.
Cambridge: Cambridge University Press.
Jespersen, Otto. (1922). Language: Its Nature, Development and Origin. London: Allen
& Unwin.
Marr, Vivian (ed.). (2008). The Chambers Dictionary. 11th edn. Edinburgh: Chambers
Harrap.
Plag, Ingo, Julia Homann, & Gero Kunter. (2017). Homophony and morphology: The
acoustics of word-final S in English. Journal of Linguistics 53, 181–216.
Seyfarth, Scott, Marc Garellek, Gwendolyn Gillingham, Farrell Ackerman, & Robert
Malouf. (2018). Acoustic differences in morphologically-distinct homophones.
Language, Cognition and Neuroscience 33, 32–49.
Yang, Charles. (2016). The Price of Linguistic Productivity. Cambridge, MA: MIT
Press.
18 Reflections on the Interface between
Word-Formation and Orthography

In Chapter 15, apparently morphophonemic rules shown in, for instance,

Shaw / Shavian were discussed and considered to be based on spelling rather
than phonology. A similar case might be made for velar softening. These are,
then, instances where word-formation has an interface with orthography. In this
chapter, we look for other cases of this interface.
One such case is with acronyms. Acronyms are abbreviations which are
pronounced as words rather than as sequences of letters. Examples are Aids
‘acquired immune-deficiency syndrome’, BASIC ‘Beginners’ all-purpose sym-
bolic instruction code’, NATO ‘North Atlantic Treaty Organization’ and WASP
‘white Anglo-Saxon protestant’. Apart from the variation in the way in which
acronyms are spelled, sometimes using all capitals, sometimes using lower-case
letters, it should be noted that there is a certain amount of laxness in what counts as
an element of the base expression from which the acronym is derived: all-purpose
is treated as providing a single letter to BASIC while Anglo-Saxon provides two
letters to WASP. Articles are often ignored. We should also consider the emergence
of the backronym, in which the appropriate surface form is chosen first, and then
a suitable phrase is invented to provide those letters: SAD ‘seasonal affective
disorder’ is probably a backronym, and perhaps the most famous example is the
USA PATRIOT Act ‘Uniting and Strengthening America by Providing Appropriate
Tools Required to Intercept and Obstruct Terrorism’, where the full version sounds
far more negative and threatening than the backronym.
The link with orthography comes from the fact that acronyms are based on
the letters in the base expression rather than the sounds. Once the letters are put
together, they are then pronounced according to the usual grapheme-to-
phoneme rules for English. Consider BASIC as cited above. If we look at the
way in which the relevant letters are pronounced in the original, we would
expect /b/, /ɔː/, /s/, /ɪ/, /k/, but we get /beɪsɪk/, where long <a> is pronounced /eɪ/
in accordance with the same rules that lead to the pronunciation of the ordinary
word basic. Similarly, if PATRIOT used the phonemes represented by the
letters in the supposed underlying form, we would get /pətriɒt/ rather than the
actual /peɪtriət/. Even when the acronym does not have the form of an already

144
 18 Word-Formation and Orthography 145

existing word, the same principles are involved. In the case of UNICEF ‘United
Nations International Children’s Emergency Fund’ we get /juːnɪsef/ rather than
/juːnɪʧɪf/, and with laser ‘light amplification by stimulated emission of radiation’
we get /leɪzə/ rather than /læser/ (which is phonotactically impossible).
These principles seem to hold in many places where there is shortening.
Consider clipping compounds, in which the elements of the compound are
clipped forms of words. Modem ‘modulator and demodulator’, pronounced
/məʊdem/, following the spelling, with an open first syllable, rather than
/mɒdiːm/ which we would get from following the values attached to the letters
in the base form. Similar comments can be made about fin-lit ‘financial liter-
acy’, hi-fi ‘high fidelity’, sci-fi ‘science fiction’. Biopic is relevant if pro-
nounced /baɪɒpɪk/ but not if pronounced /baɪəʊpɪk/. It is not clear whether
examples like e-mail and Man U count as clipping compounds, but the single
letters get the pronunciation of the name of the letter, whatever the source: we
do not find */emeɪl/ for example.
Clippings that are not parts of compounds seem to be slightly more immune
to this phenomenon, but even then, we find both respelling to maintain the
appropriate phonemes and changed pronunciation of the original (Jamet 2009).
Respelling: caff (< café), coke (< Coca-Cola or < cocaine), cos (< because),
cuz (< cousin), delish (< delicious), indie (< independent film), mike
(< microphone), natch (< naturally), Oz (< Australia), peeps (< people), ute
(< utility vehicle)
Repronouncing: merc (/mɜːk/ < mercenary /mɜːsɪnəri/), mayo
(/meɪəʊ/ < mayonnaise /meɪəneɪz/), rasp (/rɑːsp/ < raspberry /rɑːzbri/), zoo
(/zuː/ < zoological garden, at least for those who say /zəʊəlɒʤɪkəl/)
Both simultaneously: bike (< bicycle)
Note that in fan (/fæn/ < fanatic) the full vowel follows the spelling, because
a lexical item with only the vowel /ə/ would not be possible. What Bauer and
Huddleston (2002: 1636) call ‘embellished clippings’ have extra material at the
end, but this does not seem to affect the pronunciation, except that words
ending <(s)sie> seem to be pronounced with a /z/ rather than an /s/: Aussie
(/ɒzi/ < Australian), mozzie (/mɒzi/ < mosquito, note the spelling reflects the
new pronunciation), possie (/pɒzi/ < position). Just occasionally the embellish-
ment allows more consonants in the base of the clipping, which must help
recognizability: lesbo (< lesbian, with lez and lezzie as alternative forms – none
of them necessarily flattering terms).
Blends seem to be rather more constrained by the pronunciation, but spelling
is also relevant. Musicassette and magicube have /i/ rather than /ɪ/ for those
speakers for whom the two vowels are not homophonous, tigon has a full vowel
for the <o>, even though the corresponding vowel in lion is reduced. The
meaning behind evilution and medievil (Lopez Rua 2012) can be discerned only
146 Part IV Interfaces

from the spelling, as can fauxbia (Beliaeva, p.c.), and funerealm depends on the
spelling to work at all. Adams (1973: 153) comments on the use of /ɪ/ rather
than /iː/ in the first syllable of skinoe (from ski and canoe).
Back-formation occasionally shows examples where the derived form’s
spelling determines its pronunciation, especially with cases that end with <t>
such as contracept, where the <t> is pronounced /ʃ/ in the complete form
contraception. Attrit from attrition shows the same phenomenon. In this case,
though, the same alternation is found in instances which are not formed by
back-formation, such as act / action and alternate / alternation.
The most obvious place where orthography is involved is in the creation of
words – often tradenames, and often involving drugs – where letters from
a long name are put together to make a new name. An example from Barnhart
et al. (1990) is pemoline ‘a drug used to relieve depression’, supposed to come
from PhEryliMino-OxazoLIdinoNE (where the capitals indicate the letters
that made it through to the name). Although it may not be true, it looks as
though the drug could just as well have been called phyzolone. It is not clear
that words like this are formed by word-formation (as opposed, say, to word-
manufacture, assuming the two can be distinguished), but the influence of the
spelling is clear.
More surprising than these are the instances of apparently straightforward
word-formation where the spelling nevertheless has an influence. Consider the
name of the educationalist Piaget. His name, if not pronounced in the French
way, is /piæʒeɪ/. There is an adjective derived from his name, Piagetian. We
might expect this to be pronounced /piæʒeɪən/ (which is apparently the most
usual pronunciation in the USA) or perhaps /piæʒiːʃən/ (like Venetian), though
this does not seem to occur. However, we do find (Wells 1990) /piəʒetiən/. The
/t/ must arise from the spelling. We find the same phenomenon with other
French names, such as Corneille /kɔːneɪ/ giving Corneillian /kɔːneɪliən/ or
/kɔːniːliən/, Flaubert /fləʊbeə/ giving Flaubertian /fləʊbɜːtiən/, Louis /luːi/
giving Louisiana /luːiːziænə/. Rabelais giving Rabelaisian /ræbəleɪʒən/. It is
not clear how far this effect spreads. Pasteur, with /æ/, gives rise to pasteurize,
with /ɑː/, which might be an orthographic influence, but it is less clear.
What becomes clear is that word-formation is not purely a matter of creating
phonological strings from phonological representations: orthographic repre-
sentations are also important in creating new words.

Challenge
Find other examples of new inventions being named in the way illustrated by
pemoline in the text. Could you generate such names by taking the full form and
applying some mathematical formula to it (e.g. take the first letter, then
a random letter following that, then the letter in numerical position that is the
 18 Word-Formation and Orthography 147

square of the second letter, then add the number of the second letter . . .). If not
(and I assume that the answer is that you cannot) what constraints are there on
choosing letters to make up such names? For example, what consonant clusters
arise in the names? Are there always analogies with already known words?

References
Adams, Valerie. (1973). An Introduction to Modern English Word-Formation. London:
Longman.
Barnhart, Robert K., Sol Steinmetz & Clarence L. Barnhart. (1990). Third Barnhart
Dictionary of New English. New York: Wilson.
Bauer, Laurie & Rodney Huddleston. (2002). Lexical word-formation. In Rodney
Huddleston & Geoffrey K. Pullum, The Cambridge Grammar of the English
Language. Cambridge: Cambridge University Press, 1621–1721.
Jamet, Denis. (2009). A morphophonological approach to clipping in English: Can the
study of clipping be formalized? Lexis 2. https://doi.org/10.4000/lexis.884
López Rúa, Paula. (2012). Beyond all reasonable transgression: Lexical blending in
alternative music. In Vincent Renner, François Maniez & Pierre J.L. Arnaud (eds.),
Cross-Disciplinary Perspectives on Lexical Blending. Berlin: De Gruyter Mouton,
23–34.
Wells, J.C. (1990). Longman Pronunciation Dictionary. Harlow: Longman.
19 Reflections on the Interface between
Word-Formation and Borrowing

19.1 Introduction
Neoclassical word-formation creates multiple problems of description for
many European languages, not least for English. The words formed by this
process are generally called ‘neoclassical compounds’ because a word like
photograph is not a word formed in Greek and borrowed into English – the
Ancient Greeks did not have photographs. Rather, they are words made up of
Greek elements (hence the element ‘classical’), but they are modern (hence
‘neo-’); the elements are derived from Greek words (hence ‘compounds’), but
they may include affixes as well – sometimes Greek affixes (such as the -ia in
euphoria) and sometimes English affixes (such as the -er in philosopher).
Because of this, it is probably more accurate to term the relevant words
‘neoclassical formations’. What we are dealing with here is a part of the word-
formation of English (and, of course, of other languages as well) whose
fundamental elements have been borrowed, mainly from Greek but also from
Latin, and many of the problems associated with neoclassical formations arise
because of this.

19.2 Patterns of Neoclassical Formation

Canonical neoclassical formations are created from two elements each of
which is borrowed from one of the classical languages. Photograph illustrates
a word made up of Greek elements, agriculture one made up of Latin elements.
In words found in modern usage, it is not always possible to tell whether the
formation is created in English (sometimes borrowed from French or German)
or is borrowed as a unit from the classical languages. Photograph is a recent
formation, agriculture is borrowed from Latin (probably via French).
The rules of formation for these words are not always clear. The medial -o- in
words of Greek origin and the -i- in words of Latin origin are probably best
treated as a linking element, and are usually omitted when the form of the first
element brings two vowels into contact with each other, as in homonym from
homo- ‘same’ and -onym ‘name’ or nephralgia from nephr- ‘kidney and -algia

148
 19 Word-Formation and Borrowing 149

‘pain’ (contrast nephropathy ‘kidney disease’ from nephr- + o + path + y). The
linking element is normally deleted before Greek [h] (often termed ‘rough
breathing’ in the literature on the Greek language) as well. These rules are not
always followed by the time the word is found in English, so that glycaemia,
glychaemia and glycohaemia are all used synonymously in English, albeit not
equally frequently at the same period. Greek and Latin inflections are usually
deleted in both elements, so that the Latin form of agriculture had a case/
number marker -a at the end: agricultura. The stress in English is typically
determined over the complete formation, treated as a word without morpho-
logical boundaries.
In a less canonical pattern, a classical element is linked with an English word.
Demonology, Egyptology and typology provide examples of English words
linked with Greek -ology. Even though type comes from a Greek original, the
meaning shown in typology is not the meaning it had in Greek. The -ology
element, derived from a Greek element meaning ‘word’, has developed a new
meaning of ‘subject of study, science’ in modern English. The word hieroglyph
may provide a better model for this reason, though a less familiar one. Once
a foreign element becomes familiar enough to be considered an unexceptional
word, it can be treated the same way as native forms, and used in the same
constructions as native words. Although words like Egyptology are different
from words like photograph from an etymological point of view, in the eye of
people coining new terms, they are probably much the same, or the -ology is seen
as being an English affix. This also accounts for the phenomenon of a mixture of
Latin and Greek elements occurring in the same word, as in television.
As was mentioned above, neoclassical elements can also contain affixes,
either classical or English. In some cases, it may not be clear whether the affix is
classical or whether it is English. These affixes may also attach to obligatorily
bound stems that are not compounds. This means that a sequence of, say,
classical element + classical element + suffix may have the structure [[classical
element] [classical element + suffix]] or [[[classical element] [classical elem-
ent]] [suffix]], and it is not always easy to determine which structure applies.
The English affixes (some of them with etymons from the classical lan-
guages, sometimes through French) that are used tend to be the most frequent
affixes, as illustrated below.

-al exoskeletal, matriarchal, octahedral, primogenital, puerperal

-er oceanographer, philosopher
-ic chromospheric, encephalitic, haemostatic, homeopathic, matronymic,
philharmonic, photostatic, plutocratic, psychogenetic, pyrotechnic
-ism demagogism, necrophilism, pentadactylism, phototropism,
zoomorphism
-ist gymnosophist, homeopathist, hydrologist, oenologist, palaeobotanist
150 Part IV Interfaces

-ize anthologize, democratize, hypnotize

-ous androgenous, coprophagous, homophonous, phosphorous,
polygenous, zygomorphous
-y androgeny, cacophony, gastronomy, geometry, hagiolatry,
historiography, homonymy, hydrocephalous, oligarchy,
philanthropy, polygamy, rhinoscopy, tracheotomy

Greek affixes include

a-, an- atypical, aneroid

endo- endocarp
exo- exogamy
hypo- hypochondria
neo- neology
poly- polyglot
syn-, sym- syncope, sympathy
-ia phobia
-itis hepatitis
-oid anthropoid

Latin affixes include

ab-, abs- ablative, abstruse

ad-, ac-, af- etc. advent, accept, affluent
con-, col-, com-, cor- conceal, collect, compact, correct
contra- contradict
in-, im-, il-, -ir invisible, impatient, illegible, irrelevant
inter- intercede
pro- produce
-ion legion
-ix testatrix

There are two important points to note about these affixes. The first is that it
is often difficult to be sure that they should be classified as affixes rather than as
compound elements. The classification in the original language may be differ-
ent from the classification in modern English. This is particularly the case
where the element is homophonous with an adjective, adverb, preposition or
number in the original language. The second is that, especially with Latin
affixes, the form may also be used as an English affix, sometimes with
a different phonology or a different meaning from the original Latin. For
example, Latin ex- in expel, where ex- means ‘out’ is not the same element as
the ex- in ex-wife, where ex- means ‘former’; the Latin re- in refer, although it is
historically related to the English re- in re-educate, is pronounced differently
(/rɪfɜː/ versus /riːedjʊkeɪt/) and is rather more transparent in meaning.
 19 Word-Formation and Borrowing 151

Finally, it should be noted that some of these borrowed elements can have
multiple functions in English. Mega- looks like a neoclassical element in
megaphone, like a prefix in mega-deal, and like a lexeme when it means
‘excellent’.

19.3 Discussion
Borrowing from Greek and Latin (and to a lesser extent from French, although
many classical words are transferred to English via French) causes problems of
description in English. If we borrow Schadenfreude from German, we treat it as
a single unanalysable unit despite the fact that it is analysable for German
speakers. If we borrow orthodox from Greek, where it originally came from
elements meaning ‘right opinion’ we have a cluster of related words that allow us
to isolate each element: words like orthopaedic, orthographic, orthoepy, para-
dox, heterodox. The temptation is thus to treat such words as illustrating a kind of
word-formation, especially as the processes are often productive (the overall
pattern of compounding is productive, and compounds with particular elements
may be productive). Yet the meaning of the elements or of the word as a whole
may have changed over time (orthodox would probably be glossed as ‘received
opinion’ rather than ‘right opinion’ these days), the link between mage and
magic is etymological, but no longer semantic. There may be alternative expres-
sions (e.g. received wisdom) which have to be distinguished from the borrowed
word, often in very subtle ways. The analysis becomes awkward, and it may not
be possible to treat all neoclassical forms in the same way.
Overall, the fact that so many words have been borrowed into English from
the classical languages and used in formations that look more or less like
classical formations means that a coherent treatment of words of this type is
theoretically and practically very difficult. Borrowing of isolated words from
other languages has a minimal effect on the system of English morphology. We
can even borrow affixes (such as -age and -ment from French) without upset-
ting the general way in which word-formation operates. The swamping of
English vocabulary from Greek and Latin, though, upsets the processes of
word-formation by introducing new patterns which are too pervasive to be
ignored, but which are often variable, difficult to analyse (depending, for
instance, on when they are coined and by whom) and which show traces of
the gradual assimilation of the Greek and Latin elements into English, often
with formal and/or semantic changes.

Challenge
The suffix -ia in mania and phobia was cited in this chapter as an instance
of a Greek suffix. Is it used only on Greek bases, or is it used elsewhere as
152 Part IV Interfaces

well? The suffix -ic in manic and phobic, on the other hand, while of Greek
origin (boosted by Latin usage), is usually discussed as an English suffix.
Are the two really as different as this implies? How do they differ from
each other, and how can they be seen as two distinct examples of the same
phenomenon?
Part V

Patterns of Word-Formation in English

20 Reflections on the Limits of Conversion

20.1 Introduction
One of the places where the borderline between word-formation and syntax is
least clear is in the area of conversion, and specifically its borderline with what
is called ‘coercion’ (Pustejovsky 1995). Although I talk about syntax here,
coercion is really a semantic process whereby the semantics of a particular
word is adjusted in order to make sense in its syntactic/semantic/pragmatic
environment. There are many examples of this phenomenon, which typically
work within a word-class while conversion typically works over word-class
boundaries (for discussion and more examples and for a classification, see
Audring and Booij 2016).
They seem to be very British.
I began the book.
The light flashed until dawn.
There are three Rachels in my class.
I need to read some index.

Very British makes the adjective British gradable, when it would be ungrad-
able by default; begin usually implies an activity, while a book is an object, so
that beginning a book has to be interpreted as reading or writing the book (at
least in normal discourse; it could feasibly be interpreted as eating a book under
suitable conditions); flashing is a momentary event, while until dawn implies
continuity; Rachel is usually a name, but here must mean ‘people called
Rachel’ and so be a common noun; index is usually a countable noun, but in
context has to be read as uncountable. The difficult examples arise where it is
not clear whether there is a shift of word-class or not. For example, Nagano
(2018) argues that the difference between relational and qualitative adjectives
should be seen as a matter of conversion rather than coercion, so that the
difference seen in
A young professional soccer player
A (very) professional young soccer player

155
156 Part V Patterns of Word-Formation in English

(where the differing order of professional with relation to the adjective young
indicates a different class of adjective) is a matter of a shift of word-class, not
a matter of coercion, despite the fact that conversion is usually taken to imply
a change in major word-class (e.g. a change from noun to verb or noun to
adjective without any change in form – for recent discussion, see Bauer and
Valera forthcoming).
In this chapter, I look at three places where the borderline between conver-
sion and coercion may be relevant.

20.2 Example 1: The Rich and Related Constructions

Payne and Huddleston (2002: 417) point out that it is possible to have what they
term “fusion of internal modifier and head” with some classes of adjective and
some syntactic structures. Some of their examples are given below. The
importance of Payne and Huddleston’s analysis in the present context is that
they see this as neither conversion nor coercion.
Henrietta likes red shirts, and I like blue.
Henrietta likes Russian vodka, and I like Polish.
I prefer cotton shirts to nylon.

They note that “the boundaries to what is admissible are hard to define”.
They include here examples like
The rich cannot enter the kingdom of heaven.
In relation to such examples, they say that the deceased and the accused are
unusual in having the potential to be singular or plural. They further note
(Payne and Huddleston 2002: 418) that adjectives derived from nouns by
conversion cannot participate in this structure:
*The intellectual are not to be trusted. [my example, not theirs]

They reject (2002: 421) an analysis whereby the relevant word has a different
function in modifier and head positions while retaining its fundamental cat-
egory because the constructions in
I prefer cotton shirts to nylon.
I prefer cotton to nylon.

would not be distinguished if nylon were a noun in both.

There are several things to worry about here.
The first is that it is not clear whether we are dealing with the same
construction in all these cases. In
Henrietta likes red shirts, and I like blue
 20 Limits of Conversion 157

we can only interpret blue in the light of the preceding material. If we had
Henrietta likes red hair dye and I like blue
blue would have a different interpretation. This seems to imply that, in this
instance, blue gains its interpretation as a function of the syntactic environment.
This is different from what happens in
The rich cannot enter the kingdom of heaven
where the rich must mean ‘people’, independent of the syntactic environment,
and where a definite determiner is required, which is not the case in some of
Payne and Huddleston’s other examples. Payne and Huddleston cite examples
with adverbial modification of rich in such uses to show that rich remains an
adjective, but adjectival modification is also found, which implies that rich is
(or has at least some features of) a noun.
The very rich cannot enter the kingdom of heaven.
The powerful rich cannot enter the kingdom of heaven.
It is far from clear that nouns by conversion are excluded from this construc-
tion, although when they do appear they do not have access to the range of
possibilities that such adjectives usually have. Consider
I think we the intellectual are suffering from too much thinking because we
haven’t gotten rid of ignorance completely (https://philosophistry.com/arch
ives/2004/04/if-ignorance-is.html accessed 12 August 2023)

It is relatively easy to show that adjective to noun conversion like intellec-

tuals and the construction illustrated by the rich are different constructions.
One relevant question to ask, though harder to answer, is whether this means
that they belong to different word-classes.
intellectuals the rich
can take a range of determiners requires the determiner the
can take a plural marker cannot take a plural marker
relatively restricted productivity relatively free productivity
freely modified by adjectives, not can be modified by adverbs or
modified by adverbs adjectives
The productivity of the construction like the rich is problematic. Where the
rich gets plural agreement, it musts refer to humans (perhaps better, to
rational beings), while singular agreement implies that the reference is to non-
humans (except with the accused and the deceased mentioned by Payne and
Huddleston 2002).
The poor are always with us.
The improbable we will do at once, the impossible takes a little longer.
158 Part V Patterns of Word-Formation in English

The rich are different from us.

The rich is bad for the digestion.
The adjectives that can take part in this construction are numerous, but
apparently not unrestricted, as we see if we compare with what happens in
other languages. For Amade and Bécaud (1967)
L’important c’est la rose
the important is the rose
‘The important thing is the rose.’
And for Hans Andersen in The Ugly Duckling (Andersen [1835–72] 1961: 197)
[D]et Grønne er godt for Øinene
The green is good for eyes.the
‘Green is good for the eyes.’

Yet these do not allow translation into English with the same construction as
we see with the rich above. It seems that the rules for noun phrases apparently
headed by adjectives are not equivalent across languages. Payne and
Huddleston (2002: 417) classify these examples alongside other such as
The French do these things differently.
This is verging on the immoral.
They like to swim in the nude.
They see such examples, just like the rich examples, as having special
interpretations, and thus, presumably, in the terms of Construction Grammar,
being members of different constructions. However, the French are people, just
like the rich, the immoral refers to something non-human, just like the impos-
sible, and only the nude seems semantically rather marginal, and it is classified
as an idiom by Payne and Huddleston.
Just how freely this construction can be used productively, however, remains
hard to determine. Colour adjectives seem to be excluded, but the black and the
white seem possible as racial terms, but in just those contexts black and white can
be nouns by conversion (like intellectual) and have other grammatical options.
the presence of poor whites was a constant threat, challenge and embarrass-
ment to South African formations of white identity. (Thandiwe Ntshinga,
Poor Whites: A Threat, Challenge and Embarrassment to South African
Formations of White Identity, 2016. https://brill.com/display/book/9781848
883833/BP000007.xml accessed 21 March 2024.)
Adjectives which do not seem to be possible in this construction include
ample, American (unlike French!), big, basic, flawless, icy, but how far any of
 20 Limits of Conversion 159

these are just not widely used and whether there are any generalizations over
these unsystematic examples are questions that need to be considered further.
All that can be said thus far is that there appear to be sufficient lexically
determined gaps for us to say that this looks like a morphological construction
rather than a syntactic one.

20.3 Example 2: Indicators of Composition

Another type of puzzle is raised by words like brick, copper, cotton, iron,
leather, steel, stone. Such words are used as nouns, as in
Cotton is grown in India and in the USA.
We need cutlery made of steel.
The stone for Stonehenge was brought from Wales.

At the same time, they have usages which appear to be adjectival in examples like
We used to have a copper kettle.
A stone wall enclosed the field.
The facing is brick.

This means that these words cover the functions of both wood and wooden, both
wool and woollen, where one member of the pair is a noun and the other an
adjective, and so look as though they have a single form for two word-classes,
which is a typical expectation when we are dealing with conversion. However,
matters are not quite that simple.
The first point to consider is that there are words which fit neatly into neither
the category like stone nor the category like wool. Consider gold and silk. Gold
can be a noun as the name of a metal (My ring is made of 18 carat gold) and an
adjective as a colour (Their names were spelt out in gold letters), and golden
can also have both functions (a golden crown, a golden beach) or neither
(a golden oldie, a golden opportunity). A gold coin means ‘made of gold’
(or, these days, often no more than ‘the colour of gold’ – a gold coin donation
does not necessarily involve a Krugerrand), but golden in (kill the goose that
lays) the golden egg also means ‘made of gold’. It appears that each word has
two polysemes, but that the polysemes of each are the same. The case of silk is
rather easier, and more typical. The -en suffix creates an adjective, but an
adjective whose meaning is ‘resembling ~’ rather than ‘made of ~’. So silken
is usually used of hair, or of a touch rather than of dresses or other clothing. Silk,
on the other hand, can mean the material (She bought some Indian silk) or that
something is made of that material (silk stockings). Wooden can mean ‘made of
wood’ (a wooden box) or ‘resembling wood and therefore not natural for
a person’ (a wooden expression). Woollen usually means ‘made of wool’, but
160 Part V Patterns of Word-Formation in English

in an expression like woollen mill is parallel to flour mill in the sense that the
mill produces wool or flour, despite the lack of grammatical parallelism. Much
of the complexity here is due to lexicalization. The suffix -en is no longer
productive, and individual words have developed idiosyncratically, at least to
some extent. This means that the forms with silk, gold, wool, copper are more
likely to be semantically regular than the forms with an overt adjective.
The second point to consider here is that the relevant words do not behave
like adjectives in allowing comparative forms or adverbs in -ly to be derived
from them. We can see this as a function of their meaning (why would we need
an adverb from copper?) or as a sign that they are not adjectives, but nouns. To
the extent that comparatives or adverbs are required, the noun is first turned into
an adjective, and then that adjective is used as a base: silkier, silkiness,
steeliness, stonier, woollier, woollily and so on. Perhaps more importantly,
many of these words are awkward in predicative position: brick, used above, is
much better in predictive position than stone, for example.
The third point is that if we accept, for instance, brick wall, as a noun + noun
sequence, then we expect any premodifying adjective to modify the entire unit,
as in tall brick wall, where it is the wall and not the brick that is tall. However,
when the first element alone is modified, it is modified by an adjective not by an
adverb, and it is where the adjective + the first element is already fixed
expression: a red brick wall is a wall made of red brick, an Egyptian cotton
towel is made of Egyptian cotton rather than being (necessarily) an Egyptian
towel. This behaviour is typical if these words are nouns; if they were adjec-
tives we might expect *a redly brick wall, parallel to a beautifully slim figure.
My conclusion is that despite their use in some constructions which make
them look a bit like adjectives, these words are really nouns, and that we do not
have to worry about deriving an adjective from a noun (or vice versa) here at all.

20.4 Example 3: Metonymy

As a final example, consider instances of metonymy, like
The omelette in the corner hasn’t paid yet.
You don’t often see a jogging hijab.
In both these cases, a person is referred to by something closely connected to
them: the meal they ordered in a café or the headscarf they are wearing, and this
is central to metonymy. Figures of speech like metonymy are not usually
considered to be word-formation processes, because it is assumed that figures
of speech are processed by general, non-linguistic, cognitive principles, and do
not need to be dealt with explicitly in a grammar, because they transcend
language and can be found, for example, in pictures or in music (consider the
 20 Limits of Conversion 161

use of a piece of 1920s music on a film soundtrack to indicate a period in which

the action is supposedly taking place) (Littlemore 2015). Unusually, though,
conversion is often seen as a form of metonymy, even though it typically
demands a change of word-class (Dirven 1999). We thus have two contrasting
views of the link between metonymy and conversion: on the one hand, meton-
ymy is not word-formation and not fundamentally linguistic, although it has
many linguistic effects; on the other hand, conversion (which is a type of
metonymy) is a word-formation process, a linguistic process. Furthermore,
although metonymy does not usually cause a change of word-class when it
applies in language, in conversion it typically does. Clearly, we have a problem
of theoretical coherence here, one which can be solved by stipulation, although
we would prefer an alternative solution.
One way to do this is to go back and look at the instances of coercion.
Consider, for example, British versus very British. The two are cognitively
close to each other, because being seen as British can suggest that there are
several features that can give this impression, and if many of them are relevant,
we have a case for seeing Britishness as being a cline. Reading a book is an
activity in which the book plays a central and cognitively crucial role. One flash
is a constituent part, and a defining one, in a series of flashes. The form Rachel
(that is, the word itself) is what is shared by the people whose name is Rachel.
In all these cases, we might be discussing metonymy, a figure of speech where
one entity is used to provide access to another entity to which it is somehow
related (Littlemore 2015: 4). That is, there is an argument for seeing many (and
perhaps all) cases of coercion as cases of metonymy. If that is true, then
coercion and conversion are just terms for subtypes of metonymy, and the
differences between the two are just the ways in which the subtypes differ from
each other. Depending on one’s position, the differences or the similarities
might be considered the most important factors.
Ironically, the biggest objection to such a view comes from within Cognitive
Linguistics. Brdar and Brdar-Szabó (2014) claim that the term metonymy is
being applied so freely that it becomes meaningless. Personally, I find an
argument that says that metonymy is ubiquitous and so we should call it
something else in part of its range a rather strange argument. Seeing the
extensive scope of metonymy is part of its value. This does not, for me,
imply that there is no point in having specific terms for subtypes of metonymy,
but does imply that there is benefit in recognizing the overarching category.

20.5 Discussion
Although conversion is well recognized as part of word-formation, that does
not mean that there is agreement on just what is part of conversion and what is
not. The trouble with this is that it implies that while we may recognize
162 Part V Patterns of Word-Formation in English

conversion in some instances, our definition of conversion or our understand-

ing of it is not sufficient to determine just what should count as conversion and
what should not. Apart from the instances discussed here we have cases like
The enemy downed three of our planes, where the verbal uses of prepositions is
extremely rare and semantically unsystematic, cases like I don’t want to hear
you effing and blinding where the semantic regularity of the process meaning
‘saying f . . . ing or blinding’ is remarkably constrained and it is not clear
whether we have verbs or nouns in the output, because No effing or blinding
makes the relevant words look like nouns, and cases like I’m feeling very
under-the-weather where a phrase appears to be the source of what might be
considered a word (see further Section 16.3). It may be that we need better
criteria for determining the limits of conversion.

Challenge
The construction illustrated in
The rich are not like you and me
has long been a puzzle. It is sometimes referred to as ‘partial conversion’, but
that does not really answer the question of whether or not it is a type of
conversion. Calling it ‘partial’ draws attention to its syntactic restrictedness,
which raises the question of whether conversion should be defined primarily by
a shift in word-class or whether adopting the full set of features of the new
word-class is part of the primary definition. This question is probably not
resolvable. But we can consider whether the examples like the French, the
Dutch, the British (but not *the American, *the Indian), meaning ‘the people’ is
the same construction as the one with the rich or a separate one? Is being used
with plural concord and meaning ‘people’ sufficient to link the two, as was done
in this chapter? What evidence can you find in either direction?

References
Amade, Louis & Gilbert Bécaud. (1967). L’important c’est la rose. La Voix De Son
Maître [His Master’s Voice] – EGF 963
Andersen, H.C. ([1835–72] 1961). Samlede eventyr og historier. Jubilæumsudgave.
Odense: Flensted.
Audring, Jenny & Geert Booij. (2016). Cooperation and coercion. Linguistics 54,
617–37.
Bauer, Laurie & Salvador Valera. (forthcoming). Conversion: A position paper. To be
published by Oxford University Press.
Brdar, M. & R. Brdar-Szabó. (2014). Where does metonymy begin? Cognitive
Linguistics 25, 313–40.
 20 Limits of Conversion 163

Dirven, René (1999). Conversion as conceptual metonymy of event schemata. In Klaus-

Uwe Panther & Günter Radden (eds.), Metonymy in Language and Thought.
Amsterdam: Benjamins, 275–87.
Littlemore, Jeannette. (2015). Metonymy. Cambridge: Cambridge University Press.
Nagano, Akiko. (2018). A conversion analysis of so-called coercion from relational to
qualitative adjectives in English. Word Structure 11, 185–210.
Payne, John & Rodney Huddleston. (2002). Nouns and noun phrases. In Rodney
Huddleston & Geoffrey K. Pullum (eds.), The Cambridge Grammar of the
English Language. Cambridge: Cambridge University Press, 323–523.
Pustejovsky, James. (1995). The Generative Lexicon. Cambridge, MA: MIT Press.
21 Reflections on Back-Formation

A typical way of forming a new word is to take a known word and, using it as
a base, add an affix. For example, we can find the verb write, add the suffix
-er, and get writer (or, to be more precise, someone did this in the distant past,
and the word that was formed by this method is still transparent in its
formation today).
Historians of the language know that such a process is not always used. To
cite a standard example, the word editor did not arise historically by adding the
suffix -or to the verb edit. Instead, the English verb edit was formed by
a process of deletion from the noun editor, which preceded it.
If we look at the words writer and editor from a twenty-first-century viewpoint
they appear to be parallel formations: both have an agentive suffix added to a verb.
The different pattern of formation has been made invisible with passing time. This
has a couple of important results: the first is that it is difficult to find instances of
back-formation, because they look just like regular formations; the second is that
most examples of back-formation are within the domain of the etymologist.
Knowing about back-formation, once sufficient time has passed, requires expert
knowledge of a type that most speakers of the language do not possess. Moreover,
even the experts can have trouble finding the appropriate information. Marchand
(1969: 391) concludes from this that back-formation “has often diachronic rele-
vance only”. But, as is pointed out by Bauer (1983: 230), “back-formation is
a synchronically productive process in English word-formation”.
One of the results of relatively recent (or ad hoc) back-formation is that the
output is often stylistically marked. It often sounds slightly jocular, even if the
jocular overtones do not usually last particularly long – although couth from
uncouth has retained its non-serious tone for several hundred years. Some
recent examples are given below.
On Saturday we houseclean (William Kotzwinkel. 1994. The Game of Thirty.
Boston and New York: Houghton Mifflin, p. 183)
You’d . . . axe-murder the entire crew (Anne Rivers Siddons. 1980. The House
Next Door. Collins, p. 38)
Derik said, “At least we’re being . . . whatever that word is.”

164
 21 Back-Formation 165

Lucas said, “What word?”

“You know . . .”
“Proactive,” Sally said.
“So let’s proact our asses over to Dalloglio’s place.” (John Sandford. 2002.
Mortal Prey. New York: Putnam, p. 297)
“imagining your consternation,” Leo said.
“I am indeed consterned,” I said, not caring that it wasn’t a word. (Jack
Frederickson. 2013. The Dead Caller from Chicago. New York: Minotaur, p. 305)
fiddling with an electronic cigarette . . . “Apparently, it’s a cessation aid,”
Callow mumbled. “The only thing it’s cessating is my will to live.” (Jamie
Doward. 2016. Hostage. London: Constable, p. 94)
Which of these instances of back-formation individuals find jocular may
depend on how recent the formation is in the individual’s experience. For the
author, surveille, from surveillance still sounds non-serious, though it has
become a standard form for many people. Just as euphemisms lose their
euphemistic nature with regular usage over time, so old jokes regularly lose
their jocularity with repetition. Damon Runyon’s more than somewhat and
Lewis Carroll’s curiouser and curiouser were both originally jokes. We can
thus possibly link the jocularity of recent back-formations to the fact that
listeners are aware that they are unfamiliar (see Chapter 4). The unfamiliarity
draws attention to the items, and one of the ways of reacting to the unfamiliarity
is to see it as being deliberately jocular. The more familiar the back-formation
becomes, the less it has the jocular effect and the more it becomes just a normal
expression.
The logic behind the term ‘back-formation’ is that a rule of word-formation
is, in some way, undone. Either affixation is cancelled or something that looks
as though it is an affix is deleted (even if there is no real affix present, as with
the verb to cessate which is not the basis of the noun cessation). There might,
however, be a better way of viewing the process. Consider the word surveil-
lance. Surveillance was borrowed from French. In French, it is derived from the
verb surveiller, but the verb did not accompany the noun surveillance into
English. English speakers thus had a noun with no corresponding verb. English
does, however, have many words ending in -ance which do have corresponding
verbs: acceptance, allowance, assistance, assurance, clearance, conveyance,
resistance and so on. On the basis of a paradigm of this nature, the speaker can
deduce that surveillance must be a comparable noun, and that therefore sur-
veille must be the base verb. The verb surveille is already there, waiting to be
recognized, it is not created when a perceived need arises. What back-
formation really is, under this view, is an acknowledgement of a form which
already exists in the paradigms of English word-formation. When speakers first
use it, it is recognized as unfamiliar, but its unfamiliarity does not affect its
grammaticality. This does not mean that speakers do not, on occasions, make
166 Part V Patterns of Word-Formation in English

historically false analogies. Given orient and orientation they can deduce that
the noun is orient, the verb is orientate and the nominalization of that verb is
orientation. Given deduction (which can correspond to the verb deduce or to
the verb deduct), speakers can make the assumption that the verb deduct is just
another way of saying deduce. They may not limit themselves to what linguists
would call patterns of word-formation, but may see pease as the plural of pea,
and thus create a new singular noun with a regular plural, because that form was
already there in the language as a mass noun. The label of ‘back-formation’
comes from linguists who are aware that the etymology is not what the forms
superficially suggest. From the point of view of the speaker, this is simply
recognition of the patterns of morphology in the language, and using this
knowledge where there is an apparent gap in their vocabularies. Perhaps back-
formation would be better termed ‘recognition of an unfamiliar base’.
One particular type of back-formation is the type illustrated by houseclean in
the examples above. The earliest form is housecleaner or housecleaning (in some
cases it will be clear which, in others it may not be), and since both -er and -ing
are regularly (though not exclusively) attached to verbs it is assumed that there is
a verb to houseclean. In principle, we might say that [house [clean-er]] has been
reanalysed as [[house-clean] er] (Adams 1973: 106). This is one of the major
ways in which compound verbs are created in English. Other examples are
breath-test, crash-land, gate-crash, head-hunt, sky-dive, or, arising from a past
participle rather than a present participle, tailor-make and perhaps colour-code.
Adams (2001: 106–8) provides many other examples. The theoretical question is
whether the new forms are really compound verbs. Historically speaking, there
is no compound verb in house-cleaner which can be left over when the -er is
removed. If the line suggested above is given credence, though, we can say that
whatever the etymological analysis of such verbs suggests, speakers treat them as
though they are compounds to which suffixes have been added. The analyst has
to decide whether to follow history or speakers’ intuitions. In this particular case,
such a decision has downstream implications, at least in defining a compound.

Challenge
One of the major problems that back-formation creates is the one related to
expressions like house-clean, discussed above. Should such forms be con-
sidered to be verbs, or should they be considered to be instances of back-
formation and not verbs? This is related to questions about conversion, where,
for instance, a cuddle (which comes historically from the verb to cuddle) can be
seen as a morphologically simple noun or a morphologically complex instance
of conversion. By extension, we can ask whether the verb to carbon-copy (from
the noun a carbon-copy) is or is not a compound. Consider the implications of
 21 Back-Formation 167

such questions, looking at what positions you are forced to adopt when you
accept either point of view. Can you resolve the issue?

References
Adams, Valerie. (1973). An Introduction to Modern English Word-Formation. London:
Longman.
(2001). Complex Words in English. Harlow: Longman.
Bauer, Laurie. (1983). English Word-Formation. Cambridge: Cambridge University
Press.
Marchand, Hans. (1969). The Categories and Types of Present-Day English
Word-Formation. 2nd ed. Munich: Beck.
22 Reflections on Coordinative Compounds

English coordinative compounds cause continuing problems of description,

including what subtypes should be recognized, whether all of the constructions
that have been considered under the heading are words, whether they are
compounds, and how their heads are to be determined (see Chapter 13).
A full discussion of these questions is provided in Bauer (2023), but in that
paper it is left to the individual reader to determine where the lines should be
drawn. Here I want to consider whether we have evidence which may lead to
a conclusion on the range of constructions that should properly be considered to
be coordinative compounds.
At first blush, it seems odd that there should be any problem here at all. If we
take a fundamental definition of a coordinative compound, then we can say that
it is a compound (and for most authorities that implies that it is a word) whose
elements are coordinated (linked semantically by a relationship meaning ‘and’,
rarely ‘or’ or ‘but’). Wälchli (2005) adds to this that coordinative compounds
are prototypically used to denote instances of what he terms ‘natural coordin-
ation’, that is collocations of words which recur because their denotata are
frequently found together in the world, or are seen as being of importance in the
relevant culture. All languages have instances of natural coordination, but not
all languages use coordinative compounds to mark the relationship. For
instance, in English, husband and wife might be considered an instance of
natural coordination, while man and landlady would not be, although both are
marked with overt syntactic coordination. We know from Wälchli (2005) that
coordinative compounds as a mode of expressing natural coordination tend to
become more common as one moves eastwards across Eurasia. This may be
enough to make it odd that there is such a plethora of possible types available in
English.
At the same time, if natural coordination is a prototypical reading of
a coordinative compound, this implies that there is room for a great deal of
variation, so that many constructions that we might wish to call ‘coordinative
compounds’ might show some looser form of coordination. Added to this, we
have a question of how we can be sure to recognize natural coordination and,
surprisingly, how to recognize coordination.

168
 22 Coordinative Compounds 169

Consider, for example, the expression soldier boy. How are we to interpret
this? If we think it means ‘the person is a soldier and a boy’ we have a gloss
which implies coordination. If we think that it means ‘a boy who is a soldier’
we have a subordinative relationship, with an essive semantic relation holding
between the elements. Then we might see soldier boy as similar to woman
doctor (see Chapter 13 and below). But soldier boy has the gender-marking
element second instead of first, so the equivalence is not total, and while
a woman doctor is typically a woman, a soldier boy is not typically a boy,
unless figuratively speaking (although boy soldiers are talked about, they are
denoted by an expression that puts the gender/age marking first). To further
complicate the issue, coordinate structures are usually semantically equivalent
when their elements are reversed (at least in syntax). Bread and jam means the
same, has the same denotatum, as jam and bread. If both soldier boy and boy
soldier are coordinative, we would expect them to mean the same, but they do
not. We might therefore conclude that at least one of them, perhaps both of
them, is not a coordinative compound. The example indicates that identifying
coordination is not necessarily obvious, and that therefore identifying potential
coordinative compounds is not straightforward.
In what follows, various different types of construction that might count as
coordinative compounds will be considered, and the question will be raised as
to whether they really are coordinative compounds and whether they are
compounds. The types discussed may not be exhaustive, but all are mentioned
in the literature as being relevant.
The first type is a locative type illustrated by Alsace-Lorraine, Budapest and
Nelson-Marlborough. Here we find the name of a locality being made up of the
names of two pre-existing localities which, between them, indicate the extent of
the new locality. Most such names in English are foreign names borrowed from
the appropriate area, but even then alternative formulations of names built from
similar elements are common: São Tome and Principe, Czechoslovakia, SeaTac.
Relatively recently, however, some names of this kind have been established
as anglophone names: Minneapolis-St Paul, Newcastle-Gateshead, Otago-
Southland. There is no obvious reason for this change in formation pattern.
They do look like genuine coordinative compounds: as names, they act as single
words, the entire word denotes a discrete entity, the words contain no grammat-
ical markers (such as and).
The next type is the type where the names of commercial entities are made up
of the names of two or more historically earlier entities: HarperCollins,
Mercedes-Benz, Rank-Hovis. Again the new entity is named by its origins,
but it may not be clear whether anything remains of the original component
parts. What makes these pairs seem slightly suspicious (and the same could be
said of the place names) is that the output of the process is a name (definite and
applied to one particular entity) rather than a word which can denote any one of
170 Part V Patterns of Word-Formation in English

a set of very similar entities; most nouns do not have unique reference, but may
refer to any member of a set. It is not clear that this is material, since it follows
from the usage of the word as a name. It is nevertheless interesting to compare
this usage with the compound names of gods in the Sanskrit tradition. The
Sanskrit word mitrāváruṇau ‘Mitra and Varuṇa’ denotes a team made up of the
two deities, not a new entity that arises from parts of both. Even though there is
no necessity that every coordinative compound should show precisely the same
type of interpretation, and the distinction could be entirely a matter of the
pragmatics of putting two names together, nevertheless this may indicate that
such commercial coordinative structures are not prototypical compounds of the
dvandva type.
We do find some examples where we apparently have two nouns coordin-
ated. Good examples are hard to come by in English, but at least the following
are suggested in the literature as relevant examples: dinner-dance, murder-
suicide, trailer-truck. There is also a much wider range of examples which are
less clear, including vowel segment, fighter-bomber, fridge-freezer, soldier
ant. It is difficult to discuss these, since each example is slightly different
from the others. For some writers, things can only be coordinative compounds
if the two (or more) element coordinated are of the same type. A soldier ant
must thus be ‘an ant that resembles a soldier’ (a subordinative type) and not
‘an insect which is both a soldier and an ant’. This might also suggest that
a trailer-truck must be ‘a truck which has a trailer’ rather than ‘a trailer and
a truck as a unit’. Even dinner-dance could be interpreted as ‘a dance at which
dinner is served’ rather than ‘an occasion on which dinner is served and
guests dance’. This may be interpreting the constraints on coordination too
severely, though. In the field I saw a soldier and an ant seems to be a perfectly
good instance of syntactic coordination. Because there is so much disagree-
ment on which of these words is relevant, if any, it is dangerous to assume that
any of these is a genuine coordinative compound, and it is certainly clear that
being able to give a construction a gloss using and is not a reliable test for
a coordinative compound.
One type that makes this particularly clear is the type illustrated by woman
doctor, boy scout and hen pheasant. In such examples, a gloss such as ‘the
person is a woman and a doctor’ is clearly possible, and some writers see this
as sufficient for these to be viewed as coordinative compounds. But, as has
been noted since the time of Panini, the real function here is ascriptive (or
attributive), and the first element is a gender marker (in the case of boy scout,
perhaps also an age-marker). That is the function is the same as in male tiger,
she-wolf, bull-calf, and possibly also as lioness. These are thus not coordina-
tive at all.
What Bauer (2008) calls the translative type and the co-participant type seem
easier to deal with.
 22 Coordinative Compounds 171

Translative: literal journey: Paris–Rome (flight), east–west (trajectory)

Translative: figurative journey: English–French (dictionary), subject–verb
(agreement)
Co-participant: non-directional: Arab–Israeli (talks), Manchester–Chelsea
(match), father–daughter (dance)
Co-participant: directional: CEO–employee (communication)
There are two reasons why these should not be seen as coordinative com-
pounds, and the two are linked. Most obviously, the two elements in the
translative series are not equally weighted: changing the order changes the
message. Then all of these are based on prepositional usages. Paris–Rome is
a snappier form of from Paris to Rome. Manchester–Chelsea means ‘between
Manchester and Chelsea’, but allows for attributive usage. At the very least, if
these are compound nouns, they are exocentric because the prepositional mean-
ing is not shown (as well as because the expressions are not hyponyms of either
element), but it seems more straightforward to see these as syntactic expressions.
Examples such as Sapir–Whorf (hypothesis), Creutzfeldt–Jakob (disease)
are harder to exclude. The ‘team’ meaning we saw with the Sanskrit example is
present, and if we assume that they act as adjectives, they must be grammat-
ically, as well as semantically, exocentric. This does not prevent a coordinative
interpretation.
Coordinative adjectival compounds, or at least some of them, seem rather
easier to classify and to see as truly coordinative. This may be because they
have a longer history in Germanic and in Indo-European more generally.
First, we have examples like bitter-sweet, manic-depressive, obsessive-
compulsive, shabby-genteel (note the ‘but’ relationship in the last example,
rather than the expected ‘and’ relationship). English does not have many of
these, preferring, for example, sweet and sour to sweet-sour. They are not
semantically homogeneous, sometimes indicating a hybrid, sometimes the
extremes of something, and they are hyponyms of both elements.
The same is true of a series of expressions like historical–social, political–
religious, literary–historical. This pattern is clearly productive, though the use
of learned adjectives is slightly unusual.
The possibility of coordinative verbs is far more contentious. Although
many examples have been suggested, such as blow-dry, cough-laugh, drink-
drive, fly-drive, freeze-dry, sleepwalk and stir-fry, they are all problematic in
one way or another. Blow-dry could be syntactic because it can be interrupted
(blow my hair dry), drink-drive is nearly always used in the -ing form, freeze-
dry and stir-fry can be seen as being types of drying/frying rather than the
addition of two actions, and the sleep in sleepwalk might be seen as a noun.
Cough-laugh is the best hope for a genuine coordinative compound (and cry-
laugh is also attested), but even then, for medical people, there is a cough/laugh
172 Part V Patterns of Word-Formation in English

syndrome which is triggered by either a cough or a laugh, so that the meaning

here is not entirely prototypical.
Other possible categories include compromises like north-west and blue-
green (though English does not have many of either type, and prefers bluey-
green or bluish green for the colour terms). Some languages have words like
red-white meaning both red and white (e.g. in stripes), but English usually
avoids these. Then there are mathematical coordinatives like ninety-two (in
older English two-and-ninety was used), which might be syntactic because in
two hundred and ninety-two there is overt syntactic coordination.
The most difficult category to deal with is the type illustrated by actor-
director, singer-songwriter, teacher-researcher which denote a single entity
viewed from two (or more) points of view (see also Chapter 13). The traditional
Sanskrit dvandvas show two distinct entities which share some kind of func-
tion, so these are a distinct class from the semantic point of view. The overall
expression is a hyponym of both elements, which matches the potentially
coordinative adjectives, but it is not clear whether that should be a criterion,
and if so, in which direction.
It seems that there are many types here which we can exclude, many
types which are controversial member of the set, and hardly any that are
clear-cut coordinative compounds. But that does not mean that the cat-
egory cannot be applied to English. Just what counts, though, is often
a matter of how the category of coordinative compound is defined for
English.
One solution to the problems that have been raised here is simply to deny
their existence. Anderson (1992: 316) says of coordinative compounds that
they are “not really found in English”, without explaining what he would do
with the examples that have been cited here. Adams (2001: 3) is more
explicit: anything that includes coordination breaks the right-hand-head rule
(see Chapter 13) and therefore cannot be a compound, and must be a phrase.
This is neat, though it fails to explore the possibility that it is the right-hand-
head rule that is wrong, or at least incomplete. However, given the difficul-
ties that face us if we try to determine which of these apparently coordinate
structures might be compounds, perhaps this solution has much to recom-
mend it – at least as long as we are dealing with English and not Vietnamese
or Chinese.

Challenge
Consider the type illustrated by singer-songwriter or director-producer. Can
you find any arguments either for or against the notion that these are com-
pounds, that they are coordinative or that they are headed structures? What are
the implications of your consideration?
 22 Coordinative Compounds 173

References
Adams, Valerie. (2001). Complex Words in English. Harlow: Longman.
Anderson, Stephen R. (1992). A-morphous Morphology. Cambridge: Cambridge
University Press.
Bauer, Laurie. (2008). Dvandva. Word Structure 1, 1–20.
(2023). Coordinative compounds, including dvandva. In Peter Ackema, Sabrina
Bendjaballah, Eulàlia Bonet and Antonio Fábregas (eds.), The Wiley Blackwell
Companion to Morphology. Hoboken, NJ: Wiley.
Wälchli, Bernhard. (2005). Co-compounds and Natural Coordination. Oxford: Oxford
University Press.
23 Reflections on the Irregularity of Prepositions

Although nouns, verbs and adjectives have regular functions in English

word-formation, prepositions seem much more marginal in such uses, per-
haps because of their predominantly grammatical function. Nevertheless,
they do occur, at least sporadically, in a number of apparent cases of word-
formation. In syntax, prepositions in English have several different functions.
The prototypical use of prepositions is with a noun phrase, as in on the street,
in our house, down the chimney. We also find the same forms used as adverbs
in expressions such as She was looking on, They all traipsed in and We jumped
down. Some authorities (Emonds 1972) talk about transitive and intransitive
prepositions here. In verbs such as She put on an angry expression, The
soldiers fell in, I’ll look up the answer it may not be clear whether we should
call the same forms ‘prepositions’, ‘adverbs’ or something else, and so some
authorities prefer the term ‘particle’, perhaps to avoid the issue, although
there is dispute in the literature as to whether prepositions in their core use of
governing a noun phrase are a type of particle, and for some authorities
‘particles’ include the infinitival to (as in to listen), possibly words like not
(also termed an adverb) and pragmatic particles like oh, or like (She’s like
really bright). I shall provisionally use the term ‘preposition’ here, but it
should be clear that terminology in this area is not completely standardized.
For a more sophisticated argument pointing out the difficulties of classifica-
tion here, see Lee (1999).
For Chomsky (1970), the preposition is one of the four major word-classes,
but that does not mean that it is like the others. Adjectives, nouns and verbs
have lexical meaning, and typically denote states, entities and actions, while
prepositions frequently have grammatical meaning, not necessarily predict-
able from any inherent meaning. For example, We’re going down the pub
(contrast We’re going down the river) does not really make sense except
idiomatically.
The class of prepositions is not a closed class. There is a set of prototypical
prepositions, but there are also sequences of prepositions and expressions that
function as prepositions though they are derived from words which belong to
a different word-class. Examples of each type are given below.

174
 23 Irregularity of Prepositions 175

prototypical prepositions: above, across, after, at, before, by, down,

for, from, in, near, of, on, out, past, through, to, under, up, via, with
prepositional sequences: down from, into, onto, off of, out of, through-
out, up to, within
derived prepositions: concerning, excepting, excluding, following,
given, including
Given this situation, we might expect to see word-formation processes which
allow us to create prepositions, but we do not. As is shown in the few examples
above, participles are used as prepositions as required. One possible analysis of
this is that participles are changed into prepositions by conversion, but this
implies that the participles, usually thought of as being inflected forms, can
subsequently undergo conversion, usually considered a derivational process.
The analysis is therefore not universally accepted.
Yet prepositions do seem to undergo conversion on some occasions.
Examples are rare, however, and often very informal.
prepositions as verbs: to down (a drink, an aeroplane), to off (‘murder’),
to out (‘make something about a person – usu. their sexuality – public’),
to up (the price; to up and leave ‘suddenly depart’)
prepositions as nouns: a by (in cricket), a down (‘bad opinion’), an
in (‘access’, the ins and outs ‘the details’), the off (ready for the
off ‘ready to start’), an over (in cricket), up (ups and downs ‘vicissi-
tudes’); on the up and up (‘being increasingly successful’)
prepositions as adjectives: down (‘depressed’)
prepositions in attributive position: the down train (‘away from
London’), the up train (‘towards London’), no through road
There are also instances where the preposition might be analysed as coming
from a homophone in another word-class: above (< adverb), before (< conjunc-
tion), near (< adjective), past (< noun), round (< adjective).
Even less frequently, we find prepositions being used as bases in overt
derivational morphology.
creating adjectives: backward, downward, inward, offish (‘aloof’),
outward, upward
creating adverbs: afterward(s), overly (‘excessively’)
creating nouns: afters (‘course after the main course’), downer (‘bad
experience’, ‘drug that makes you relax’), innie (‘type of navel’),
insider, outie (‘type of navel’), outing (‘excursion’), outsider, overage
(‘surplus’), rounders, underling, upper (‘drug that gives you energy’)
176 Part V Patterns of Word-Formation in English

Forms like upper (of a shoe, for instance) might be viewed as a comparative
form or as an -er derivative (historically, it is a comparative, but it appears to
have undergone conversion to a noun, which puts derivation further from the
root than inflection), in uppity the <it> is not easily explicable.
What is striking about these examples is how few of the prepositions seem
to be used in these ways. It is as if there is a core set of prepositions (not
necessarily the most basic prepositions in grammatical usage) which can be
used in these ways. It is also noticeable that many of these are not only informal
but idiomatic.
This raises the question of whether there is any productive use of preposi-
tions as bases in word-formation. Perhaps, rather than saying that there is
conversion here, we should just say that word-classes are very fluid with this
set of words, and should view the set of overt derivations from prepositions as
being sporadic creations rather than the result of a strong pattern or rule of
word-formation. This might help explain the low profitability of these patterns.
However, if that is the case, we need to explain why prepositions occur so
much more frequently in compounds (and, not necessarily the same thing,
when used as prefixes). Consider the indicative examples given below.
as modifiers to adjectives (usually participial adjectives): down-
graded, inborn, near-sighted, outsourced, overearnest
as modifiers to nouns: afterlife, by-product, by-road, down-light, off-
side, overcoat, undershirt, upbeat, up-side
as modifiers to verbs (some of these might be considered prefixes):
downgrade, underplay, outsource, overlook, upchuck (‘vomit’),
upload
creating exocentric adjectives: after-dinner, in-depth, near-death,
off-centre
creating exocentric adverbs: off-key, overnight
creating exocentric nouns: afternoon, at-home, off-spring
creating exocentric verbs: overnight, withstand
If we put these things together, we get the distribution set out in Table 23.1.
Looking at the material in this light (even though we could add more forms)
gives a different perspective. It appears that some forms are pure prepositions,
with no other use, while those words which have adverbial function also have
the possibility of appearing in other functions. In other words, the core
function here is adverbial, with some of the prepositions sharing form with
adverbs. The use of with in examples like withhold is unexpected, but it is
becoming more widely used as an adverb in Are you coming with? (originally
from German). The forms used as bases for further overt word-formation are
 23 Irregularity of Prepositions 177

Table 23.1 Uses of forms that appear as prepositions

Modifier in Base
Form Adverb In PhrVb Conv to N or V Compound Modifier in NP for w-f
above ✓ above address
across ✓ come
across
after ✓ look after afters afterlife afters
at
before ✓ beforehand
beside
by ✓ stand by by-line
down ✓ put down to down downhill down payment downer
from
in ✓ drop in an in inlay in thing innie
near ✓ to near near-side near miss
of
off ✓ swear off to off, the off off-side off day offish
on ✓ get on on-side
out ✓ put out to out, an out outstay outie
outside ✓ the outside outside outsider
broadcast
over ✓ hold over an over overpay overly
past ✓ the past past time
round ✓ to round, a round round table
through ✓ pull through train
through
to ✓ come to
under ✓ go under underground under surface underling
up ✓ look up to up, ups and upkeep upper
downs
with withhold

more limited, apparently to spatial adverbs, though not all of them. Note that
using the term ‘intransitive preposition’ in place of ‘adverb’ does not help
here, since we still have to be able to refer to a class of prepositions which
have both transitive and intransitive usage.
Because prepositions have a function as much as a meaning, it is adverbs
which are used in word-formation, even when the adverbs share a form
with prepositions. This goes some way to explaining some of the apparent
irregularities in the use of things that look like prepositions in English word-
formation. Not all gaps can be expected to be filled in a derivational process,
and the use of these adverbs is still rather limited, more in some constructions
than in others.
178 Part V Patterns of Word-Formation in English

Challenge
The adverbs used as bases for further derivational morphology seem to be
particularly restricted. Look for further examples, and determine whether this
particular set can be defined in semantic or functional terms and whether the
affixes used with them are restricted in any way.

References
Chomsky, Noam. (1970). Remarks on nominalization. In Roderick A. Jacobs &
Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar.
Waltham, MA: Ginn, 184–221.
Emonds Joseph. (1972). Evidence that indirect object movement is a structure-preserving
rule. Foundations of Language 8, 546–61.
Lee, David. (1999). Intransitive prepositions: Are they viable? In Peter Collins & David
Lee (eds.), Clause in English: In Honour of Rodney Huddleston. Amsterdam:
Benjamins, 133–47.
24 Reflections on Reduplication

Reduplication, the meaningful repetition of all or part of a word, is sometimes

considered to be one of the few candidates for a universal process of word-
formation. As we might expect, therefore, English does show some patterns of
reduplication, but relatively few in comparison with some other languages. We
can distinguish between various types of reduplication: the type where a whole
word is repeated, and the types where a part of a word is repeated. When part
of a word is repeated we find cases where consonants change and cases
where vowels change. There are formal or semantic subtypes of each of these
types.

24.1 Whole-Word Reduplication

Whole-word reduplication is rare in English, and restricted to a few semantic or
stylistic types, mostly childish words, sometimes extending to slang words.
Some examples are given below.

boo-boo ‘mistake’
choo-choo ‘train’
chin-chin a toast, ‘cheers’ (originally from Chinese)
chuff-chuff ‘train’
do-do ‘faeces’
housey-housey ‘bingo’; also without reduplication
mama ‘mother’
no-no ‘something forbidden’
papa ‘father’
pee-pee ‘urine, an act of urination’
poo-poo ‘dismiss (e.g. an idea)’
putt-putt ‘a form of mini-golf in which only putts are used’
quack-quack ‘the call of a duck; a duck’
so-so ‘indifferent, neither good nor bad’
wee-wee ‘urine, act of urination’
woo-woo ‘involving the supernatural’
woof-woof ‘dog’
yo-yo ‘child’s toy’

179
180 Part V Patterns of Word-Formation in English

This list may include more than one type, since quack-quack, for example,
has stress on the first element if it means ‘duck’ but stress on the second element
if it means ‘the noise a duck makes’. Mama and papa have second-element
stress, at least in isolation and naming parents, unlike, for instance, do-do.
Some loan words like dik-dik, gogo, juju, muumuu, pompom, tomtom may not
count as instances of reduplication in English, although their adoption may be
due to the attraction of the reduplicative structure.
Repeated element compounds may also be included here, though it is not
clear that the same morphological process is involved, since phrases as well as
word-like elements may be repeated (see Chapter 11). These examples are
rarely established. Some examples are given below.
Think me up a way to earn some blunt, would you? Not earn-earn it, but come
into it proper-like. (Grace Burrowes. (2017). Too Scot to Handle. New York:
Forever, pp. 236–7)
But I didn’t know her-know her. (Kirsten Lepionka. (2017). The Last Place
You Look. New York: Minotaur, p. 86)
It is often hard to draw the line between reduplication and syntactic repeti-
tion. Some examples of structures which are at least arguably syntactic (and
thus not word-formation) are given below, but the classification is not always
clear-cut.

aye-aye positive response to an order

blah blah indicating meaningless talk
bye-bye ‘goodbye’
fifty–fifty ‘equally likely’
ha ha indicating laughter, or mock laughter
there there used to comfort
yes yes indicating emphatic agreement

Clearly syntactic repetition includes things like It’s very very interesting, It’s
a long long way, That’s really really stupid.

24.2 Repetition of Part of a Word: Consonants Change

Typically, what happens in words like this is that the initial consonant or
consonant cluster of a word changes, so that the two parts of the new word
rhyme. These constructions are therefore often called ‘rhyme-motivated’.
(For rhyme more generally, see Chapter 17.) In some instances, there is an
ordinary word to which a rhyming element is added, but in many cases
neither element is an independent word of English. Some examples are given
below.
 24 Reduplication 181

argy-bargy ‘an argument’

arty-farty ‘exaggeratedly artistic’
bow-wow ‘dog, sound of a dog’
boogie-woogie ‘style of jazz’
fuddy-duddy ‘very conservative’
hanky-panky ‘improper behaviour’
helter-skelter ‘fairground ride’
hocus-pocus magic word
hoity-toity ‘very haughty’
hubble-bubble ‘hookah’
itsy-bitsy (or itty-bitty) ‘very small
lovey-dovey ‘demonstratively affectionate’
mumbo-jumbo ‘gobbledygook’
namby-pamby ‘lacking vigour’
piggy-wig(gy) ‘nice pig’
roly-poly ‘rolled-up, round’
super-duper ‘excellent’
teeny-weeny ‘very small’
wagger-pagger-bagger ‘waste paper basket’ (old Oxford University slang)
walkie-talkie ‘hand-held radio’
willy-nilly ‘whether you want to or not’

It is not entirely clear which element is the central element here and which
element is the reduplicated element, and neither is it entirely clear what
consonant is likely to occur in the reduplicated element. Labial consonants
(including /w/) seem to be slightly overrepresented in the construction as
a whole, but other consonants are certainly not excluded. Similarly, there
seems to be some preference for plosives. The use of disyllabic elements is at
least strongly preferred, if not required (consider wham-bam). The construction
is not very productive.
Whether expressions like (the screaming) hab-dabs, the heebie-jeebies fit
into this pattern or not may be controversial, since the plural marking seems to
be an important factor in these words.
One very specific pattern, used almost exclusively in American English,
where it derives from Yiddish, is the use of an introductory /ʃm/ cluster (note
that this cluster is not generally allowed in English). The word with this cluster
is usually written as a separate word from its base (often separated by
a comma), which differentiates this structure from the others illustrated
above which seem to be added to a base to create single words. Some examples
are given below.
Well, daughter/shmaughter, know what I mean? (Robert Wilson. 2015.
Stealing People. London: Orion, p. 174)
182 Part V Patterns of Word-Formation in English

“So you think she’s dead?”

“Think, schmink. Her blood was in the kitchen.” (John Lescroart. 2014.
The Keeper. New York: Atria, p. 20)
Coincidences, smincidences. (Kate Medina. 2016. Fire Damage. London:
Harper, p. 292) [note the unusual form, from a British publisher]

24.3 Repetition of Part of a Word: Vowels Change

Where the vowel in a stressed syllable changes, the constructions are said to be
‘ablaut motivated’. Unlike the rhyme-motivated examples, these do not strongly
prefer disyllabic elements, though they certainly occur. There are two major
formal patterns, one where /ɪ/ alternates with /æ/ and one where /ɪ/ alternates
with /ɒ/ (with appropriate replacements in American English). In some instances,
all three can be found in the same expression. Some examples illustrating all three
patterns are given below.

/ɪ/ ~ /æ/
chit-chat ‘idle talk’
dilly-dally ‘delay’
flim-flam ‘trick’
kit-cat ‘originally the name of a London club’
knick-knack ‘small ornament’
mish-mash ‘medley’
pitter-patter ‘a faint sound, esp. of footsteps’
riff-raff ‘undesirable people’
shilly-shally ‘act indecisively’
tittle-tattle ‘gossip’
zig-zag ‘turn alternately left and right’

/ɪ/ ~ /ɒ/
clip-clop ‘sound of a horse walking’
ding-dong ‘fiercely contested, sound of bells’
flip-flop ‘change direction, rubber sandal’
ping-pong ‘table tennis’
sing-song ‘an event where everyone joins in singing,
lively pitch movement in speaking’
tick-tock ‘sound of a time-piece’
tip-top ‘very best quality’
wishy-washy ‘weak, of poor quality’

/ɪ/ ~ /æ/ ~ /ɒ/

slip slop slap slogan for avoiding skin-cancer
 24 Reduplication 183

Although we can find occasional instances with different vowels (e.g. bim
bam boom), the vowels in this construction are usually very restricted. Note,
however, a few expressions such as

gewgaw ‘bauble’
hee-haw ‘sound made by a donkey; to make such a sound’
see-saw ‘playground equipment’
fee-fi-fo-fum expression attributed to fairy tale giants
ho-hum ‘boring, ordinary; said to express boredom’
(oh) me, oh my expression of amazement

24.4 A More Complex Pattern

There is a similar, if more complex, construction in English which has appar-
ently remained unnoticed. It is unusual partly because it does not create nouns,
adjectives or verbs, it is partly unnoticed because it is fairly rare, and it is not
clear how regular it is.
Examples of the relevant construction(s) are provided below.

bippety Disney’s Cinderella

boppety boo
blankety blank a replacement for an expletive
bumpety bump (BNC)
bump
clickety clack repeated clicking sound (e.g. of typing on a typewriter)
clippety clop ‘sound made by horses’
dinkety bonk a type of music (BNC)
flippety flop a vocal representation of a tumbling movement
hippety hop a representation of a repeated hopping movement
hoppity hop as hippety hop
hoppity hoppity hoppity hop extended version of hoppity hop
lickety split ‘very quickly’
skippety-hop a representation of the rhythm of the heart (BNC)
snippety snap used to encourage speed (Urban Dictionary);
a representation of the sound made by scissors
tickety boo ‘in the best possible condition’

What these share formally is a first element with dactylic form ending in /əti/
(which in some varieties becomes [əɾi]) and a monosyllabic second element,
with occasional reduplication of one element or the other. Sometimes the
stressed vowels match, sometimes they fit with the usual ablaut patterns
discussed above. Some of these forms have alternatives without the /əti/,
such as clip clop, flip flop, but while clip clop remains a representation of
a sound, flip flop does not necessarily remain a representation of movement, but
184 Part V Patterns of Word-Formation in English

may become a noun. The general construction type appears to be productive,

but it is hard to find written examples.

24.5 Discussion
These constructions remain rather marginal in English, despite the familiarity
of reduplication in many languages. The patterns of reduplication are also quite
restricted, with initial consonants or consonant cluster and stressed vowels
involved but not, for example, reduplication of initial or final syllables or the
use of random vowels or consonants, which are common in many other
languages. These constructions can be considered from the point of view of
what remains the same, or from the point of view of what changes. What is
particularly surprising in the English examples is how often the base is not an
independent form. Finally, note that the same patterns of ablaut, at least, are
found in fixed syntactic phrases such as in dribs and drabs ‘a little at a time’.

Challenge
Choose any one of the patterns of reduplication discussed here, and consider
whether it is better viewed as a matter of what is retained or better viewed as
a matter of what is changed. What leads you to suppose you can tell? Have any
patterns of English reduplication been omitted in this presentation?
Part VI

Historical Questions
25 Reflections on Dead Morphology

25.1 Introduction
There are certain patterns of word-formation which, although they once
allowed the creation of new forms, now no longer seem to have that
ability, despite the fact that some of the words created by the pattern
still exist. Typically, speakers are not aware of these patterns, but occa-
sionally there is some residual awareness, for at least some of the words
involved. In what follows, these patterns are divided according to the
word-class they once created, although alternative classifications would
obviously be possible.

25.2 Creating Nouns

There are a few examples of ablaut-linking verbs with a corresponding
nominal (Wescott 1970). Examples are provided below. Some of these are
very old, and may never have been felt to be morphologically linked in
English.

Verb Corresponding noun

abide abode
bind band
bleed blood
break breech
breed brood
drink drench
feed food
ride road
shoot shot
sing song
sit seat
write writ

There are other examples where ablaut is a concomitant to affixation, which

are not considered here.

187
188 Part VI Historical Questions

There is a Germanic diminutive suffix -ock which can still be found in

a number of English words, though its diminutive meaning has been lost in
most of them. Some relevant words are listed below.
bollock, bullock, buttock, haddock, hillock, paddock, pillock

It is not always clear what words contain this suffix from an etymological
point of view, as there are similar-sounding suffixes from other sources, and
because the relevant base is not always clear. Many of the words in the OED
that are said to contain this suffix are now obsolete or extremely rare.
There is a Germanic diminutive suffix which comes down to English in the
form -en. There are few remaining words with this suffix, and even then, it is not
always clear whether this suffix is involved etymologically, or whether there is
simply a coincidence of form. Potentially relevant words are given below.
chicken, kitten, maiden

Another diminutive suffix, this one from French, has the form -erel or -rel,
and is rarely recognizable as a suffix at all. Surviving words in which this
etymological item may appear (although there is frequently some doubt)
include those below. The newest derivatives with this suffix are from the
seventeenth century.
cockerel, dotterel, gomerel (Scottish), hoggerel, kestrel, mackerel, mongrel,
pickerel, scoundrel, wastrel
There is a Germanic suffix -t which creates nouns. It is generally accepted as
being a variant of the suffix -th, either positionally motivated (especially after
<gh>) or regionally motivated (-t is a northern form). Alternation between -t and
-th in the word height/highth may still be found regionally, and is used by Milton.
over head up grew
Insuperable highth of loftiest shade,
Cedar, and Pine, and Firr (John Milton. 1667. Paradise Lost, Book 4)
The Germanic nominalizing -t has various etyma, which have merged by the
time we get to English. Examples of -t on adjectives and verbs are given below.
Theft is an unusual example on a nominal base.
suffix -t on adjectival bases: drought, height
suffix -t on verbal bases: draught, drift, flight, gift, might, plight,
shrift, sight, thirst, thought, thrift, weight
The suffix -th (as in truth and warmth) is usually considered to be dead today.
It is certainly difficult to find new forms, and where they are found, they are
often made up by analogy with existing forms or are resurrections of older
forms. Some speakers feel that coolth is innovative, but has been in continual
 25 Dead Morphology 189

use, usually jocularly, since the sixteenth century. There are many words which
are etymologically the result of -th affixation which are not recognize as such
today, or are only recognized by language professionals. Relevant words
include birth, death, dearth, filth, health, mirth, stealth, youth. Part of the
reason that such words are so obscure is that the vowel sounds have changed
so much, and no longer reflect the original base.

25.3 Creating Adjectives

There is a very small set of adjectives related to a verbal base by ablaut.
Verb Adjective
heal hale, whole
fill full

There is an etymological suffix, today spelled -le, which has left very few traces
in English and which can no longer been seen as a suffix because its bases are no
longer recognizable. Examples include brittle, fickle, idle, little and nimble.
The suffix -ac as in cardiac and demoniac is based on classical models, and
shows some productivity in English, with words such as hypochondriac and
maniac. The more recent brainiac is probably coined directly from maniac,
and does not illustrate genuine suffixation at all.

25.4 Creating Verbs

There are patterns of ablaut that link transitive (or causative) and intransitive
verbs, but these patterns are not predictable in the modern language and are
very restricted in the number of items covered. There are other examples where
ablaut is a concomitant to affixation, which are not considered here.

Intransitive Transitive
fall fell
lie lay
rise raise
sit set

The suffix -le on verbal bases creates verbs which denote repetitive move-
ments of small scale. Sometimes, it is sounds which are repetitive. There are
many words in this class still in use, but many of them no longer have (in some
cases, may never have had) an analysable base, which supports the notion that
verbs that end this way may be viewed as monomorphemic. In some cases, the
190 Part VI Historical Questions

putative base has a long vowel, and the derivative a short vowel, which also
masks the historical relationship. Selected examples, divided between verbs of
movement and verbs of sound, are given below.
verbs of movement: drizzle, scuttle, shuffle, sidle, sparkle, suckle,
twinkle, waddle, waggle, wriggle
verbs of sound: babble, cackle, crackle, giggle, gobble, mumble,
prattle, rattle, rustle, sizzle, sniffle, tinkle, warble
There are some words which end in -le which probably do not contain the
same suffix. Some examples are given below.
bamboozle, puzzle, wheedle
A similar meaning is provided by -er in some verbs, particularly in words
denoting vocalization of some type or the play of light. Examples are given
below. In most of these, no base is currently recognizable, which explains the
loss of productivity. In many cases, there has never been an independent base in
English (although there may have been in older Germanic), and the -er is of
questionable morphological value although it seems to have had the ability to
carry meaning, perhaps as a phonaestheme.
vocalization: chatter, gibber, jabber, mutter, snicker, snigger, splutter,
stammer, titter
play of light: flicker, glimmer, glitter, shimmer
other: clamber, flutter, quiver, shudder, waver
There are several verbs of English which end in -ish, and which are derived
from French -iss (in turn derived from Latin -isc). It is not clear that this was
ever a suffix in English, but it is derived from a suffix. There are very few words
where the removal of -ish leaves a recognizable base, and it is not always clear
that any such base is relevant. Some examples are given below, some attached
to what may once have been a base, some with no apparent base.
possible base: banish, brandish, burnish, flourish
no recognizable base: abolish, accomplish, demolish, establish, finish,
furnish, languish, perish, punish, ravish
The prefix be- can be found in several sets of words, and it is marginally
productive only when it co-occurs with a final -ed (as in be-trousered), and
otherwise not productive. In some of its uses it is one of those prefixes which
appears to be the head of its construction in the sense that it creates verbs from
other bases. Examples of various constructions with this prefix are given below.
Many of the words with prefix be- are now rare or archaic (e.g. bedrench, bego,
bespeak, bethink) and some persist only – or mainly – as participles (e.g.
belated, beloved).
 25 Dead Morphology 191

semantically opaque formations: become, befall, beget, behave,

behold, belay, beseech, beset
adjectival bases: becalm, bedim, belate, belittle, benumb
connotation of scattered result or thoroughness of result: bedaub,
bedraggle, besmear, bespatter, bespeckle, besprinkle
nominal bases giving rise to verbs: bedevil, befog, befoul, bejewel,
benight, besiege, bewitch
with the meaning ‘remove’: behead
creating transitive verbs: bedazzle, belie, bemoan, berate, bestir, bewail
The opaque words listed above are all formed on the basis of Germanic verbs,
probably before the English period. Strictly, begin belongs here, but gin is no longer
found in isolation in English. Beseech shows English palatalization of the final
consonant in seek. Behave has become regular, suggesting the loss of connection
with have; the others maintain their irregular past tense and past participle.

25.5 Generalizations
Some generalizations about why morphological patterns die out can be seen in
the examples provided here. With patterns of apophony, there was no predict-
ability in the alternating vowels by the time English emerged from Germanic so
that new forms could not be created with confidence. The result was that,
although ablaut remains a pattern which can be seen in conjugation of verbs, it
is seen as mainly irregular and the overall idea of apophony as a way of forming
words in English disappears. Many patterns of apophony are not even recog-
nized as creating morphophonemic variation, only historical links.
If bases become unrecognizable, then affixation dies because the word
cannot be analysed. We see this with the various -le suffixes, and with many
examples of -th.
If the meaning of an affix becomes unclear, then the affix is likely to stop
being used. The prefix be- provides a good example of this.
One very important point is that until the use of a particular morphological
process exists only in non-transparent words, there is always the possibility that
the process will return. Although it seems unlikely that -th will be resuscitated,
while there are a few transparent instances such as truth and warmth, it remains
a possibility, even though it could not then be used with to reconstruct estab-
lished words like depth and width.

Challenge
Many English derivational suffixes disappeared at the end of the Old English
period. For those who have studied Old English, consider the suffixes which
192 Part VI Historical Questions

vanished, and attempt to determine why they ceased to be productive. For those
with little familiarity with Old English, look at those suffixes which persisted
into modern English (use an etymological dictionary to discover relevant
forms), and suggest reasons why those suffixes should have persisted. Are
there general principles which mean that affixes are lost or persist?

References
OED. The Oxford English Dictionary [online]. oed.com
Wescott, Roger W. (1970). Types of vowel alternation in English. Word 26, 309–43.
26 Reflections on Compounds in English
and in Wider Germanic

English is a Germanic language, but one which for various historical reasons
has been strongly influenced by Latin and French. As a gross oversimplifica-
tion, we might say that the basic vocabulary of English is Germanic, but that the
refined vocabulary of English comes from French and Latin (with some of
the technical vocabulary coming from Greek). The general patterning of how
the language works, though, is Germanic. One part of that is that things that
are called ‘compounds’ in English tend to function like the compounds in
Germanic languages rather than like the compounds in Romance: they are
right-headed (a postage stamp is a kind of stamp, just as in German a
Brief·marke ‘letter stamp’ is a kind of Marke, while in French a timbre-poste
‘stamp post’ is a kind of timbre, following the left-hand element), the typical
Romance structure of verb + noun as in French garde-robe ‘keep dress’ (i.e.
‘wardrobe’), is not widely used in English and not used in German. In this
chapter, the structure of compounds in the Germanic languages will be con-
sidered in rather more detail, and compared with the structure of compounds in
English. It will be shown that the match is not as close as these brief introduc-
tory remarks might imply. For clarity of presentation in non-English examples,
the decimal point is used to show boundaries between elements where this is
not clear from the orthography, as was done in the Briefmarke example just
above.
We can begin by considering compounds made up of an adjective and
a noun, and the difference between a compound and a syntactic phrasal
structure. Throughout Germanic, nouns have inherent gender (masculine,
feminine or neuter in German or Icelandic, common or neuter in Dutch and
Danish) and in indefinite phrases an attributive adjective agrees with its head
noun, and we find so-called strong forms of adjectives.

‘A big house’ ‘A big man’ (nominative case, where this is significant)

Danish German
Et stor·t hus En stor mand Ein groß·es Haus Ein groß·er Mann

When there is a definite determiner, we find the so-called weak form of the
adjective.

193
194 Part VI Historical Questions

‘The big house’ ‘The big man’

Danish German
Det stor·e hus Den stor·e mand Das groß·e Haus Der groß·e Mann

Due to various linguistic changes, the patterns do not work exactly the same
way in all the Germanic languages, and there are exceptions to the general
pattern, but this is a reasonable overall pattern for noun phrases. However,
when the sequence of adjective and noun is a compound, the adjective appears
in the stem form.

‘A big city’
Danish German
En stor·by Eine Groß·stadt

Furthermore, any town which is big could be en stor by or eine große Stadt,
but en storby or eine Großstadt puts it in a different class, it is, in English terms,
a city. The compound also gets stress on the first element of the compound
(Dutch has some exceptions: Don 2009: 375), while the syntactic phrase gets
the stress on the noun. Compounds are written as single word, phrases as
multiple words. In Dutch (Don 2009: 374) and German (Neef 2009: 388) the
adjectives that can occur in these constructions are limited, largely to monosyl-
labic Germanic adjectives, although the constraint is hard to specify accurately.
In Danish forms like polar·eskimo ‘polar Eskimo’ and privat·kunde ‘private
customer’ are found, though they are rare. In Danish, but not in German, we
also have another sign that the compound is a single word. We say den store by
‘the big town’ but stor·by·en ‘big·town·the’, where the postposed definite
article is used only where the noun phrase contains just one word.
Some adjective + noun compounds are exocentric, that is they do not denote
the entity normally denoted by the noun. In English, for example, a redcap is
not a kind of cap, but is a kind of person/goblin/bird (depending on the dialect
of the speaker or familiarity with the entities thus named). Examples from other
languages are given below.

Language Compound Gloss of Translation Source

elements
Danish gul-bug yellow-belly ‘warbler (bird sp.)’ (Bauer 2009)
Dutch rood-borst red-breast ‘robin’ (Don 2009)
Frisian grou·kont fat·arse ‘person with a fat arse’ (Hoekstra 2016)
German Rot-kehlchen red-throat·dimin ‘robin’
Norwegian raud·spette red·spot ‘flounder (fish sp.)’

Now compare this with what happens in modern English. We do not have
strong and weak forms of adjectives in English, only a base form, so that the
 26 Compounds in English and in Wider Germanic 195

syntactic distinctions cannot be shown in English. The only evidence we might

have that English does not allow inflected adjectives in compounds, is that we
cannot have inflected comparatives and superlatives in this position: a blacker
bird is fine in syntax, but *a blackerbird is not a possible compound.
Interestingly, both Danish and German have compounds with superlatives in
them: Danish største·belastning ‘maximum (lit. biggest) load’, German
Höchst·geschwindigkeit ‘top (lit. highest) speed’.
There is, in English, a mismatch between the classificatory semantics and the
stress (and orthography). While blackbird has stress on the adjective and
classificatory semantics, blue whale has stress on the noun (as indicated by
the orthography, though the orthography is not entirely reliable) and classifica-
tory semantics. English also has a set of words with non-Germanic, non-simple
adjectives with initial stress such as dental hospital, dramatic society, primary
school, solar system (Bauer 2020).
As in the other languages, exocentrics are found in English.
The situation in English does not match the situation in the rest of
Germanic because the patterns are more complex and less predictable in
English. The productivity of the pattern is also not the same in English and
in some of the other languages. Nevertheless, the central examples look very
similar indeed.
We now move to look at noun + noun compounds. Noun + noun
compounds are the default compounds in Germanic, and have been there
since the earliest days of Germanic. Historians of the language generally
divide these compounds into ‘proper’ and ‘improper’ compounds (in
German echt ‘genuine’ and unecht or eigentlich ‘real’ and uneigentlich,
with a range of translations in other languages, including English). The
proper compounds (which appear to be considered genuine because they
were the earliest type) may originally have had a noun-class marker
following the stem (Harbert 2007: 30), but this was lost early, leaving
stem + word compounds. The improper compounds carried an inflectional
marker on the first element, typically a genitive marker. In later Germanic,
this genitive marking becomes very much more common, but also less
obviously genitive (Harbert 2007: 30). In some instances, the marker
looks, to modern eyes, much more like a plural marker (perhaps especially
in Dutch – Booij 2002: 179), or like an unmotivated form. Some of the
apparent genitive markers are also used in places where they could not
now occur as genitives. For example, German Liebe·s·lied ‘love·le·song’
could not have an s-genitive on the end of a feminine noun, and Liebe is
a feminine noun. Modern practice is thus to avoid classifying these as
genitives, but to see them as semantically empty linking elements (hence
the gloss ‘le’ above). Once these are seen as semantically empty, it is not
clear that there is any difference in type between the proper and the
196 Part VI Historical Questions

improper compounds (Chambers and Wilkie 1970: 60–1). The formal

distinction is still alive in Faroese (Thráinsson et al. 2004: 204–8), and
Icelandic with genitive singular and plural both found, but there still
seems to be little semantic difference between proper and improper. In
Frisian, the genitive compounds have second-element stress and the first
element refers (that is a koken·s·doar ‘kitchen·gen·door’ is the door of
a specific kitchen, not a general type of door: Hoekstra 2016). It may be
best not to consider these to be compounds. Apart from this example, the
pattern of occurrence of the various linking elements (including no linking
element) is typically a matter for discussion in teaching and descriptive
grammars of nearly all the Germanic languages. Despite occasional
examples which appear to show contrast (German Land·mann ‘land man
= farmer’ vs. Land·s·mann ‘land·le·man = compatriot’ and the same
distinction holds in Danish) linking elements are likely to be predictable
only on a probabilistic basis, influenced by various analogies (Krott et al.
2001).
One interesting use of the linking elements is that a linking element, perhaps
most often an s-link, can be used to mark the constituency of a three-element
compound. In compounds of the form [[[A][B]] C] a link may be added
between elements B and C. This is reported from at least Danish, Faroese,
German and Swedish (Bauer 2017: 140–1). Bauer (2009: 407) points out that at
least for Danish this is no more than a tendency.
Semantically, Germanic compounds are typically right-headed, Danish
by·bus ‘town bus = urban bus’ denotes a bus and not a town. This may or
may not be definitional. Gender is typically also derived from the head element,
although occasional exceptions are found. Similarly, the plural marker is
usually, but not universally, inherited from the head. Otherwise, the semantic
link between the elements is extremely free, except that the modifying element
does not refer: a Land·mann is not a man connected with a specific piece of
land, but a man connected with the land in general. Neef (2009: 395, citing
Heringer) points out that German Fisch·frau ‘fish woman’ could mean any one
of the following or others:

a woman who sells fish

a woman who has bought a fish
a woman standing close to a fish
a woman eating a fish
a woman who looks like a fish

He concludes that, although some compounds have a more established

meaning than others, in principle the compound cannot be understood without
 26 Compounds in English and in Wider Germanic 197

reference to the context. All that is given is that a compound of the form [A B]
means ‘a B having something to do with A’.
As with adjective + noun compounds, noun + noun compounds may also be
exocentric. Some examples are given below.
Language Example Gloss Translation Source
Danish storke·næb stork beak ‘pantograph’ (Bauer 2009)
German Bücher·wurm bookworm ‘bookworm’ (Neef 2009)

They can also be coordinative, though these constructions are often contro-
versial (see Chapter 22).
Compounds can be elements in compounds, occasionally leading to very
long words, especially in technical language. This is discussed for English in
Chapter 12.
Phonologically, compounds are stressed on the first element, and ortho-
graphically they are typically presented as a single word, though a recent
spelling reform in Danish allows some separated spellings, so that the
name of Odense Banegård Center ‘Odense Railway Station’ was contro-
versial when the facility was opened. This convention is relatively new in
German.
The greatest difference between the rest of Germanic and English is that
there is no simple way to define a noun + noun compound in English.
Stress on noun + noun sequences, while largely predictable from many
factors, is not unified, the spelling is not consistent, and the semantic
relationships between the elements are just as diverse as they are in the
other Germanic languages. The various tests for a compound are difficult
to apply and give inconsistent outcomes (Bauer 1998). Formally, a noun
modifying another noun (whether we think of this as a compound or as
a syntactic structure) is usually uninflected, but there is an increasing use
of plural marking on the modifying noun. When this is found, some
scholars see it as a sign that the construction is not a compound (Pinker
1999: 178–87), while others see unusually marked compounds (Bauer
et al. 2013: 443). In either case, Bauer (2017: 143) concludes that the
English forms are not linking elements. Where a genitive is found modi-
fying a noun with an appropriate semantic reading, the tradition is not to
call this a compound (but see Rosenbach 2006), but to see it as a syntactic
construction. To add to the confusion, the use of an apostrophe to distin-
guish between a plural and a genitive is not standard: we find girls school
or girls’ school, dogstooth or dog’s-tooth. The result is that it is hard to tell
what constructions in English should be seen as comparable to compounds
in other Germanic languages. This is despite the fact that there is much in
common between the various languages, notably the semantics, including
exocentricity.
198 Part VI Historical Questions

Verb + noun compounds are standard in many Germanic languages.

Language Example Gloss Translation Source

Danish fortælle·kunst tell·art ‘narrative art’ (Bauer 2009)
Danish koge·punkt boil·point ‘boiling point’ (Bauer 2009)
Dutch eten·s·tijd eat·le·time ‘mealtime’ (Booij 2002)
Dutch leer·boek learn·book ‘textbook’ (Don 2009)
Faroese renn·i·skógar run·le·shoes ‘running shoes’ (Thráinsson et al. 2004)
German Web·fehler weave·error ‘flaw in weaving’ (Neef 2009)
German Wende·punkt turn·point ‘turning point’

Note that in Danish the verb is in the infinitive form, in German it is in the
stem form, and in Dutch and Faroese it can have a linking element. While
English does have compounds with base form verbs such as think-tank, wait-
time, it is typical, especially in British English, to have an -ing form in such
cases, as illustrated in the examples above. These forms are then most fre-
quently interpreted as being nouns rather than verbs. Note that, for instance,
British draining board is American drain board, and British frying pan is
American frypan (Bauer et al. 2013: 477). Because of widespread conversion
in English, the proper analysis of such forms is often difficult: is boomtime
a case of a verb + noun compound or not?
Compound adjectives are quite common across Germanic, but there are
relatively few patterns. We have a noun + adjective pattern which can be
glossed as ‘adjective like a/the noun’. As a subtype here we have the type
where the noun acts as an intensifier, and the comparison is obscure or non-
existent. Hoeksema (2012) calls these ‘elative compounds’. Since these
forms have right-element stress, they may not be compounds. We have
a noun + adjective pattern where the noun is an argument of the adjective.
We have a pattern with an adverb + noun (although the adverb often has
adjectival form, but there are good reasons for thinking that adjective and
adverb may be part of the same word-class in much of Germanic). There are
verb + adjective compounds. Then there are various exocentric patterns
where the head of the word is not an adjective, but the compound is used
adjectivally. The various types are illustrated below, including English
examples. Where no source is given, the words are sourced from dictionaries
and experience.

Overt comparison
Danish himmel·blå sky blue ‘sky blue’
Dutch fluister·zacht whisper·soft ‘as quiet as a whisper’ (Don 2009)
English blood red
German himmel·blau sky blue ‘sky blue’
 26 Compounds in English and in Wider Germanic 199

Elative compounds
Danish snot·dum snot·stupid ‘very stupid’ (Bauer 2009)
Danish sten·rig stone·rich ‘very rich’
Dutch ape·trots ape·proud ‘very proud’
Dutch kei·leuk boulder·funny ‘very funny’ (Booij 2002)
English shit hot ‘very hot [usually not in
temperature, but in
approval]’
German blut-arm blood poor ‘very poor’
Nominal argument
Danish mad·glad food·happy ‘fond of good food’
Danish morgen·frisk morning fresh ‘refreshed by a good (Bauer 2009)
night’s sleep’
Dutch lood·vrij lead free ‘lead free’ (Booij 2002)
Dutch milieu-vriendelijk milieu-friendly ‘green’ (Booij 2002)
English ankle-deep
English word-final
German arbeit·s·scheu work·le·shy ‘afraid of hard work’
German herz·zerreisend heart·tearing ‘heart-breaking’ (Neef 2009)
German leben·s·gefährlich life·le·dangerous ‘perilous’
Adverbial
Danish fri·t·stående free·adv·standing ‘free -standing’ (Bauer 2017)
Dutch dicht·bevolkt thick·peopled ‘densely populated’ (Don 2009)
Dutch licht·grijs light·grey ‘light grey’ (Booij 2002)
English dark blue
Faroese blið·mæltur soft·spoken ‘softly spoken’ (Thráinsson
et al. 2004)
German voll·automatisch full automatic ‘fully automatic’
Verb + adjective
Danish stryge·fri iron·free ‘non-iron’ (Bauer 2009)
Dutch koop·lustig buy·cheerful ‘acquisitive’
English go-slow
Faroese renn·vatur drip·wet ‘dripping wet’ (Thráinsson
et al. 2004)
German treff·sicher hit·sure ‘accurate’ (Neef 2009)

Where coordinative adjectival compounds are concerned, there are some

common patterns in modern Germanic, and some interestingly different ones.
The coordinated learned adjectives (typically Latinate or Greek) occur in
those languages that have such adjectives. Coordinated colour adjectives in
most Germanic languages show the presence of both colours, while in English,
coordinated colour adjectives usually show a hybrid between the two colours.
Both sides have exceptions. But in Danish, Dutch and German there is
a difference in stress between the two types, with initial stress reflecting the
200 Part VI Historical Questions

hybrid reading, and final stress reflecting the presence of both colours. In
English, while red-white may occasionally be used to discuss the flags of
Denmark, Austria or Poland, conjoined adjectives are usually preferred, so
that the Danish flag is red and white. In English, the blended colour terms are
rare, with adverbial modification of the colour term being preferred (bluey-
green, greenish-blue) (Bauer 2010). Coordinated native adjectives (other than
colour adjectives) are more common in the other Germanic languages than in
English, where overt coordination seems to be preferred.

Coordinated learned adjectives

Danish dialogisk-lyrisk dialogical- ‘dialogical- (Bauer 2009)
lyrical lyrical’
Dutch pedagogisch- pedagogical- ‘pedagogical- De Haas &
didactisch didactic didactic’ Trommelen
(1993)
English philosophical-
historical
German wissenschaftlich- scientific- (Fleischer & Barz
technisch technical 2007)
Colour adjectives
Danish rød-hvid red-white ‘red and white’
Dutch rood·bruin red·brown ‘reddish (Bauer 2010)
brown’
Dutch rood-groen red-green ‘red and green’ (Bauer 2010)
English blue-green
German blau-grün blue-green ‘blue and (Bauer 2010)
green’
Coordinated native adjectives
Danish døv·stum deaf·dumb ‘deaf and (Bauer 2009)
dumb’
Danish bitter-sød bitter-sweet ‘bitter-sweet’ (Bauer 2009)
Dutch doof·stum deaf·dumb ‘deaf and (De Haas &
dumb’ Trommelen 1993)
Dutch zoet·suur sweet·sour ‘sweet and (De Haas &
sour’ Trommelen 1993)
German taub-stumm deaf-dumb ‘deaf and (Fleischer & Barz
dumb’ 2007)
German süß·sauer sweet·sour ‘sweet and (Fleischer & Barz
sour’ 2007)

Compound verbs are generally considered rare in Germanic, with the con-
struction being apparently unknown in Gothic and sometimes controversial in
other languages (Harbert 2007: 30). Nevertheless, examples are attested in
older states of Germanic and can often be found in modern languages. The
exception to the general rule is the phrasal verb (or particle verb), which is
 26 Compounds in English and in Wider Germanic 201

common. Some languages have a distinction between separable and insepar-

able verbs. Bauer (2009) cites Danish til·falde ‘to·fall = fall to [e.g. it falls to us
to complete this]’ versus at falde til ‘fall to = settle [e.g. to settle in a new
place]’. Hammer (1991) gives examples of tot·schlagen ‘dead·hit = kill’ and
fest·stehen ‘firm·stand = be certain’, where elements other than particles can be
separated from the verbal stem in German. I shall not consider such complica-
tions here, but rather look at examples of compound verbs.

Danish råd·spørge counsel·ask ‘seek advice’ (Bauer 2009)

Danish små·bjæffe small. ‘yap’
pl·bark
Dutch fijn·hakken fine·chop ‘chop up finely’
Dutch slaap·wandele sleep·walk ‘sleepwalk’ (De Haas &
Trommelen
1993)
Dutch snel·schrijven quick·write ‘take (De Haas &
shorthand’ Trommelen
1993)
Dutch stof·zuigen dust suck ‘vacuum clean’ (De Haas &
Trommelen
1993)
English cold-call
English pistol-whip
Faroese góð·kenna good·know ‘accept’ (Thráinsson et al.
2004)
Faroese leið·beina way·lead.to ‘direct, instruct’ (Thráinsson et al.
2004)
German frei·halten free·hold ‘keep unencumbered’ (Fleischer & Barz
2007)
German kopf·stehen head·stand ‘stand on one’s (Fleischer & Barz
head’ 2007)
German rad·fahren bike·drive ‘ride a bike’ (Neef 2009)
German spül·bohren rinse·drill ‘to drill while (Fleischer & Barz
rinsing with 2007)
water’
German stehen·bleiben stand·remain ‘keep still’ (Fleischer & Barz
2007)
Old Scandinavian bók·setja book·set ‘record’ (Haugen 1982)
Old Scandinavian full-nøgja full-satisfy ‘satisfy’ (Haugen 1982)

As will be clear from some of the examples here, one of the problems with
things that look like compound verbs in Germanic is that they are formed either
by conversion from a compound noun or by back-formation, often from an
202 Part VI Historical Questions

agentive form. For some scholars (e.g. Marchand 1969, Booij 2002: 161) this
means that verbal compounding is not productive. Where it is productive, it is
often restricted. Adams (2001: 100–9) provides good coverage of compound
verbs in English, but does not really see them as being created as compounds;
Bauer (2017: 138), on the other hand, suggests that “verbal compounding is no
longer marginal in current English”.
Once a category of compounds has been established in the grammar of
a language, things which may be marginal to the category are easily incorpor-
ated within it. Some typical examples are given below, sometimes included as
compounds, sometimes not, depending on the analyst, but much more easily
seen as compounds in wider Germanic than in English, where orthographic and
stress criteria are not consistently relevant. The categories illustrated in
Table 26.1 are not intended to be exhaustive, and gaps in the table do not
necessarily indicate that relevant forms are not possible, nor does the presence
of an example necessarily make the claim that it is (or should be considered)
a compound rather than something else (perhaps a case of univerbation of
commonly collocating elements).
The fundamentals of the compound are similar across Germanic, with
differences of productivity in different traditions, and a few exceptional types
that I have not dwelled on. One notable exceptional type is illustrated by mann-
skratti ‘deuce of a fellow, devil of a man’, karl·tötrið ‘old.man·poor.fellow’ in
Icelandic. This type is left-headed and is used only in emotionally charged
language (Einarsson 1945: 180).
Much of what we find in English is recognizable from the dominating
Germanic paradigms. The fact that compounds are classifying, that the seman-
tics of noun + noun compounds is fixed only at the most general level, the fact
that we find noun + noun constructions in great numbers, and also find adjec-
tival compounds (much less in some of the languages) but find relatively few
compound verbs, that the fundamental pattern is for stem + word with no
inflection on the first element and the modifying first element does not refer
to specific members of the class, all these are things that English shares with the
Germanic pattern. But English has adjective + noun constructions like red
squirrel which have the classifying semantics of Germanic compounds, but the
superficial form of syntactic noun phrases, that English has many instances
where learned adjectives are used as modifiers rather than using compounds
(English renal artery, Danish nyre·blod·åre ‘kidney·blood·artery’, English
bovine tuberculosis, Danish kvæ·tuberkulose ‘cattle·tuberculosis’, English
structural engineer, Danish bygning·s·ingeniør ‘building·le·engineer’), the
fact that the preferred interpretation of colour adjective + colour adjective
compounds is different in English and in the other languages (and we can
add conjoined place-names like English London–Paris and the equivalents in
other Germanic languages where the English output is used attributively and
Table 26.1 Other less frequent compound patterns in a range of Germanic languages

Construction Danish Dutch English Faroese German

Phrase+N hvorfor-skal-man-op-om-morgenen- kat-en-mus·spel soon-to-be-divorced Sieg·ist·möglich·Strategie
stemme cat-and-mouse-game wife victory-is-possible-
why-must-one-up-in-the-morning- ‘game of cat and strategy
voice mouse’
Pronoun+N hun·ræv she-devil Ich·form
she·fox I·form
‘vixen’ ‘first person singular’
Numeral atten-årig drie·dimensionaal four-dimensional drei·stöckig
+Adj 18-year.ly three-dimensional three·floor.ed
18-year-old ‘three storied’
Numeral+N tre·dækker drie·tand six-pack fimm·króna Fünf·kamp
three-decker three·tooth five·kroner five·struggle
‘triplane’ ‘trident’ ‘a five-kroner coin’ ‘pentathlon’
Numeral+V tre·dele tví·býta
three·divide two·divide
‘trisect’ ‘divide in two’
Prep+N efter·år after-care Vor·teil
after·year before·part
‘autumn’ ‘advantage’
Prep+Adj til·bøjelig boven·natuurlijk off-white unter·irdisch
to·flexible above·natural under·terrestrial
‘inclined to’ ‘supernatural’ ‘subterranean’
Prep+V op·råbe uit·gaan download zu·setzen
up·call out·go to·put
‘announce’ ‘go out; originate ‘add’
with’
Prep+Prep til·med into dort·hin
to·with there·to
‘moreover’ ‘thither’
204 Part VI Historical Questions

the wider use in Germanic is for adverbial use – Bauer 2010), these all suggest
systems which are importantly different. The same is true of the importance of
stress in most of Germanic for defining a compound, as opposed to English, the
clear perception of compounds as single words in most of Germanic, and the
frequent analysis of forms like forestry worker as syntactic and not morpho-
logical in English (although there are also many who disagree). All these
factors make it seem that while there may be central cases of overlap, there
are more peripheral places where the two traditions diverge, and what is
included as a compound may differ from one tradition to the other.
The importance of this is that general Germanic scholarship on compounds
may not always transfer to the treatment of English, and it is hard to know
whether this has done damage to the description of English or not. At the very
least, it has meant that there is dispute in the literature on English about
precisely where the boundaries of compounding lie. It is tempting to suggest
that the differences arise through the influence of French on English vocabu-
lary. For French, stress is not a relevant factor in determining whether some-
thing is or is not a compound; French has differing preferred patterns of
marking the classifying relationship – either by using adjectives rather than
nouns as modifiers or by using post-modifying prepositional phrases, as in
French chemin de fer ‘path of iron’ corresponding to German Eisen·bahn
‘iron·way’ for ‘railway’. While this seems eminently plausible, it is hard to
prove, largely because some of the same descriptive problems arise in other
Germanic languages as well. Whatever the cause of the differences, they mean
that the analyst must not simply assume that what is said of other Germanic
languages automatically applies to English.

Challenge
Constructions like blue whale, red squirrel, white cell seem to be halfway
between compounds and syntactic structures: they are spelled and stressed like
syntactic constructions, but their meaning is like the meaning of compounds,
and like compounds they do not allow coordination of the adjective with
another adjective while retaining their specialized meaning. Can you find any
other aspects of their behaviour which might be helpful in determining their
status? Whether you can or not, how do you think such constructions should be
analysed, and how would you justify your position?

References
Adams, Valerie. (2001). Complex Words in English. Harlow: Longman.
Bauer, Laurie. (1998). When is a sequence of two nouns a compound in English?
English Language and Linguistics 2, 65–86.
 26 Compounds in English and in Wider Germanic 205

(2009). IE, Germanic: Danish. In Rochelle Lieber & Pavol Štekauer (eds.), The
Oxford Handbook of Compounding. Oxford: Oxford University Press, 400–16.
(2010). Co-compounds in Germanic. Journal of Germanic Linguistics 22, 201–19.
(2017). Compounds and Compounding. Cambridge: Cambridge University Press.
(2020). Blackbirds and blue whales: Stress in English A+N constructions. English
Language and Linguistics 25, 1–20.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Booij, Geert. (2002). The Morphology of Dutch. Oxford: Oxford University Press.
Chambers, W. Walker & John R. Wilkie. (1970). A Short History of the German
Language. London: Methuen.
Don, Jan. (2009). IE, Germanic: Dutch. In Rochelle Lieber & Pavol Štekauer (eds.), The
Oxford Handbook of Compounding. Oxford: Oxford University Press, 370–85.
Einarsson, Stefán. (1945). Icelandic: Grammar, Texts, Glossary. Baltimore: Johns
Hopkins Press.
Fleischer, Wolfgang & Irmhild Barz. (2007). Wortbildung der deutschen
Gegenwartssprache. 3. Auflage. Tübingen: Niemeyer.
Haas, Wim de & Mieke Trommelen. (1993). Morfologisch handboek van het
Nederlands. ’s-Gravenhage: SDU.
Hammer, A.E. (1991). Hammer’s German Grammar and Usage. 2nd ed., revised by
Martin Durrell. London: Arnold
Harbert, Wayne. (2007). The Germanic Languages. Cambridge: Cambridge University
Press.
Hoeksema, Jack. (2012). Elative compounds in Dutch: Properties and developments. In
Guido Oebel (ed.), Intensivierungskonzepte bei Adjektiven und Adverben im
Sprachenvergleich / Crosslinguistic Comparison of Intensified Adjectives and
Adverbs. Hamburg: Verlag dr. Kovač, 97–142.
Hoekstra Jarich, F. (2016). Frisian. In Peter O. Müller, Ingeborg Ohnheiser, Susan Olsen
& Franz Rainer (eds.), Word-Formation: An International Handbook of the
Languages of Europe. Berlin: De Gruyter Mouton, 2451–65.
Krott, Andrea R., Harald Baayen & Robert Schreuder. (2001). Analogy in morphology:
Modeling the choice of linking morphemes in Dutch. Linguistics. 39: 51–93.
Marchand, Hans. (1969). The Categories and Types of Present-Day English Word-
Formation. 2nd ed. Munich: Beck.
Neef, Martin. (2009). IE, Germanic: German. In Rochelle Lieber & Pavol Štekauer
(eds.), The Oxford Handbook of Compounding. Oxford: Oxford University Press,
386–99.
Pinker, Steven. (1999). Words and Rules. London: Weidenfeld & Nicolson.
Rosenbach, Anette. (2006). Descriptive genitives in English: A case study on construc-
tional gradience. English Language and Linguistics 10, 77–118.
Thráinsson, Höskuldur, Hjalmar P. Petersen, Jógvan Í. Lon Jacobsen & Zakaris Svabo
Hansen. (2004). Faroese: An Overview and Reference Grammar. Tórshavn:
Føroya Fróðskaparfelag.
Part VII

Questions Involving Inflection

27 Reflections on Inflection inside
Word-Formation

27.1 Introduction
There are two apparently disconnected factors that seem to disallow inflectional
affixes from occurring inside word-formation. The first is the general rule that
derivational affixes appear closer to the root than inflectional affixes. To the extent
that this rule holds, it bans any sequence of inflectional suffix + derivational suffix.
The second is the observation that the first element in compounds is not inflected.
We know this to be false, in the sense that it is an overgeneralization, so that
discussions have been focused on explanations for the exceptions to the rules. It
should be said that the constraint of inflection within word-formation is supposed
to hold of English (and probably other languages as well), but not to be a universal.
Many languages are reported to have internal inflections in compounds, including
Finnish and Tariana.
Finnish
auto·n·ikkuna ‘car·gen·window’ (Karlsson 1999: 242)
maa·lta·pako ‘country·abl·flee.nmlz’ ‘rural depopulation’ (Sulala and
Karjalainen 1992: 362)
Tariana
maːnakadɾu-ni ‘açai.tree.parrot-poss’ ‘scarlet macaw’ (Aikhenvald 2003: 132)

Many languages are reported to have inflectional affixes closer to the root than
diminutive markers at least, but also with suffixes which are more obviously
derivational (see Bauer 2003: 100).
German
Kind-er-chen ‘child-pl-diminutive’
Welsh
merch-et-os ‘girl-pl-diminutive’
Dutch
muzikant-en-dom ‘musician-pl-nmlz’

209
210 Part VII Questions Involving Inflection

Of course, any theory that makes predictions about the sequencing of

inflectional and derivational markers assumes the split morphology hypothesis
(see Section 2.4), and that provides a useful starting point.

27.2 Inflection and Derivation in English

A good definition of inflectional morphology in English would presumably
provide some behaviour that is shared by all inflectional affixes and not shared
by any derivational affixes. While many such criteria have been suggested (see
Plank 1994), it is noteworthy that Bauer et al. (2013: 29), like others dealing
with English, stipulate where the boundary between inflection and derivation
lies rather than providing a definition of the distinction. Their list of inflectional
affixes (with which we might want to disagree) is as follows:
plural and possessive marking on nouns, third-person singular marking on
present-tense verbs, past tense and past participle marking on verbs, present
participle marking on verbs, comparative and superlative marking on adjec-
tives and adverbs.
While this position reflects a certain consensus, it is far from a definitive list.
Questions that could be raised about the list include: is the plural really inflec-
tional (Beard 1982)? Is the comparative in English really inflectional (Matthews
1974: 48–9)? Is the possessive marker inflectional, or is it a clitic? Is adverbial -ly
really derivational (Giegerich 2012)? Is ordinal -th (as in fifteenth) really deriv-
ational (Bauer et al. 2013: 536–7)? Are the participles still instances of inflection
when the forms are used as adjectives or other word-classes? Is n’t an inflectional
affix (Zwicky and Pullum 1983)? More generally, we might ask whether it makes
sense to ask about the application of the inflection–derivation divide to such
forms at all.
These are all relevant questions, and they may not be questions which can be
given a definitive answer. Furthermore, this does not exhaust the list of poten-
tial problems. This means that it would be possible to answer the whole issue
by saying that it is not clear that the question makes any sense at all, and it
therefore cannot be answered. However, in the present state of scholarship, the
question does make sense, and the apparent restriction has been noted several
times, so it is worth considering. In doing so, the ostensive definition of
inflection used by Bauer et al. (2013) will simply be accepted as a starting
point for the discussion. Future research may help refine the question a bit.

27.3 Inflection inside Compounds

We can illustrate the constraint by looking at comparative and superlative
marking. If we accept that words like blackbird and tallboy are compounds
 27 Inflection inside Word-Formation 211

(which is generally accepted, but not necessarily beyond question), then the
constraint tells us that we should not be able to mark the comparative or
superlative on the adjective within the compound. Since we cannot have
*blackerbird or *tallestboy (or indeed other parallel words) the constraint
seems to hold, and these inflectional forms are not allowed within compounds.
In this particular case, though, the constraint is not (or may not be) restricted to
compounds. Given black market and tall order ‘difficult assignment’ (which
most would take not to be compounds), we cannot have *blackest market or
*taller order, either. At least, we cannot have these if the semantic idiosyncra-
sies of black market and tall order are to be retained. They might be possible
in environments in which the elements are entirely compositional, for example
He wrote three orders on the wall, in ever-increasing size. The tallest order was
that no one was to go outside. Such examples would not normally be seen as
compounds, though. Just what the limits of the constraint really are may not be
clear, but it does seem to hold within compounds.
But that is just one example. We also have to consider plural markers,
possessive markers, third-person singular -s, -ed (and its irregular con-
geners) and -ing. Most of these seem to appear in compounds in some
form or another.
First we must consider instances where the modifying element in the com-
pound is a piece of syntax, as in dad-needs-a-new-sports-car syndrome, grass-
is-greener syndrome (Bauer et al. 2013: 488). Here we find needs and greener
inside the compounds. Are these counterexamples? Strictly speaking they are,
though it depends on the care with which the constraint is formulated. But since
the inflection is not a marker of the role of the first element, but arises instead
from the piece of syntax that is transplanted inside the compound, such forms
will be treated as irrelevant here.
The easiest of these to deal with may be the third-person singular -s on
present-tense verbs. As far as I know, this never occurs within anything that
might be thought to be a compound. Again, the constraint seems to work. The
same is probably true of the past tense -ed (or irregular forms), though it may be
hard to tell whether a past tense or past participle form is involved. We seldom
find forms like *sangsong or *spentthrift.
The others are all problematic in one way or another. Some potential
counterexamples are given below.
plural: mice droppings, sales manager, sports bar, systems analyst
possessive: bullseye, cat’s-eye, foolscap, lambswool, menswear
past participle: freedman
present participle: humming bird, sleeping pill, washing machine
. . . in case anything social-media-worthy gets dragged up. (Gilly
Macmillan. 2017. Odd Child Out. New York: Morrow, p. 159)
212 Part VII Questions Involving Inflection

Of all of these, perhaps the plural is the most interesting and the most
frustrating to deal with. One of the problems here is that within Lexical
Phonology and Morphology it was suggested that irregular plurals could
occur in the first element of compounds, but not regular ones, so that mice-
catcher was permissible but not *rats-catcher, because the irregular plural was
at Stratum I and thus available as a compound element, while regular plurals
were added after compounding had applied (see Kirchner and Nicoladis 2009
for some discussion). The difficulty with this theoretical stance is that only
some irregular plurals seem to be used in this way.
“Missy, what are you looking for back there?” And she said, “I’m looking for
mice manure.” (Tony Hillerman. 1999. The First Eagle. London: Hodder
Headline, pp. 60–1)
Mice-free; mice population (used on Radio New Zealand National’s Morning
Report, 22 March 2018)
Lice infestation is a parasitic skin infestation caused by tiny wingless insects.
(www.msdmanuals.com/en-nz/home/skin-disorders/parasitic-skin-infec
tions/lice-infestation accessed 25 November 2023)
As Kirchner and Nicoladis (2009: 95) point out, we do not find *teethbrush
or *childrencare, even though we expect their referents to be used for more
than one tooth or more than one child, and we can add that *feetpath, *geese-
bumps and *menpower seem equally impossible, though menfolk, womenfolk
are standard forms (these may be parallel to gentlemen farmers where both
elements are plural). Latin and Greek plurals do not seem to be widely used in
this way (we find, for instance, bacterial growth rather than *bacteria growth),
but it is hard to be sure, partly because many such forms are unstable anyway.
Alumni association is at least one common exception, and see media-worthy
above. Plurals with voiced fricatives (e.g. calf/calves) do not appear to be used,
but with zero plurals such as deer-hunting and sheep-farming it is impossible
to tell whether there is a plural form involved or not, and the singular form
seems more likely.
A rule which says that irregular plurals may be used in the first element of
compounds as long as they end in /s/ seems far less motivated than a general
rule based on Stratal Morphology. Moreover, it is not an obligatory rule:
mouse-droppings and louse-infestations are equally possible forms. Overall,
it seems that the use of irregular plurals in compound first elements has been
overstated.
Regular plurals in the first element of compounds seem to be far more readily
available. One set of these, perhaps the earliest examples, are those where the
form in -s is not a simple plural of the unmarked noun. For instance, if there is
such a thing as an arm deal, it might be a deal to buy prosthetic limbs, while an
arms deal is one made to purchase weapons. This phenomenon accounts for
 27 Inflection inside Word-Formation 213

forms such as brains trust, clothes drier, draughtsboard, grassroots movement,

mains power and savings bank, among many more. On the other hand, it is by
no means automatic that a final -s will be used in such cases. Refreshment
trolley and refreshments trolley can both be found, trouser press is used rather
than a form with plural trousers, a card game is the same as a game of cards,
wage bill and wages bill are both found. Moreover, there is no clear justification
for the variation that is found in such instances. Even in this particular set of
circumstances, a form with no plural marking is, overall, more usual than might
be thought.
In the mid-1960s reports start to appear in print of a new usage (or a rapidly
increasing usage) of plural attributives or plural modifiers (Mutt 1967,
Dierickx 1970) in English. Just when the trend began is difficult to say, but
it was noticed at that point. These are usages such as burns unit, careers
officer, companies legislation, drugs problem, letters column, textiles indus-
try, which today do not seem at all out of the ordinary. If these are compounds,
they break the rule about not allowing inflected forms in the first element of
compounds. Even within this set of words, some are less innovative than
others (and may seem more compound-like). The more established pattern
includes many apparent compounds ending in man: craftsman, marksman,
salesman, sportsman, tradesman, yachtsman and some with wider usage such
a sports jacket. These do not seem to involve any genuine plural meaning
(a craftsman may master only one craft, a yachtsman may sail on only a single
yacht), and in most cases the singular and plural forms do not have different
referents. Yet there does seem to be some tendency to use the plural form
when a plural meaning is intended, so that companies legislation involves
many or all companies, not just one, while company policy involves just one
company. Previously, though, that would not have been enough to call forth
a plural form. Whether this is the justification for the new usage or not, it is not
clear why it should suddenly have gained traction. (For further discussion of
this type, see Bauer 2017: 140–8, 2022: 157–62.)
If we are to uphold the postulated constraint against inflection in compounds,
we would have to say that the relevant structures here are not compounds.
Pinker (1999) does just that, in effect claiming that sport jacket and sports
jacket illustrate two different constructions. Unfortunately, the only sign of the
distinction may be that one has a plural noun and the other does not. However,
in principle this remains an option (whether Pinker’s analysis is used or some
other), although it is one which looks as though it is circular.
The same thing may be true with possessives. The examples cited above
look like compounds, and look like possessives whether or not the standard
spelling involves an apostrophe. The spelling, though, as in other compounds,
is a poor guide. A quick survey of four dictionaries gives the following range
of spellings.
214 Part VII Questions Involving Inflection

bugbane
cowbane
dogbane dogsbane dog’s-bane
fleabane
flybane
foxbane
hare’s bane
henbane hensbane
ratsbane rat’sbane
wolf’sbane wolf’s-bane, wolf’s bane

These various spellings suggest that some genitive modifiers can be

included in compounds, and that they are in variation with non-genitive
modifiers. They also suggest some insecurity about such items, although
the uninflected modifier remains the preferred option in most cases, and the
frequency of the modifying noun may play a role. The underlying problem is
that the genitive is not simply a marker of possession in English, but has
multiple functions (as in many other languages). It can mark a determiner
showing possession, it can show things that are not possession, and it can be
used as a classifier for a following noun. Since an unmarked noun can also be
used as a classifier, there is an overlap of functions, and the overlap of forms
between a determiner and a classifier can also leave forms ambiguous on
occasions. For example, dog fennel and dog’s fennel are alternative names for
‘mayweed’, and this woman’s magazine is classifying in This woman’s maga-
zine contains no recipes or sewing tips but possessive in Every woman was
told to bring a magazine to read, and this woman’s magazine was The
Economist. It is the classifying genitives which might be thought to be
compounds, perhaps especially those which are stressed on the first element
(like dog’s fennel and menswear).
It can be denied that genitives in attributive position create compounds
(in which case there is some obligation to explain why these things are not
compounds and what they are), or it can be denied that the genitive is an
inflection. Either or both may be valid. But it certainly looks as though
genitives create constructions which behave in a very similar manner to
compounds. (For a more detailed discussion, see Rosenbach 2006.)
The participles may be less difficult to dismiss. It is absolutely true that the
morphological process whereby -ed and -ing create adjectives and nouns is
a puzzle (see Chapter 7), and equally puzzling just what the relationship is
between those forms and the inflected forms of the verb is. But we may be able
to ignore that puzzle, and say that nouns and adjectives can occur in the first
elements of compounds, and that their form is irrelevant to this generalization.
 27 Inflection inside Word-Formation 215

Then freedman and hummingbird are just normal compounds, and we do not
need to look at their internal constituency.
There is one potential problem here. While compounds like hummingbird are
common, the -ing form is always a noun in them (in hummingbird, the -ing form
could be interpreted as an adjective, but such cases are rare). Compounds with
past participles in the first element are extraordinarily rare (as opposed to noun
phrases with past participles in attributive position). Some of this could be to do
with the problem of defining an adjective + noun compound in English (Bauer
2021), but we may not even have to look into that. Most English compounds of
the form adjective + noun contain a monosyllabic adjective. Most past participles
and all -ing forms are disyllabic or longer. In the spirit of Optimality Theory, all
we need to say is that the length constraint overrules other constraints. However,
that is not the end of the matter. First, we have apparent adjective + noun
compounds with learned words in the first element, although they do not permit
participles either: these are forms like dental hospital, medical school, musical
box, nervous system, operatic society, solar system (Bauer 2021). The other
problem is that irregular past participles are often monosyllabic and must be
listed at Stratum I in the Level Ordering system, yet they do not seem to appear in
compounds either: we do not find *blown-hair, *drunk-customers, *dug-
gardens, *read-papers, and where we do find premodifying participles, we do
not find a compound: burnt skin, grown children, hand-fed piglets, home-grown
vegetables, interesting lives. Knitwear is a possible exception.
Overall, if we accept the presuppositions in the question of whether inflec-
tional affixes can appear in the first elements of compounds, the answer is that
they are limited rather than they do not occur. Why they are limited, and
perhaps why those that can occur are exceptional, remain open questions.

27.4 Inflections inside Derivational Affixes

Inflectional affixes are found inside derivational affixes quite frequently, but
not usually in productive usage and, in any case, not in many patterns. Some
examples are provided below.
with meaning-changing or obligatory -s: folksy, gutser, gutsful, gutsy,
gutsiness, newsy, sudsy, woodsy
with irregular plural: (see below)
with plural -s: (see below)
with possessive -s: not found
with -ed: excitedness, presumedly, supposedly
with -ing: interestingly, lovingness
with comparative: betterment, betterness, lessen, worsen
with superlative: mostly
216 Part VII Questions Involving Inflection

One of my plainclothesers did get a little excited. (Jonathan Kellerman. 2010.

Deception. New York: Ballantine. E-version, p. 244)
Her mom had taught her how to . . . deal with handsy men. (Laura Griffin.
2019. Desperate Girls. London: Headline Eternal, p. 91)
both . . . conceding the regular Unitedites’ prior claim to the true dejection of
defeat. (Martyn Bedford. 1997. Exit, Orange and Red. London: Bantam,
p. 261)

Here, nearly all of the instances can be justified without a claim that inflec-
tion can precede derivation: the forms are not simply inflectional, the inflected
form is irregular, the -ed and -ing create adjectives whose internal structure is
irrelevant. Exceptions are found, but are rare.

27.5 Discussion
Although it could be argued that, as phrased, a ban on inflection inside word-
formation in English makes too many assumptions and is too superficial to be
a coherent constraint, it holds up fairly well. We might, for instance, argue that
a simple dichotomy between inflection and derivation is not sufficient (Booij
1996), that compounds are not well enough defined to support the claim, that
inflection is not well enough defined to be a usable category in such a context,
and all those points are important, though not necessarily an exhaustive list of
objections. Nevertheless, the limited instances of inflection inside word-
formation that we find in English suggest that even if the constraint is not
absolute, it holds as a default. The exceptions that require further exploration
are the use of the plural and the genitive in the first elements of compounds. The
matter of the genitive might come down to a definitional problem: if we have
a classifying genitive does it form a compound or something else? The plural is
much harder to dismiss. The development of plurals in this position may arise
from instances like arms deal, where the -s is arguably not just a plural marker,
but shows some features of a derivational affix, and instances like sportsman,
although it is not clear why such a minor pattern should expand. We do not yet
appear to have a good account of how such forms developed, although their
increase seems to be a twentieth-century phenomenon (see, for example, the
graphs in Bauer 2017: 145–7). Without a better appreciation of precisely what
has happened, it will be hard to understand what motivated the development.

Challenge
It is clear that plural nouns can occur in modifying nouns in English, and in
some cases the resultant construction is a compound. The case with foreign
nouns (containing foreign plural markers) is less clear, probably because
 27 Inflection inside Word-Formation 217

examples are rarer. What examples can you find of Latin or Greek plurals used
attributively? (Other languages may be considered, but they are even
more difficult to find and to analyse.) Are the examples you find compounds
or not? How can you tell? Do your findings influence your view of plurals
within word-formation? If so, in what way and why?

References
Aikhenvald, Alexandra Y. (2003). A Grammar of Tariana. Cambridge: Cambridge
University Press.
Bauer, Laurie (2003). Introducing Linguistic Morphology. 2nd ed. Edinburgh: Edinburgh
University Press.
(2017). Compounds and Compounding. Cambridge: Cambridge University Press.
Bauer, Laurie. (2021). Blackbirds and blue whales: Stress in English A+N
constructions. English Language and Linguistics 25, 581–600.
(2022). An Introduction to English Lexicology. Edinburgh: Edinburgh University
Press.
Bauer, Laurie, Rochelle Lieber & Ingo Plag. (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Beard, Robert. (1982). The plural as a lexical derivation. Glossa 16, 133–48.
Booij, Geert. (1996). Inherent versus contextual inflection and the split morphology
hypothesis. Yearbook of Morphology 1995, 1–16.
Dierickx, Jean. (1970). Why are plural attributives becoming more frequent? In
Jean Dierickx & Yvan Lebrun (eds.), Linguistique contemporaine: homage à
Eric Buyssens. Brussels: Éditions de l’institut de sociologie de l’université libre,
39–46.
Giegerich, Heinz J. (2012). The morphology of -ly and the categorial status of ‘adverbs’
in English. English Language and Linguistics 16, 341–59.
Karlsson, Fred. (1999). Finnish: An Essential Grammar. 2nd ed. London: Routledge.
Kirchner, Robert & Elena Nicoladis. (2009). A level playing-fieldː Perceptibility and
inflection in English compounds. Canadian Journal of Linguistics/Revue
Canadienne de linguistique 54, 91–116.
Matthews, P.H. (1974). Morphology. Cambridge: Cambridge University Press.
Mutt, O. (1967). Some recent developments in the use of nouns as premodifiers in
English. Zeitschrift für Anglistik und Amerikanistik 15, 401–8.
Pinker, Steven. (1999). Words and Rules. London: Weidenfeld & Nicolson.
Plank, F. (1994). Inflection and derivation. In Ron E. Asher (ed.), The Encyclopedia of
Language and Linguistics. Oxford: Pergamon, vol. 3, 1671–8.
Rosenbach, Anette. (2006). Descriptive genitives in English: A case study on construc-
tional gradience. English Language and Linguistics 10, 71–118.
Sulkala, Helena & Merja Karjalainenen. (1992). Finnish. London: Routledge.
Zwicky, Arnold M. & Geoffrey K. Pullum. (1983). Cliticization vs. inflection: English
n’t. Language 59, 502–13.
28 Reflections on Canonical Form

Canonical form is not something that is much discussed in present-day

morphology, though it was familiar in the 1950s (see Hockett 1958). Part of
the reason for the notion being ignored is probably that it has been replaced by
arguments based on prototypes. Canonical form, though, may be worthy of
some rehabilitation, since it allows us to focus on form rather than on the range of
factors that make something prototypical in its category. A reformulation in other
terms would be perfectly possible, but might not add much.
The notion of canonical form is that certain word-parts have typical shapes in
different languages. For instance, although there are other patterns, there is
a familiar pattern in Arabic in which the root is made up of three consonants,
so that katab ‘he wrote’, maktuub ‘office, writing place’ and kitaab ‘book’ all
share the root KTB, to do with writing.
Part of the reason that so many ‘embellished clippings’ (Bauer and
Huddleston 2002) are two syllables long in English may be that there is
a canonical form of two syllables for a basic word in English (though we
would need rather more evidence to elevate this assumption to the nature of
a hypothesis). Relevant words are Aussie, cauli, drongo, Maccas (this is what
the hamburger restaurant McDonald’s is called in Australia), muso, piggie,
preggers and so on. However, there is one place in English where the idea of
a canonical form really does seem to play a clear role.
Prefixes in English contain at least one whole syllable. They may contain
more than one syllable (epi-, mega-, mini-, super- and, if you think these are
prefixes, cardio-, electro-, syntactico-). Prefixes do not contain less than
a syllable. In case this seems obvious, consider suffixes in English, which
can contain just consonants, in words like cats, loved, tenth. It is worthy of
comment that we do not have prefixes of this form.
Therefore, some morphological puzzles of English which will be illustrated
immediately below do not contain single-consonant prefixes, because a prefix
must contain a whole syllable and not just a consonant. The most often cited
example here is that if we look at ear and hear, it might be thought that there is
a prefix h- which creates verbs from nouns (even though there is no verb *heye).
Similarly, the relationship between roquet and croquet is not a matter of a prefix

218
 28 Canonical Form 219

(it is not even clear what the prefix would mean in this instance). The relationship
between melt and smelt in the sense of smelting ore) is also excluded, although
there is an etymological link here. And the relationship between ethane and
methane (or ethanol and methanol) is not a matter of a prefix m- (etymologically,
there are distinct Greek elements eth- and meth- here). The four pairs cited here
do at least have some meaning to relate them (however obscure); an indefinite
number of other pairs of English have a putative prefix of non-canonical form
and no serious meaning correlation: owl and cowl; oast and toast; ray and pray;
awe and flaw. Perhaps most interesting, though, is that if we accept this conclu-
sion, then the shm- that we find in (American English) school-shmool, professor-
shmofessor (see Bauer et al. 2013: 413 and Section 24.2 in the present book) is
not a prefix, but some other kind of word-formation – probably fixed consonant-
ism associated with reduplication.
Another place where we might make appeal to canonical form is in the
consonants used in suffixes. The consonant phonemes of English are set out
below, as if they were set out in the chart of the International Phonetic
Association, with places of articulation shown by the columns, manners of
articulation shown by the rows, and voicing shown by position in the cell,
voiceless consonants on the left, voiced on the right.

Now I want to consider some sets of affixes of English. Precisely what the
inflectional affixes of English are is slightly controversial (see Section 27.2). The
adverbial -ly in words like apparently is much discussed, and the ordinal -th in
words like ninth is a possibility. The -st in lovest and the -t in willt have now
vanished, as has -eth in the third-person singular of the present tense. The suffixes
below are generally agreed to be inflectional.
-ed (/t/, /d/, /ɪd/), -s (/s/, /z/, /ɪz/), the plural marker, -er (/ə(r)/), the compara-
tive marker, -est, the superlative marker, -en and -ren, plural markers, -en,
a past participle marker, -s, the third-person singular of the present tense,
and -ing
220 Part VII Questions Involving Inflection

Just below, the consonant table is repeated, but this time those consonants
found in the inflectional affixes are highlighted, and those mentioned in poten-
tially inflectional suffixes are dimmed.
Consonants from inflectional suffixes

A glance at this version of the chart makes it perfectly clear that the /ŋ/ in -ing
is way out of line. All the other consonants in inflection are alveolar – and even
if we included old forms like -eth in maketh, they would be coronal (articulated
with the tip and/or blade of the tongue).
What happens if, instead of looking at inflectional suffixes, we look at
derivational suffixes? Here we must distinguish between those which are native
to English, and those which have been borrowed, usually from French, some-
times from Latin or Greek (even then, often via French). The latter list is the
longer one, and probably not a closed list. Both types are listed below.
major native derivational suffixes: -dom (kingdom), -en (leaden), -er
(killer), -ful (hopeful), -hood (neighbourhood), -ish (greenish), -less
(useless), -ling (duckling), -ly (friendly), -ness (brightness), -ship
(citizenship), -some (fulsome), -ster (youngster), -ter (laughter), -th
(warmth), -ward(s) (backward(s)), -y (chilly)
major borrowed derivational suffixes: -able (presentable), -age
(marriage), -al (parental), -an (African), -ana (Victoriana), -ant
(attendant), -ar (uvular), -ation (civilization), -ee (employee), -ese
(journalese), -esque (Disneyesque), -ess (duchess), -ette (usher-
ette), -ic (telepathic), -ify (justify), -ist (specialist), -ity (serenity),
-ive (active), -ize (lionize), -let (piglet), -nik (folknik), -or
(conductor), -ous (zealous), -y (telepathy)
If we now repeat the consonant table again with the consonants from these
two lists highlighted, it will become clear that rather more consonants have
 28 Canonical Form 221

been allowed in, and that there is no longer a clear-cut restriction. However, if
we weight the various consonants by the number of times they recur in affixes,
the same set would still be the most heavily weighted ones.
Consonants from native suffixes

Consonants from borrowed suffixes

If we move on and look at prefixes, there is even less pattern. It is not

obviously possible to give a definitive list of prefixes (even native ones, because
the status of things like by- may be unclear), but there do not appear to be
motivated gaps in the consonants that may appear in them. Some prefixes are
listed below to make the point, and only /ð/ seems to be systematically excluded.
This is just one way in which prefixes are more word-like than suffixes.
Some illustrative prefixes of English
a- (atelic), ambi- (ambidextrous), ante- (antediluvian),
arch- (arch-enemy), be- (bewitch), circum- (circumpolar), crypto-
(crypto-communist), deca- (decapod), dis- (dislike), em- (embed),
endo- (endocentric), extra- (extraordinary), giga- (gigabyte),
222 Part VII Questions Involving Inflection

hemi- (hemisphere), hepta- (heptagon), hyper- (hyperactive), multi-

(multilingual), para- (para-legal), proto- (proto-language), quadr-
(quadruped), step- (step-father), theo- (theocentric), vice- (vice-
regal), with- (withhold), yotta- (yottabyte), zetta- (zettabyte)
If we conclude that the initial s- in smelt is not a prefix because of its form, we
must also ask whether -ing is an inflectional suffix, since its form makes it
appear an unsuitable member of that class. The argument for -ing being
inflectional is that it meets the criteria for an inflectional affix: it is formally
regular, it is semantically regular, it is productive, it does not change word-
class, and it realizes one of the markers of tense/aspect, which are otherwise
considered to be inflectional because they are required by the grammatical
structure of the sentence. But this characterization is limited to the use of -ing
that we find in examples like the following.
I am enjoying retirement.
She is managing well under the circumstances.
It is raining.
The suffix -ing also occurs in two very different functions: creating adjec-
tives and creating nouns.
Creating adjectives
A very interesting book.
A rather surprising suggestion.
Breaking news.
This wine is disgusting.
Creating nouns
This building was designed by Sir Christopher Wren.
His building so close to the boundary was illegal.
Their constant moaning annoys me.
Their accidentally killing the bird upset me.

The best way of dealing with these forms in a grammar is not agreed upon.
There are several possibilities, not all equally attractive. Among the possibil-
ities are (a) assume three homophonous affixes -ing, (b) view -ing as being
a word-class changing inflectional affix, (c) see the adjectival and nominal uses
of -ing as being derived by conversion from the verbal -ing form, (d) see the
participle as being a separate word-class from noun, verb or adjective. All of
these solutions have problems associated with them.
Among the problems is that the various -ing adjectives and nouns do not
behave in the same way. Some of the adjectives allow premodification and give
rise to -ly adverbials, others do not; some of the nouns allow modification by
adjectives, while in other constructions they are modified by adverbs; some of
the nouns act like action nominalizations, others do not. Some of this is
 28 Canonical Form 223

predictable grammatical behaviour, some of it is lexical. For instance, adjec-

tival -ing is less productive than nominal or verbal -ing.
An alternative way of looking at this is that -ing is fundamentally nominal (as
is its origin), and fundamentally an action nominalization. Nouns can be used
attributively and in predicative position, where they overlap in function with
adjectives. The use of -ing to mark the continuous form then has to be described
as a construction involving a nominal which has become idiomatized and so is
not readily analysable in its modern form. Under such an analysis, -ing would
no longer be an inflectional marker, and the lower productivity of adjectival
uses of -ing than on nominal and verbal uses would be explained. This is a lot to
derive from the canonical form of inflectional affixes, but the analysis might
nevertheless have some internal, as well as historical, merit.
We can also look at the canonical form of words. It was suggested above that
it might be the case that words are typically two syllables long in English. If we
look at the words containing just one meaningful element (that is, a stem, but no
affixes, so monomorphic or monomorphemic words), and look at the most
common words in English, the average length of such a word is 1.4 syllables,
with adjectives and verbs slightly shorter than nouns. Given that, by Zipf’s law,
rarer words will be longer than common ones, an approximation of two-
syllable words being some sort of default length makes sense. If we look at
derived forms (not compounds) in the first 1,000 most frequent words, in which
a separation into base and affix makes clear sense in modern English, then the
average length of the derivative is 2.94 syllables. We can deduce that, in
common words, derivatives of the required nature usually contain only one
affix, adding about one syllable to the length. Again, rarer words are likely to be
longer. If inflected forms are also counted, the length of the words observed will
also be slightly longer. These figures are made less reliable by the number of
words, usually of Greek and Latin origin, which do not conform to the criteria
used for derivatives here, and which are often longer. Given the existence of
words like institutionalization, or antidisestablishmentarianism, it is clear that
the average or expected length can be exceeded by a considerable amount, but
the rarity of such forms indicates that the expected range is between about one
and six syllables, with very few words going beyond those limits.

Challenge
Make a case for any one of the following possibilities: (a) that all -ing forms are
verbal, and other word-classes are derived from that; (b) that all -ing forms are
adjectival, and all other uses are derive from that; (c) that all-ing forms
are nominal and that all other uses are derived from that; (d) that the three
-ing suffixes are homophonous but distinct suffixes; (e) that one of the three
224 Part VII Questions Involving Inflection

uses of -ing is derived by conversion from one of the others. Compare and
contrast any two of these solutions.

References
Bauer, Laurie & Rodney Huddleston. (2002). Lexical word-formation. In Rodney
Huddleston & Geoffrey K. Pullum, The Cambridge Grammar of the English
Language, Cambridge: Cambridge University Press, 1621–1721.
Bauer, Laurie, Rochelle Lieber & Ingo Plag, (2013). The Oxford Reference Guide to
English Morphology. Oxford: Oxford University Press.
Hockett, Charles F. (1958). A Course in Modern Linguistics. New York: Macmillan.
29 Reflections on the Spread of Regular Inflection
to Simple and Derived Forms

29.1 Introduction
If you have a noun like mill and from that you make a compound like windmill,
then you expect that both mill and windmill will make their plural in the same
way, by adding an -s: mills, windmills. What may be less clear is why you
should have this expectation, except, perhaps, that it is clearly the majority
pattern. Does the -s on windmills arise because the plural of mill is mills, does it
arise because it is the plural of a compound, does it arise because complex
words are inflectionally more regular than simpler words, or is there something
else going on?
We can answer some of these questions fairly simply by giving examples. If
we have one tooth and two teeth, but one sabre-tooth and two sabre-tooths, then
it is clear that the plural of the compound does not follow automatically from
the plural form of the second element. If we have one wolf, two wolves and one
timber wolf, two timber wolves, it is clear that compounds do not automatically
take a regular plural, or that complex words are automatically inflectionally
more regular than simple ones. There is something more complicated going on.
Various explanations have been given which are supposed to cover at least
some of the cases. But if we look at what is going on here in detail, we find that
there are no rules, and scarcely even tendencies. Somehow, our morphological
rules (even the inflectional ones) have to be able to cope with irregularity.

29.2 Figurative Uses

When the personal computer was invented and was driven by an instrument
known as a mouse, it immediately became important to know what the plural
of mouse was. In the first years after the introduction of these tools, they were
often referred to as computer mouses. More recently, computer mice seems to
be preferred. The explanation that is often offered here is that figurative uses
of a word can be inflected regularly, even if the word usually inflects irregu-
larly. For this to be a reasonable explanation, there must be other supporting
examples, and there are.

225
226 Part VII Questions Involving Inflection

Bauer (2009) cites the following examples:

Todd . . . didn’t need wilder gooses to chase. (Kevin J. Anderson and Doug
Benson. 1995. Ill Wind. New York: Tom Doherty, p. 381)
If it turns out that Qatar is innocent . . . our gooses could be cooked. (John
Sandford. 2001. Chosen Prey. London: Simon & Schuster, p. 300)
And that’s when those louses / Go back to their spouses / Diamonds are a girl’s
best friend. (Jule Styne, from Gentlemen Prefer Blondes, 1953)
The difficulty with these examples is that they do not seem to cover all
the possible nouns. Although we cannot prove much from failing to find
a particular example, it would seem odd to say The new laws have no
tooths, where tooth is being used metaphorically, or to talk about
Chessmans. Yet Walkmans was the plural of Walkman for many speakers
when the device was first introduced. All we can say is that there are
examples where figurative usage seems to license a regular plural for
a normally irregularly inflected noun, certainly not that this is a rule of
English morphology.

29.3 Names
If you visited Disneyland or Disney World, you might find yourself in
a shop surrounded by stuffed toys, all representations of Disney characters.
Under such conditions, you might say something like Look at all these
Donald Ducks!, but what would you say if the character was Mickey
Mouse? Speakers do not know. Both Mickey Mice and Mickey Mouses
are found. We might argue that the regular plural is possible because these
are not real mice, but involve a figurative use of the word mouse. A more
likely explanation is that the relevant factor is that Mickey Mouse is
a name.
Pinker (1999: 171–2) cites the example of the Toronto Maple Leafs,
a hockey team, and sees this as a matter of seeing accumulations as different
from the set of individuals that make up the accumulation. But he also cites
the Timberwolves, a basketball team. He suggests it depends where you start:
if you start from Maple Leaf, you get Maple Leafs, but if you start from
Timberwolves, you can get one Timberwolf without having a team of
Timberwolfs. The argument seems to allow for whatever outcome you find,
which may be what is needed because there is certainly variation. Pinker
(1999: 162) cites an example from J.R.R. Tolkien’s The Fellowship of the
Ring, where characters argue as to whether the Proudfoot family should be
called the Proundfoots or the Proudfeet. This leads us to an alternative
 29 Regular Inflection in Simple and Derived Forms 227

explanation, namely that since a Proudfoot is not a foot, that is that the
compound is exocentric, it is the exocentricity that is crucial.

29.4 Exocentric Constructions

There is a now old-fashioned slang term meaning ‘policeman’, a flatfoot. But if
you have multiple police officers, what do you then call them if you want to use
that term? You can have one sabretooth, but usually two or more sabretooths,
even if that animal has many teeth. Some real examples are provided below,
some using irregular morphology, some using regular morphology.
She’s running around with lowlives. (Bill Pronzini. 2008. The Laughter of
Dead Kings. London: Constable, p. 35)
Native Americans have been living in this region for over ten thousand years.
The Cheyenne, Kiowa, Shoshone, Blackfeet and, more recently, the Crows.
(James Rollins. 2011. The Devil Colony. New York: William Morrow, p. 372)
Can you do others? . . . I mean animals, you know, or still lifes. Lives. (A.J.
Finn. The Woman in the Window. New York: William Morrow, p. 90)
“Isn’t that what they call cops”, he continued. “Flat foots?” (Nora Roberts.
1990. Public Secrets. New York: Bantam, e-version, p. 192)
Bauer (2009) gives examples from dictionaries – the OED and Gove
(1966) – with umlaut and fricative-voicing plurals. The examples are repro-
duced below.

broadleaf broadleafs
cloverleaf cloverleafs, cloverleaves
coltsfoot coltsfoots
cottonmouth cottonmouths /ðz/
flatfoot flatfoots, flatfeet
frogmouth frogmouths /θs/, /ðz/
goosefoot goosefoots (‘plant’), goosefeet (‘hinge, junction, etc.’)
lowlife lowlifes, lowlives
tenderfoot tenderfoots, tenderfeet
waterleaf waterleafs

Again, we see variation, with the regular plural a possibility, but not
a necessity in such environments.

29.5 Conversion/Zero-Derivation as a Circuit Breaker

The verb ring, as of a bell, is irregular and its past tense is rang. However, if
ring means ‘to put a ring round’ as in ring the city or ring a pigeon, then the past
228 Part VII Questions Involving Inflection

tense is regular, and is ringed. Kiparsky (1982: 12) draws attention to such
verbs, and suggests that when the verb is derived from a noun (by a process of
conversion/zero-derivation), the verb is always regular. That is, we have
a process which we can write as follows, where the Ø-symbol is used to
indicate the intervening conversion/zero-derivation, and where the output
verb is always regular:
ring]n + Ø > ring]v
Kiparsky cites verbs such as ink, link, wing as being particularly significant
here, since there is an opposing tendency to make verbs of this phonological
make-up irregular. Pinker (1999: 158) adds more examples, but seeks out
examples where the noun in the base of the process has a homophonous verb
which is usually irregular. It is irrelevant to the question whether to deflea (an
animal) sounds like flee which has an irregular past. On this basis, Pinker adds
denominal verbs like brake (‘apply the brakes’), spit (‘put on a spit’), string (as
in to string beans ‘to remove the strings from’ or to string a violin ‘add strings
to’) all of which have regular past tense and past participle forms in line with
the prediction. (There is a problem that the verb string as in string a violin also
comes from a noun, but has a past strung; in the seventeenth century the verb
was regular.) Note also the implication that the noun ride must come from the
verb ride, rather than vice versa, because ride is an irregular verb. The same is
true of the nouns bid, bite, break, drink and so on, and this probably feels right
in terms of the semantics of the noun–verb pair.
Kiparsky pushes this further. Most complex verbs (whether made up of
a preposition, which may be viewed as a prefix, and a verb or a noun and
a verb or of two verbs) inflect in the same way as the verbal (second)
element. Outgrow has outgrown because grow is the head of the new verb
and inherits its inflection; sunburn inflects like burn; cough-laugh inflects
like laugh. So understand has understood, like stand has stood, even
though there is no literal standing. The metaphorical use of stand in
understand is not enough to give it a regular inflection, and we do not,
except possibly in child language, get *understanded. Yet the past tense of
the verb grandstand is grandstanded (Kiparsky 1982: 12). But the verb
grandstand is based on the noun grandstand, and so it functions just like
ring (the city) mentioned above. Its past tense is grandstanded. In other
words, we get the following pattern, where the Ø resets the inflection to
regular inflection:
stand]v > stand]n + grand > grandstand]n + Ø > grandstand]v

Unfortunately, Kiparsky does not give other examples. But there are relevant
cases.
 29 Regular Inflection in Simple and Derived Forms 229

blue-light]n + Ø > blue-light]v past participle blue-lighted

free-fall]n + Ø > free-fall]v past participle free-fallen
highlight]n + Ø > highlight]v past participle highlighted or highlit
jacklight]n + Ø > jacklight]v past participle jacklighted
moonlight]n + Ø > moonlight]v past participle moonlighted ‘worked
extra hours’
or moonlit ‘illuminated by the moon’
searchlight]n + Ø > searchlight]v past participle searchlighted
spotlight]n + Ø > spotlight]v past participle spotlit or spotlighted
test-drive]n + Ø > test-drive]v past participle test-driven
The fact that there is not a wider set of appropriate verbs is unfortunate, and
may invalidate the data; but the examples here (all from the OED) do not
provide strong support for Kiparsky’s position.

29.6 Back-Formation
One potential cause, which does not seem to have received previous comment,
is back-formation. The verb creep out has a regular past tense and past parti-
ciple, creeped out, which might be considered odd in that the past tense and past
participle of creep are crept. But creep out does not seem to derive from the
verb creep, from which its meaning cannot be deduced; rather it seems to derive
from the noun creeps (as in It gives me the creeps) or the adjective creepy. Since
a final suffix is deleted to give creep out, we appear to have back-formation
here. The same might be true of moonlight (‘work extra hours off the books for
cash in hand’). Although the OED sees this as being derived from the noun
moonlight, an alternative analysis is that it comes from moonlighter, by back-
formation. This accounts for the meaning and for the usual regular form of the
verb in this sense, although moonlit is sometimes found in this reading. Since
back-formation is a relatively rare formation pattern, examples where this
hypothesis can be tested are hard to find, but it seems plausible.
Clury, who moonlit as a private investigator, had been loved by Jake. (William
Bayer. 1994. Mirror Maze. New York: Villard, p. 36)
Tracey would have match-maked her socks off. (Milly Johnson. 2019. The
Magnificent Mrs Mayhew. London: Simon & Schuster, pp. 260–1)

29.7 Jokes
The native words of English that take ablaut plurals are all monosyllabic, if we
take it that woman is etymologically a compound of man. The monosyllabic native
230 Part VII Questions Involving Inflection

words that end in <ouse> /aʊs/ are few in number, and most of them have irregular
plurals.

grouse grouse (‘game bird’)

grouse grouses (‘complaint’)
house houses (/haʊzɪz/ in standard English)
louse lice
mouse mice

Blouse (pronounced with /z/ in Britain) is French. Scouse is usually an

adjective, but can also be a noun; its origin in lobscouse is obscure, but may
be from Dutch. This leaves spouse, which, while it is also French in origin, no
longer has the French form.
Given these parallels, we would not necessarily expect a new monosyllabic
word with a spelling ending in <ouse> to take an umlaut plural, but the umlaut
is the most common plural within this very restricted paradigm. Finding spice
as the plural of spouse is thus, not entirely surprising, and the fact that spice has
a homophone meaning ‘condiment, like pepper’ is irrelevant. That spice is not
the usual plural of spouse is not entirely surprising, either. When it is used, it is
always a joke plural (like meeces as the plural of mouse). As meeces shows,
jokes do not have to conform to ordinary rules or patterns of formation, though
they are better jokes the nearer they come to conforming to some pattern:
conjugating think as think / thank / thunk is funny because it reflects an existing
pattern and clashes with other lexemes in different forms.
This is different from the other examples in that the regular plural is not the
one chosen for the joke, but we do find regular and irregular morphology in
alternation with each other.

29.8 Conclusion
Although the regular inflection of forms which might be expected to show
irregular inflection can be explained in many instances, by the factors mentioned
above, all of them show that we cannot see this as an absolute rule, only as
a tendency of different strengths in the different cases. These factors may allow
regular inflection, but they do not demand it. This allows us to see a pattern, but
does not seem to be compatible with the notion of a rule, since rule usually
implies that what the rule describes is, at the very least, always available.

Challenge
There are not many verbs like ring which are irregular in some meanings, but
regular when they are derived from a noun. How many can you find (including
 29 Regular Inflection in Simple and Derived Forms 231

compound verbs)? Are there any with an overt nominalization marker? Are
there any which become adjectives rather than nouns?

References
Bauer, Laurie. (2009). Facets of English plural morphology. In Ročenka textů
zahraničních profesorů / The Annual of Texts by Foreign Guest Professors.
Prague: Philosophical Faculty of Charles University Prague, 9–21.
Gove, Philip (ed.). (1966). Webster’s Third New International Dictionary. Springfield,
MA: Merriam-Webster.
Kiparsky, Paul. (1982). Lexical morphology and phonology. In Linguistic Society of
Korea (ed.), Linguistics in the Morning Calm. Seoul: Hanshin, 3–91.
OED. The Oxford English Dictionary [online]. oed.com
Pinker, Steven. (1999). Words and Rules. London: Weidenfeld & Nicolson.
30 Conclusion

30.1 Introduction
This book has covered a wide variety of topics, not necessarily always from the
same point of view. Some features will have been recurrent, though: a certain
scepticism about the way in which theories assume languages work, where
more data can show that things are not as straightforward as the theory suggests
(see, for example, Chapters 5 and 15); a desire to clarify what individual
technical terms mean (see, for example, Chapters 9 and 22), not necessarily
to be prescriptive, but to point out that the same term does not always mean the
same thing and that clarification may be helpful; an attempt to distinguish
between things that always happen and things that may happen (see
Chapter 29) because of a conviction that word-formation is extremely flexible
in the way that it works; an attempt to point out areas where accurate descrip-
tions are difficult. A handful of topics are so general that they cut across
individual presentations and recur, and some of these are considered below.

30.2 Rules
What does a linguistic rule look like? In the late 1950s a rule was phrased as an
instruction: given a particular input, this procedure is carried out, with an
inevitable output. But such a view cannot hold for long. There are too many
places where there are options as to what the output might be. Given a verb like
present, we can say He presented her with a medal, He presented a medal to
her, A medal was presented to her, She was presented with a medal and so on.
Knowing what the verb is, what was presented and the gender of the presenter
and presentee is not sufficient to determine the syntactic output. And while we
might attempt to narrow matters down by looking at the information flow, for
instance, a simple input–output view of a rule is of limited value. So then we
say that the output can be anything that is permitted by the rules. Active and
passive are both permitted, the subject of the passive sentence can be the person
to whom something is presented or the object which is presented, and so on. As
long as the output is licensed by a rule, it is fine. In either case, though, the rules

232
 30 Conclusion 233

are coercive. They restrict the possible outputs. Such a view makes sense: you
cannot do whatever you like in organizing words into sentences, there are
strong constraints, and errors can lead to incomprehensibility.
One of the things we learn by considering word-formation is that the outputs
need not be something that is predetermined. The word catanality (‘what cats
have instead of personality’) cannot be predicted from what we know about
extant patterns of English word-formation, yet is attested. The verb well-pay
(see Section 2.5) may not even be grammatical. Variability in outputs used to be
described in terms of variable rules (following the work of Labov), but variable
rules are not predictive, they are simply a descriptive statistical analysis of
observed patterns in the past, so they cannot be used to tell us which form can
arise, which means that, beyond a certain point, they do not help with what we
find in word-formation. Analogy (see Chapter 8) might work, but nobody
knows precisely how it works or how it can be instigated or constrained. At
this stage, I see analogy working within paradigms as being a very promising
way of making progress in word-formation and explaining the developments
that arise, but that, too, will have weaknesses.

30.3 Ludicity, Technicality and Productivity

Catanality and well-pay illustrate the very strong ludic element that pervades
the creation of new words, but of course there is also a lot of innovation in the
technical and scientific areas, as new words are invented for new entities, new
procedures, new theoretical concepts and the like. What these have in common
is that the formations are considered and consciously formed. There is a line of
thought in morphological studies, going back to Schultink (1961), which says
that consciously formed words do not illustrate productivity. But sometimes the
consciously formed words use the patterns that are also available for uncon-
scious formations so that the distinction is hard to apply. There are certainly
many unconsciously formed words produced by speakers, although the process
by which they operate is not clear – which brings us back to the whole question
of rules and analogies (see Sections 2.6 and 2.8).
These examples follow the normal anglophone tradition in word-formation
studies of looking at the patterns of word-formation and what they mean. In
the onomasiological approach to word-formation the view is different. As
neatly encapsulated by Carstairs-McCarthy (2010: 258), the onomasiological
approach turns
the world upside down. That is, the preoccupation of [scholars in this tradition] is not
with structures and the meanings associated with them, but rather with meanings and the
structures used to express them.
234 Part VII Questions Involving Inflection

This throws a different light on productivity, and may also imply that all
word-formation is consciously produced in that there is always a choice
between a derivative, a compound, a phrase, etc. as a way of naming something.
It also suggests a reason why rhyme (see Chapter 17) or alliteration may be
relevant forces in word-formation, though it still requires the speaker to know,
in some way or another, what patterns can be used to produce new words.

30.4 The Instability of Theoretical Constructs

Our inevitable experience of education leads us to believe that theoretical
constructs and processes associated with them are well defined and fixed. It
would probably be counterproductive for teachers to say we have a particular
entity which we can define in any one of four different ways with different
theoretical implications. Yet that is precisely the reality. It just requires
sophistication in dealing with the subject matter to be able to see the differ-
ences, and evaluate the different variants. For the morpheme, we find Mugdan
(1986) providing several different views of the nature of the morpheme, and
we can still ask questions about whether further constraints are needed. There
is less in the literature on the nature of the lexeme, but still no clear agreement
(see Chapter 9). And a process like blocking raises almost as many questions
as it provides answers (see Chapter 5). Theoretical constructs evolve, are
refined, are sometimes rejected, and even then refuse to go away. Some of this
is simply a by-product of our naming techniques. We do not want to go
through the hassle of creating a new term for every minor adjustment we
make to an existing idea, since a plethora of quasi-synonymous technical
terms is difficult to deal with, and sometimes, when scholars do create new
terms, it is not helpful. To my mind, this is the case for Allen’s (1978) term the
ISA criterion (a side door ISA door) to replace the established linguistic term
of ‘hyponymy’ and related terms.

30.5 The Difficulty of Dealing with Data

Nigel Fabb cites the late Morris Halle as having said that “data is overrated”
(https://linguistics.mit.edu/hallememories/). I’m not sure I agree, but
I certainly think that data is (to continue the number agreement) extremely
difficult to evaluate. The huge increase in the amount of linguistic data
available to us through the internet has advantages and disadvantages. On
the plus side, there is far more unedited material available for perusal than
there used to be, so that what can be found is far less uniform in its style level.
On the downside, every piece of data now has to be evaluated: who wrote it,
was the writer being serious (and does it matter), is a typographical
error involved (see Section 2.5), was a lapse of memory involved? In
 30 Conclusion 235

word-formation we have to ask whether attestation implies existence,

whether a new formation of a particular type indicates productivity, whether
a pattern has been misinterpreted or not, and so on. Some of the examples
I have cited in this book come from publishing houses that seem to be
particularly innovation-friendly, and it might not be clear to what extent
a particular editor is letting innovative uses through while another might
remove them. Of course, it might just be that I enjoy reading authors who
happen to publish with certain publishing houses. Whatever we might con-
clude, it is very comforting if the collected examples tell a coherent story,
even if it is one which is restricted to particular genres.

30.6 Insecurity about Processes

Although we can see repeated patterns in word-formation, and we can make
hypotheses about the way in which the members of the pattern arise, it is
remarkable how often we do not really know how the processes work. Back-
formation (see Chapter 21) is a classic example. We can see examples which fit
our prejudices about the process of back-formation, but there is so much vari-
ation in what we see that it is difficult to be sure how speakers actually process it,
either in production or in analysis. Conversion (see Chapter 20) is another such
process. A rather different case is provided by coordinative compounds (see
Chapter 22). We can find plenty of examples which the analyst can gloss as
instances of coordination, but it is not always clear whether this gloss is justified
in the broader scale of things. Is every instance that can be glossed with
a coordinating conjunction necessarily a matter of coordination, or is there
genuine ambiguity between glosses such as ‘A and B’ and ‘an A which is
a B’? And this is before we start to ask whether something is or is not
a compound, a matter where the answer may depend upon the definition we
happen to have chosen to work with – which may be inevitable, but casts doubt
on the ability of linguists to judge matters of classification. The insecurity might,
in the short term, be a good thing, in that it proves that scholars are trying to solve
a problem, but we must assume that the ultimate aim is to remove the insecurity.
Some do this by changing the definition (see Chapter 7), others do it by changing
our view of how to classify (see Chapter 7). Both can be helpful. What we need is
a coherent view of how the various parts of the whole fit together.

30.7 Fuzzy Grammar

This book is largely written from the point of view of someone who believes that
grammatical categories need to be clearly distinct from each other: we should
know whether we are dealing with a noun or an adjective, with inflection or
derivation, with a compound or something else. At the same time, dissatisfaction
236 Part VII Questions Involving Inflection

with such a system has been expressed because it does not seem possible – either
within English or across languages – to determine which side of the dividing line
we find ourselves on or even, in some instances, if there is a dividing line.
Whether we get to this point of view by dealing with prototypes, canonical
categories, problems with defining categories or some other route, the result is the
same: we may be clear about what the most central, typical, canonical examples
of a category (noun, word, compound, inflection) are, but we have no good way
of circumscribing the categories clearly when it comes to the more peripheral
members or potential members of the category. We have entered the area of fuzzy
grammar (Aarts et al. 2004). All the problems we have with categories and their
inconsistent behaviour push is in the direction of fuzzy grammar. As Sapir (1921:
38) puts it, “all grammars leak” (although people interpret that comment in
different ways). What makes this awkward is that it is hard to see just how to
deal – whether in a formal way or in a more intuitive way – with a system that
works in this way. Yet it is crucial that we should know how to do this if we are to
make progress with describing the way language works.

30.8 Summing up
It will probably be obvious that, although this book happens to deal with word-
formation, the concerns that arise when we consider word-formation recur in
almost any other study of linguistic structure. Calude and Bauer (2022), for
instance, deal in depth with cases where it is not clear what is happening to
certain English constructions diachronically, and where we do not know how
best to describe the structure of the language from a theoretical point of view.
The examples considered are different, but the overarching concerns are often
very similar. In this, the study of word-formation is simply the study of
linguistic structure in one specific field, and studying English word-formation
merely a way of restricting the scope of the enterprise.

Challenge
Choose any issue that has been raised in this book that you find to be particu-
larly contentious. Why is it contentious? If you wanted to provide a solution to
the issue, what information would you need? Does a definitive evidence-based
answer seem possible? How should your answer help define a fruitful approach
to the study of word-formation?

References
Allen, Margaret R. (1978). Morphological investigations. Unpublished PhD disserta-
tion, University of Connecticut.
 30 Conclusion 237

Bas Aarts, David Denison, Evelien Keizer & Gergana Popova (eds.). (2004). Fuzzy
Grammar: A Reader. Oxford: Oxford University Press.
Calude, Andrea S. & Laurie Bauer. (2022). Mysteries of English Grammar. New York:
Routledge.
Carstairs-McCarthy, Andrew. (2010). Review of Rochelle Lieber & Pavol Štekauer
(eds.), 2009. The Oxford Handbook of Compounding. Oxford: Oxford University
Press. Word Structure 3, 252–60.
Mugdan, Joachim. (1986). Was ist eigentlich ein Morphem? Zeitschrift für Phonetik,
Sprachwissenschaft und Kommunikationsforschung 39, 29–52.
Sapir, Edward. (1921). Language. London: Hart-Davis.
Schultink, Henk. (1961). Produktiviteit als morfologisch fenomeen. Forum der Letteren
2, 110–25.
 Index of Topics

abbreviation, 22 token, 47
ablaut, 184, 187, 189, 191, 229 type, 47
accidence. See inflection borrowing, 47, 123, 141, 148–151
acronym, 16, 22, 30, 37, 90, 91, 144 bound variant. See allomorph
affix, 5, 57, 64, 71, 84
apparent deletion of, 165 canonical form, 218
borrowing of, 151 canonicity, 6, 236
boundary, 122 Carroll, Lewis, 165
competing, 45 Chinese, 172
derivational, 107 clipping, 16, 30, 37, 145
homonymy of, 28, 84, 98 embellished, 145, 218
repetition of, 100 coercion, 155, 161
sequence of, 97 Cognitive Linguistics, 6, 14, 84, 161
affixation, 28, 29, 37, 59, 122 coinage, 45, 51, 64
cancellation of, 165 collocation, 81
to syntactic bases, 131–132 competition, 46–48
affixoid, 57 composition. See compounding
affix-telescoping, 97, 99 compound, 22, 36, 37, 58, 71, 110, 168,
alliteration, 142, 234 193–204, 209, 210–215, 225, 235
allomorph, 4, 5, 19 appositional, 104, 172
allomorphy, 4, 5, 20, 29, 127 clipping, 145
A-morphous Morphology, 13, 86 coordinative, 103, 168–172, 197, 199, 235
analogy, 53, 64–68, 85, 196, 233 co-participant, 170
analysability, 18, 99, 189 elative, 198
Andersen, Hans C., 158 genuine. See compound, proper
apophony, 16, 191 identical constituent or repeated element,
Arabic, 90, 218 100, 180
assonance, 142 proper vs improper, 195
availability, 29, 39, 43 synthetic, 37
tautological, 91–92
back-formation, 16, 22, 44, 146, 164–167, 201, translative, 170
229, 235 verbal, 104, 166
backronym, 144 compounding, 1, 30
bahuvrihi, 26 connotation, 83
base, 57, 64, 67 conscious formation, 65, 66, 233, 234
blend, 22, 105, 145 Construction Grammar, 14, 158
blocking, 28, 43–46, 48, 50, 51, 234 conversion, 1, 16, 22, 60, 155–162, 175, 201,
domain or type, 44 227–229, 235
individual or token, 43 partial, 162
inflectional, 43 coordination, 110
lexical, 44 natural, 168

238
 Index of Topics 239

corpus, 2, 20, 22, 81, 113, 139 hyponymy, 26, 58, 59, 81, 82, 103, 104,
cranberry morph. See morph, unique 171, 234
creativity, 64 hypostatization, 81

Danish, 27, 100, 133, 193, 194, 195, 196, 197, Icelandic, 193, 196, 202
198, 199, 200, 201, 202 idiom, 72, 87, 174, 176
decomposition, 111 infixation, 2
degemination, 123 inflection, 15–16, 30, 44, 69, 88, 209–217, 219,
derivation, 1, 15–16, 30, 88, 175, 215–216, 220 225–230
derivative, 1 initialism, 1, 16, 24, 69, 91
diachrony, 30, 54, 164 institutionalization, 17, 31, 43, 45, 51
dictionary, 2, 20, 22, 41, 77, 78 internal modification, 16
diminutive, 140–141, 188, 209 Irish, 141
Distributed Morphology, 6, 13, 83 Italian, 50, 140
Dutch, 93, 111, 138, 140, 193, 194, 195, 198, item-familiarity, 39, 65, 99
199, 200, 201, 209, 230
dvandva, 170, 172 Japanese, 2
Jerome, Jerome K., 28, 132
elsewhere condition, 43
endocentricity, 58, 102, 104 Kristofferson, Kris, 78
etymology, 19, 32, 164, 166, 188, 219
euphemism, 165 Latin, 69, 71, 86, 97, 148, 150, 151, 190, 193,
Exemplar Grammar, 14 199, 212, 220, 223
exocentricity, 26, 102, 103, 171, 194, 197, Lear, Edward, 81
227 lemma, 16
expletive insertion, 24 level ordering, 43, 119, 215. See also Lexical
extender, 5, 125 Morphology
lexeme, 1, 14, 16, 69–73, 88, 127, 134, 234
familiarity, 47 phrasal, 71
Faroese, 196, 198, 199, 201 lexical item, 72, 145
figure of speech, 6, 22, 31, 37, 48, 103, 160, Lexical Morphology, 6, 97, 119, 120, 212
169, 225 lexicalization, 17, 119, 128, 160
Finnish, 209 lexicography, 16, 31
formative, 71 linking element, 148, 195, 196, 197
French, 91, 92, 104, 133, 134, 146, 148, 149, listedness, 72
151, 165, 188, 190, 193, 204, 220, 230 listeme, 71, 72
frequency, 20, 23, 46, 47, 78, 122, 125, ludicity, 41, 52, 66, 99, 142, 164, 189, 233
139, 141
Frisian, 194, 196 malapropism, 78
fuzzy grammar, 235–236 Mandarin, 35
Māori, 16, 93
German, 24, 27, 30, 50, 107, 111, 140, 148, metonymy, 84, 103, 160–161. See also figure of
151, 176 speech
Germanic languages, 100, 133, 171, 193–204 Milton, John, 188
Gothic, 200 morph, 4, 5, 123
grammeme, 71 unique, 27
Greek, 69, 120, 138, 148, 149, 150, 151, 193, morpheme, 3–5, 35, 102, 139, 223, 234
199, 212, 219, 220, 223 morphology, 1, 19, 30–31, 51, et passim
expressive, 24
headedness, 6, 27, 58, 91, 100, 102–108, marginal, 24
110, 135, 158, 190, 193, 196, 202, 228. morphome, 128
See also right-hand head rule morphophonemics, 21, 119–130, 144
homonymy, 78, 84, 98 motivation
homophony, 150, 230 ablaut, 182
hyperonym, 102. See also superordinate rhyme, 180
hyphen, 133 multiple word expression (MWE), 17, 31, 48,
hypocoristic, 140 69, 134
240 Index of Topics

name, 61, 93, 226 Shakespeare, William, 120

Natural Generative Phonology, 19 slang, 78
neoclassical compound, 16, 148. See also neo- sound symbolism, 138
classical formation Spanish, 93, 104, 140
neoclassical formation, 36, 67, 84, 113, 148–151 spelling, 14, 15, 20, 24, 61, 121, 122, 123, 126,
no-phrase constraint, 28 129, 133, 144–146
norm, 50–51, 53–54, 128 split morphology, 15, 17, 210
Norwegian, 194 stratum, 119
novelty, 39–41 structuralism, 13
suffix and suffixation, 1, 2, 16, 22, 71, 97, 107,
onomasiology, 48, 233 113, 209, 218, 219. See also affixation
onomatopoeia, 138 superordinate, 25, 81, 82, 102
opacity, 18 suppletion, 43
Optimality Theory, 215 Swahili, 2
overabundance, 44 Swedish, 27, 133, 138, 196
synchrony, 17, 164
Panini, 13, 170 synecdoche, 103
paradigm, 6, 14, 53, 78, 165, 233 synonymy, 43
participle, 15, 21, 62, 127, 210, 222
particle, 174 Tariana, 209
periphrasis, 45 tautology, 90, 100
phonaestheme, 16, 138–140, 190 Tolerance Principle, 139
phrasal verb, 16, 24, 69 Tolkien, J.R.R., 27, 226
plural modifier, 213 transparency, 18
polysemy, 84 transposition, 36, 87
potentiation, 29
pre-emption. See blocking unitary base hypothesis, 28, 59
prefix and prefixation, 1, 2, 14, 16, 20, 22, 29, unitary output hypothesis, 28
99, 124, 218. See also affixation univerbation, 67
productivity, 2, 17, 20–21, 32, 39, 47, 50, 99, Usage-Based Linguistics, 14
132, 139, 151, 190, 233. See also avail-
ability, profitability velar softening, 122, 144
marginal, 20 Vietnamese, 172
profitability, 39
prototype, 5, 58, 218, 236 Welsh, 93, 209
West Greenlandic, 35
reanalysis, 166 Wodehouse, P.G., 78
recursion, 97–101 word, 1, 14–15, 35, 69, 168, et passim
redundancy, 20, 47, 90 actual, 51
reduplication, 2, 16, 107, 179–184, 219 grammatical, 71
repeated morph constraint, 28 lexical, 71
resonance, 139 manufacture, 146
rhyme, 141–142, 234 morphosyntactic, 14
right-hand head rule, 28, 102, 103, 105, 172, 196 phonological, 14, 112
root, 209 potential, 51–53
bound, 112, 113 virtual, 51
rule, 14, 17, 25, 31, 52, 120, 129, 139, 230, word manufacture, 52
232–233 Word-and-Paradigm, 86
occupied slot. See blocking word-based hypothesis, 29
redundancy, 19 word-based morphology, 86
variable, 233 word-class, 59, 84, 102, 156, 198
via-rule, 19 word-form, 1, 70, 71, 72
Runyon, Damon, 165 word-formation, 1, 13, 14, 15, 32, 48, 160, 175,
Russian, 50 232, et passim

Sanskrit, 170 Yiddish, 181

separable vs inseparable verbs, 201
separation constraint, 29, 87 Zipf’s law, 37, 223
 Index of English Word-Forming Elements

a-[a, adv, 29, 114 -en]dimin, 188

a-/an-[negative, 221 -en]participle, 219
-able, 98, 220 -en]pl, 93, 219
-ac, 189 -en]v, 18
-age, 50, 85, 87, 151, 220 endo-, 115, 221
-al]a, 84, 97, 98, 149, 220. See also -ial]a epi-, 218
-al]n, 84 -er]a, 28, 45, 120, 122, 215, 219
ambi-, 221 -er]dimin, 140
-an]a, 98, 125, 220 -er]n, 5, 17, 28, 29, 60, 64, 65, 71, 84, 87, 98,
-ana, 220 107, 108, 121, 148, 149, 164, 166,
-ant, 220 176, 220
ante-, 115, 221 -er]pl, 93
anti-, 112, 113, 114 -er]v, 19, 190
-ar, 84, 98, 220 -(e)rel, 188
arch-, 221 -ese, 220
-ary, 98 -esque, 220
-ate]v, 19, 29, 71, 97, 98 -ess, 84, 220
-ation, 83, 97, 98, 220 -est, 120, 219
-eth, 219, 220
be-, 114, 190, 221 -ette, 140, 220
by-, 84, 221 ex-, 150
exo-, 115
cardio-, 218 extra-, 221
-ce, 21, 45, 46, 47, 165
circum-, 221 -fold, 84
con-, 124 for-, 114
counter-, 60 -free, 112, 114
crypto-, 221 -ful, 98, 220
-cy, 21, 45, 46, 47
giga-, 221
deca-, 221 grand-, 124
dis-, 112, 124, 221
-dom, 30, 220 hemi-, 222
hepta-, 222
-ed]a, 127, 128, 214, 215 -hood, 84, 220
-ed]participle, 127, 190 hyper-, 222
-ed]past, 127, 211, 219 hypo-, 141
-ee, 64, 65, 83, 84, 220
-een, 141 -ial]a, 98
electro-, 218 -ian]a, 98, 125, 126
em-/en-, 4, 114, 221 -ian]n, 98, 123
-en]a, 160, 220 -ic, 17, 45, 47, 97, 98, 121, 123, 149, 220

241
242 Index of English Word-Forming Elements

-ical, 45, 47, 97, 99 post-, 84, 112, 114, 115

-ie, 140, 141 pre-, 84, 112, 114
-ify, 98, 123, 220 pro-, 112, 114
il-, 124 proto-, 222
im-, 124 pseudo-, 115
in-, 20, 53, 114, 124 psycho-, 113
-ing]a, 39, 214, 215, 222, 223
-ing]n, 166, 211, 214, 215, 222, 223 quadr-, 222
-ing]v, 39, 61, 90, 120, 219, 220, 222 quasi-, 115
-ion, 19, 50, 84, 97
-ish]a, 47, 57, 98, 100, 112, 121, 220 re-, 84, 100, 112, 150
-ish]v, 190 -ren, 219
-ism, 30, 98, 121, 123, 149 -ric, 30
-ist, 97, 121, 123, 149, 220
-istic, 97, 99 -s]3sg, 211, 219
-itis, 84 -s]adv, 84
-ity, 84, 98, 121, 123, 220 -s]dimin, 84, 140
-ive, 220 -’s]gen, 214
-ize, 84, 97, 98, 123, 150, 220 -s]pl, 213, 215, 216, 219, 225
-scape, 84
-kin, 140 -ship, 98, 220
shm-, 181, 219
-le]a, 189 socio-, 113
-le]v, 121, 189 -some, 220
-less, 84, 112, 113, 220 -st, 219
-let, 140, 220 step-, 222
-like, 47 -ster, 220
-ling, 140, 220 sub-, 99, 115, 124
-ly]a, 3, 35, 59, 124, 220 super-, 100, 112, 114, 115, 218
-ly]adv, 16, 30, 124, 127, 160, 210, 219 supra-, 112, 114
syntactico-, 218
maxi-, 114
mega-, 100, 151, 218 -t, 188
-ment, 21, 29, 98, 151 -t]2sg, 219
meta-, 99 -teen, 84
micro-, 100, 112, 114, 115, 141 -ter, 50, 220
mini-, 100, 112, 114, 115, 141, 218 -th]n, 17, 18, 19, 21, 98, 188, 220
mono-, 114 -th]ordinal, 210, 219
multi-, 222 theo-, 222

-n’t, 210 un-, 3, 20, 28, 100

nano-, 141 -ure, 29, 98
-ness, 84, 98, 113, 124, 127, 220
-nik, 220 vice-, 222

-o, 140 -ward(s), 220

-ock, 188 -wise, 112
-ola, 141 with-, 222
-ology, 149
-or, 98, 164, 220 -y]a, 121, 220
-ous, 84, 121, 150, 220 -y]n, 121, 150, 220
over-, 84 yotta-, 222

para-, 222 zetta-, 222

poly-, 114 -zza, 140