1

Word-Formation in English
An introduction

Mihaela Tănase-Dogaru
 2

Contents

Chapter 1 Words and morphemes

1. The domain of morphology

1.1. Inflection vs. Derivation
2. What’s in a word?
2. 1. Words as signs
2.2. Words, word-forms and lexemes
3. Morphologically complex words
3.1. Constituent signs. Simple versus complex signs
3.2. Roots, stems, bases
4. Reviewing questions and exercises

Chapter 2 On Complex Words

1. The morpheme
2. Allomorphy
2.1. Conditioning
3. Constituent signs. Simple versus complex signs
3.1. Constituent morphemes
3.2. Morpheme identification and classification
3.2.1. Substitution
3.2.2. Distributional Patterning
3.3.3 Morph classification. Morpheme identification
4. Reviewing questions and exercises

Chapter 3 Productivity
1. Problems with the morpheme
2. What is productivity?
3. Possible vs. actual words
4. The mental lexicon
 3

5. How do we measure productivity?

6. How do we constrain productivity?

6.1. Pragmatic restrictions
6.2. Structural restrictions
6.2.1. Phonological restrictions
6.2.2. Morphological restrictions
6.2.3. Semantic restrictions
6.2.4. Syntactic restrictions
7. Blocking
7.1. Homonymy blocking
7.2. Synonymy blocking
8. Reviewing questions and exercises

Chapter 4 Structuralist approaches to word structure

1. Structuralist methods
2. The Structure of Compounds
2.1. Compound classification
2.1.1. Endocentric Nominal Compounds
2.1.2. Exocentric compounds
3. Affixation
3.1. Prefixation
3.2. Suffixation
3.3. Identifying Immediate Constituents in multi-morphemic words
4. Conclusions
5. Reviewing questions and exercises

Chapter 5 Generativist approaches to morphology

1. Introduction
2. Early Generative Grammar and morphology
3. Lexicalism
 4

4. Aronoff (1976)
4.1. Word formation rules (Word Formation Rules)
4.2. Restrictions on Word Formation Rule
4.2.1. Syntactic restrictions
4.2.2. Semantic restrictions
4.2.3. Phonological restrictions
4.2.4. Morphological restrictions
4.2.5. Restrictions on the output
4.3. Conditions on Word Formation Rules
4.4. Adjustment rules
4.4.1. Phonological adjustment rules
4.4.2. Morphological adjustment rules
4.4.3. Syntactic adjustment rules
5. Reviewing questions and exercises

Chapter 6 Generativism and lexicalism. The X-Bar Schema. Affixes as Lexical Items
1. Introduction. The X-Bar Schema
2. Selkirk’s Model.
2.1. Selkirk’s Lexicon
2.2. Affixes in Selkirk’s model
2.2.1. Syntactic information
2.2.2. Semantic information
2.2.3. Phonological information
3. Neutral and non-neutral affixes. The Affix Ordering Generalization
4. Dual membership in class I and Class II.
5. Category-Changing Prefixes
6. Reviewing questions and exercise

Chapter 7 Generativism and lexicalism. The Permanent Lexicon and Percolation

1. Introduction
2. Similarities and differences
 5

3. The Permanent Lexicon (Lieber 1983)

3.1. Lexical entries
4. Lexical structure. The Feature Percolation Conventions
5. Conclusions
6. Reviewing questions and exercises

Chapter 8 Argument structure and affixation

1. Introduction
2. Subcategorization
3. Argument structures
3.1. Theta-roles
3.2. Arguments structures exemplified
4. Changes brought by affixation to argument structures
4.1. Passivization
4.2. Dative Shift
4.3. Incorporation
5. Morphological operations on argument structure
6. Reviewing questions and exercises

Chapter 9 Thematic inheritance

1. Introduction
2. The argument structure of affixes.
3. Syntactic approaches to inheritance.
3.1. Thematic affixes
4. Reviewing questions and exercises

Chapter 10 Nominalizations
1. Introduction
2. Nominalizations: events and results
3. One non-ambiguous event nominalization
4. Disambiguating between event and result nominals
 6

5. How do we analyze nominalizations?

6. Reviewing questions and exercises
Chapter 11 Compounds
1. Introduction. What are compounds?
1.1. What makes the difference between compounds and syntactic phrases
2. Towards a classification of compounds
2.1. A classification according to semantic and phonological factors
2.2. A classification according to morpho-syntactic factors
3. Root compounds
4. Synthetic compounds
5. Left-headed Compounds
6. ‘Outbreak’ / ‘breakout’
7. Reviewing questions and exercises
 7

Foreword

This book will hopefully be of use to anyone who is interested in morphology, in the relation
between morphology and syntax in particular and in how the two can shed light on one another.
The perspective is both chronological - from early structuralist analyses of morphemes and
word-formation processes to latter-day generativist theories of the morphology-syntax interface -
and synchronic, both theoretical and applied - each chapter contains a section of reviewing
questions and exercises with answers. I dedicate the book to all my students of morphology, who
more than once showed true dedication and a solid sense of humour.
 8

Chapter 1 Words and morphemes

1. The domain of morphology

Coming from Greek, the term ‘morphology’ is derived from the word ‘morphe’, which
means ‘form’ or ‘structure’ and ‘logie’, which means ‘account’ or ‘study’. in this respect, the
word ‘morphology’ can apply to any domain of human activity that is concerned with the study
of the structure or form of something (Baciu 1998: 1).
In linguistics, morphology is defined as that subdiscipline that accounts for the internal
structure of words. The question why anyone would need to possess knowledge of word
structure is answered by Baciu (1998:1) in the following terms:
“The knowledge of word structure is in many respects of a kind with the knowledge of
sound structure and sentence structure. It is part of what a language user has to know in order to
be regarded as a speaker of the language and, for that reason, it is part of that knowledge of
language which linguists regard as properly linguistic. Hence, it is something which linguistic
theory has to account for, in the same way that it accounts for knowledge of phonological
patterns or knowledge of syntactic structures” (Baciu 1998: 1)
Word-structure has been dealt with traditionally from two major points of view: one type of
complexity is thought to be due to the presence of inflections while the second type of
complexity of word structure is thought to be due to the presence of derivational elements (Baciu
1998: 4, Plag 2003: 14).
The term ‘inflection’ (conjugation and declension in traditional grammar, Plag 2003: 14)
covers grammatical markers for number, gender, case, person, tense, aspect etc. It is defined in
classical grammatical theory as ‘a change in the form of a word to express its relation to other
words in the sentence’ (Baciu 1998: 4).

(1) sings vs. sing - variants of ‘go’

students vs. student – variants of ‘student’

As the examples in (1) indicate, inflectional operations do not cause a word to change its
lexical category. In other words, inflectional operations, such as adding the plural -s or the
 9

present simple 3rd person -s, do not create new lexemes, but new word-forms, i.e. grammatical
words (Plag 2003: 14).

The term ‘derivation’ covers word-formation processes such as affixation, compounding and
conversion, whereby new words are formed from existing words.

(2) teach + er = teacher (a noun with the meaning of agent)

animate + ation = animation (a noun with abstract meaning)

In the examples in (2), suffixes are added to the stem, having as a result the derivation of
new words. In (2a), the agentive suffix -er is added to the verb ‘teach’, yielding the noun
‘teacher’; in (2b), the nominalizing suffix -ation is added to the verb ‘animate’, a process which
derives the nominalization ‘animation’.
Therefore, derivational processes typically induce a change in lexical category and add
further elements of meaning. Word-formation processes are restricted to the domain of the major
lexical categories: nouns, verbs, adjectives, (adverbs).

2.1. Inflection vs. Derivation

In what follows, we will be looking at a few major differences between inflectional affixes
and derivational affixes in English. The examples in (3), taken from Plag (2003: 14), illustrate
both inflectional and derivational affixes:

(3) a. Derivation b. Inflection

Worker (she) works
Useless (the) workers
Untruthfulness (is) colonializing
Interview (we) picked
Curiosity (the) children
Passivize John’s (house)
Terrorism Emily’s (job)
 10

When looking at the examples above, the following differences between the category of
inflection and the category of derivation emerge:

1. Inflectional morphemes encode grammatical categories, such as plural (4a) , person (4b) ,
tense (4c) or case (4d). In Plag’s words (2003: 15), ‘these categories are relevant for the building
for sentences and are referred to by the grammar’. The relation between a noun like ‘students’
and the predication ‘are studying’ is coded in terms of subject-verb agreement, which is
syntactically relevant and, therefore, the domain of inflection.

(4) a. workers
b. she/he works
c. picked
d. John’s

2. Derivational affixes can occur at either side of the base, while inflectional morphemes can
only be expressed by suffixes. There is no inflectional prefix in English. The only exception to
this rule comes in the form of irregular inflection, which however irregular it may be, it is still
not a prefix (5). Moreover, words like ‘workers’ and ‘colonializing’ in (3b) indicate that
inflectional morphemes always occur outside derivational morphemes and they close the word
for further affixation (*workers-hood, *colonializing-er) (Plag 2003: 15). In the examples in (3a),
‘un-truth-ful-ness’ shows that derivational affixes can occur inside other derivational affixes.

(5) Mouse - mice - irregular plural inflection

Ring - rang - irregular tense inflection

3. On of the most important differences between derivational and inflectional suffixes is

whether they change the lexical category of the base. The suffixes in (3a) change the lexical
category of the base. The suffixation of -less turns a noun into an adjective (6a). The suffixation
of -ize turns an adjective into a verb (6b). Inflectional suffixes, on the other hand, never change
the lexical category of the base. The suffixation of the plural -s to a noun like ‘worker’ yields yet
a noun. Similarly, the suffixation of the 3 rd person singular -s to a verb like ‘work’ yields a verb.
 11

However, there are instances of derivational affixes that do not change the lexical category
of the base. In the examples in (3), the suffix -ism illustrates the claim. It is added to a noun and
it yields a noun. To an example like -ism, we can add the majority of prefixes (see the examples
in 7):

(6) a. use - useless

b. passive - passivize

(7) de-bug a computer

non-scientific approach
post-apocalyptic movie

4. The process of derivation exhibits semantic opacity. For example, the meaning of the
word ‘interview’ in (3a) is not related to the meanings of ‘inter-’ and ‘view’. One cannot infer
the meaning of the whole based on the sum of the meanings of constituent parts, in other words,
the meaning is opaque or non-transparent. ‘Non-transparent formations are quite common in
derivational morphology but rare in inflection’ (Plag 2003: 15)

5. Inflectional categories tend to be fully productive while derivational categories show

restrictions as to the possible combinations. A productive suffix can be attached to any word of
the appropriate class. For example, the 3 rd person singular present suffix -s can (and must) be
attached to any verb; the past tense suffix -ed is attached to any (regular) verb. Similarly, plural
morphology in English is fully productive in that all countable nouns take the plural -s, with the
exception of a few irregular plurals, such as ‘oxen’, ‘mice’, etc). In contrast, the adjectival suffix
-ive cannot be attached to just any verb (8a); similarly, the adjectival suffix -al cannot be
attached to just any countable noun (8b):

(8) a. *walk-ive exhaust-ive (Plag 2003: 15)

*read-ive operat-ive
b. *desk-al department-al
*child-al phras-al
 12

The examples in (9a) illustrate a derivational paradigm (Bauer 1983: 11) while those in (9b)
illustrate an inflectional paradigm in Romanian, which is an examples of a language with rich
verbal and nominal inflection:

(9) a. National (6) b. Lucr-ez

Nation-hood Lucre-ezi
Nation-al Lucr-ează
Nation-al-ize Lucr-ăm
Nation-al-ist Lucr-ați
Nation-al-ist-ic
Nation-al-ity

3. What’s in a word?

The notion of ‘words’ is much more problematic than it might be assumed at first sight,
having benefited from multiples definitions, depending on the stance of the linguist.
One possible definition is that of the orthographic word, namely an uninterrupted string of
letters which is preceded by a blank space and followed by either a blank space or a punctuation
mark (Plag 2003: 4).
In (10), we count seven orthographic words, one of which is followed by a full stop.

(10) That is no country for old men. (from Matthews 1991:24)

However, things are sometimes more complicated. Take a look at (11) and try to determine
how many words there are:

(11) Benjamin’s girlfriend lives in a high-rise apartment building. (taken from Plag 2003: 5)
 13

The number of the words will depend on whether we consider the apostrophe as a
punctuation mark, which would turn ‘Benajmin’s’ into two words, on whether you consider the
hyphen to be a punctuation mark, in which case ‘high-rise’ represents two words and on whether
you sense a similarity between ‘apartment building’ and ‘girlfriend’, despite their different
orthography.
What other definitions can we find for ‘word’? If we take the phonological stance, the word
is defined as a unit in speech surrounded by pauses and carrying one instance of main stress, the
syllable carrying main stress being the most prominent one in the word:

(12) random
analysis
photography
understand

Closely related to the phonological criterion is the integrity criterion, stipulating that the
word is an indivisible unit into which no intervening material can be inserted. This intervening
material can only intervene at the edges:

(13) friend-s / *frie-s-nd

un-friend-ly / *friend-un-ly

The semantic definition of the word views it as expressing a unified semantic concept, a
definition which may run into problems once we start questioning what a unified semantic
concept really is.
Finally, from a syntactic point of view, words are considered syntactic atoms (the smallest
elements in a sentence) belonging to certain syntactic classes, known as parts of speech (nouns,
adjectives, verbs, prepositions), word classes or syntactic categories. One can test if a certain
item is a word by checking if it belongs to a certain word class. Fir instance, ‘the’ belongs to the
category ‘Determiners’ because it selects a noun and it always appears before the noun.
The chart in (14) summarizes the properties of words (apud Plag 2003: 8):
 14

(14) Properties of words:

a. Words are entities having a part of speech specification
b. Words are syntactic atoms
c. Words (usually) have one main stress
d. Words (usually) are indivisible units

2. 1. Words as signs

From the definitions above it was noticeable that words are not easily defined. However, one
important feature of words emerged, namely words are not just forms but meaningful forms,
which means that words behave as signs.
According to Ferdinand de Saussure (1916), the sign is the union of the semantics (the
concept, the signified, ‘signifie’) and the phonetics (the acoustic image, the signifier,
‘signifiant’).
The signified is not the ‘object of reality/the referent’ but a concept, therefore the
relationship between a linguistic expression and its ‘referent/object of reality’ is a mediated
(indirect) one (Baciu 1998: 10).

(12) concept (thought) Ogden and Richards (1931)

symbol(expression) referent (object)

Besides defining the linguistic sign as the association between a signifier and a signified, de
Saussure pointed out a few properties of the linguistic sign:

a. Arbitrariness.
There is no relation between the signifier and the signified, i.e. there is nothing in the nature
of the concept to justify the choice of a particular string of sounds. Conversely, there is nothing,
 15

for instance, in the sequence c-a-t that suggests the domesticated furry and cute animal. The
speakers of the same community use the same signifiers for the same signifieds as a matter of
linguistic convention.

b. Immutability
When looked at in synchrony, i.e. in a particular moment of social development, the relation
between the signifier and the signified cannot be altered, precisely by virtue of its highly
conventional nature

c. Mutability
When looked at in diachrony, i.e. along its history, the relation between the signifier and
signified can suffer specific mutations.

(15) ‘meat’ in Old English (‘mete’) used to designate ‘food’ or ‘an item of food’
‘minister’ in Middle English (‘ministre’) used to designate a ‘servant’
‘silly’ in Old English (‘seely’) used to mean ‘happy’

d. relative value
A sign has a relative value according to the contrasts it forms with other signs. According to
Saussure, the value of a linguistic sign does not come from its intrinsic signification, nor does it
come from its acoustic image. In fact, this arbitrarily chosen ‘signifier’ has no value, and the
concept – ‘the signified’ – does not have true value by itself because it exists within a language
system. Conceptually, linguistic meanings are not independent. Rather, they are dependent on
other linguistic signs within their language system to determine what they are. Therefore, the
actual idea or concept that the sign expresses can be understood by what it is not – by its
differences to other linguistic signs.
As Saussure explained “Concepts are purely differential and defined not by their positive
content but negatively by their relations with the other terms of the system. Their most precise
characteristic is in being what they other are not.” (1959:117)
For example, if something isn’t ‘hot’ or ‘cold,’ it must be ‘warm’. If something is ‘good,’ it
means it is not ‘average’ or’ excellent’.
 16

(16) ‘pork’ has the same referent as the French ‘pork’; however, it has a different value because
it contrasts with the English ‘sheep’.
‘plural’ has a certain value in a language in which there is also a dual number but a
different value in a language without it. (apud Baciu 1998, Săvescu 2014)

Hjelmslev (1961) further intoduced a division between substance and form. Each side of
the linguistic sign - content and expression - can in its turn be analysed into substance and form.
The linguistic sign is a system of relations, seen as the connection between the form of
expression (signifiant) and the form of content (signifie).

(17) Substance (sounds = phonetics)

Signifiant
Form (phonemes = phonology)
language = form
Form (concepts, sense)
Signifie
Substance (reality)

The relevance of the discussion about the nature of the sign will become evident in our
discussion on the internal structure of words. For now, let us return to simple words.

2.2. Words, word-forms and lexemes

We have established that ‘eat’ is a word. What are ‘eats’ and ‘eating’, then? Are they
different words or the same word? We will refer to things like ‘eat’, ‘eats’, ‘ate’, ‘eating’ as
word-forms.
Is there a concept that groups together all forms of the verb ‘eat’? The set of forms the verb
‘eat’ may have is referred to by the term lexeme. A lexeme is defined as a unit
of lexical meaning that exists regardless of the number of inflectional endings it may have or the
number of words it may contain. It is a basic abstract unit of meaning. (The Cambridge
 17

Encyclopedia of the English Language 1995:118). Therefore, ‘eat’, ‘ate’, ‘eats’ and ‘eating’ are
forms of the same lexeme, which we represent as EAT.

3. Morphologically complex words

3.1. Constituent signs. Simple versus complex signs

A simple sign is defined as a sign which is not analysable into two or more smaller
constituent signs.

(18) chair
neighbour
matter
brow
great
promise
discuss (Plag 2003: 10)

A complex sign is defined as a sign which is formed of and analysable into at least two
smaller constituent signs.

(19) a. employee
inventor
inability
meaningless
suddenness
unhappy
decolonialization (Plag 2003:10)
b. apartment building
greenhouse
team manager
truck driver
 18

blackboard
son-in-law
pickpocket

The examples in (19) contain morphologically complex words: ‘employee’ is analysable as

‘employ’ + ‘ee’, ‘inventor’ is analysable as ‘invent’ +’or’, etc. These words can be decomposed
into their smallest meaningful units, which are called morphemes.
The words in (18) cannot be decomposed into more meaningful components. They are called
monomorphemic simple signs.
Some morphemes can occur on their own; they do not need to attach to a base, nor do they
need an affix. Such morphemes are called free morphemes.

(20) employ
invent
chair

Some morphemes can never occur on their own; they always need to attach to a base. Such
morphemes are called bound morphemes.

(21) -ity
-ive
-ly

Some bound morphemes must always attach before the base (such as ‘-un’). They are called
prefixes. Some bound morphemes must always attach after the base (such as -ity, -ize, etc). they
are called suffixes. The term affix covers both suffixes and prefixes.
English also has bound roots, usually of Latin origin (see Plag 2003: 11):

(22) later- only appear in combination with the adjectival suffix -al
circul- only appears in combination with -ation, -ar, etc
approb- only appears in combination with -ate, -ation
 19

simul- only appears in combination with -ant, -ate, -ory

3.2. Roots, stems, bases

The part of a word to which an affix is attached is called a base. Roots are bases that can no
longer be decomposed into morphemes. The term stem is generally used for bases of inflections
but it can also cover derivational bases. According to Baciu (1998: 12), the term stem is defined
as being made up of a root morpheme plus any other non-root morpheme except the last added
morpheme.

(23) a. Manliness - ‘manly’ is the stem; ‘man’ is the root

b. manly - ‘man’ is both the root and the stem

Infixes are intervening affixes:

(24) abso-bloody-lutely

Compounding is a process that combines two bases, without the need for affixes.

(25) water-lily
school-boy

Conversion, zero-suffixation or transposition is a process whereby a lexical category

comes to be used as a different lexical category without recourse to affixation:

(26) Give me water.

John waters his flowers every day.

Apart from affixation, compounding and conversion, which are sometimes referred to as the
major word-formation processes, there are other processes that involve deletion of material of the
base.
 20

Truncation or clipping deletes a part of the base. It can be found with proper names (25) or
with common nouns (26):

(25) Aaron - Ron

Elizabeth - Liz
Mike - Michael
Trish - Patricia

(26) condominium - condo

disco - discotheque
laboratory - lab
demonstration - demo

Sometimes, especially with proper names, truncation and affixation work together to yield
diminutives:

(27) Amanda - Mandy

Andrew- Andy
Patricia - Patty
Robert - Robbie

Blends are amalgamations of parts of different words:

(28) smoke + fog = smog

breakfast + lunch = brunch

Acronyms combine the initial letters of compounds or phrases into a pronounceable new
word:

(29) NATO
UNESCO
 21

We will conclude this section by going back to the domain of morphology and
conceptualizing it as in (30), apud Plag (2003: 17)

(30) Morphology

Inflection Word-formation

Derivation Compounding

(31) Derivation

Affixation Non-affixation

Prefixation Suffixation Infixation Conversion Truncation Blending

 22

Reviewing questions and exercises:

Exercise 1.
Explain and illustrate the notions of:
a. Orthographic word
b. Word-form
c. Lexeme

Start from the following examples:

a. I have been to Paris twice.
b. I will never be afraid.
c. Any human being would frown upon such a deed.

Exercise 2
Define the following terms and give three examples for each of them:
a. Morpheme
b. Prefix
c. Suffix
d. Affix
e. Compound
f. Root
g. Truncation

Exercise 3
Identify the morphemes in the words listed below and specify whether they are free or bound
morphemes, prefixes, suffixes or roots:
a. Unbelievable
b. Intersperse
c. Teachers
d. Numerous
e. Living-room
 23

f. Computerize

Exercise 4
a. List all morphemes in the sentence below. How many morphemes have you found?
b. List all complex words and state which type of morphological process (inflection, derivation,
compounding) has affected them.

High-school students sometimes don’t have problems with friendly advice.

Exercise 5
Define and illustrate the main differences between inflectional and derivational affixes.

Exercise 6 (Plag 2003:19)

Determine whether the adverbial suffix -ly in English is a derivational or an inflectional suffix,
by applying all the criteria discussed in the lecture.

Slowly, aggressively, smoothly, hardly, rarely, intelligently, purposefully

 24

Chapter 2 On Complex Words

1. The morpheme

In the previous chapter the notion of morpheme was introduced. It was defined as the
smallest meaningful unit.

(1) teach-er
-er has the meaning ‘agent’

Traditionally, the morpheme is conceived of as a unit of form and meaning.

(2) un-happy

‘un-’ is an entity which consists of the meaning and the sounds or letters expressing the
meaning (Plag 2003: 20). Therefore, the morpheme is a sign, in Saussurian terms.
Lyons (1968: 181), Matthews (1974: 11), Bauer (1983: 14) all define the morpheme as the
minimal unit of grammatical analysis.

(3) un-touch-able-s (Bauer 1983: 13-14)

Each of the segments has its own form, its own meaning and its own distribution. ‘un-’ has
the phonological form /ʌn/ and a meaning of negation; ‘touch’ has a fixed phonological form and
a fixed meaning; ‘-able’ sometimes appears as ‘-ible’, has a fixed meaning ‘something that can
be V-ed’; ‘-s’ has a range of phonetic forms but a constant meaning of plurality.

2. Allomorphy

Morphemes do not occur as forms in a language, in a similar fashion to lexemes and

phonemes. Morphemes are abstract elements of analysis and what occurs is a phonetic form
which realizes the morpheme. (Bauer 1983: 15).
 25

(4) This cow is eating grass.

These cows are eating grass.

The examples in (4) show that ‘cow’ and ‘cows’ have different distributions. This difference
happens because ‘cows’ is made up of two morphemes: {COW} and {plural}.

(5) The sheep is eating grass.

The sheep are eating grass.

Although the word-form ‘sheep’ is identical in (5), agreement indicates that ‘sheep’ is
singular in (5a) and plural in (5b). Therefore, ‘sheep’ in (5b) is made up of two morphemes:
{SHEEP} and {plural}.
Therefore, the morpheme {plural} was shown to have at least two variants. We will call
these variants allomorphs.
When the phonetic or orthographic strings which realize morphemes are segmentable, they
are called morphs. A morph ca be defined as a segment of a word-form which represents a
particular morpheme (Matthews 1974: 83, Bauer 1983: 15).

(6) was

In (6), we can recognize the morphemes {BE}, {preterite} and {singular} but the word-form
cannot be segmented into morphs.

(7) tree-less

In (7), we can recognize the morphemes {TREE} and {-LESS}, which map the way the
word-form is segmented into morphs. In other words, there are two morphs which realize the
morphemes {TREE} and {-LESS}.
 26

When different morphs represent the same morpheme, the phenomenon is called
allomorphy, and the morphs are called allomorphs. Let us take another example, from Plag
(2003: 27).

(8) The indefinite article in English

[ǝ] question [ǝn] answer
[ǝ] book [ǝn] author
[ǝ] fence [ǝn] idea
In isolation: [ei]

The examples in (8) show that there are three distinct realizations of the indefinite article: [ǝ],
[ǝn] and the stressed variant [ei] in isolation. These are different morphs that represent the same
morpheme and are called allomorphs.

2.1. Conditioning

When do we know when to use [ǝ], [ǝn] or [ei] for the indefinite article? By looking at the
examples, we induce the rule that [ǝ] is used before a consonant, [ǝn] is used before a vowel and
[ei] is used when nothing follows. It is, therefore, the sound structure that conditions the
distribution of the allomorph. This is called phonological conditioning.
Let us take another example (9):

(9) The plural morpheme -s in English

Books dogs churches
Cats bags horses
Deaths boys badges

In English, the plural morpheme has three distinct phonological realizations. It is realized as
/iz/ after sibilant consonants (as in ‘horses’ and ‘churches’); it is realized as /s/ after any other
voiceless obstruents (as in ‘books’ and ‘cats’); and it is realized as /z/ anywhere else (as in ‘dogs’
 27

and ‘bags’). These forms depend on the phonetic environment and are, therefore, phonologically
conditioned. (Bauer 1983: 15)
Let us take another example of an English plural:

(10) Ox - oxen

OX is the only lexeme which makes its plural by adding -en. This variant of the plural
morpheme is conditioned by the lexeme. It is, therefore, lexically conditioned.
There is a third type of conditioning, called morphological conditioning. It is exemplified
in (11):

(11) Explain maintain courage (Plag 2003: 28)

Explanation maintenance courageous
Explanatory

The pronunciation of the base EXPLAIN varies according to the kind of suffix attached to it.
The attachment of ‘-ation’ causes three different effects. First, stress is shifted from the second
syllable of the base plain to the first syllable of the suffix. Second, the second syllable of the base
is pronounced [Ɛk] instead of [ik]. Third, the first syllable of the base receives secondary stress.
The attachmment of ‘-atory’ to explain leads to a different pronunciation of the second
syllable of the base: [æ] instead of [ei].
The attachment of ‘-ance’ and ‘-ous’ to ‘maintain’ and ‘courage’ leads to vowel changes.
In all the cases in (11), there is more than one base allomorph, and the appropriate allomorph
is dependent on the kind of suffix attached to it. In this case, we can talk about morphological
conditioning.

3. Constituent signs. Simple versus complex signs

In the previous lecture, a simple sign was defined as a sign which is not analyzable into two
or more smaller constituent signs, while a complex sign was defined as a sign which is formed of
and analyzable into at least two smaller constituent signs.
 28

The question we turn to now is how we decide if an element is a simple or complex sign.
Following Baciu (1998), we will discriminate between simple and complex signs by using the
commutation test and the following three theoretical statements:
(i) a is a constituent sign of a complex sign X if and only if in at least one other context apart
from X there occurs an element which has the same form and the same sense which a has in X;
(ii) a can be identified as a constituent sign of a complex sign X if and only if all other
constituents of X can be identified as constituent signs;
(iii) the sense of a complex sign is at least in part predictable from the senses of the minimal
signs which form it (i.e. the sense of the complex sign is a function of the senses of the
constituents). (Baciu 1998: 15)
Therefore, a constituent sign always appears with a constant sense and a constant form in all
the complex signs in which it can be identified as a part.
Let us take the following example (Baciu 1998: 16):

(12) cranberry reassure

boysenberry reconsider
huckleberry rebuilt

(13) reassure
reconsider
rebuilt

In the examples in (13), by commutation or substitution, we uncover the element re, which
has a constant sense of its own, namely ‘again’. Therefore, re in (13) has a constant sense and a
constant form in all the contexts in which it occurs. All the other constituents uncovered by
commutation in (13): ‘assure’, ‘consider’, ‘build’ also have a constant sense and a constant form.
In conclusion, ‘re’ and ‘assure’ qualify as bona fide signs. All the elements under (13) are
complex signs.
Now let us look at the following examples:
 29

(14) cranberry
boysenberry
huckleberry
gooseberry
raspberry
strawberry

In (14), by applying commutation, we identify the sequence ‘berry’ in all the elements,
which has a constant form and sense (kind of fruit). We need to ask whether ‘cran’, ‘boysen’,
‘huckle’, ‘goose’, ‘rasp’ and ‘straw’ have constant sense? By asking this question, it’s easy to see
that it’s not ‘cran’ that has the sense of ‘acid berry of the heath family’ but the whole expression
‘cranberry’. Therefore, though we can identify a form ‘cran’, ‘boysen’, ‘huckle’ in ‘cranberry’,
‘boysenberry’ and ‘huckleberry’, respectively, we cannot identify a sign ‘cran’, ‘huckle’ and
‘boysen’ since condition (i) is not met.
Consequently, the elements ‘cran’, ‘boysen’, ‘huckle’, ‘goose’, ‘rasp’, and ‘straw’ do not
qualify as signs since they have no clearly identifiable sense. Moreover, if ‘cran’, ‘boysen’ and
‘huckle’ are not signs neither is ‘berry’ a sign in these elements. Thus, ‘cranberry’, ‘huckleberry’
and ‘boysenberry’ are simple signs.

3.1. Constituent morphemes

In the previous section, we saw that the elements ‘cran’, ‘berry’, ‘huckle’ and ‘boysen’ have
constant form, which makes them morphemes, but not signs.
The ‘morpheme’ is therefore a recognizable unit of analysis (Harris 1955, quoted in Baciu
1998:16) defined as ‘an element endowed with constant form and certain constant phonological
and morphological properties’ but with ‘no sense independently of the individual words in which
they occur’.
Constituent signs are also morphemes but not all morphemes qualify as signs. Let us take
anotther example (Baciu 1998: 16)

(15) a. understand
 30

withstand
b. undertake
partake

Since we cannot relate the form ‘-stand’ in ‘understand’ or ‘withstand’ with the free
occurrence ‘stand’ we can notice that both items have the same form /stand/, but the free
occurrence ‘stand’ has a number of senses that cannot be related to any of the interpretations of
‘-stand’ in (15). The same line of reasoning applies to the items ‘with’, ‘under’, ‘take’ etc.
Therefore, the examples in (15) are simple signs but bimorphemic.
Harris (1955, quoted in Baciu 1998: 18) defined the morpheme as ‘a sequence of phonemes
endowed with constant form and which evince an  degree of independence’.
Consequently, we can distinguish between morphemes that:

(i) show a low degree of independence;

(ii) show partial independence;
(iii) show a high degree of independence.

(i) Morphemes exhibiting a low degree of independence occur in a limited, closed number of
words (the ‘cranberry’ type). These morphemes never qualify as signs since they have no sense
of their own independently of the words in which they occur. ‘Cranberry’ type items are simple
signs but complex in point of morphemic structure.

(16) cranberry
boysenberry
huckleberry
gooseberry
raspberry
strawberry

(ii) Morphemes that show partial independence are represented by the so-called class of Latinate
verbs (Aronoff 1976):
 31

(17) refer remit resume receive reduce

defer demit presume deceive deduce
prefer submit consume conceive induce
confer permit subsume preceive conduce

The word in (17) are complex forms made up of constituent morphemes, i.e. constituents
with a constant form and constant phonological properties. They exhibit a greater combinatorial
independence than the ‘cranberry’ morphemes but their independence is not unlimited. The
morphemic segments ‘re-’, ‘-mit’, ‘de-’, ‘-sume’, ‘-ceive’ etc. are independent of one another but
not independent of the set as a whole, i.e. the combinatorial possibilities are limited to the
members in the set.

(iii) Morphemes with a high degree of independence are represented by morphemic segments
like ‘re-’ in ‘rebuild’ or ‘mis-’ in ‘misunderstand’ or ‘misuse’, or free morphemes like ‘boy’,
‘take’, etc.
These morphemes have a constant sense independent of the particular words or
combinations in which they occur, and always qualify as signs.

(18) rebuild
reassure
reconsider
misunderstand
misuse
misguide
boy
student
book
 32

3.2. Morpheme identification and classification

There are two structuralist procedures for setting up morphemic segments (Harris 1955
quoted in Baciu 1998: 19): substitution and distributional patterning.

3.2.1. Substitution

According to the substitution test, a phoneme sequence is decomposable into morphemic

segments if and only if one part of the respective sequence occurs without the other part in the
same total environment.

(19) That’s our roomer A B

room A D
London C D
Londoner C B

The examples in (19) show that the tentative morphemic segments ‘-er’, ‘room’, ‘London’
may occur without the other in the same total environment: ‘That’s our ____’.
‘if in the total environment X, the combination AB occurs and AD occurs and CB occurs
and CD occurs where A,B,C,D are each phonemically identifiable sequences, then it is possible
to recognize A, B, C, D as being each of them (tentatively) discrete morphemic segments in the
environment (context, frame) X’ (Baciu 1998: 21).
In (19), this translates as:

a. X is “That’s our____”
b. AB is roomer (where B is -er)
c. AD is room (i.e. D is zero)
d. CD is London (where D is zero)
e. CB is Londoner (where B is -er)
 33

If we apply substitution to the following to the environment in (20):

(20) a. That’s our roomer

room
hammer
ham
b The stars were gleaming.
glittering.
glowing.

In (20 a) substitution identifies the tentative morphemic segments: ‘room’, ‘ham’ and ‘–er’;
‘hammer’ would be analyzed as made up of two morphemic segments: ‘ham’ and ‘-er’.
In (20 b) the application of substitution would yield for each element three morphemic
segments because ‘gl-’ and ‘-ing’ occur independently in other combinations in the same total
environment.
This problem is known as ‘oversegmentation’. What we need to do is back up the results by
applying distributional patterning.

3.2.2. Distributional Patterning

Distributional patterning is a test that adds to substitution the notion of varying the context.
A particular independent phonemic sequence is a morphemic segment only if it has identical
relations to many other environments.

(21) The announcer is no good

announcement
governer
government
assigner
 34

assignment

What happens is the following: by substitution, the phoneme sequence ‘announcer’ is made
up of two morphemic segments since ‘announce’ may be substituted by ‘govern’, ‘assign’,
‘reinforce’ and ‘-er’ can be substituted by ‘-ment’.
However. the results of the application of substitution are to be validated by the procedure of
distributional patterning, which requires the occurrence of the identified morphemic segments in
other contexts as well.
If we find that there is any other general characteristic that distinguishes the group
‘announce’, ‘govern’, ‘assign’ as compared to other sequences which occur before ‘-er’, we also
find that all the morphemic segments that occur before ‘-er’ also occur in the contexts below:

(22) a. I cannot ________ it

b. Let’s try to _______.
c. While ______ ing it.

Therefore, the phonemic sequence ‘ham’ in ‘hammer’, which substitution alone would
analyse as made up of two morphemic segments ( ham+er), is actually made up of just one
morphemic segment, alongside elements like ‘anger’, ‘spider’, ‘quiver’, ‘stammer’, ‘shimmer’,
‘glimmer’, etc., (Baciu 1998: 22):

(23) The *hamment is no good

I cannot *ham it
Let’s try to *ham it
While *hamming it

In conclusion, morphemic boundaries in an utterance are determined not on the basis of

considerations interior to the utterance but on the basis of comparison with other utterances.
 35

3.3.3 Morph classification. Morpheme identification

We have seen that the distinction morph-morpheme overlaps the distinction variant-invariant
and is patterned on the distinction phone-phoneme.
But is phonemic resemblance alone enough for two morphs to belong to the same invariant
morpheme?

(24) do the phonemically identical morphs -er in ‘writer’ and and -er in ‘larger’ belong to one
and the same invariant morpheme [ER]?

The grouping of morphemes is based on the notion ‘partial resemblance of morphs’. This
notion refers to two distinct properties:

a) resemblance in phonemic form;

b) resemblance (i.e. identity) of context of occurrence.

(25) governor
assigner
announcer

The words in (25) are made up of two morphs: a verbal base ‘govern’, ‘assign’ and
‘announce’ and a suffix ‘-er’. How do we that the three occurrences of ‘-er’ can be classified
under one single morpheme [ER]?
The answer lies in the fact that the stems to which they attach belong to the same
distributional class, namely the class traditionally identified as verb. Therefore, the three
occurrences of ‘-er’ are variants of the same invariant morpheme ER. Moreover, the complex
elements ‘governor’, ‘assigner’, ‘announcer’ share the same context of occurrence.

(26) stronger
larger
 36

The words in (26) are again made up of two morphs: an adjectival base ‘strong’ and ‘large’
and a suffix ‘-er’. The question is now the following: can this ‘-er’ be grouped together with the
‘-er’ in ‘teacher’, ‘assigner’, ‘governor’?
Given (ii), we can now understand why -er in stronger and -er in teacher cannot be grouped
under one single invariant morpheme. The two phonemically identical morphs are not variants of
the same morpheme because they have distinct contexts of occurrence:

(27) This /X/V #er/N is perfect (Baciu 1998: 25)

(28) The /X/A #er/A man.

Applying the two principles we mentioned: phonemic resemblance and identity of context of
occurrence, the following criteria for morph classification can be set up.

(i) Morphs which are phonemically identical and have also identical environments constitute
repetitions of one and the same invariant morpheme. This is the case of ‘-er’ in ‘governor’,
‘announcer’, ‘assigner’.

(ii) Morphs that are phonemically different but display absolutely the same distribution i.e. have
identical contexts of occurrence, are classified under one single morpheme and are said to be free
variants of the same morpheme. The two free variant morphs represent two phonemic
non-conditioned realizations of one invariant morpheme. A well known example is economics
with the free variants: /ekonomiks/ vs. /iyknomiks/

(iii) Morphs which have phonemically different forms and the same general environment, except
that the environment of one morph always contains some element X which does not occur in the
environment of the other morphs and which accounts for the different phonemic form of the
respective morpheme, belong to one invariant morpheme.
Such morphs are said to be in complementary distribution and are known under the name of
allomorphs. As said before, the different phonemic realization of the morphs can be conditioned
morphologically or phonologically.
 37

(iv) Morphs that do not have the same phonemic form and, moreover, do not have the same
distributional contexts are instances of different morphemes.
 38

Reviewing questions:

1. State the definition of the sign in Saussurian terms

2. Saussure characterized the sign as being arbitrary and having relative value. Comment on this.
3. Define the notion of simple and complex linguistic signs, constituent sign.
4. Give the definition of morpheme, allomorph, conditioning.
5. Mention the basic properties of inflectional and derivational morphemes.

Exercises

1. Are the following simple or complex signs?

(1) Blackberry
Blueberry
Gooseberry
Butterfly
Withstand
Understand
Undergo
Undertake
(2) Misguide
Misinform
Misplace
Mislead
(3) Mistake
Mischief
Mistress
(4) Discolour
Disclaim
Dismantle
Disagree
 39

Disable
(5) Disciple
Discreet
Discuss
(6) Retake
Repay
Rearrange
Retry
Rephrase
Reformat
Restate
(7) Receive
Refer
Resume
Rest
Retain
Remain
(8) Recover
Refuse
Remove
Report
Refrain
Regard

2. Identify the suffixes in the underlined words. To what word-class (lexical category) do the
words to which the suffixes are added belong, and what word-class results? Decide whether the
suffixes are inflectional or derivational. Justify your decision.

(9) I ducked He was sheepish He ducks

Two ducks three ducklings
 40

He is humourless You are ducking the issue

3. State the lexical category of the base to which ‘en-’ is attached. What is the lexical category of
the new word? What is the meaning (or meanings) of ‘en-’ in these words? Is there one ‘-en’ or
are there two?

(10) Base New Word Base New word

cage en-cage noble en-noble
large en-large rich en-rich
robe en-robe rage en-rage
danger en-danger able en-able

4. What is the plural morpheme in English? What is the allomorph of the plural morpheme that
occurs in the group of words below? Is the choice phonologically, morphologically or lexically
conditioned?

(11) addenda, data, strata, media, desiderata

stimuli, radii, fungi, alumni
asses, fishes, badges, leeks, moths, cups, bards, mugs, rooms, keys, shoes

5. Study the following examples. Do the underlined morphemic segments below belong to an
invariant morpheme? Explain.

(12) Inexpensive
Impossible
Illogical
Irrational
Irregular
Ingenious
Irresistible
Immobile
 41

Inconsistent
Incomprehensible
Illegal
Illiterate
Incompetent
Irresponsible
Indifferent
Inevitable

(13) Baker
Teacher
Driver
Survivor
Beggar
Governor

6. Multiple affixation: identify the root morpheme and the affixes that make up the complex
words below:

(14) diction, dictate, dictatorial, contradict, contradictory, contradiction, contradictoriness,

benediction

7. Take the free morpheme ‘nation’ and add as many prefixes and suffixes as you can.

8. Examine carefully the following sentence:

(15) He had begun to read the novel a few days before. He had put it down because of some
urgent business conferences, opened it again on his way back to the estate by train; he permitted
himself a slowly growing interest in the plot, in the characterisations. That afternoon, after
 42

writing a letter giving his power of attorney and discussing a matter of joint ownership with the
manager of his estate, he returned to the book in the tranquillity of his study which looked out
upon the park with its oaks. (Julio Cortazar, Continuity of Parks)

List five free and three bound morphemes that occur in the sentence
List three functional morphemes in the sentence

9. Explain the notion of morphologically-conditioned (base) allomorphy starting from the

following examples (Plag 2003: 42)

(16) Active - activity

Affect - affection
Curious - curiosity
 43

Chapter 3 Productivity

1. Problems with the morpheme

When we discussed the notion of morpheme, we assumed that, more often that not, there is
a clear mapping of form and meaning. However, there are phenomena that may prove to be
problematic for our notion of morpheme.
The first phenomenon to be discussed in this vein is conversion. Conversion is a process by
means of which words are derived from other words without any visible marking:

(1) to walk - a walk

to throw - a throw
water - to water
book - to book (Plag 2003: 22)

In the case of conversion, we could speak of a zero-morph, which is added to the base to
derive a new word. For example, the verb ‘water’ is form from the noun ‘water’ by adding to the
base noun a zero-morph with the meaning ‘apply X’. (Plag 2003: 22)
The second phenomenon that poses problems to an analysis of morphemes as perfect
form-meaning mappings is truncation. In this case, it is not clear what the morph is and what
semantic difference there is between truncated and non-truncated forms. According to Plag
(2003: 22), the most natural way to account for truncation is to assume that the morph is the
process of deleting material. As for the difference in meaning, truncated forms seem to signal
familiarity of the speaker with the entity referred to by truncated forms.
A third problematic phenomenon is the existence of vowel alternations that add some
meaning to the base, as in causative alternations (2), paste tense formation (3) and plural marking
on nouns:

(2) to fall - to fell

(3) sing - sang
take - took
 44

ring - rang
(4) foot - feet
goose - geese
mouse - mice

Although there is no separate affix that could convey the causative meaning, as is the case
with ‘pur-ify’, for instance, it could be argued that the vowel carries the causative meaning in (2),
the past tense meaning in (3) and the plural meaning in (4).
A fourth problematic phenomenon is the existence of discontinuous morphemes.
Sometimes, there is more than one form for one meaning:

(5) the progressive ‘be - ing’

the perfective ‘have - en’

A last phenomenon that poses problems to an analysis of morphemes as ‘things’ is the

existence of meaningless morphs:

(6) infer
confer
prefer
refer
transfer

Even if the sequence ‘-fer’ qualifies for the status of morph, it carries no meaning for
speakers of Modern English. They are Latinate verbs and the morph ‘-fer’ has a low degree of
independence. In Latin, the root ‘ferro’ carried the meaning ‘to carry’.
We have looked at a number of problems that are faced by an analysis of morphemes as
minimal units of analysis mapping form and meaning. The notion of morpheme has been
enriched with an interpretation of morphemes as processes rather than as ‘things’.
In what follows, we will analyse the notion of productivity in word-formation.
 45

2. What is productivity?

The term productivity refers to the property of an affix to be used in order to coin new
complex words. Productivity varies with different affixes - some affixes are productive, other are
less so.
For example, the suffix ‘-th’ can only attach to a limited number of words (7) but it cannot
attach to words beyond that set. :

(7) long
wide
broad
deep
warm
grow
strong
true

We could not form *thickth (for ‘thickness’), for instance. The form ‘coolth’ is attested in
dictionaries but it could never compete with ‘coolness’ (see Matthews 1991: 70, Plag 2003: 62).
Therefore, the suffix ‘-th’ is unproductive.
On the other hand, there are suffixes that approach full productivity. In English, the suffix
‘-er’ can be added to any new verbal base to form a new lexeme which means ‘the person who
carries out the meaning of the verb (Bauer 1983: 62).

(8) teacher
driver
reader
eater
writer
producer
….
 46

Similarly, the suffix ‘-ful’ can be added to the name of any container to form a noun (Bauer
1983: 63):

(9) canful
pocketful
spoonful
handful
fistful
….

Other examples of productive affixes include ‘-able’, which attaches to verbal bases to form
adjectives, with the meaning ‘can be V-ed’ (10) and ‘-ness’ (11), which attaches to adjectives to
form nouns (Matthews 1991: 70).

(10) readable
writable
doable
edible
drinkable
understandable
teachable
….
(11) Carefulness
Attentiveness
Carelessness
Redness
…..
 47

What we need to ask ourselves now is what makes an affix more productive than another?
What makes the suffix ‘-ness’ more productive that ‘-th’? In order to attempt answering this
question, we need to take a closer look at an issue central to morphology - the issue of possible
words.

3. Possible vs. actual words

Very often, the restrictions on forming complex words are unclear. Why does ‘-ity’ attach to
some adjectival bases ending in ‘-ous’ but not to all? (Aronoff 1976)

(12) Curious - curiosity

Capacious - capacity
Monstrous - monstrosity
Glorious - *gloriosity
Furious - *furiosity

Moreover, some affixes occur to a large number of bases; other are a lot less productive (like
‘-th’).
If we want to have some idea why this happens, we need to understand the notion of
possible word. A possible word is defined (Plag 2003: 46) as a word whose semantic,
morphological structure is in accordance with the rules and regularities of the language but is not
attested in dictionaries.

(13) a. Affordable, manageable, readable

b. Cannibalizable (Plag 2003: 23)
c. Capsization (Matthews 1991: 62)

We can add the suffix ‘-able’ to all transitive verbs to form adjectives with the meaning ‘can
be V-ed’. In (13 a), the words formed by adding ‘-able’ are actual words, one can find them in
dictionaries. (13 b) illustrates a possible word, one that is derived according to the rules of the
language but is not to be found in dictionaries. It is important to state that we can understand
 48

‘cannibalizable’ precisely because it is formed according to the rules of the language, which
means that its meaning is compositional.
Therefore, one important property of possible words is predictability of meaning. As hinted at
in the examples in (13), one can predict the meaning of any adjective derived with the suffix
‘-able’ from transitive verbs. The same thing happens in Romanian, where the suffix ‘-ibil’ can
be attached to transitive verbs and the resulting derived adjectives show predictability of
meaning:

(14) A înțelege - înțelegibil (spontaneous speech produced by a non-native speaker of

Romanian)

As you can see from the example in (14), we also need to ask ourselves how speakers store
their words - actual and possible - and where. We will come back to this in section 4.
We turn now to what an actual word is. Actual words are words attested in texts, or used by
speakers in conversation and which are understood by other speakers. The class of actual words
contains both morphologically complex and simplex words (what we have called simple and
complex signs). In this class we find well-behaved complex signs, namely words that are derived
following the rules of the language:

(15) affordable
readable
manageable - ‘can be V-ed’

However, we may find words that do not behave according to the rule:

(16) Knowledgeable - does not mean ‘can be knowledged’

Probable - does not mean ‘can be probed’ (Plag 2003: 47)

We call words such as those in (16) idiosyncratic and we mean by that words that are not in
accordance with the word-formation rules of English. At the same time, we arrive at a second
 49

important property of possible words. Possible words cannot be idiosyncratic; only can actual
words can.

4. The mental lexicon

Words are stored in the mental lexicon for speakers to retrieve them when needed. The
mental lexicon is a kind of dictionary that is stored in the brain instead of being written on paper.
Idiosyncratic words must be stored as such in the mental lexicon, because speakers cannot
derive them following the rules of word-formation.
The question is now whether regular complex words are stored in the mental lexicon or
whether speakers derive those words whenever they need to use them.
Linguists have conceived of the mental lexicon in different ways. In one model of analysis,
the lexicon only contains information which is not predictable, which means that only simplex
words, roots and affixes would be present in the lexicon. In the words of Di Sciullo and Williams
1987:3), the lexicon is “like a prison - it contains only the lawless”.
If we adopt this view of the lexicon, it would only contain ‘manage’, read’, ‘afford’, ‘-able’
and speakers would derive the words ‘manageable’, ‘readable’, ‘affordable’ whenever they
needed them in conversation, namely speakers would combine the roots and affixes whenever
needed.
One important reason why many linguists have adopted this view of the lexicon is economy
of storage. According to this argument, the lexicon should be minimally redundant, meaning
that no information should be listed more than once in the mental lexicon and everything that is
predictable by rule should not be listed (see Plag 2003: 48).
However, it seems that speakers need to access the lexicon very quickly when they are
engaged in conversation (speakers utter about three words per second Plag 2003:48). therefore,
the brain must be optimized with respect to the processing of words. The necessity for quick
access may be in conflict with the necessity of economy of storage.
An alternative conception of the lexicon is that where complex words are listed, too, for
speakers to retrieve them whenever necessary. However, this view of the lexicon is in conflict
with the principle of economy of storage.
 50

How can we balance the two principles? Apparently, access to morphologically complex
words works in two ways. There is direct access to whole words (the whole-word route) and
there is also access to decomposed elements (the decomposition route). When a new complex
word is stored in the lexicon, it is simultaneously processed in two ways: it is both decomposed
and looked up as a whole. The faster route wins and the item is retrieved in that way (Plag 2003:
49)
When the words are frequently used, speakers tend to store and retrieve them via the
whole-word route. When the words are less frequent, speakers tend to store and retrieve them via
the decomposition route.

5. How do we measure productivity?

We have seen that some affixes are considered to be more productive than others. The
question is now how we can measure the productivity of affixes, how we can tell for sure that a
certain affix is productive and how productive it is.
One way of measuring ‘the statistical readiness with which an element enters into new
combinations’ (Bolinger 1948: 18) is to count the number of attested different words derived
with a certain affix at a given point in time. The problem with this way of measuring productivity
is that there can be many words derived with an affix but speakers do not use the affix to make
up new words. For example, the suffix ‘-ment’ created many new words which are still in use but
it does no longer create new words and it is therefore considered unproductive (Bauer 2001:
196).
Counting the number of words derived with a certain affix may work if we only count the
number of neologisms in a given period of time. For example, in the Oxford English Dictionary,
one can find 284 new verbs in -’ize’ in the 20th century, which shows that this is a productive
suffix (Plag 2003: 53).
However, we should also take into account the criterion of frequency, that is how often a
neologism formed with a certain affix is used at a given time. Linguists do that using large text
corpora. This is how we find out that ‘-ness’ is the most productive suffix of English, although
the number of instances in which it appears in OED equals that of ‘-ize’.
 51

6. How do we constrain productivity?

In answering this question, we must first look at the functions of word-formation. In other
words, we must try to find out why people create new words.

a) The referential function of word-formation

The referential function of word-formation refers to those instances when speakers coin new
words to designate new concepts. In such cases, a new word is created in order to give a name to
a new thing or concept.

(17) The Time Patrol also had to unmurder Capistrano’s great-gramdmother, unmarry him from
the pasha’s daughter in 1600, and uncreate those three kids he had fathered. (Kastovsky 1986:
594, quoted in Plag 2003: 59)

b) The syntactic recategorization function of word-formation

This function of word-formation refers to those instances when speakers use syntactic
recategorization to condense the information. Longer phrases or whole clauses can be replaced
by complex words. ‘His clumsiness’ may replace ‘the fact that he is clumsy’.

(18) Yes, George is extremely slow. But it is not his slowness that I find most irritating.

c) The attitudinal function of word-formation

This function of word-formation refers to those instances when speakers want to convey a
particular attitude.

(19) Come here, sweetie.

What a good kitty.
 52

We have seen that there are three main reasons for which speakers create new words. But
not all possible words are created. The question is now what restrictions operate on
word-formation. There are mainly two major types of restrictions on word-formation: pragmatic
restrictions and syntactic/structural restrictions (see Bauer 1983: 84-87, Plag 2003: 60-63)
6.1. Pragmatic restrictions

Plag (2003: 60) names fashion as one of the most obvious cases of pragmatic restrictions:

(20) mega-
giga-
mini-

A second restriction on word-formation is known as the requirement of existence (Bauer

1983: 85). Words are formed to denote only items, actions or qualities that exist. Existence may
cover fictional or mythological domains.

(21) magpie-brake

There is no such word as the one in (21) because there is no thing that could be described as
such. However, one can conceive of a story in which ‘flying pumpkins were brought to a
standstill by having magpies pull against them’ (Bauer 1983: 85).
A third pragmatic restriction on word-formation is known as the nameability requirement.
To put it very simply, this restriction requires that the concepts encoded by derivational
categories should be very simple and general:

(22) un- ͇ ‘not X’

-en ͇ ‘make X’

The concepts should not be highly specific or complex, as illustrated in the example of a
possible denominal verb-forming category (Rose 1973: 516): “grasp NOUN in the left hand and
 53

shake vigorously while standing on the right foot in 2.5 gallon galvanized pail of
corn-meal-mush”.

6.2. Structural restrictions

Structural restrictions in word-formation may refer to restrictions operating on all levels of

linguistic analysis: phonology, morphology, syntax, and semantics.

6.2.1. Phonological restrictions

Phonological restrictions may refer to individual sounds or to syllable structure or stress.

(23) Noun-forming ‘-al’ (from Plag 2003: 61)

a. arrive - arrival
betray - betrayal
construe - construal
deny - denial
propose - proposal

a. enter - *enteral
promise - *promiseal
manage - *manageal
answer - *answeral
forward - *forwardal

In (23), the phonological restriction that operates is stress-related. Nominal ‘-al’ attaches to
verbs that end in a stressed syllable, so verbs ending in an unstressed syllable are excluded as
possible bases. This restriction does not mean that all verbs ending in a stressed syllable can take
 54

‘-al’ (*delayal, *explainal), which means that there are more restrictions operating in
word-formation at the same time.

6.2.2. Morphological restrictions

Affixes can be sensitive to the morphological structure of their base words.

(24) a. colon-ize-ation
parameter-ize-ation
global-ize-ation
b. *colon-ize-ment
* parameter-ize-ment
* global-ize-ment
c. *colon-ize-al
*parameter-ize-al
*global-ize-al

What (24) shows us is that almost every verb derived with ‘-ize’ can be nominalized by
adding ‘-ation’. All other nominalizing affixes (-ment, -age, -al) are excluded, once ‘-ize’ is
attached.

6.2.3. Semantic restrictions

Sometimes, there are semantic restrictions operating on word-formation:

(25) a. employ-ee
detain-ee
absent-ee
 55

b. I’d discovered that if I hugged the right side of the road, drivers would be more reluctant
to move to their left thereby creating a squeezee play with me being the squeezee. (from
Plag 2003: 63)

c. In a card game of the whist family, a squeezee is a player who is squeezed, in other words
he/she is forced to play a card (from the internet1)
d. After making himself a glass of grapefruit juice, John threw the *squeezees away. (from
Barker 1998: 710)

What seems to be at work in the case of ‘-ee’ affixation is that ‘-ee’ derivatives must refer to
sentient entities.

6.2.4. Syntactic restrictions

Word-formation rules are generally restricted to members of a certain syntactic categories.

To what category does the base to which ‘-un’ attaches to have to belong?

(26) a. Table *untable (from Plag 2003: 30)

Car *uncar
Pillow *unpillow
b. Available unavailable
Broken unbroken
Aware *unaware
c. (to) sing (to) *unsing
(to) walk (to) *unwalk
(to) tell (to) *tell

The examples in (26) tell us that ‘-un’ has to attach to adjectival bases (and certain verbs,
such as ‘un-earth’, ‘un-fold’, ‘un-dress’).

1
[Link]
eHr45e&sig=5DFMo-C35QULSUeDv8fuqBrdhUk&hl=ro&sa=X&ved=2ahUKEwiNn-yjgpreAhWoqIsKHWE7Ce
c4ChDoATAJegQIABAB#v=onepage&q=squeezee&f=fals
 56

Another example for syntactic restrictions would be the suffix ‘-able’, which attaches to
verbs, and the suffix ‘-al’, which attaches to nouns.

7. Blocking

‘Blocking’ is the name given by Aronoff (1976: 43) to the phenomenon where a complex
form is not derived because another form already exists.
The term ‘blocking’ may be used to describe two types of phenomena: homomymy blocking
and synonymy blocking (see Plag 2003: 63-65).

7.1. Homonymy blocking

Homonymy blocking refers to those case where an already existing form blocks the creation
of a semantically or phonologically identical derived form:

(27) liver ‘inner organ’

?liver ‘someone who lives’

7.2. Synonymy blocking

Synonymy blocking refers to those cases where the creation of a new word is blocked
already existing synonymous lexemes.

(28) *dog-ess
*horse-ess (from Matthews 1991: 75)
bitch
mare

In (28), the creation of new words by adding the feminine suffix ‘-ess’ is blocked because
English already possesses specific non-derived words ‘bitch’ and ‘mare’.
 57

(29) *stealer / thief (from Bolinger 1975: 109)

(30) *ungood / bad (from Bauer 1983: 87)
*unbig / small

Reviewing questions:

1. Make a list of possible problems that arise when thinking about the definition of morphemes
as ‘things’ that perfectly map form and meaning. Illustrate them
2. Define the notion of productivity and give examples of productive and unproductive affixes in
English.
3. Define the notions of possible and actual words and illustrate them. Give two properties of
possible words.
4. Define the notion of mental lexicon. How are lexical items stored (and retrieved from) in the
mental lexicon?
5. Define productivity.
6. How do we measure productivity?
7. What are the functions of word-formation? Illustrate.
8. Define pragmatic restrictions on word-formation. Illustrate.
9. Define structural restrictions of word-formation. Illustrate.
10. Define blocking. Illustrate

Exercises
1. What property of actual words can we infer from the following examples?

(1) Readable
Teachable
Writable
Understandable

(2) Knowledgeable
 58

Probable

2. Starting from the two ways in which items are stored and retrieved from the mental lexicon
(the whole-word route and the decomposition route), describe the two possible ways in which the
word ‘insane’ is stored in and retrieved from the lexicon. Which route do you think is likely to be
used by most speakers? Why?

3. Take a look at the following table (from Plag 2003: 51). Which are the words for which the
whole-word route is likely to be taken by speakers and why?

(3) Frequencies of ‘-able’ derivatives in the British National Corpus

-able derivative Frequency
Abominable 84
Absorbable 1
Abstractable 2
Abusable 1
Acceptable 3416
Accountable 611
Accruable 1
Achievable 176
Acid-extractable 1
Actable 1
Actionable 87
Actualizable 1
Adaptable 230
Addressable 12
Adjustable 369
Admirable 468
Admissable 2
Adorable 66
 59

Advisable 516

4. Using the discussion on phonological restrictions operating in word-formation, explain the

impossibility of the following formations.

(4) Noun-forming ‘-al’ (from Plag 2003: 61)

Arrive - arrival enter - *enteral
Betray - betrayal promise - *promiseal
Construe - construal manage - *manageal
Deny - denial answer - *answeral
Propose - proposal forward - *forwardal

5. Using the discussion on phonological restrictions operating in word-formation, explain the

impossibility of the following formations.

(5) a. Black - blacken (from Plag 2003: 62)

Fat - fatten
Long/length - lengthen
Loose - loosen
Wide - widen

b. Fine - *finen
Dull - *dullen
Long - *longen
Low - *lowen

c. Candid - *candiden
Equivalent - *equivalenten
Expensive - *expensiven
Hilarious - *hilariousen
Valid - *validen
 60

6. The nominal suffixes ‘-ation’, ‘-ication’, -ion’, ‘-ance’, ‘-al’, ‘-age’, ‘-y’, and ‘-ment’ are
roughly synonymous. The obvious question is which mechanisms govern their distribution, I.e.
which verb takes which suffix. we will try to answer this question for a subset of verbs, namely
those derived by the suffixation of ‘-ify’, ‘-ize’ and ‘-ate’. The data below exemplify the
nominalization of the pertinent verbs ‘magnify’, ‘verbalize’ and ‘concentrate’. State the
restrictions that constrain the selection of nominalizing suffixes with derived verbs of these types.
(Plag 2003: 70)

(6) Magnification verbalization concentration

*magnify-ation *verbalize-cation *concentrate-ation
*magnify-ion *verbalize-ion *concentrate-cation
*magnify-ance *verbalize-ance *concentrate-ance
*magnify-al *verbalize-al *concentrate-al
*magnify-age *verbalize-age *concentrate-age
*magnify-ment *verbalize-ment *concentrate-ment

7. The verb-forming suffixes ‘-ify’ and ‘-ize’ impose severe phonological restrictions on their
possible base words. There seem to e three classes of words involved, one class taking
obligatorily ‘-ize’, one class taking obligatorily ‘-ify’, and one minor third class which can take
both suffixes. Try to establish the pertinent phonological restrictions, using the following data,
which are all 20th century neologisms from the OED (Oxford English Dictionary). Hint: consider
the number of syllables and the stress patterns for all derivatives and try to find the appropriate
generalization (Plag 2003: 70-71)

(7) a. -ize derivatives

Academicize accesorize absolutize acronymize adjectivize
Aerosolize anodize anthropologize bacterize Baskonize
Bolshevize Bonderize bovrilize cannibalize capsulize
*artize *massize *bourgeoisize *Japanize *speechize
 61

b. -ify derivatives
Artify bourgeoisify gentrify jazzify Karstify
Massify mucify mythify Nazify negrify
*randomify *federalify *activify *modernify *Germanify

Chapter 4 Structuralist approaches to word structure

1. Structuralist methods

In linguistics, structuralist theory relies on the notions of distribution, formal definitions

and constituent. Generative linguistics, which is an offshoot of structuralism, also employs these
concepts in linguistic analysis.
The distribution of an element represents the totality of environments where that element
appears. The grammar determines a set of positions where the element may appear:

(1) Students hate teachers.

Teachers love students.

The examples in (1) indicate that, in English, the distribution of bare plurals, i.e. nouns that
systematically appear determinerless, includes the subject and the object positions.
Formal definitions are definitions that exclusively rely on the position of the element in the
sentence. Formal definitions were meant to replace notional definitions, which were thought to
be unreliable.

(2) The subject - notional definition:

The Subject is the part of the sentence that shows who / what performs the action or what
the sentence is about.
 62

(3) a. All my students hate word-formation.

b. All my students rushed into the room.
c. It is raining heavily.
d. There is a problem with your analysis.

The definition in (2) may work for the examples (3a,b) but not for the examples (3 c,d). By
relying on meanings and notions, notional definitions easily run into difficulties. If we replace
the definition in (2) with the definition in (4), which is a formal definition, we can see that, this
time, it is not so easy to come up with counter-examples:

(4) The Subject - formal definition:

[__ VP]

The notion of constituent is to be understood in opposition with the notion of string. A

string is any sequence of two or more adjacent elements; a constituent is a string that has formal
features (see Cornilescu 1995).

(5) Women students are tempting new subjects.

In (5), ‘tempting new’, ‘students are’ are examples of strings. ‘Women students’ and ‘new
subjects’ are constituents, because they show a degree of internal cohesion. This can be proven
by the substitution test, by means of which we replace ‘women students’ with ‘they’.
While developing the concept of constituent, Structuralist Theory also developed
procedures of uncovering the structure of complex forms, such as the analysis into immediate
constituents (ICs) and the analysis into endocentric and exocentric compounds.
Immediate Constituent analysis represents a technique of breaking up complex forms (such
as sentences) into successive components (phrases/constituents); it proceeds in binary steps
down to the level of the ultimate immediate constituents, i.e. morphemes (apud Baciu 1998).
The basic assumptions of this procedure are the following:
a) a linguistic form is either simple or complex
b) a simple form is a morpheme
 63

c) a complex form consists of two or more I(mmediate) C(onstituents) standing in a construction.

Each Immediate Constituent occupies a certain position in the construction.

IC Grammar aims at establishing the following facts about the analyzed language:
a) a list of the constructions in the language (at all levels of analysis)
b) the positions of the construction and the form (simple or complex) that may occupy these
positions
c) the list of the simple forms (morphemes) classified as to their occurrence in position classes.

In IC grammar, the constructions were classified as endocentric and exocentric.

A construction is endocentric when the distribution of the whole construction is the same as
the distribution of one member, which is viewed as the head of the construction.

(6) Schoolboy
The head is ‘boy’; the distribution of ‘schoolboy’ is the same as the distribution of ‘boy’.
The boy ate his dinner.
The schoolboy ate his dinner.

A construction is exocentric when the distribution of the whole is not equal to the
distribution of either member.

(7) Lazybones
The head is ‘bones’; the distribution of ‘lazybones’ is not the same as the distribution of
‘bones’
The dog dug a hole to bury his bones.
??The dog dug a hole to bury his lazybones.

2. The Structure of Compounds

Therefore, endocentric compounds are compounds whose constituents follow the syntactic
pattern of a head and an attribute to the head, i.e. a determinant and a determinatum. In the case
 64

of endocentric compounds, the compound has the same distributional properties and selection as
the head.

(8) Black+bird
Bird = head
Black = attribute to the head
‘A blackbird is a kind of bird’

Exocentric compounds have different selections for the distributional context of occurrence
than any of the component parts. Exocentric compounds are frozen constructions with
metaphor-like usage:

(9) chatter+box
box = head
chatter ≠ attribute to the head
‘A chatter-box is not a kind of box’.

The general distributional class of blackbird is the same as that of ‘bird’, hence ‘bird’ is the
head of the construction since its structural properties are inherited by the whole construct: bird
and blackbird have the same distributional properties, and the construction is characterized as
endocentric.
To indicate the complexity of the construction we can use either labelled bracketing (10a) or
a tree like representation (10b) (apud Baciu 1998)

(10) a. [ [black]A [bird]N]N

b. N
(blackbird)
3
A N
(black) (bird)
 65

In the case of chatter-box, chatterbox and box do not belong to the same subclass of nouns.
The head noun ‘box’ belongs to the subclass of nouns for which the pronominal substitute is the
pronominal ‘it’ while the construction ‘chatterbox’ would always substitute for nouns having the
feature [+animate; +human], i.e. ‘she’ or ‘he’.
Chatterbox and box are not mutually substitutable within the same distributional contexts,
i.e. the two items are members of different subcategorial classes of nouns, therefore the
construction is exocentric.

2.1. Compound classification

Compounds are subcategorized according to the type of construction. Within each subtype
several sub-classes are identified depending on the class they belong to (nominal, adjectival or
verbal compounds) and the class of the constituent parts.

2.1.1. Endocentric Nominal Compounds

This class includes the sub-classes N+N (earthquake), V+N (drawbridge), N+Ving
(home-coming) Ving+N (sleeping-car).

(11) N
housekeeping
3

N N
house keeping
3

V suffix
(keep-) (-ing)

2.1.2. Exocentric compounds

Exocentric compound can be nominal compounds N+N (birdbrain), A+N (dimwit), or verbal
compounds (N+V) (brain-wash)). All exocentric compounds are highly lexicalized expressions.
Exocentric nominal compounds of the type N+N like birdbrain, butterfingers, chatterbox or
 66

A+N like dimwit, bluebeard, redcap have also been referred to as Bahuvrihi compounds by
analogy to the old Sanskrit compounds that denoted a person by one of his characteristic features
of behaviour. (apud Baciu 1998, Lieber 2004)

3. Affixation

Affixation is defined as a process that attaches an affix (prefix or suffix) to a base, with or
without change of class.
One important distinction to be operated in the class of affixes is that between affixes that
change the pronunciation or the shape of the base words and those that do not. While prefixes do
not change the stress pattern of the base, suffixes generally have an effect on the stress pattern of
the base.

(12) contextualize - decontextualize (Plag 2003: 79)

organize - reorganize
modern - postmodern
modify - premodify
argument - counterargument

(13) feminine - feminize - loss of two sounds

mercury - mercurate - loss of final vowel
selective - selectivity - main stress shift to the syllable preceding the suffix
signify - signification - main stress to the suffix
employ - employee - main stress to the suffix

Not all suffixes trigger phonological changes. Among these ‘well-behaved’ suffixes ‘-ness’
and ‘-less’ feature prominently:

(14) propaganda - propagandaless

radiation - radiationless
manager - managerless
 67

(15) adventurous - adventurousness

articulate - articulateness
openminded - openmindedness

Generally, suffixes that trigger phonological changes of the base begin in a vowel.
Consonant-initial suffixes do not generally trigger phonological changes of the base.

(16) Phonological properties of some suffixes (apud Plag 2003: 80)

Suffixes that Examples Suffixes that do not Examples
trigger alternations trigger alternations
-(a)tion Alternation -ness Religiousness
-y Candidacy -less Televisionless
-al Environmental -ful Eventful
-ic Parasitic -hood Companionhood
-ize Hypothesize -ship Editorship
-ous Monstrous -ly Masterly
-ive Productive -ish Introvertish
-ese Japanese -dom martyrdom

A property of affixes that can be associated with their phonological structure is their
etymology. Latinate suffixes prefer Latinate bases and often have bound roots as bases. Native
suffixes (suffixes of Germanic origin) are indifferent to this kind of distinctions. The same
Germanic suffix can take both a Latinate and a native base. For example, in (17) ‘ident-’ is a
bound root and all the bases in the second column are of Latin/Greek origin.
 68

(17) Latinate and native suffixes (apud Plag 2003: 84)

Latinate Examples Native suffixes Examples with Examples
suffixes Latinate base with native
bases
-ify Signify -ship Citizenship Friendship
-ate Investigate -ness Attentiveness Sweetness
-al Colonial -ful Beautiful Helpful
-ity Identity -hood Companionhood Brotherhood

From the structuralist perspective, affixation can also result in exocentric and endocentric
constructions.

3.1. Prefixation

Prefixes are subdivided into two sub-classes function of whether they do or do not alter the
form class to which the base belongs:

A. prefixes that do not change the form class of the base

(18) mis-
dis-
un-
over-
out-

Such prefixes derive endocentric constructions of the form AB = B, i.e. ‘mislead’ = lead, in
which the morpheme sequence AB ‘mislead’ formed by prefixation has the same distributional
properties as B ‘lead’ has.

(19) mislead
 69

mis- lead

B. prefixes that change the form class of the base

(20) be- behead

en- endanger
de- dethrone
a- asleep

In this case, the morpheme sequence AB ‘behead’ formed by prefixation does not have the
same distributional properties as B ‘head’ has.
These constructs are of the form ABB which indicates that the relation between the two
morphemes is exocentric.
The change of form class of the base is accounted for by assuming that before prefixation
the base has undergone a process of conversion analyzed as zero suffixation.

(21) V
(behead)
2
Prefix V
be- 2

N Suffix
head 

3.2. Suffixation

From the point of view of structuralist linguistics, suffixation more often than not result in
exocentric formations, since most suffixes change the lexical category of the base.
 70

(22) kingdom
stardom
wisdom

These constructs are of the form ABB which indicates that the relation between the two
morphemes is exocentric. The distribution of ‘kingdom’ does not equate either the distribution of
‘king’ or the distribution of ‘-dom’.
We will come back to the issue of determining the head in the case of suffixation in the
following chapters.

3.3. Identifying Immediate Constituents in multi-morphemic words

The question we are trying to answer now is how we correctly identify the immediate
constituents of constructions that are made up of more than two morpheme sequences.
The principle to be followed is still that of the binary steps the problem is that, given a
sequence of morphemes abc, the immediate constituents of the sequence can be constructed in
two ways:

(23) [[ab]c]
[a[bc]]

(24) unfriendly
un#friend#ly .
[[un#friend] #ly]
[un#[friend#ly]]

“The order of the steps in which the immediate constituents are linked should be supported
by the total structure of the language, i.e. the constructs identified by the analysis should
correspond to patterns of general relevance for the structure of the language as a whole.” (Baciu
1998: 32)
 71

The decision depends on the structural patterns (distributional frames) that the two different
combinations evince. The construct friendly is easily substitutable in the frames in (25), which
indicates that it has the distribution of an adjective

(25) This _________ response

friendly
godly
cowardly

All the constructs in (25) also substitute for monomorphemic adjectives in the same minimal
context:

(26) This__________ response

kind
true
just
fair

Moreover, all the elements in (26) allow prefixation with un#

(27) unkind
unfair
unjust.

The other potential combination *unfriend does not evince any class of similarly patterned
constructs. (Baciu 1998:32)
However, in recent years, unfriend has permeated at least some English sociolects:
 72

(28) If you unfriend someone, you'll be removed from that person's friends list as well. If you
want to be friends with this person again, you'll need to add them as a friend again.2

If unfriend can be easily explained as part of Facebook language, which means that it is not
part of everyday language, examples like (29-32)) may prove problematic for our analysis of
immediate constituency.

(29) The raven himself is hoarse

That croaks the fatal entrance of Duncan
Under my battlements. Come, you spirits
That tend on mortal thoughts, unsex me here,
And fill me from the crown to the toe topful
Of direst cruelty! (Macbeth, Act 1, Scene 5, 38-43)

(30) Tears of joy, as opposed to fright, threatened to unman him. (Jennifer Fallon, Prison Keep)3

(31) Unchain my heart

Baby, let me be
(Joe Cocker, Unchain my heart, 1987)

(32) Unbreak my heart

Say you'll love me again
Undo this hurt you caused
When you walked out the door
And walked out of my life
Uncry these tears
I cried so many nights
(Toni Braxton, Unbreak my heart, 1996)

2
[Link]
3 [Link]
 73

While still maintaining that ‘unfriendly’ is first constructed by adding ‘-ly’ to ‘friend’ and
only then, is ‘-un’ prefixed to the form ‘friendly’, we will explain forms like ‘unman’, ‘unsex’,
‘unchain’, etc as products of language creativity. Language creativity is best explained in the
framework of generative grammar, which is the focus of the next chapters.

4 Conclusions

Structuralism has offered the first model of formal analysis of linguistic structure. Formal
definitions delimit linguistic units by virtue of their place in relation to other linguistic units,
similarly taken as linguistic primitives, and not by virtue of their sense. The linguist establishes
the invariant units of each linguistic level using ‘discovery procedures’ conceived as rigorous
inductive methods of language analysis capable to uncover the structure of language.
It was structuralists who developed the concept of morpheme, by means of which word
formation came to be viewed as the disposition of morphemes in a word. Therefore, during
structuralism, morphology came to be dominated by word analysis rather than word formation.
 74

Reviewing questions

1. Define ‘distribution’. Give the distribution of bare plurals, bare singular count nouns and bare
mass nouns in English and Romanian.

2. Define ‘formal definitions’. Give formal definitions for the direct object in English and for the
category N(oun).

3. Complete the following:

a) IC analysis is a technique of...............…
b) Morpheme sequences may display two types of construction:....…
c) A construction is endocentric when...............
d) A construction is exocentric when..……….

Exercises:

1. Identify three strings and three constituents in the following sentence:

Derivational morphology will never be my cup of tea.
 75

2. Identify the ICs and discuss the type of construction evinced by the following morpheme
sequences:

a) old man
the old man
cooking butter
punchcard
searchlight
time-saving
redskin
bluebeard
butterfingers
asleep
endanger
dethrone
commercialize
Discovery
unbearable.
b) fish blood system
sad fiction writer
old car dealer

3. Examine the words below. What is the syntactic and semantic relationship between the
constituents that make up the compound? Identify the word-class of the two constituent words
and the word-class of the compound; identify the head; -identify the type of compounds

Schoolboy
Bedroom
Teapot
Water-lily
Hothouse
 76

Oversight
Greenfly
Skyline
Outskirts
World-wide
User-friendly
Short-lived
Good-natured
Outspoken
Near-sighted
Outstay
Undersell
Passer-by
Notary public
Mother-in-law
Attorney general
lazy-bones
Dimwit
Butterfingers
Dare-devil
Blockhead
Turncoat
Loudmouth,
Girlfriend
Bitter-sweet
Moneylender
Bookseller
Handwritten,
God-fearing
Awe-inspiring
 77

4. How do you know that ‘head’ is a noun or verb in the following:

The head of the village school has arrived

She will head the village school
The heads of the village schools have arrived
She headed that school

5. Mark the syllable that receives main stress in the following pairs. What do you think is the
morpheme that converts verbs into nouns in these cases?

VERB NOUN NOUN VERB

Torment torment pattern pattern
Survey survey advocate advocate
Protest protest patent patent
Digest digest lever lever
Progress progress

6. Consider the data in the following sentences. Which of the four morphemes would you
consider bound and morphemes and which would you consider free morphemes? If you decide
that some of them could be considered free morphemes, what would be the more accurate
treatment of the formations: derived words or compound words?

a) There was never an error-free text, Cropper said.

b) Now the lanes were carless, lawless.
c) Arriving on her broomstick at the prison-like school gates, Mildred peered through the railings
into the misty playground.
d) She had been a teacher and she had made sure the girl went to a good school: “my granny had
more influence on me education-wise. (Plag 2003: 72)

7. Look at the italicized morphemes below. Are they affixes or bound roots? Are the formations
the results of derivation or compounding?
 78

a) Biochemistry
Biorythm
Biowarfare
b) Photograph
Photoionize
Photoanalysis
c) Geology
Biology
Neurology

8. Look at the following data on gapping. Gapping refers to a phenomenon where it is possible
to coordinate two words by leaving put one of the constitutive elements. Try to find the
appropriate generalization to account for the data.

a) Possible gapping in compounds

Word and sentence structure
Computer and cooking courses
Word-structure and -meaning
Speech-production and -perception
b) Possible gapping with prefixes
De- and recolonialization
Pre- and postwar
Over- and underdetermination
c) Possible gapping with suffixes
Curious- and openness
Computer- and internet-wise
Child- and homeless
d) Impossible gapping with suffixes
*productiv(e)- and selectivity (productivity and selectivity)
*feder- and local (federal and local)
 79

*computer- and formalize (computerize and formalize)

9. What do the suffixes ‘-ion’ and ‘-ure’ have in common, apart form their being nominalizing
suffixes? Look at the following data and state the generalization. Focus on the
morpho-phonological side of the matter (Plag 2003: 105)

a) Erode - erosion
Conclude - conclusion
Confuse - confusion
Persuade - persuasion
b) Compose - composure
Erase - erasure
Close - closure
Dispose - disposure

10. What do the suffixes ‘-ity,’, ‘-ize’, ‘-ify’ and ‘-ism have in common? Look at the following
data and state the generalization. Focus on the morpho-phonological side of the matter (Plag
2003: 105)

Atomic - atomicity
Iambic- iambicity
Historic - historicity
Opaque - opacity
Historic - historicize
Classic - classicize
Erotic - eroticize
Opaque - opacify
Classic - classicism
Romantic - romanticism
 80

11. No consider the following forms and relate their behaviour to the behaviour of the words in
exercises 9 and 10.

Anarchy - anarchism
Monarchy - monarchism
Masochist - masochism

Chapter 5 Generativist approaches to morphology

1. Introduction

In the late 1950s, Noam Chomsky revolutionized the usual notion of grammar as something
that exists in books, by stressing that a grammar is more than a system of rules that puts together
sounds, words or sentences; it is a mental system.
In other words, a grammar is the linguistic knowledge represented in the speaker’s mind.
This sense of the term ‘grammar’ can be more precisely described as linguistic competence.
What does it mean to know and understand a language? Take, for instance, the following
sentences:

(1) John is easy to please.

John is eager to please.
(2) It is easy to please John.
*It is eager to please John

It is our linguistic competence enables us to differentiate what patterns of grammar are

possible in our language and what patterns are impossible. This knowledge is mostly below the
level of conscious awareness of ordinary language users, without any training in linguistics.
For example, all native speakers of English intuitively know that (3a) is a well-formed word
in English but not (3b) and that (4a) is a grammatical sentence in English but not (4b):
 81

(3) a. textbooks
b. *textsbook

(4) a. Did John go to school today?

b. *Went John to school today?
c. What did Jon eat?
d. *What did John eat pizza and?

Another thing our linguistic competence endows us with is an ability to understand the
grammatical sentences of a language, even though the sentences do not make sense:

(6) Colorless green ideas sleep furiously

Yet another ability provided by our innate linguistic competence is an ability to understand
what the sentence means, whether it is ambiguous or not:

(7) I like peanut butter and ham sandwiches

(8) We sell gasoline to anyone in a glass container

This idea of linguistic competence as represented in the speaker’s mind has become very
influential since the 1950s. Most descriptive grammars are now written assuming this notion of
linguistic competence.
More technically, a linguist grammar is said to be a model of the speaker’s competence, a
model of the speaker’s internalized grammar, of his tacit knowledge of language.
Now, since grammar formalizes the native speaker’s intuitions, his competence, the
proposed model for a linguist writing a grammar is somewhat similar to that of a child learning
his native language. Both the linguist and the child are confronted with external data in the given
language (the linguist formalizes the ‘rules’ of the language, the child will ultimately construct a
mental representation of the grammar of his language).
 82

In this framework, the child possesses a language faculty, which is part of his genetic
endowment. The child’s mind is not, as Chomsky notes a ‘blank state’, but it possesses a
blueprint for language, a device that allows him to acquire/develop language.
It is important to emphasize the fact that grammars are learnable in the sense that, by the age
of three, any child is able to fully understand the language that is being spoken around him, he
will be able to hear sounds, he will be able to figure out the constituents that sentences are made
up of, he will be able to assign meaning to these words and sentences.

2. Early Generative Grammar and morphology

With the advent of early Generative Grammar, morphology lost relevance in the general
organization of linguistic theory. If in the framework of structuralism, morphology played a
central part, in generativism syntax is the main focus. The question morphologists have striven to
answer for decades is whether word-formation can be accounted for in terms of syntactic rules or
whether word-formation rules belong to a separate component, such as the lexicon.
In the view of early Generative Grammar, the lexicon contained only simple words, without
including compounds or derived words, which could only be constructed by the transformational
component. All possible variations in form that words and morpheme might show were assigned
to the phonological component.
Generative Grammar assumed no morphological rules. All morphological relations were
accounted for by phrase structure rules, which were allowed to operate on words and
morphemes.
For example, in ‘Aspects of the Theory of Syntax’, Chomsky proposed to use
‘nominalization transformations’ to account for the relation between word pairs such
destroy-destruction, stating that ‘the phonological rules will determine that nom+destroy
becomes destruction’ (Chomsky 1968:11). Therefore, the morphological relation between
destroy and destruction was accounted for by a combination of syntactic and phonological rules.
Inflectional morphology was assumed to operate in a similar fashion. For example, both
regular and irregular inflected verb forms like sang and mended were analyzed as sing+past and
 83

mend+past, ‘where past is a formative with an abstract feature structure introduced by syntactic
rules’ (Chomsky and Halle 1968: 11 quoted in Scalise and Guevara 2005: 149).
However, transformations could only account for productive, transparent and regular
phenomena. However, complex words tend to be less regular (9), they may undergo idiosyncratic
lexicalization (10) and they may not be fully productive (11) (Scalise and Guevara 2005: 149):

(9) Destroy - *destroy-ation

(10) Transmission - ‘the act fo transmitting’ or ‘gearbox of a car’
(11) Read - *read-ation
Therefore, the need for having morphological rules that operate in the lexicon was gradually
acknolwedged.

3. Lexicalism

The view according to which the processes that form complex words - derivation and
compounding - are accounted for by a set of Lexical Rules, which are independent of and
different from the syntactic rules of the grammar is known as lexicalism. In the lexicalist
framework, Lexical Rules are assumed to operate in a presyntactic component, the Lexicon. (see
Scalise and Guevara 2005: 147)
In later generative work, the Lexicon is much more than a simple list of words to provide
syntax with input. Besides a list of underived lexical entries and idiosyncratic complex units, it
also contains an explicit internal computation, the word formation component.
It was Chomsky’s Remarks on Nominalizations that suggested the presence of lexical rules
to account for morphological facts of the type in (9)-(11). ‘Fairly idiosyncratic morphological
rules will determine the phonological form of refuse, destroy, etc., when these items appear in
the noun position’ (Chomsky 1970: 271).

4. Aronoff (1976)

Considered by many morphologists the foundational work in lexicalism (see Baciu 1988,
Scalise and Guevara 2005, Plag 2003), Aronoff (1976) was the first monograph in morphology
 84

in the framework of generative grammar. The model of Word-formation proposed by Mark

Aronoff (1976) marks a ‘watershed in the development of morphological theory within
generative grammar’ (Spencer 1991). A good deal of the work done subsequently is an
extension of, or reaction to Aronoff’s theory of word-formation (see Baciu 1998).
Aronoff’s model assumes the existence of a separate morphological component (lexicon)
which includes:
(a) a list of the atomic and complex words;
(b) a set of word formation rules (Word Formation Rules) that generate potential words on
the pattern of actual words;
(c) a set of obligatory adjustment rules.
The most important feature of Aronoff’s theory is the assumption that word-formation rules
operate over words and not morphemes, i.e. Aronoff adopts a theory of word-based
morphology. A second feature is that Aronoff restricts himself to derivational morphology. All
other aspects of morphology including inflection are regarded as syntactic. No reference is made
to compounding (see Baciu 1998).
In Aronoff’s view, morphemes cannot serve as the basis of word-formation processes
because the very notion of morpheme is problematic. ‘If words are to be analyzed exhaustively
in morphemes, one is left with isolated strings that cannot be interpreted as ‘meaningful
elements’’ (Scalise and Guevara 2005: 157).

(12) Cranberry, huckleberry, strawberry, blackberry

When we isolate the morpheme berry, we are left with cran and huckle, which do not exist
independently. In this framework, these words are not formed by regular morphological
processes. They are listed in the Lexicon as such.

(13) Refer, defer, prefer, transfer

(14) Remit, demit, transmit

If we adopt a theory based on morphemes instead of words, we would need to recognize the
stems fe and mit and the prefixes re-, de-, pre-, trans-. however, as Aronoff points out, ‘neither
 85

the prefix nor the stem has any fixed meaning’ (Aronoff 1976: 12). These units without meaning
cannot be considered minimal linguistic signs. Consequently, Aronoff’s proposal is to look at
morphology as word-based, because words are indeed minimal signs. “All regular
word-formation processes are word-based. A new word is formed by applying a regular rule to a
single already existing word. Both the new word and the existing one are members of major
lexical categories” (Aronoff 1976: 21)

4.1. Word formation rules (Word Formation Rules)

The main features of word formation rules in Aronoff’s framework are the following (see
Baciu 1998).
a) Word Formation Rules apply only to words; these words must be existing words. Thus,
possible but non-existent words cannot be the base of a Word Formation Rule.
b) Word Formation Rules can take as bases only single words; they do not apply to units lower
than or higher than words (i.e. morphemes or phrases)
c) both the input and output of a Word Formation Rule must be members of the categories N,
V, A, i.e. the ‘open-ended’ lexical classes.
d) Word Formation Rules are intrinsically ordered; that is, they apply each time their structural
description is met in the sense that the condition for the rule application is specified on the
base of the rule.
Aronoff (1976: 22) describes word formation rules as follows: “[Word Formation Rules]
specify a set of words on which [they] can operate. This set we will term the base of that rule.
Every Word Formation Rule specifies a unique phonological operation which is performed on
the base. Every Word Formation Rule also specifies a syntactic label and subcategorization for
the resulting word, as well as a semantic reading for it, which is a function of the reading of the
base.”
In Scalise and Guevara’s view (2005: 158) word-formation rules can have the following
form:

(15) Suffixation
[WORD]X [[WORD]X + Suf]Y
 86

Happy happiness
(16) Prefixation
[WORD]X [Pre+[WORD]X ]X
Happy unhappy
(17) Compounding
[WORD]X, [WORD]Y [[WORD]X + [WORD]Y]Z
Apron, string apron string

In this formulation, suffixatio changes the lexical category of the base (X Y), while
prefixation does not (X X).
The semantic reflex of the application of word formation rules is composionality of meaning.
The meaning of the whole word is ‘composed’ by the partial meanings of the constituents and it
is usually given as a paraphrase:

(18) happy+ness ‘the state of being happy’

un+happy ‘not happy’
apron, string ‘string of the apron’

In Aronoff’s framework, words and affixes belong to different levels. Words are in the
lexicon while affixes are parts of rules, which means that they are not lexical items (see Baciu
1998, Scalise and Guevara 2005).

4.2. Restrictions on Word Formation Rule

In order to define the class of possible words of a language, Aronoff provided word
formation rules with a series of restrictions: syntactic, semantic, phonological and
morphological.

4.2.1. Syntactic restrictions

 87

Word formation rules have access to the syntactic properties of the base. The base is one of
the major syntactic categories: N, V, A. word formation rules do not apply to functional
categories: articles, pronoun, etc.

(19) - able attaches to V

Word formation rules are sensitive to the subcategorization frame of the base:
(20) -able attaches to transitive V

Subcategorization frames indicate the narrow syntactic context where an element may
appear:

(21) Eat [__NP]

Run [__]
Depend [__PP]
Inform [__NP, PP]
Act [__AdvP/PP]

4.2.2. Semantic restrictions

In the words of Scalise and Guevara (2005: 160), derivational affixes select the base to
which they attach also with respect to its meaning.
(22) -ee attaches to [+animate, +sentient]

4.2.3. Phonological restrictions

An example of a phonological restriction is the suffix -al, which attaches only to verbs with
main stress on the last syllable:

(23) Try - trial

Propose - proposal
 88

Arrive - arrival
Deposit - *deposital
Recover - *recoveral
Promise - *promisal

4.2.4. Morphological restrictions

Some word formation rules are sensitive to the morphological makeup of the base, crashing
the derivation unless the base has the right internal structure.

(24) Colon-ize-ation *colon-ize-ment *colon-ize-al

Parameter-ize-ation *parameter-ize-ment *parameter-ize-al
Global-ize-ation *global-ize-ment *global-ize-al

What (24) shows us is that almost every verb derived with ‘-ize’ can be nominalized by
adding ‘-ation’. All other nominalizing affixes (-ment, -age, -al) are excluded, once ‘-ize’ is
attached.

4.2.5. Restrictions on the output

All words that are created by word formation rules must be members of the major lexical
categories: N, V, A, P. The category of the output is specified by the word formation rule.

(25) [[happy]A + ness]N

The semantic restriction on the output requires that its meaning should be compositionally
derived from the meaning of the base:

(26) [[X]V + er]A ‘one who Xs habitually, professionally

Drinker ‘one who drinks habitually’
 89

4.3. Conditions on Word Formation Rules

In Aronoff’s view, an affix cannot attach to any lexical category, it selects words of one and
only one category. When dealing with an affix that selects more than one category, we are forced
to assume that there are two or more homophonous affixes. This condition is known as the
Unitary Base Hypothesis.

(27) Accept+able V+able

Charity+able N+able

According to the Unitary Base Hypothesis, there are two -able suffixes, one that attaches to
verbs and one that attaches to nouns. Although it may seem strange or redundant, this fact is
supported by the fact that denominal -able adjectives may be further derived with -ness but not
with -ity, while deverbal -able adjectives select both -ness and -ity (Scalise and Guevara 2005:
163):

(28) Charitableness *charitability

Acceptability acceptableness

The Unitary Base Hypothesis is also supported by the fact that formations using
homophonous suffixes have different meaning paraphrases, which means that their semantics is
different.

(29) V + able = capable of being Xed

A+able = characterized by X

Another condition formulated by Aronoff (1976) is the Binary Branching Hypothesis. This
states that morphological structures are binary, regardless of how compplex they are.
 90

(30) Indutrialization

Industry -al -ize -ation

4.4. Adjustment rules

The application of the word formation rule to the base triggers different changes sometimes
of a fairly drastic kind (see Baciu 1998). These changes are phonological, morphological and
syntactic in nature and are dealt with in the word-formation component by adjustment rules.

4.4.1. Phonological adjustment rules

Phonological adjustment rules relate to the phonological changes brought about by the
operation of a word formation rule, particularly in terms of the stress pattern of the newly
derived word. Some affixes appear to be sensitive to the stress pattern of the base. As mentioned
before, the suffix -al only attaches to bases that end in a tressed syllable. Moreover, the bases
have a stressed vowel followed (optionally) by a consonant (sonorant/anterior). After -al
attachment, the stress patter of the new word changes:

(31) trial, denial

*constructal, *resistal

4.4.2. Morphological adjustment rules

 91

Morphological adjustment rules relate to the allomorphy rules that account for the
allomorphic variation which is lexically or morphologically conditioned by certain specified
morphemes (see Baciu 1998).
Allomorphy rules explain the alternation -fy /-fic with the nominalization suffix -ation as a
case of partial suppletion, by writing a rule which states that when -ation is added to a base
ending in -fy, the latter is replaced by the -fic allomorph.

(32) electrify –electrification

4.4.3. Syntactic adjustment rules

Syntactic adjustment rules refer to changes in the subcategorization frame of the base
brought about by the application of a word formation rule. The application of the word formation
rule can affect the subcategorization frame of the base words in three different ways:

a) frames are inherited from the base word

b) frames are deleted from the base word
c) frames are added from the base word

a) inheritance

(33) The enemy destroyed the city

The enemy’s destruction of the city

The subcategorization frame of the verb is inherited by the noun with minor changes due to the
nominal properties of the noun destruction. The subject and object of the verb become noun
specifier and prepositional complement respectively.

b) deletion
Deletion is the most frequent adjustment triggered by the application of word formation
rules:
 92

(34) They broke the glass into pieces.

The glass is breakable *into pieces
(35) I ate the potato
I overate *the potato

c) addition
In the case of addition, one element is added to the subcategorization frame of the base:

(36) He ran
*He outran.
He outran Mary

Sample of Word Formation Rule: RE#Attachment (from Baciu 1998)

a) [X ]V → [re# [X ]V ]V
SD: Conditions on the base:
(i) Morphological Conditions: X = Vtransitive (entirely productive)
X = Vintransitive (mainly French loans)
(ii) Semantic condition:
X implies a ‘change of state’ in the object upon which X applies

(b) Semantic operation:

[re# [X]V ]V = ‘DO AGAIN/ANEW X’
c) Phonological operation:
re# is stress sensitive; no change of stress pattern; the prefix is stressed : /`re `punch /

Syntactic Adjustment Rules:

(a) deletion of subcategorization frame of the base:
(i) deletion of sentential complement :
I think that he is innocent
 93

*I rethink that he is innocent

I rethink the problem
(ii) deletion of double object frame:
He wrote a letter to me
He wrote me a letter
*He rewrote me a letter
He rewrote a letter
(iii) deletion of preposition:
He thought about the problem
*He rethought about the problem
He rethought the problem

(b) inherited subcategorization frame:

He adjusted to civilian life
He reajusted to civilian life
They modelled/remodelled the education system.

Reviewing questions:

1. Discuss the innateness hypothesis.

2. What role did morphology have in early Generative Grammar?

3. Define ‘lexicalism’.

4. What are the main tenets of Aronoff’s theory?

5. Are words and affixes treated on the same par by Aronoff?

6. Discuss restrictions on word formation rules.

 94

7. Discuss adjustment rules.

8. Discuss the Unitary Base Hypothesis

Exercise:

1. How do the following examples argue against a morpheme-based morphology?

Cranberry, huckleberry, strawberry, blackberry

Refer, defer, prefer, transfer

2. Taking Scalise & Guevara’s formulation of Word Formation Rules, formulate Word
Formation Rules for the following:

disallow
mislead
rearrange
formulation
careless
boyhood

3. Give subcategorization frames for the following:

They always eat their dinner in the kitchen.

Tom broke the window with a hammer.
He runs in the park every day.
They work hard for their families.
You may depend on me to help you.
Please refrain from smoking in the auditorium.
Can you inform me about it right now?
 95

I will fill you in on it immediately.

4. Formulate syntactic restrictions for the attachment of -ness and un-.

5. Formulate semantic restrictions for the attachment of -er.

6. Formulate phonological restrictions for the attachment of en-.

7. Look at the following examples of en- derivation. State the lexical category of the base to
which ‘en-’ is attached. What is the lexical category of the new word? What is the meaning (or
meanings) of ‘en-’ in these words? Is there reason to state that there are two en- homophonous
prefixes? (Katamba & Stonham 2006)

Base New Word Base New word

cage en-cage noble en-noble
large en-large rich en-rich
robe en-robe rage en-rage
danger en-danger able en-able

8. Formulate word formation rules for -less attachment in English and re- prefixation in
Romanian.
 96

Chapter 6 Generativism and lexicalism. The X-Bar

Schema. Affixes as Lexical Items

1. Introduction. The X-Bar Schema

In the 80s, after Aronoff’s (1976) groundbreaking monograph in morphology, both

morphological theory and syntactic theory were much concerned with the principles by which
both word structure and sentence structure are generated. The question was whether both word
structure and sentence structure are generated by the same principles.
In the standard model of generative theory the notion that each phrase had a head is of
paramount importance. This was captured in the principle of endocentricity, according to which
each phrase is built around a head of the same category as the whole.

(1) [NP the books about the war]

[VP never eat a hamburger]
[AP quite certain about Mary]
[PP almost in the house]

As shown in (1), there seem to be important similarities in the internal structure and
behavior of phrases of different categories. The aim of X-Bar Theory was to generalize over the
data and formulate cross-categorial or category-neutral rules. These rules should indicate the
general principles of phrase structure and should define parameters of variation regarding the
organization of phrasal categories within a language or across languages.
 97

The intuitive idea behind X’ Theory is that phrases are built around lexical or functional
heads so that phrases are projections of structures round categories.
The head is defined as the nucleus of a phrase. It is a noun for NPs, a verb for VPs, a
preposition for PPs, and an adjective or adverb for APs. The distribution of the phrase is given
by the distribution of the head. In other words, where books may appear, books about the war
may also appear.

(2) I never seem to cease reading books.

I never seem to cease reading books about the war.

The principle of endocentricity is formalized in (3):

(3) a. Every phrase XP has an X0 head

b. Every head X0 projects to a maximal projection XP

The first projection of some X0, called X’ (X-bar), contains all and only subcategorized
constituents, called complements. It is important to stress that the notion of complement is
mainly a configurational notion, defined as the element that appears to the right of the head and
has little relation to the syntactic function of direct object.
Complements provide information about entities and locations, events, etc., whose existence
is ‘implied’ by the meaning of the head; they appear in the subcategorization frame and are
obligatory constituents or arguments. Complements make a phrase ‘complete’ or ‘saturated’.
(see Săvescu 2014)

(4) [NP the books about the war]

[VP never eat a hamburger]
[AP quite certain about Mary]
[PP almost in the house]
 98

We have identified in the examples in (4) both the heads and the complements. We are left
with ‘the’, ‘never’, ‘quite’, ‘almost’. These are specifiers. Semantically, specifiers help to make
the meaning of the head more precise; syntactically, specifiers mark a phrase boundary.
The X-Bar schemata are given in (5):

(5) The XP Rule

XP (X”) → (Specifier) ˆ X’
X’ → X0 ˆComplements

XP
2
Spec X’
2
X0 Compl

The theory of word structure has followed a parallel development towards a theory in which
rules of word-structure follow from more general principles of grammar.
As we have seen, Aronoff (1976) states separate word formation rules for every affix; each
rule included, along with the syntactic and semantic information, a series of morphological
restrictions on possible bases. Cast in this way, word formation rules appear to be very different
from phrase structure rules in any form.
One assumption that has been prevalent in theories of derivational (and inflectional)
morphology is that complex words have a hierarchical constituent structure which can be
represented by tree-diagrams.
Linguists like Lisa Selkirk (1982), Edwin Williams (1981), Rochelle Lieber (1980, 1983)
proposed that word formation rules be recast in a more general form which makes them look
more like syntactic rules of deriving phrases.
 99

2. Selkirk’s Model.

According to Baciu (1998:73), the main tenet of Selkirk’s theory is that the grammar needs
to be able to represent general features of word structure in a language. Selkirk claims that it is
necessary to have direct representation in the grammar of the fact that a language is, for instance,
exclusively suffixing.
Consequently, it is necessary to identify a category of suffix, which means that the grammar
will contain a rule such as that in (6), where the affix slot can be filled by different suffixes.

(6) N→N + aff

N→N + dom
N→N + hood
N→N + less

In this respect, Selkirk’s model from the theory of Aronoff in which such affixes are all
introduced by their own word formation rules.
In Selkirk’s theory word structure is viewed as having the same general formal properties as
syntactic structure. Moreover, word structure is generated by the same rule system.
The hierarcchical projections of the X-Bar schema are extended to morphology. The type of
analysis Selkirk assumes is a variant of X-bar syntax which defines the structure of
W(ord)-syntax.
All Word-syntactic categories, be they of the category Word or ‘lower’ than Word, are in the
X-bar hierarchy, since the rules of W-syntax and the rules of S(entence)-syntax share the Word
level categories (see Baciu 1998: 73).
In W-syntax the ‘maximal morphological projection’ is identical to the X0-level projection
in S-syntax, i.e. words., while below X0, two purely morphological projections exist: Root (X-1)
and Af(fix) (see Scalise and Guevara 2005: 169).

(7) Word (N0)

2

Word (A0) Affix

 100

g g

Root (A-1) -ness

g

happy

(8) Word (A0)

g
Root (A-1)
2
Root (N-1) Affix
g g
Monstr- -ous

Selkirk argued that in W-syntax the other categories involved are:

(i) X stem (where Stem is simply a convenient term for the type X−1 that is one down in the X-bar
hierarchy from Word (X0)),
(ii) the category Xroot (or X−2) contained within Stem, and
(iii) the category Xaff which is somewhat special in that it is not ordered within the hierarchy (it
cannot be assigned an integer).
W-syntax differs from S-syntax in that the R(oot) level category cannot dominate the
W-level category. In S-Syntax, it is common that non-maximal projections dominate other
maximal projections (see 9) but this cannot happen in W-Syntax. Selkirk proposes a universal
principle that only in the domain of W-Syntax, no constituent can dominate a constituent of
higher X-Bar level (see Selkirk 1982, Scalise and Guevara 2005: 169)
 101

(9) VP
2

V’
2

V0 NP
eat g

N’
g
N0
hamburgers
The trajectory followed by morphology and syntax is illustrated in (10) (apud Baciu 1998):

(10) XP
X Syntax
X0
X−1
X−2 Morphology
Xaff

The features that play a role in Word-syntax fall into to two classes:
(i) the syntactic category features [± Noun], [± Verb], etc., which represent the distinctions
among Noun, Verb, Adjective, Preposition. (Chomsky (1970), Jackendoff (1977, Cornilescu
1995))
(ii) diacritic features, which include those relevant to the particulars of inflectional and
derivational morphology: conjugation or declension class markers, features for tense (e.g. [±
Past]), gender (e.g. [± feminine]), person, number etc. The derivational features may include
ones such as [±latinate].
If we believe that words have the same structure as sentences, in other words if we believe
that Word-Syntax icomplies woth the same principles as Sentence-Syntax, we need to also adopt
a fundamental concept, that of a head. The notion of morphological heads is controversial, both
 102

for derivational and inflectional morphemes. Remember the discussion about exocentric and
endocentric formations and the discussion about prefixes, which generally do not change the
lexical category of the word, while suffixes generally do.
Williams (1981) proposed that all words are headed and identified the head of
morphologically complex words as in (11):

(11) In morphology we define the head of a morphologically complex word to be the righthand
member of that word. Call this definition the Right-hand Head Rule (RHR).

Selkirk (1982) adopts the idea of headed words. In her view, a morphologically complex
word will have a head, which will have the same syntactic category as its mother. Other features
(morphological features, diacritics of various sorts and so on) may percolate up the word tree.
The X-bar schema proposed for generating the word structures of languages conform to the
general format in (12):

(12) a. Xn → Ym Xaff
b. Yaff Xm

Rule (12a) refers to suffixes and (12b) to prefixes. In the (a) case the affix is the head, in the
(b) case the base is the head.

2.1. Selkirk’s Lexicon

In Selkirk’s theory of word-structure the rules of W-syntax are housed in the lexical
component or lexicon.
The lexical component contains a variety of subcomponents.
a) a list of freely occuring lexical items (assumed to be atomic and complex words) which
represents the dictionary or lexicon in the restricted sense
b) a list of the bound morphemes, which together with the dictionary constitutes the extended
dictionary
 103

c) the set of word structure rules (WSR) characterizing the possible morphological structure of
the language.
The dictionary also includes those phrases that are usually called idioms and those having
idiosyncratic meaning but not those words that are entirely regular or compositional in form and
meaning.
Unlike Aronoff’s model, in Selkirk’s model, the word structure rules are regarded as
redundancy rules or well-formedness conditions on lexical items defined over the permanent
lexicon, but which can also be used to create new words (see Baciu 1998: 74).
In Selkirk’s model, inflectional morphemes are treated on a par with derivational
morphemes. Selkirk (1982) assumes that both derivational and inflectional morphology are
‘lexical’ in the sense that they are treated by the rules of the morphological base and not by the
rules of the syntactic component. Similarly, compounding is regarded as a morphological rather
than a syntactic phenomenon, since affixes intermingle with compounds, in the sense that affixes
can be found both ‘outside’ and ‘inside’ compounds:

(13) [[pickpocket]hood]
[[[ neigbour]hood] planning]]].

2.2. Affixes in Selkirk’s model

Since in Selkirk’s model, affixed words are believed to be headed, affixes are treated as
lexical items. They are assigned to a category and have a lexical entry, like any other unbound
morpheme or morphologically complex item, the main distinction between lexical categories
proper and affixes being the fact that affixes are always bound.
Affixes being treated as lexical items entails their having a lexical entry. The lexical entry
of an affix contains three types of information: syntactic, semantic and phonological.

2.2.1. Syntactic information

 104

A particular affix displays two syntactic properties. The first includes the name (feature
bundle) and type (X-bar level) of the affix’s sister category, and whether the affix is suffixed or
prefixed (see Baciu 1998: 75).
For instance, the morpheme ‘-less’ is a suffix and attaches only to a nominal category of the
type N, as in ‘treeless’. The second syntactic property of the affix is the name of the category
which dominates the affix and its sister. The category dominating -less, for example, is always
adjectival.

(14) a. -less = a suffix attaching only to a nominal category of the type N. The category
dominating –less is always adjectival.
b. subcategorization frame = [Noun__ ]

The second syntactic property involves the categorial make-up of the affix itself. ‘-less’ is
assigned the categorial status ‘adjectival affix’ (Aaff ) and the subcategorization frame [Noun__],
by means of which the grammar encodes the information that the category dominating ‘-less’ in
a syntactic representation is an adjective.

(15) -less: [Noun__] + Aaff = A[N[tree]N Aaff[-less]Aaff]A

the category dominating -less in a syntactic representation is an adjective

2.2.2. Semantic information

The semantic information present in the lexical entry of an affix refers to the semantic
analysis of an affix as a function involving a change in lexical form (see Baciu 1998: 76) or to
the meaning contribution of the affix itself.
For example, the affix ‘-able’ has associated with it the pair of lexical rules: (i) Obj → Subj
and (ii) Subj → by Obj/ø. Moreover, the semantic analysis of the suffix ‘-able’ involves also
some characterization of the notion ‘able to be Ved’.

(16) Nobody can accept you proposal.

Your proposal is not acceptable (by anybody).
 105

Obj → Subj
Subj → by Obj/ø
‘-able’ = can be Ved

Other derivational affixes are simply modifiers making a meaning contribution of their own.

(17) a. piglet
townlet
Rivulet
‘-let’ = ‘small N’
b. statuette
bachelorette
undergraduette
‘-ette’ = ‘small/young N’

2.2.3. Phonological information

The phonological attributes of affixes refer to the pronounciation of the affix itself and to
whether they bring about changes in the pronounciation of surrounding morphemes.

(18) Magnify + -ation = magnification

Electrify + -ation = electrification

Therefore, the lexical entry of an affix will have the form in (19):

(19) Lexical Entry of an Affix

a) Category (including type (always Affix), syntactic category features,and diacritic
features)
b) Subcategorization frame
c) Semantic functions
d) Phonological representation
 106

3. Neutral and non-neutral affixes. The Affix Ordering Generalization

English affixes fall into distinct classes with respect to their phonological properties.
Chomsky and Halle (1968) call them neutral and non-neutral affixes.
Neutral affixes are ‘ignored’ or not taken into account by the principles determining the
stress pattern. Neutral affixes are therefore stress-neutral. Non-neutral affixes are not ignored by
the stress pattern principles and enter into the canonical stress patterns of English words (see
Baciu 1988:79).
Siegel (1974) provides the important insight that the affix classes motivated on the basis of
phonology also play a crucial role in the description of English word structure, more specifically
in the distribution of affixes within English words. Siegel (1974) claims that there is a pattern
to the distribution that the non-neutral and neutral affixes have with respect to each other. Siegel
(1974) uses the terms Class I and Class II respectively to refer to these affixes.

(19) Class I suffixes : +ion, +ity, +y, +al, +ic, +ate, +ous, +ive
Class I prefixes: re+, con+, de+, sub+, pre+, in+, en+, be+
Class II suffixes: #ness, #less, #hood, #ful, #ly, #y, #like
Class II prefixes: re#, sub#, un#, non#, de#, semi#, anti#

Siegel (1974) claims that class II affixes may appear outside class I affixes but class I affixes
may not appear outside class II affixes. Selkirk (1982:91) calls this claim the Affix Ordering
Generalization (AOG).
For example (see Selkirk 1982, Baciu 1998), the suffixes ‘-ous’ and ‘-ity’ are Class I
[non-neutral] affixes, while the suffixes ‘-ness’ and ‘-less’ are Class II [neutral] affixes.

(20) [[monstr-os1]-ity1]
[[procliv-it1]-ous1]
(21) [[fear-less2]-ness2]
[[tender-ness2]-less2]
 107

The examples in (20) and (21) show us that, if the two suffixes in the complex structure of
words both belong to either Class I or Class II, they may appear in any order.
However, the order of the two classes themselves is not free. Suffixes of Class I precede
those of Class II:

(22) [[danger-ous1]-ness2], [[activ-ity1]-less2])

(23) *fear-less2-ity1 ,
*tender-ness2-ous1.

The Affix Ordering Generalization is also valid with combinations of prefixes and suffixes.
The examples in (24) show that the class I prefix ‘in-’ may appear either ‘inside’ or outside’ the
Class I suffixes ‘-ive’ and ‘-ity’ as in:

(24) [in1 -[sensit]-ive1]]

[in1-[[sensit]-iv1]-ity1].

The examples in (25) show that the prefix ‘in-’ may also occur inside Class II suffixes, but
not outside of them:

(25) [[in1-[hospitable]-ness2]
*[in1-[glutton]-ish2]].

These facts suggest that the rules of the morphological component are extrinsically ordered.
Specifically, the rules attaching Class I affixes apply before rules attaching Class II affixes.
In this way Class I affixes will always precede Class II affixes (see Selkirk 1982, Allen 1978,
Baciu 1988).
Elaborating on this idea Allen (1978) proposes that the rules of the morphological
component are organized into extrinsically ordered blocks or levels (known as the Level
Ordering Hypothesis), the rules within each block being unordered with respect to each other.
For English, the order of application, the blocks (levels) are:
 108

(21) Level I (Class I Affixation)

Stress Rule
Level II (Class II Affixation)
Level III (Compounding)

Selkirk (1982) remarks that some derivational affixes may appear outside compounds.

(22) Adjectives: a) un-self-sufficient, non-weather-related

b) painstaking-ly,

(23) Nouns: a) ex-frogman, non-earthquake

b) pickup-ful, laidback-ness

(24) Verbs: a) re-overthrow, mis-backdate

Selkirk (1982) rewrites the Affix Ordering Generalization as follows: Class II affixes may
appear inside or outside a compound, while Class I affixes appear only inside compounds. This
generalization is known as the Compound-Affix Ordering Generalization (CAOG).
In the analysis proposed by Selkirk (1982), besides the category Affix, English word
structure involves two category types relevant for derivational morphology, namely, Word and
Root. The term Stem is reserved for the level relevant to inflectional morphology.
Affixes are subcategorized for words or roots, or both. The two columns below list the
subcategorization frames, as seen by Selkirk, of some of the affixes (from Baciu 1998: 82):

(25) Non-neutral/ClassI/Root Neutral/ClassII/Word

-ous: [Nr ___] -less: [N__ ]

-ity : [Ar __ ] -ness: [A__ ]
-ive : [Vr __ ] -er : [V___ ]
-ate: [Ar__ ] -ize: [N___ ]
-ory: [Vr___ ] -y: [N___ ]
 109

[V___ ]

The distribution of affixes is guaranteed by these subcategorizations. The rule schemata for
Affixation is given in (26) (see Baciu 1998: 82):

(26) a) X→Yaff X b) Word→Affix Word

X→Y Xaff Word→Word Affix

b) Xr→Yaff Xr b) Root → Affix Root

Xr→Yr Xaff Root→ Root Affix

The term Root is chosen for the category type lower than, or contained within, the Word in
word structure. The status of an item as a Root does not imply that it is bound. Every
monomorphemic non-affix morpheme is redundantly a Root, and in principle it may also be a
Word.
(27) sad
A[Ar [sad]A r ]A
W
g

R
g
sad

Bound roots (28) are assigned a subcategorization frame, in a similar fashion to affixes.

(28) De-ceive, per-mit

[Affix__ ]

(29) Moll-ify
[__ Affix]
 110

The complete system of rule schemata for English derivational morphology, as suggested by
Selkirk, is given below (apud Baciu 1998: 84):

(30) a. Xn → Yn Xaff
b. Xn → Yaff Xn
c. X → Xr
(where n = Root or Word)

The rule schemata generates an (infinite) array of structures, into which the items listed in
the lexicon may be inserted, and provides all of the grammatical orderings of Class I and Class II
affixes.

(31) W

scarce
nation

R Aff

scarc -ity
 111

nation -al

W Aff

scarce -ness
nation -hood

W Aff

R
R Aff
danger -ous1 -ness2

4. Dual membership in class I and Class II.

There are some affixes in English that belong to both Class I and Class II. The fact that an
affix may be a member of both classes can be expressed in the subcategorization frame of the
respective affix, in the sense that the type of the category for which it subcategorizes is simply
left unspecified.
 112

(32) [ __ An ] where n = Word or Root.

An example of an affix with dual membership is the English prefix ‘un-’. It is a negative
prefix, like ‘in-’ and ‘non-’. The prefix ‘in-’ is a Class I affix, while ‘non-’ is a Class II affix. The
prefix ‘un-’ belongs to both classes:

(33) un - real - ity1

un - recept -iv1- ity1
(34) un - health - y2
un - mind -ful2

In (32) the prefix ‘un-’ appears inside or outside Class I affixes (such as -ity, or -ive). In (32),
‘-un’ appears outside class II suffixes (such as -y, or -ful). Moreover, ‘-un’ may appear outside
compounds:

(35) Unreality
N
g
Nr
2
Ar Naff
2 g
Aff Ar
un- real -ity

In Selkirk’s model, the difference exhibited by ‘in-’ , ‘non-’ and ‘un-’ is represented in the
lexicon as follows:

(36) in- [ __ Ar ] e.g. in - act- ive1, in- toler - ant1

non- [ __ A] e.g. non - profit-making; non- smoke -er2 ; non-alcohol-ic1
 113

un- [ __ An ] ( n = Word, Root), un-real-ity1 , un-health-y2

There are other prefixes, along with some suffixes, that occupy a variable position in word
structure (see Selkirk 1982).

(37) noun prefixes: arch-, vice-

arch-offend -er2, arch - imperialist -ic1
vice-chairperson, vice - president- ial1

(38) verb prefixes: re- , mis- , de-

re-broadcast; re-defin-ition1
mis-underline, mis-calcul-ation1
de-outflank; de-popul-ation1

5. Category-Changing Prefixes

Most prefixes in English are non-category changing affixes. Given the Right hand head rule
none of these prefixes will qualify as heads but will be sisters to a category identical in features
with the mother category.
Therefore, there is no basis for assigning them one or another array of syntactic features i.e.
the category of these prefixes is listed merely as Affix. The lexical entry for these non-category
changing affixes will reflect this lack of category feature assignment:

(39) in-: [ Aff Vr ]Vr in-flate; in-duce;

con-: con-tribute; con-stitute
pre-: pre-dict; pre-cede
per-: per-mit; per-sist;

However, a small number of prefixes in English are category-changing. They are sister to
categories whose features are not those of the mother node. These prefixes include:
 114

(40) a- Ar [a- Vr [sleep ]Vr ]Ar; A r [ a - Nr [kin ]Nr ]A r

be- V [be- N [ cloud ]N ]V; V [ be- A[calm ]A ]V
en - V [en- A [noble]A ]V; V[en- N[slave]N]V
de- V[de- N[bug]N]V

These affixes are considered marked cases and they will be assigned the category features of
their mother node. For instance, ‘a-’ is Aaff and the others are Vaff.

Reviewing questions:

1. Name two important principles of X-bar Theory. What is the general form of the X-bar
schema? In what sense can the X-bar system be extented to word-formation?

2. Define the notion head in syntax. Can we determine the head of a word in the same way?
Explain by giving the word tree markers for the following:

happiness; unhappy; legalize; relegalize;

lighthouse; redcoat; drawbridge.

3. What is the structure of the lexicon in Selkirk’s model of the morphological component? What
information does it comprise?

4. What are the identifying properties of affixes? What kind of information does the lexical entry
for affixes contain? What is the general form of affixation rules in Selkirk’s model?
 115

5. Define neutral/non-neutral affixes. In what sense is the distinction relevant the the distribution
of affixes? What is the AOG? Explain by using the examples below:

monstrosity; proclivitous; fearlessness; tendernessless

*fearlessity; *inselfsufficient; *disbackdate

6. Pronounce carefully the following pair of words. Place the stress mark ` before the syllable
that receives main stress in each word. Are ‘–ic’ and ‘–ee’ neutral or non-neutral affixes?

(i) strategy strategic

morpheme morphemic
photograph photographic
democrat democratic
(ii) detain detainee
absent absentee
pay payee
employ employ

7. Pronounce carefully the adjectives ‘wide’, ‘long’ and ‘broad’. Pronounce carefully the nouns
‘width’, ‘length’ and ‘breadth’, which are derived from the above adjectives by suffixing –th.
Transcribe phonetically the adverb widely and broadly which are derived from the above
adjectives by suffixing –ly. Determine whether –th and –ly are neutral or non-neutral affixes.
 116

Chapter 7 Generativism and lexicalism. Lieber (1980,

1983)

1. Introduction

A second important model of morphological analysis, which is couched in the generative

framework, is provided by Lieber (1980, 1983). Lieber’s treatment of word structure is similar in
spirit to that of Selkirk in that it overrides the use of particular rules for each affix. However, it
follows the progress of X-bar theory in syntax one step further by exclusively relying on
subcategorization frames from affixes and thereby, inating an unwanted redundancy in Selkirk’s
framework.
Lieber (1980, 1983) assumes that all theoretical mechanisms needed for a generative
morphology are available to both inflectional and derivational processes, which means that both
types of affixes are given the same treatment. (Lieber 1980: 60).

2. Similarities and differences

 117

As seen in the discussion of Aronoff’s and Selkirk’s models, generative frameworks of

morphological analysis generally assume that the word formation component of a grammar
includes a dictionary, a lexicon or a list of lexical entries.
In Aronoff’s model, there are lexical entries for unanalyzable words in English and for
lexicalized complex words, like ‘transmission’ - part of a car, which have complex structure but
non-compositional meaning. However, affixes do not have lexical entries, but are atatched by
word-formation rules which contain the kind of information which, for words, is specified in
lexical entries: semantic representation, category information, etc.

(1) ‘Re-’ attachment [X ]V → [re# [X ]V ]V

X = Vtransitive (entirely productive); X = Vintransitive (mainly French loans)
X implies a ‘change of state’ in the object upon which X applies
[re# [X]V ]V = ‘DO AGAIN/ANEW X’
re# is stress sensitive; no change of stress pattern; the prefix is stressed : /`re `punch /
In Selkirk’s model, all lexical items, including words, stems, and affixes have lexical entries,
containing phonological, syntactic, and semantic information. Selkirk also uses a set of rewriting
rules, similar to the phrase structure rules of early generative grammar, to generate trees into
which affixes and roots are inserted.

(1) S NP VP
NP Det N
Det the
N student
VP Aux MVP
Aux T
T -s
MVP V NP
V hate
NP N
N syntax
 118

(2) Xn → Ym Xaff - suffixation

Yaff Xm - prefixation

The theory proposed in Lieber (1980, 1983) adopts a framework similar to that of Selkirk
(1978) in that all lexical elements, stems and affixes alike, are given entries in a dictionary and
then undergo insertion into lexical trees. However, the theory differs considerably from that of
Selkirk (1978, 1982) in that the lexicon, called the Permanent Lexicon (Lieber 1980: 62) is
ascribed a different organization and rewriting rules that produce trees are ascribed a different
nature.

3. The Permanent Lexicon (Lieber 1980: 61-69)

3.1. Lexical entries
All unanalyzable morphological elements are called lexical terminal elements and are
assigned lexical entries. A lexical entry specifies all information about a terminal element, in the
form of:

a) The category and conjugation or declension class of an item

b) Phonological representation
c) Semantic representation
d) Subcategorization. Affixes differ from non-affix morphemes in that affixes have frames
indicating the category of items to which they attach and the category of items produced.
Subcategorization frames indicate category (N, V, A) and other diacritic features of the items to
which they attach.

(5) un- attaches to adjectives

[A __ [A
(6) -ive attaches to verbs with the diacritic feature [+Latinate]
Preventive, abusive, *understandive, *findive
[+Lat]]V__] A
 119

e) Diacritics, such as [+/- Latinate] or [+/- Neutral]

f) Insertion frames. Lexical terminals are also specified for the syntactic frames into which they
can be inserted. Affixes also have insertion frames.

(7) ‘go’ takes a single NP argument, which is the subject

[NP___]
‘throw’ takes two arguments
[NP___ NP]
‘put’ take three arguments
[NP___ NP PP]

(8) ‘-ize’ forms verbs with two-place argument structures

‘The riot factionalized the city’ (Lieber 1980: 65)

Below are sample lexical entries given by Lieber (1980: 65-66):

(9) PREFIX: in- phonological representation /in]/

semantic representation: negative
category / subcategorization: [A __ [A
insertion frame: [___NP; [Link].__]
diacritics: Level I (Class I)

(10) SUFFIX: -ize phonological representation /ayz/

semantic representation: causative
category: ]N__]V
insertion frame: NP__(NP)
diacritics: Level II (Class II)

(11) STEM: run phonological representation

semantic representation
category: V[___]V
 120

insertion frame: NP__(NP)

diacritics: [-Latinate]

As shown, the lexical entries for affixes are identical to lexical entries for non-affix
morphemes except for the presence of subcategorization information in the entries for affixes.
Affixes or bound morphemes require the presence of some other lexical constituent.
A stem is defined as a morpheme whose lexical entry does not subcategorize another
morpheme. (Lieber 1980: 66).
An affix is a defined as a morpheme whose lexical entry specifies another morpheme to
which it can attach. (Lieber 1980: 67).

4. Lexical structure. The Feature Percolation Conventions

Starting from the assumption that morphemes listed in the permanent lexicon are fitted
together in a hierarchical structure, Lieber (1980, 1983) shows that, in order to generate complex
words, there is a single context-free rule which generates unlabeled binary branching tree
structures.

(12) 2

2
1

3
1

3
1 1
 121

Lexical terminals are inserted into these tree structures subject to their subcategorization
restrictions:

(13) 2

happy]A ness]N

2 ness]N
Grime]N y]A

3
3 s]V

Standard]N ize]V

Tree structures need to be labeled on the basis of information about individual morphemes
inserted into those structures. Trees are labeled by means of four Feature Percolation
Conventions.

(14) Convention I (Lieber 1983: 252)

All features of a stem morpheme, including category features, percolate to the first
nonbranching node dominating that morpheme.

3
N -ize]V
g
[[standard]N

(15) Convention II (Lieber 1983: 253)

 122

All features of an affix moprheme, including category features, percolate to the first
branching node dominating that morpheme.

V
3
N -ize]V
g
standard]N

(16) Convention III (Lieber 1983: 253)

If a branching node fails to obtain features by Convention II, features from the next lowest
labeled node automatically percolate up to the unlabeled branching node.

V
3

[counter V
g

[attack]V]V

The prefix ‘counter-’ attaches to nouns (counterweight), verbs (coutersign) and adjectives
(counterintuitive). The resulting word is a verb when the base is a verb, a noun when the base is
a noun, and a adjective when the base is an adjective, i.e. the resulting words seem to have the
features of the bases amd not the features of the outermost morpheme. ‘counter-’ is transparent to
category in the sense that it lacks category features of its own.

(17) Convention IV
If two stems are sisters (they form a compound), features from the right-hand stem percolate
up to the branching node dominating the two stems.

firetruck
 123

N
3

Fire]N truck]N

Convention IV will be discussed in more details in the chapter devoted to the analysis of
compounds.
It is important to stress that it is not only category membership that passes from a lexical
morpheme to a node, but the entire feature content of a morpheme, including gender features (for
languages that have gender, like German or Romanian) or the diacritic distinction [+/- Latinate]
for English. Derived words adopt all features of their outermost morphemes, along with their
category.

(18) N
-Lat
3

A -ness]N
+Lat -Lat
2
V -able]A
g +Lat
break]V

In English, a word will adopt the feature [+Latinate] or [-Latinate] from the outermost affix.
‘Breakable’ is [+Latinate] by virtue of the feature value of the affix ‘-able’. When ‘breakable’
undergoes further affixation, it may take an affix that is subcategorized for [+Latinate] forms,
‘-ness’ or ‘-ity’. (Lieber 1980: 85)

5. Conclusions
 124

In Lieber’s framework, word structures are projected from the Lexicon, so there is no need
to repeat lexical information in a separate set of word structure rules such as those proposed by
Selkirk.
By proposing a symmetrical system of feature percolation which allows the outermost affix
at any stage of word formation, whether suffix or prefix, to determine the category and features
of the word as a whole, Lieber does away with the need for Williams’ Right-Hand Head Rule.
The rejection receives independent support from two types of data:
(i) the existence in English of a few category changing prefixes
(ii) the existence of languages (e.g. Vietnamese) which have left-headed morphology. (see Baciu
1998)
In both frameworks proposed by Selkirk (1978, 1982) and Lieber (1980,1983), the word lies
at the interface of syntax and morphology. Complex words have a hierarchically ordered
constituent structure, just as phrases in syntax have a hierarchically ordered constituent structure.
Both models of morphology we discussed represent instances of the Strong Lexicalist
Hypothesis, because they lump inflection together with derivation and regard them as reflexes of
the same phenomenon, namely affixation. (see Baciu 1998). Morphology is seen as a thoroughly
lexical phenomenon, meaning that morphology is essentially a property of the lexicon, while
word structure, though similar in some respects to sentence structure, obeys different principles.
However, certain aspects of word structure are accessible to syntactic rules, such as for instance
the rules of agreement.
 125

Reviewing questions:

1. Discuss the main similarities between Lieber’s framework and Selkirk’s model.

2. Discuss the structure of the Permanent Lexicon.

Exercises:

1. Give the lexical entry for the suffix ‘-ize’ and the prefix ‘in-’

2. State and exemplify the Feature Percolation Conventions.

3. Show the structure of the complex word ‘breakableness’.

 126

Chapter 8 Argument structure and affixation

1. Introduction

In our discussion of adjustment rules, we saw that affixation may produce changes to the
subcategorization frame of the verb. All the elements in the subcategorization frame of the base
can be inherited by the derived word, elements in the subcategorization frame may be deleted or
added. Now it is time to look at these phenomena in more detail.

2. Subcategorization

One of the most important issues in the theories of syntax and morphology concerns the
valency of verbs, that is, the kinds of complements a verb takes.
In the Aspects model of syntax, valency is represented by subcategorization frames. This
means that the traditional notions transitive, intransitive, etc. are encoded in distributional
frames:

(1) a. meet: V, [___ NP ]

I will meet [my students]NP
b. sneeze: V, [ ____ ]
They are all sneezing.
c. give: V, [ ___NP, PNP ]
I will give [good grades]NP [to my students]PNP.

The application of word-formation rules usually involves a change of the subcategorization

frame of the base. This happens mainly because a word formation rule implies a shift of lexical
category. We saw that, in the Aronovian framework, frames can be inherited (2a), frames can be
deleted (2b) or frames can be added (see Baciu 1998):
 127

(2) a. He criticized this book. His criticism of this book.

b. I told her that Lucy was coming. *I retold her that Lucy was coming.
c. He grew. He outgrew his electric train set. (from Bauer 1983:177-179)

Aronoff (1978) accounts for these changes by means of the adjustment rules that are
triggered by the application of the word formation rule and operate the modifications on the
output of a word-formation rule until the respective word ‘observes’ all the phonological,
morphological and syntactic constraints that characterize the respective language (Baciu 1998:
94).

3. Argument structures

Lieber (2009: 144) defines the valency of a verb as the number of arguments it takes.
Arguments are defined as those phrases that are semantically necessary for a verb or are implied
by the meanings of the verb. In general, arguments are obligatory:

(3) a. My students are yawning.

*are yawning.
b. Fenster devoured the pizza (Lieber 2009: 145)
*Fenster devoured.
c. I put the books on the shelf.
* I put the books.
* I put on the shelf.
* I put.

The verb ‘yawn’ has one argument, its subject noun phrase. Verbs that have only one
argument are traditionally called intransitive. The verb ‘devour’ has two arguments, its subject
and its object; it is a two-place predicate and it is traditionally called transitive. The verb ‘put’
has three arguments, the subject, the object and a phrase expressing location; it is a three-place
predicate and it is traditionally called ditransitive.
 128

The arguments of a verb are generally obligatory. However, there are cases, when the
arguments are optional:

(4) My goat eats tin cans.

My goat eats. (from Lieber 2009: 145)

Sometimes, besides arguments sentences contain some extra phrases that are not necessary
to the meaning of the verb. These are called adjuncts and often express the time or place of an
event, the instrument with which the action is performed or the manner in which it is performed:

(5) He eats bananas in the kitchen, at midnight, with a fork.

3.1. Theta-roles

A predicate and its arguments engage in semantic relations. These semantic relations are
captured by theta-roles (also known as abstract case relations or thematic relations). theta-roles
specify the parts played by the arguments representing different participants in the action or the
state denoted by the verb.
Below is a list of theta-roles and their definitions (adapted from Baciu 1998, Katamba &
Stonham 2006).

(6) Agent is the case of the individual (usually animate) that instigates the action identified by
the verb.
Peter kicked the ball.
Instrumental is the case of the inanimate instrument used to bring about the state of affairs
described by the verb.
She wrote with a pencil.
Patient is the case of the entity or individual that undergoes the process or action described by
the verb.
I patted the cat.
 129

Benefactive is the case of the individual who gains from the action or process described by the
verb.
I made a dress for my sister.
If the ‘gain’ is interpreted as negative, then the Benefactive becomes Malefactive.
The court gave him a ten-year sentence.
The Goal is the entity towards which motion takes place.
I sent a letter to the Dean.
The Source is the entity from which motion originates.
He comes from New York.
The Experiencer is the case of the (passive) recipient of a sensation or mental experience.
My students absolutely love morphology.
Locative is the case that indicates the location, direction or spatial orientation of the event or
state denoted by the verb.
The book is on the desk.
Theme is semantically the most neutral case (Katamba & Stonham 2006: 269). It is also very
difficult to define. It is an entity that undergoes movement with verbs of movement; it is the
thing given with verbs of giving; it is the entity that experiences the state with state verbs.
The balloon rose.
I gave books to my students.
Mary believes in unicorns.

It is important to point out that there is no systematic correspondence between roles and
morphologic cases. Neither is there a systematic correspondence between roles and syntactic
functions (Baciu 1998:96).

(7) a. Tom opened the door with his key. (Agent)

b. Judith loves Tom. (Experiencer)
c. The window broke. / The balloon rose up (Patient/Theme)
d. This key opened Harriet’s door. (Instrument)
 130

In (7a) the subject function corresponds to an Agent; (in 7b) it corresponds to Experiencer;
in (7c), it corresponds to a Patient/Theme; in (7d), it corresponds to an Instrument.
Moreover, sometimes one role may appear with different syntactic functions.

(8) a. Judith loves Tom.

b. It seems to me that you are tired.
c. You surprised your teachers with your theory.

In (8a), the Experiencer is a Nominative Subject; in (8b) it is a Dative Indirect Object; in (8c)
it is an Accusative Direct Object in (4c).
Because of this lack of correspondence between theta-roles and either morphologic case or
syntactic function, English make use of prepositions to identify a semantic role.
The Agent preposition is ‘by’, as found in the Passive (9a), the Instrument preposition is ‘by’,
if there is no Agent (9b), or ‘with’, if the Agent is present (9c), the Source preposition is
‘from’(9e), the Goal preposition is ‘to’ (9f).

(9) a. The declaration was given by the Prime Minister.

b. I was impressed by his declaration.
c. The victim was struck by the burglar with a heavy object.
d. It appears to me that you are very tired.
e. He comes from London.
f. I sent a letter to the Dean.

3.2. Arguments structures exemplified

If we go back to our traditional distinction between transitive and intransitive verbs, we are
now in a position to illustrate their argument structures (apud Katamba & Stonham 2006: 270).
Intransitives are one-place predicates and occur in frames with one argument, such as:

(10) Mary ran. <AGENT>

 131

Transitive verbs are two-place predicates and occur in frames with two arguments, such as:

(11) Mary kicked the ball. <AGENT, PATIENT>

Ditransitive verbs are three-place predicates. There are several different types of
ditransitives which can be distinguished on the basis of their theta-roles.
Verbs of giving (give, send, lend, post) have an agent, a goal, and a theme:

(12) John sent Mary flowers. <AGENT, GOAL, THEME>

Verbs of placing (place, put, position) have an agent, a theme, and a locative.

(13) He put the goldfish in the tank. <AGENT, THEME, LOCATIVE>

Instrumental verbs require an agent, a patient, and an instrument.

(14) John sliced the bread with a knife. <AGENT, PATIENT, INSTRUMENT>

The lexicon must specify the theta-roles needed by a verb, for the verb to appear in the right
syntactic frames. Therefore, the lexical entry for verbs must contain the following information
(apud Katamba & Stonham 2006: 270):

(15) RUN V <Agent>

KICK V <Agent, Patient>
SEND V <Agent, Goal, Theme>
CUT V <Agent, Patient, (Instrument)> - the Instrument is optional

The earliest representations of the roles associated with a verb assumed the form of an
unordered set of roles. Fillmore (1968) proposed the following lexical entry for the change of
state verb ‘break’(apud Baciu 1998: 97):

(16) break: <(Agent), Patient, (Instrument)>

 132

The thematic grid in (16) indicates that the verb ‘break’ takes three arguments, out of which
the Patient is obligatory while the other two may be optional, as the examples in (17) indicate:

(17) a. John broke the window with a hammer (<Agent,Patient,Instrument>)

b. The hammer broke the window. (<Instrument,Patient>)
c. The window broke. (<Patient>)
d. Windows break easily. (<Patient>)

4. Changes brought by affixation to argument structures

4.1. Passivization

The most obvious example of changes brought to the argument structure of verbs is the
passive voice (see Lieber 2009: 145-146, Katamba & Stonham 2006: 279-280).

(18) a. The professor scolded the students.

b. The students were scolded (by the professor).

In the active sentence (18a), the verb has two arguments, the agent and the patient. The agent
functions as the subject and the patient as the object. In the passive sentence, an agent is
unnecessary but if it occurs, it appears in a prepositional phrase introduced by the preposition
‘by’. in the passive sentence, the subject is the patient and the object is erased, so to speak, so the
passive form of the verb has one fewer argument than the active form.
Passive voice is signaled by passive morphology on the verb. The arguments are re-ordered
and the change in the grammatical function of the object is marked morphologically in the verb,
which appears in its past participle form V-en / V-ed and follows the auxiliary ‘be’. Passive is
stated as (19) by Katamba & Stonham (2006: 280)

(19) subject oblique (or null)

object subject
 133

4.2. Dative Shift

When dative shift applies (20), the order of the direct and indirect object is reversed and the
indirect object is realized as a direct object:

(20) John gave flowers to Mary.

John gave Mary flowers.

A second effect of dative shift is the disappearance of the preposition ‘to’.

4.3. Incorporation

Incorporation represents a process by means of which a semantically-independent word is

moved by syntactic rules to a new position which is ‘inside’ another word.

(21) a. t@-pelark@n qoran@ (Chukchee, a paleosiberian language spoken in

I-leave reindeer
b. t@ -qora-pelark@n
‘I'm leaving the reindeer’

This movement is accompanied by a change in nominal morphology. Although incorporation

looks as an exotic phenomenon, it happens in English too. Many English verbal compounds are
formed using a process similar to incorporation (Katamba & Stonham 2006: 296):

Verbing Nobj Nobj-Verbing Example

Hunting (for) Souvenirs Souvenir-hunting Go
souvenir-hunting
Picking Potatoes Potato-picking Go potato-picking
Rattling Sabres Sabre-rattling They are
sabre-rattling
 134

5. Morphological operations on argument structure

Within the pool of generative approaches to derivational morphology, Williams (1981) was
the first to propose that it is the argument structure of the base that plays an important role and
not the subcategorization frame (see Baciu 1998: 97).
Williams (1981) drew an important distinction between two types of arguments, the external
argument and internal arguments. In this framework the external argument is indicated by
underlining it.

(22) break: <Ag, Pat, Instr>

Williams argued that many alternations in the syntactic valency of verbs are the result of
rules affecting the argument structure of the verb. One such alternation is caused by
morphological processes that are brought about by category changing affixation. Williams
illustrates these alternations with the help of the suffixes ‘-able’, ‘-ee’, ‘-ize’.

(23) a. Tom read a book to children.

b. This book is readable.
c. *Tom is readable.
d. *The children are readable.

(24) a. These books can fit on this shelf.

b. *These books are fittable (on this shelf). (Baciu 1998: 98)

(25) a. Tom knows how to swim.

b. *Tom is swimmable.

The examples in (23-24) show that the suffix ‘-able’ attaches to transitive verbs to form an
adjective with the meaning ‘such that can be V-ed’, which is predicated of, i.e. it refers to the
Theme argument. The newly derived adjective cannot be predicated of the Agent, as the
 135

examples in (23c) and (25b) show. It cannot be predicated of the Theme of an intransitive verb
(24b), or of a non-Theme argument of a transitive verb such as Goal (23d).

(26) a. employ someone

b. employee

The examples in (26) show that the suffix ‘-ee’ attaches to a transitive verb and creates a
noun that refers to the Theme argument.
Now, let’s take a look at the changes in argument structure brought about by ‘-ize’
suffixation:

(27) a. The factory is modern.

b. They modernized the factory.

In (27), the adjective ‘modern’ has one single Theme argument. It is, therefore, a unary
predicate.

(28) modern <Theme>

The attachment of the affix ‘-ize’ creates a verb in whose argument structure there is a new
external argument, an Agent. The adjective’s original external argument, the Theme, is turned
into an internal argument.

(29) modernize <Agent, Theme>

Williams (1980) calls this process by means of which the adjective’s external argument
becomes an internal argument internalization of the external argument.

(30) modern <Th> → modernize <Ag, Th = Th> → modernize <Ag, Th = Th>

 136

Internalization has two stages. First, a new external argument is added, and then the former
external argument is demoted to internal position.
According to Williams’ theory of predication a predicate is prevented from having two
external arguments, so in a sense we are talking about the addition of a new argument, rather
than simply internalization (see Baciu 1998).
Williams (1980) argues that the two processes represented by ‘-able’ affixation and ‘-ize’
affixation practically exhaust the morphological rules which operate on argument structure. He
analyses ‘-able’ affixation as an instance of externalization, which is represented:

(31) read: <Ag,Th> → readable <Ag, Th>

In the case of ‘-able’ affixation, the suffix seems to control the external argument of the verb
and what surfaces is the internal argument of the verb (e.g. This material is washable).

(32) I washed the material.

This material is washable.

When the suffix ‘-ee’ is attached, it controls the internal argument of the verb and what
surfaces is the external one.

(33) Tom’s employee - Tom is the agent

When the suffix ‘-er’ is attached, it controls the external argument of the verb and the
internal one surfaces.

(34) Tom’s employer - Tom is the theme

Williams’ (1980) framework is insufficient when trying to account for inheritance, deletion
or addition phenomena. In the cases of inheritance, the suffixal head does not seem to control
any argument of the non-head and the entire argument structure of the non-head is inherited.
There have been several approaches to this question and they are the focus of the next chapter.
 137

Reviewing questions:

1. Correct and/or complete

a) In the Aspects model of GTG the syntactic valency of verbs is represented by .....
b) In GB Theory the syntactic valency of verbs is represented by...…
c) There is systematic correspondence between thematic roles and morphologic case and
syntactic function. Refute or support this statement relying on the following examples; identiy
the theta grid of the predicates:

(i) Tom opened the door with a key/ The key opened the door
(ii) Judith loves Tom/ It seems to me that you are tired/ You surprised your
teachers with your theory
(iii) The balloon rose up

Exercises:

1. State the subcategorization frames for the predicates below:

Mary is sleeping
Tom opened the door
They put the baby in the crib
They thought about the problem.

2. State the argument structure of the predicates in the sentences below; identify the external
argument and the internal argument of the predicates; say whether all arguments are obligatory:

Tom opened the door with a key/The key opened the door
Peter sang a lullaby for the baby
 138

John broke the window with a hammer/The hammer broke the window/The window broke.
A new-born baby weighs between 3 and 4 kilograms.
John positioned his car by the park.
He dropped the stone through the water on to her toe.
He thought about the problem for a very long time.
Everybody loves linguistics.

3. Study the sentences below and identify the changes that have occurred after affixation (if any):

(i) The City Council developed the city

The City Council’s development of the city
The development of the city by the City Council
(ii) The grammar generates compounds
The grammar is generative*of compounds
(iii) He ran
He outran Mary
(iv) Tom calculated our time of arrival
when we would arrive
Tom miscalculated our time of arrival
*when we would arrive
(v) He thought about the problem
He rethought *about the problem
He rethought the problem

4. Discuss the impact of the affixation of -able, -en, -ed, -ee -ify, on the argument structure of the
lexical item to which they attach in terms of Williams’ hypothesis (externalization,
internalization); represent the two processes schematically:

(i) They read the novel / The novel is readable

(ii) This shirt is wide/ He has widened the shirt
(iii) The water is pure/ They purified the water
 139

(iv) They have arrested the criminal/ the criminal has been arrested (by them)

Chapter 9 Thematic inheritance

1. Introduction

One of the morphological issues that have been on the minds of linguists for the last sixty
years at the very least is the issue of inheritance. Adequate mechanisms were sought to account
for the way in which arguments of verbs were preserved by their respective derivations.

2. The argument structure of affixes. Williams and Di Sciullo (1987)

Williams and Di Sciullo (1987) adopt many ideas from Selkirk (1982), such as the fact that
words have heads and the fact that affixes are viewed as lexical items. At the same time, their
framework adopts the notion of ‘external’ argument from Williams 1981.
Williams and Di Sciullo (1987) analyze cases of affixation where the affix seems to be
referring to one of the arguments of the base (1). They assume that such affixes are to be
regarded semantically as functors, in the sense that they have argument structures consisting of
just one argument notated with (R), representing a ‘referential role’ (see Baciu 1998).

(1) My employer vs. My employee

Williams and Di Sciullo adopt the following mechanisms to account for the syntactic
valency changes induced by word-formation rules:

a) the mechanism of percolation

b) the mechanism of function composition
c) the mechanism of control.

In analyzing the phenomenon of inheritance, Williams and Di Sciullo (1987) assume the
distinction between the external and internal arguments, the external argument being indicated
 140

by underlining. Another key assumption of the theory is that the head of a word derived by
affixation determines the external argument of the word derived by affixation.
To account for the cases of pure inheritance (2), Williams and Di Sciullo assume that the
argument structures of the head and non-head compose (by the mechanism of function
composition). The argument (R) of the affix becomes the external argument of the newly formed
word while the arguments of the base become internal arguments:

(2) The mayor developed the city.

The development of the city by the mayor.

(3) develop <Ag, Th>

-ment <R>
develop ment ----> development (R, Ag, Th)
<Ag,Th> <R>

Let us look at the way the arguments compose when a deadjectival nominalization is
involved (4-5). The adjective has a unique (external) argument, which is a Theme. When
combining with the argument of the suffix ‘-ness’, it is turned into an internal argument, while R
is the external argument.

(4) The book is complete.

The completeness of the book impressed me.

(5) Complete <Th>

-ness <R>
complete ness ------> completeness <R, Th>
<Th> <R>

Another case analyzed by Williams and Di Sciullo is the case where one of the arguments of
the base does not surface (6).
 141

(6) I employed Tom.

<Ag, Th>
Tom’s employer.
<Th>

To account fo such cases, Di Sciullo and Williams introduce the mechanism of control. The
argument (R) of the affix, after composing with the argument structures of bases, controls (or
binds) one of the arguments of the base, which is no longer expressed in syntax.
According to Baciu (1998), the underlying assumption is that these affixes attach to verbs
that have in their argument structure the arguments they control, external or internal, respectively.
In each case the control relation satisfies the controlled argument so that it cannot be expressed
in syntax independent of its controller:
The relation of control is expressed by placing the same index on the controller and the
controllee:

(7) employ er R controls external argument

<Agi,Th> (Ri)
(8) employ ee R controls internal argument
<Ag, Thi> (Ri)

3. Syntactic approaches to inheritance. Roeper (1987)

Roeper (1987) proposed a syntactic approach to inheritance which is linked to the notion of
implicit arguments.
The notion ‘implicit argument’ is defined by Roeper as ‘a thematic role that fails to appear
in explicit positions but retain syntactic functions’ (Roeper 1987: 267). An implicit argument can
be detected in two ways:
a) it can be made explicit by means of a prepositional phrase:

(9) The ship was sunk.

The ship was sunk by Bill.
 142

If we can add a by-phrase, we demonstrate the existence of the implicit Agent, turning the
by-phrase into a syntactic diagnostic.

b) it can control the null subject of an adverbial clause of purpose.

(10) The ship was sunk [PRO to collect the insurance].

The implicit Agent, which can be made explicit by the addition of a by-phrase, controls the
null subject of the infinitival clause, which means that it is coreferential with the null subject.
The null subject is notated with PRO, which is an empty pronominal category.
Implicit agents also control the null subject of an adverbial of purpose in the case of -ing
compounds, nominalizations, and -able adjectives:

(11) a. meat-eating [PRO to gain weight]

b. the eating of meat [PRO to gain weight]
c. Goods are exportable [PRO to improve profits]

There is a distinction between the notion of implicit agent - which is thematic - and the
notion of inferred or ‘cognitive agency’. Implicit arguments (agents) pressupose the existence of
a theta grid, while inferred thematic roles - the mere notion of agency - do not. This distinction
accounts for the following contrasts:

(12) a. *the thief of the bank /*the taxman of hidden assets

b. the robber of the bank / the taxer of hidden assets

‘thief’ and ‘taxman’ entail a cognitive notion of agency but do not have a thematic grid and
that is why the ‘of’- phrase, which refers to the theme argument makes (12a) ungrammatical.
The verbs ‘tax’ and ‘rob’ have a thematic grid that is maintained by the suffix’ -er’. ‘Robber’
and ‘taxer’ have complements and it is to these complements that the ‘of’-phrases refer.

(13) a. the grammar is learnable by the child / the view is defendable by anyone
 143

b. *John is reliable by anyone / *The view is defensible by anyone.

The words ‘learnable’ and ‘defendable’ contain the suffix ‘-able’, which carries an implicit
Agent. This implicit agent can appear in a ‘by - phrase’.
The words ‘reliable’ and ‘defensible’ (which are lexicalized forms) do not allow a
‘by’-phrase because a formal Agent is not present in the theta-grid.

3.1. Thematic affixes

To explain such differences, Roeper puts forth the idea that affixes have their own theta grid.
The theta-grid of the affixes percolates to the new node that it creates. The idea that affix
heads can carry thematic information receives support from contrasting suffixes like ‘er’ and
‘-ful’.

(14) a. John is the writer of these books / the player of games.

b. The grammar is generative *of compounds / *playful of games.

The suffix ‘-er’ carries thematic roles, which makes it able to license an internal argument in
the form of an ‘of’-phrase. On the other hand, the suffix ‘-ful’ carries no thematic information
and it cannot license an internal argument, in the form of an ‘of’-phrase.
A second condition is that the thematic roles on the theta grid of the affix must match the
roles in the theta grid of the base it attaches to. This requirement accounts for the selectional
restrictions imposed by the affix on the base it attaches to, expressed by Aronoff in terms of the
‘conditions on the base’ (Baciu 1998).
Therefore, the suffix ‘-er’, which creates nouns referring to an Agent only attaches to verb
bases which have the Agent role. ‘-er’ has the theta-grid <Ag,Th> and attaches to verbs that have
an identical theta grid.

(15) *a trier, *an intender, *a hoper

 144

The words in (15) are ungrammatical because the thematic grid of the verb does not match
the thematic grid of the affix.
The suffix ‘-able’ has the theta-grid <Ag,Th>, and attaches only to transitive verbs.

(16) learnable, defendable, pushable

The words in (16) are well-formed since the theta-grid carried by ‘-able’ matches the
theta-grid on the verb-base.

(17) *comeable

The adjective *comeable is ill-formed because ‘-able’ is a transitive affix but it attaches to
an intransitive verb, so there is a mismatch between their theta-grids.
Affixes that carry a theta-grid are called thematic affixes. In this framework, there are three
types of affixes:

(18) a. Affixes that match the thematic roles on verbs (-able,-er,-ed)

b. Affixes that inherit the thematic roles on verbs (-ing,)
c. Affixes that block the thematic roles on verbs (-ive, -ful,-ist)

‘-able’ illustrates the first case in requiring a transitive input. ‘-able’ has its own thematic
grid <Ag,Th> which must match the grid of the verb-base.

(19) readable, pushable

(20) *comeable, *dieable,* fallable (where the affix is attached to an intransitive verb)
(21) *tryable, *decidable

In (19), there is a match between the theta-grid of the affix and the theta-grid of the verb
base. In (20), the affix is attached to an intransitive verb, so there is a mismatch between the
theta-grids. (21), there is also a mismatch between the theta-grid of the affix and the theta-grid of
the base, because the verbs ‘try’ and ‘decide’ take sentential complements.
 145

The second case is illustrated ‘-ing’. ‘-Ing’ attaches to everything: transitive and intransitive
bases, and bases with different subcategorization features:

(22) The second coming

the putting of men in jail
the reading of books

Such cases if pure inheritance are accounted for by assuming that the affix has an empty
thematic grid: ( -ing [ ]).
The third case is illustrated by affixes such as ‘-ive’, ‘-ful’, ‘-ist’, ‘-some’, ‘-like’. These
affixes cannot take any ‘of’-phrase because they cannot inherit the argument structure of the
base:

(23) *generative of compounds

*playful of games
*flautist of good music

It is assumed that the affix has no grid at all and, therefore, there can be no match between
the theta-grids. Such affixes are called non-thematic affixes.
 146

Reviewing questions:

1. Account for the following nominalizations in Williams and Di Sciullo’s framework:

The mayor developed the city.

The development of the city by the mayor.

The barbarians destroyed the city.

The destruction of the city by the mayor.

The professor examined the students.

The examination of the students by the professor.

2. Explain the notion of control in the following pairs:

Employer vs. Employee

Detainer vs. Detainee
Cheater vs. Cheatee

3. Explain the notion of implicit argument starting from the following examples.

The light was turned off.

The light was turned off [PRO to scare the children]
The light was turned off by the clown [PRO to scare the children].

4. Can implicit agents control the null subject of an adverbial of purpose with -ing compounds,
nominalizations, and -able adjectives?
 147

a. meat-eating [PRO to gain weight]

b. the eating of meat [PRO to gain weight]
c. Goods are exportable [PRO to improve profits]
5. Explain the notion of inferred cognitive agency starting from the contrast below:

a. *the thief of the bank /*the taxman of hidden assets

b. the robber of the bank / the taxer of hidden assets

Exercises:

1. Explain the ill/well-formedness of the following ‘-er’ nominalizations in Roeper’s approach:

*a trier
*an intender
*a hoper
*a disappearer
teacher
runner
swimmer
writer

2. Explain the ill/well-formedness of the following ‘-able’ adjectives in terms of Roeper’s

approach:

*tryable
*comeable
*dieable
*fallable
*decidable
readable
understandable
 148

loveable

8. ‘-er’ nominalizations are called ‘agentive nouns’ because ‘-er’ controls the agent argument of
the base verb. Analyze the following and determine whether Roeper’s model can account for
these cases.

a. a lover of French cuisine

b. the freezer is in the kitchen
c. a protein .... that is a potent inducer of new blood vessel growth
d. owners of intellectual property
 149

Chapter 10 Nominalizations

1. Introduction

There are four major parts of speech: N (noun), A (adjective), V (verb), P (preposition /
postposition). These major parts of speech are defined along two dimensions: +/-N and +/- V
(Chomsky 1991, Cornilescu 1995).
The feature [+N] signals an item’s ability to carry phi-features (gender, number, person) and
the ablity of the respective item to be selected as an argument by a verb and to be assigned case
and a theta-role by that verb. The feature [+V] signals an intem’s ability to select an argument
and to assign it case and a theta-role.
The characterisations of the parts of speech will, therefore, be the following. The noun will
be [+N, -V], the adjective will be assigned the features [+N, +V], the verb will be characterized
as [-N, +V] and the preposition - [-N, -V].
This characterisation of the noun as a [+N, -V] category entails the noun’s impossibility to
have argument structures, this being a feature of [+V] categories. However, in the case of a
specific type of nominalizations (1), there is clear evidence of an argument structure, containing
external (Agents) and internal arguments (Themes).

(1) The assessment of communicative competencies by the mentor took two days.

In what follows, we will take a closer look at nominalizations with argument structures and
try to integrate these nominalizations into the larger nominal domain. We will start with a
classification of nominalizations and move on to their analysis.

2. Nominalizations: events and results

In a larger sense, nominalizations are defined as nouns derived from any other category, such as
verbs (2) or adjectives (3):
 150

(2) drive - driver

(3) bitter - bitterness

However, we are interested in deverbal nominalizations derived by -ing, -ation, -ment, -ance,
-al, since it is these nominalizations that are more likely to possess an argument structure.
A while ago, we introduced the phenomenon of inheritance, as a phenomenon where the
argument of the base verb are preserved with some modifications in the syntax of the deverbal
nominalization.

(4) The barbarians destroyed the city.

The destruction of the city the barbarians.
The barbarians’ destruction of the city.

What the examples in (4) show is that the verb ‘destroy’, which has the argument structure
<Ag, Th>, somehow passes these arguments on the nominalization ‘destruction’. ‘the city’,
which is the Theme argument in the argument structure of the verb ‘destroy’ preserves its Theme
status with ‘destruction’. ‘the barbarians’, which is the Agent argument in the argument structure
of ‘destroy’, preserves its Agent status. This status is obvious when looking at the preposition
‘by’, which introduces Agents in the passive. In the third example, ‘the barbarians’ appears in the
Genitive case, which is usually associated with the Agent. We will call this type of
nominalization a complex event nominalization (following Grimshaw 1990, Baciu 1998,
Alexiadou 2001, Borer 2003, Moulton 2013).
Complex event nominalizations describe an event, in the same way as the verb from which
they are derived does. An important feature of complex event nominalizations is that they ta ke
internal arguments, with the help of the preposition ‘of’.
However, not all nominalizations inherit the argument structure of the base verb. Sometimes,
the same nominalization may get a different interpretation, depending on the presence of
different grammatical elements, such as the plural marker ‘-s’. In (5), ‘destructions’ no longer
has an argument structure. This is shown by the impossibility of the internal argument ‘of the
city’.
 151

(5) The destructions (*of the city) were painful to look at.
(6) The examination was on the table.

The nominalization in (5-6) will be referred to as a result nominalization. Result

nominalizations do not describe events, but a variety of things related to the verb, such as the
result state of the event denoted by the verb (5) or a related concrete object (6).
There is a third type of nominalizations, which also describe an avent which is associated
with the verb from which they are derived, but do not have an internal argument. These
nominalizations are called simple event nominalizations.

(7) The destruction took a long time.

Therefore, we classify nominalizations into three types: complex event nominalizations (8a),
simple event nominalizations (8b), and result nominalizations (8c):

(8) a. The examination of the students lasted a long time.

b. The examination lasted a long time.
c. The examination was photocopied on green paper. (Moulton 2013: 4)

As the examples show, complex event and simple event nominalizations are able to appear
in the frames ‘lasted x time / took x time’, which sets them apart from result nominalizations.
What differentiates between complex event and simple event nominalizations is the presence vs.
absence of the internal argument. Complex event nominalizations require their internal
arguments, just like their verbal counterparts.

(9) a. The instructor’s deliberate examination of the patient took a long time.
b. *The instructor’s deliberate examination took a long time
c. *The instructor’s deliberate examination was on the table (Grimshaw 1990: 51-52)
 152

In what follows, we will need to know how we can be sure what type of nominalization we
are dealing with, complex event nominalizations or result nominalizations.

3. One non-ambiguous event nominalization

While the vast majority of nominalizations are ambiguous between an event reading and a
result reading, there is one type of nominalization which always designates complex events:
‘-ing’ nominalizations. Therefore, by investigating the behavior of ‘-ing’ nominalizations, we
can extract features that will help disambiguate between event and result nominalizations.

First, ‘-ing’ nominalizations require obligatory internal arguments, just like their verbal
counterparts:

(10) They felled trees.

*They felled.
The felling of the trees.
*The felling.
(11) They destroyed the city.
*They destroyed.
The destroying of the city.
*The destroying. (Baciu 1998: 128)

Secondly, ‘-ing’ nominalizations can control the PRO subject of an adverbial of purpose
clause. ‘-ing’ nominalizations can also co-occur with a ‘by-phrase’. This means that they contain
an Agent in their argument structure:

(12) The fracturing of a leg [PRO to realign the bones]

The running of the store by Bill [PRO to get better profits] (Baciu 1998: 129)

Thirdly, ‘-ing’ nominalizations only allow the definite article

 153

(13) The shooting of rabbits is illegal

*A/*one/*that shooting of rabbits is illegal. (Baciu 1998: 129)

Fourthly, ‘-ing’ nominalizations do not take the plural marker:

(14) *The mowings of the lawn were interminable. (Baciu 1998: 129)

When ‘-ing’ nominalizations occur in the plural, the ‘of’-phrase is not interpreted as
connected to the argument structure of the underlying verb. In (15), ‘buildings’ refers to objects
or results of the activity.

(15) There are many buildings in this city.

Fifthly, nominalizations never occur in predicative position, which means that they cannot
occupy the position across the copula:

(16) *That was the / an eating of apples.

Sixthly, ‘-ing’ nominalizations co-occur with the same aspectual modifiers as their verbal
counterparts. As you may remember, accomplishments take ‘in’-phrases while activities take
‘for’-phrases.

(17) a. John built the house in two years.

b. The reading of the book in two days.
(18) a. They ate apples for two hours.
b. The eating of apples for two hours.

Therefore, ‘-ing’ deverbal nominalizations always designate complex events. However, the
vast majority of nominalizations are ambiguous between an event and a result interpretation.

(19) The examination was long.

 154

The exam was long.

The examination was on the table.
The examin was on the table.
(20) The examination took a long time.
The *exam of the patients took a long time. (Baciu 1998: 130)

In (19), the noun ‘examination’ refers to a concrete entity and it has a result intrepretation. In
(20), the noun ‘examination’ refers to an event. The impossibility of the clipped form ‘exam’ is
explained by its being unambiguously a result nominal, which excludes it from the contexts of
event nominals.
It is time we turned to ways of disambiguating between the event and the result
interpretation of nominalizations (apud Grimshaw 1990, Borer 2003, Baciu 1998, Alexiadou
2001).

4. Disambiguating between event and result nominals

First and foremost, the presence of argument structure is a visible and straightforward way
of identifying complex event nominals. In (21a), of the students is a Theme argument and the
nominalization examination has argument structure. In (21b), the lack of argument structure
points to an interpretation of the nominal in terms of a result nominalization:

(21) a. The examination of the students took two days.

b. The results of the examination were announced on the spot.

One way of disambiguating between event and result interpretations of nominalizations is by

using aspectual modifiers like ‘constant’ and ‘frequent’. If the nominalization accepts these
aspectual modifiers, the presence of the internal argument is obligatory and the result reading is
excluded.

(22) a. This assignment is to be avoided

b. *The frequent / constant assignment is to be avoided
 155

c. The frequent / constant assignment of unsolvable problems is to be avoided

In (22a) there is a result nominal which does not require an internal argument. When we add
the aspectual modifiers ‘constant’ or ‘frequent’, the result interpretation is no longer available
(22b). When the event interpretation becomes available, the presence of the internal argument
becomes obligatory (22c).
Aspectual modifiers like ‘constant’ and ‘frequent’ are compatible both with complex event
and simple event nominals:

(23) a. The constant destruction of the city took a long time.

b. The constant destruction took a long time.
(24) a. The frequent examination of students exhausted the teacher.
b. Only frequent examination by the doctors kept John healthy.

What differentiated between complex event and simple event nominalizations is aktionsart
(Moulton 2013: 6-7). Only complex event nominals can take aktionsart modifiers, like ‘in two
days’ or ‘for two days’, while neither simple event nor result nominals can.

(25) a. Construction of buildings in two days / for days annoyed everybody.

b. *Construction in two days / for days.
c. *The construction in two days / for days was widespread.

As far as the presence of the aspectual modifiers is concerned, we can notice that event
nominalizations do not co-occur with sentential complements, since the latter cannot be present
in an argument structure. When a nominalization occurs with such a complement it receives a
result interpretation (26a). The impossibility of ‘constant’ and ‘frequent’ in this structure
reinforces the result interpretation.

(26) a. The announcement that the position had been filled was a surprise.
b. *The constant/frequent announcement that the position had been filled was a surprise.
 156

A second major way of disambiguating between the event and result reading is by using a
Genitive with an Agent role, accompanied by Agent-oriented modifiers like ‘deliberate’ or
‘intentional’, which determine a genuine Agentive reading of the possessive subject.

(27) a. The examination took a long time

b. The doctor’s examination took a long time
c. *The doctor’s deliberate / intentional examination took a long time
d. The doctor’s examination of the patients took a long time

In (27 a), there is an ambiguous nominal, which does not need an ‘of’-phrase. In (27 b), a
possessive is added and the the sentence is ambiguous, because the possessive may be
interpreted as a ‘possessor’, which means that the doctor is the one who is examined, or as the
agent, the instigator of the action. In (27 c), the addition of agent-oriented adjectives turns the
sentence into an ungrammatical one, because ‘the doctor’ is now unambiguously the Agent, the
nominalization is an event nominalization but there is no internal argument. Only when the
internal argument is added (27 d) does the sentences become grammatical, because the argument
structure of the nominalization contains both the Agent and the Theme.
A third important way of disambiguating between the event and the result interpretations of
nominalizations is looking at the different determiners they take.
Result nominalizations are free in the use of determiners. They co-occur with the definite
article ‘the’, the indefinite article ‘a / an’, with the numeral ‘one’, or with demonstratives like
‘this / that’.

(28) She graded the / an / one / that examination.

Another feature of result nominals is that they freely take the plural morpheme.

(29) She had graded so many examinations that she was blue in the face.

Event nominalizations do not pluralize.

 157

(30) *The constant examinations of the students made her faint.

Event nominalizations admit only the definite article, or it is possible to have no determiner
at all.

(31) (The) assignment of difficult problems has caused problems.

When we compare the distribution of result and event nominalizations, we can notice that
result nominalizations pattern like countable nouns, while event nominalizations pattern like
mass nouns (Baciu 1998: 132, Harley 2008: 337). Very intuitively speaking, this indicates the
fully nominal nature of result nominals but the hybrid nature of event nominalizations, which are
verbs incorporated into a nominal layer.
Another observation related to the distribution of nominalizations is that complex event
nominals do not occur predicatively while result nominals do:

(32) a. That was the / an assignment.

b. *That was the / an assignment of the problem. (Baciu 1998: 132)

A fourth major way of disambiguating between complex event nominalizations and result
nominalizations is by looking at the presence of postnominal genitives and temporal
possessives. These elements always indicate result nominals:

(33) This assignment of John’s drove him nuts.

(34) This week’s assignment drove all my students nuts.

The presence of the postnominal possessive ‘of John’s’ in (33) and of the temporal
possessive ‘this week’s’ in (34) indicates the result interpretation of the nominalization
‘assignment’.

(35) *This semester’s constant assigment of difficult problems drove all my students nuts.
(36) The constant assignment of unsolvable problems this semester led to disaster
 158

In (35), the presence of the temporal possessive ‘this semester’s’, which is modifier, is
incompatible with the event interpretation of the nominalization. In (36), the presence of ‘this
semester’, which is an adverbial and, therefore, an adjunct, is compatible with the intended event
interpretation of the nominalization ‘assignment’. This difference between the acceptability of
modifiers versus adjuncts reinforces the difference between the nominal status of result nominals,
which accept modifiers, just like any nouns, and the hybrid verbal-nominal status of complex
events nominalizations, which accept adjuncts, just like any verb.
A fifth way of disambiguating between complex event ad result nominalizations is by
looking at the acceptability of infinitival clauses functioning as adverbials of purpose. While
Roeper (1987) assumed the existence of an implicit argument that can control the PRO subject of
the infinitiive clause, in this framework (Grimshaw 1990), the existence of infinitival clauses
with nominalizations is explained by the presence of the argument structure with event nominals,
an argument structure which contains an Agent. Therefore, complex event nominalizations can
control the PRO subject of the infinitive clause, while result nominals cannot.

(37) The assignment of difficult problems [in order PRO to drive my students nuts.]
(38) *The assignments of difficult problems [in order PRO to drive my students nuts].

Finally, complex event nominals admit the same aspectual modifiers as their verbal counterparts.

(39) The bombing destroyed the city in two days.

The total destruction of the city in two days.
(40) They observed the patient for a few days.
The observation of the patient for a few days.

Result nominals, or simple event nominals do not allow aspectual modifiers:

(41) *The destruction in a day

(42) *The trip in a week / for a week (Baciu 1998: 135)
 159

Below is a summary of the criteria that differentiate between complex event and result
nominals, as originally proposed by Grimshaw (1990) (from Magionos 2016: 20, following
Alexiadou 2001):

(43)
Complex event nominals Result nominals
Complex events take arguments obligatorily Result nominals do not take arguments
obligatorily
Event reading Non-event reading, but result or outcome of the
event
Modification by agent-oriented modifiers is Modification by agent-oriented modifiers is
allowed disallowed
Their subjects are arguments Their subjects are possessives
By-phrases are arguments By-phrases are non-arguments
Implicit argument control No implicit argument control
Complex events can be modified by aspectual Results cannot be modified by aspectual
modifiers modifiers
Modifiers like frequent or constant appear with Modifiers like frequent or constant appear with
singular plurals
Must be singular May be plural

5. How do we analyze nominalizations?

Nominalization may be analyzed as a lexical process, where one word is turned into another
word with a different meaning and category by means of the addition of one morpheme or
several morphemes.

(44) teachV + -erN teacher

destroyN + -ationN destruction
 160

Nominalization may also be analyzed as a syntactic process, where a verbal piece of

structure is embedded under a nominal layer. The idea that there is a syntactic connection
between the a and b sentences below was captured by Chomsky (1970) in X-Bar Theory by
means of a choice of feature N or V.

(45) a. The enemy destroyed the city

b. The enemy’s destruction of the city.

The choice of a feature (N, V) determines the behavior of the label on an XP. If a phrasal
node becomes a VP, then it is assigned case (45a). If the phrasal node becomes an NP, the it
receives an affix or a bracket that blocks the assignment of case to the an object and, therefore,
requires insertion of a preposition to assign case (45 b).

XP
2
Spec X
Enemy 2
N PP
V NP
destroy city

There have been proposals along this line since the earliest models of generative grammar,
converging towards what was stated as the Functional Nominalization Thesis (Kornfilt and
Whitman 2011:1298).

(46) The Functional Nominalization Thesis: Nominal properties of a nominalization are

contributed by a nominal functional projection. The nominalization has verbal properties below
the nominal functional projection, nominal properties above it.
 161

This basically means that there is real verb inside the nominalization (see Roeper 2005: 138).
a simplified version of the syntactic structure of nominalizations is given below (from Roeper
2005: 139)

(47) Nom
2

-tion VP
g

V
g

destroy

The meaning contribution of the ‘-tion’ node is that of a complex semantic entity EVENT.
The verb sttem moves to the left attracted by the -tion node. There is syntactic evidence in favor
of the view that there is a real VP under the -tion node, coming from adverbs (47) and
VP-ellipsis (48):

(47) While the removal of evidence purposefully is a crime, the removal of evidence
unintentionally is not.
His resignation so suddenly gave rise to wild speculation.

(48) John’s destruction of the city and Bill’s doing so too.

Under the view that do so refers to VPs and adverbs also modify VPs, it is clear that there is
a real VP under the nominalization node. This explains the inheritance of the argument structure
of the verb by the corresponding nominalization.
 162

Reviewing questions:

1. Explain which are the possible interpretations of nominalizations.

2. Is there a nominalization which is non-ambiguous? Explain.

3. Give examples of the tests that help us disambiguate between different readings of
nominalizations.

4. Is there evidence showing that there is a verbal head inside nominalizations?

Exercises:

1. Discuss the following sentences. Identify the interpretation of the nominalizations and point to
the elements that disambiguate the interpretation.

The assignment of difficult problems bothers the students.

The transformation of the catterpillar was complete.
*the change of the catterpillar was complete.
The frequent assignment of difficult problems bothers the students.
*The frequent problem bothers the students.
Their performance of the play in only two hours surprised the critics.
*Their dance in only two hours surprised the critics.
Transformation of the catterpillar into a butterfly as rapidly as possible is essential for the
survival of the insect.
*The catterpillar’s change into a butterfly as rapidly as possible is essential for the survival of the
insect.
* A complete creolization of a pidgin can often occur in a single generation.
(from Harley 2018: 337)

The construction of the building lasted a long time.

 163

The construction was frequent and lasted a long time.

The new construction he built was tall.
The destruction of the building in an hour was hard to watch.
The destruction was constant and took place all day long.
The destruction they left was widespread.
Lisa’s explanation of the problem in two minutes flat impressed me.
*Lisa’s explanation that there was a problem in two minutes flat impressed me.
John’s suggestion of that possibility for so many years got tiresome.
*John’s suggestion that he father a child for so many year got tiresome.
(from Moulton 2013: 6-8)

Bill’s cultivation of the tomatoes

The bomb’s destruction of the town
The volcano’s fortuitous burial of Herculaneum
The emperor’s restoration of monarchy
The proposal’s creation of controversy
The sun’s illumination of the room.
Bill’s suspension of the activity
(from Pesetsky 1995)
 164

Chapter 11 Compounds

1. Introduction. What are compounds?

A prototypical compound is a word made up of at least two bases that can also occur as
independent words (Katamba & Stonham 2006: 304). For example, bedroom is made up of bed
and room, which can occur as words in their own right.

(1) bedroom
The bed is in his room.

There are examples of compounds involving more than two elements, but they can be
analyzed as essentially binary structures (Plag 2003: 133-134):

(2) [[communication technology] equipment]

[[power source] requirement]
[[[university [teaching award]]committee] member]

The examples in (2) illustrate a fundamental property of sentence structure and, indeed, of
natural languages, namely recursivity. Recursivity refers to the fact that words (in the case of
compounds) or clauses (in the case of sentence structure) can be repeatedly stacked on an
existing compound or clause to form new compounds or clauses:

(3) [[[university [teaching award]] committee] member] training

[[[[Tom says [that John knows [that Mary thinks [that Sue believes [that…

Although there is no structural limitation on the recursivity of compounds or sentences, the

longer they become, the more difficult it is to process them.
What other elements can form compounds, i.e. other than words that can be used on their
own?
 165

(4) a. Astrophysics
Biochemistry
Photoionize
b. Parks commissioner
Teeth marks
Systems analyst
c. Pipe-and-slipper husband
Off-the-rack dress
Over-the-fence gossip
d. * husband pipe-and-slipper
* dress off-the-rack
Gossip over the fence

The examples in (4a) contain bound roots; in the examples in (4b) the first element is clearly
a word since it is in the plural; in the examples in (4c) the first element is a syntactic phrase.
Syntactic phrases cannot appear as the right-hand constituent, however, as the examples in (4d)
show.
In the light of such examples, we can define compounds as words that consist of two
elements, the forst of which is either a word, a root, or a phrase and the second of which is either
a root or a word. (Plag 2003: 135)

1.1. What makes the difference between compounds and syntactic phrases

This subsection looks at the difference between compounds and syntactic phrases as
illustrated in (5).

(5) Greenhouse vs. Green house

The question is: when do we really know that we are dealing with a compound? In other
words, what makes the difference between compounds and syntactic phrases?
 166

According to quite a few linguists, it is not easy to come up with criteria that clearly
distinguish compounds in English (see Marchand 1969, Bauer 1978, 1983, Lieber 1992, 2005,
a.o.). there are, however, several criteria that have been proposed to distinguish compounds from
syntactic phrases, such as: stress, spelling, lexicalized meaning, unavailability of the first stem to
syntactic processes such as inflection, anaphora, and coordination, and inseparability of the first
and second stems. (Liever 2005: 376)
English spelling is the least reliable criterion for distinguishing compounds. One and the
same compound is sometimes written as two separate words, two words hyphenated, and one
word.

(6) daisy wheel

daisy-wheel
daisywheel - The Oxford Dictionary of New Words (Bauer 1998: 96)

Lexicalization is also a poor criterion for distinguishing compounds. Although many

compounds have lexicalized meaning (7a), there are many other compounds with compositional
meaning (7b).

(7) a. Greenhouse
blackboard
b. Bedroom
teapot

Stress is a criterion that often distinguishes compounds from phrases. While phrases tend to
be stressed phrase-finally, that is on the second element, compounds are usually stressed on the
first element:

(8) [the green cárpet]

[this new hóuse]
[such a god jób] - noun phrases
(9) [páyment problems]
 167

[installátion guide]
[spáce requirement] - compound (Plag 2003: 137)
(10) dóg bowl
ápple cake
trúck driver
móth eaten (Lieber 2005: 376)

This difference between stress in syntactic phrases and stress in compounds is captured by
the nuclear stress test, which basically means that stress is on the last element of the phrase and
the compound stress rule, which says that stress is on the left-hand member of a compound
(Chomsky & Halle 1968: 17).
Sometimes it is only the stress pattern that distinguishes between compounds and syntactic
phrases (Plag 2003: 137-138).

(11) Noun compound Noun phrase

a. bláckboard a black bóard
‘a board to write on’ ‘a board that is black
b. gréenhouse a green hóuse
‘a glass building for growing plants’ ‘a house which is green’
c. óperating instructions operating instrúctions
‘istructions for operating something’ ‘instructions that are operating
d. instálling options installing óptions
‘options for installing something’ ‘options that install something or the
installing of options’

However, not all compound behave alike from the point of view if the stress pattern.
Adjectival compounds consisting of two adjectives have level stress (12) and some noun-noun
compounds are stressed on the second element (13).

(12) icy cold

blue-green (Lieber 2005: 376)
 168

(13) apple píe (Plag 2003: 138)

apricot crúmble
Madison Ávenue
Penny Lane
Aluminium fóil
Silk tíe

Another criterion of differentiating between compounds and syntactic phrases is

unavailability of the first element to inflection, anaphora and coordination. While generally, the
first element in a compound does not carry plural inflection, there are cases when they do:

(14) children’s hour

girls’ club

Similarly, while the first element in a compound is generally non-referential (in bedroom,
there is no specific bed we refer to), it can sometimes be a discourse antecedent for pronouns:

(15) So I hear you’re a real cat-lover. How many do you have now? (Bauer 1998: 72)

Thirdly, while, generally, first elements of compounds cannot be coordinated (16), this can
sometimes be done, when the two elements belong to the same semantic domain (17):

(16) *bed and bathrooms

*foot and basketball
(17) cat and dog show
medical and life insurance

The last criterion is inseparability of the two elements in a compound and it is one of the
strongest tests of compoundhood. The elements of a compound cannot be separated by a
modifier:
 169

(18) *truck fast driver

*blue light green (Lieber 2005: 377)
*[Chinese dirty [jade figure]]

To briefly sum up, the line between compounds and syntactic phrases is often difficult to
draw. To make matters worse, there are also phrasal compounds, where the line between phrases
and compounds is even more difficult to draw:

(19) A who’s the boss wink

An if-you-only-want-to-know sneer (Bauer and Renouf 2001: 108, quoted in Lieber 2005:
337)

To show that the examples in (19) really are compounds (that possibly include phrases),
inseparability proves instrumental. No modifier can be inserted between the phrasal first element
and the second element:

(20) *a who’s the boss filthy wink

a filthy who’s the boss wink
*an if-you-only-want-to-know insinuating sneer
a insinuating if-you-only-want-to know sneer

2. Towards a classification of compounds

2.1. A classification according to semantic and phonological factors

Most English compounds are interpreted in a modifier-head structure, where the first
element modifies the second element.

(21) Film society - a kind of society, one which is concerned with films
Parks commissioner - a kind of commissioner who is occupied with parks
Deep-fry - a kind of frying
(Plag 2003: 136)
 170

The head in compounds is the element which is modified by the other member. From a
semantic point of view, this entails that the set of entities denoted by the compound is a subset of
the entities denoted by the head. This is known as hyponymy. For example, the set of entities
denoted by bedroom is a subset of the entities denoted by room. Headed compunds based on a
modifier-head structure are also known as endocentric compounds.

(22) Water-lily, bookcase, motor-car, skyline, India-rubber

Hothouse, sour-dough, greenfly, high-court, wet-suit
Undergraduate, near-sightedness, outskirts, underdog, oversight
(Katamba& Stonham 2006: 318)

The head of the compounds usually occurs on the right-hand side, which is captured by the
right-hand head rule, as formulated by Williams (1981). The compound gets its semantic and
syntactic information from its head, which determines its category, by means of feature
percolation, as formulated by Lieber (1983). If the head is a verb, as in deep-fry, the compound
will be a verb, if the head is a countable noun, the compound will be a countable noun, as in beer
bottle, if the head has feminine gender, the compound will have feminine gender, as in head
waitress (Plag 2003: 135).
As stated before, plural marking occurs on the head, that is on the second constituent of the
compound. Therefore, parks commissioner, even if exhibiting plural marking on the first element,
is not a plural compound, the plural form being parks commissioners.

(23) NSingular (Plag 2003: 136)

2

Parks[Noun, Plural] commissioner[Noun, Singular]

NPlural
2

Parks[Noun, Plural] commissioner[Noun, Plural]

Here is a list of compounds with plural marking on the first element (from Selkirk 1982: 52):
 171

(24) Overseas investor

Arms merchant
Buildings inspector
Sales receipt
Arms race
Weapons analysis

As you may remember, there are, however, a number of compounds that do not observe the
head-modifier relation. They are not based on a relation of attributive hyponymy, but designate
types of people. They are called exocentric compounds or bahuvrihi compounds.

(25) pickpocket
cut-throat
spoilsport

Their exocentricity refers to the semantic non-compositionality of the compound and not to
their syntactic behavior. Even in these cases, the compounds seem to have a head which
percolates its nominal features to the compound, turning it into a noun.
Exocentric compounds like those exemplified in (25) are called possessive compounds (Plag
2003: 146), because they denote which is characterized by the property expressed by the
compound. It may be said that a loudmouth is a person possessing a loud mouth, a redneck a
person possessing a red neck, etc.

(26) Loudmouth, greybeard, redneck

Besides endocentric, exocentric, and possessive compounds, there is another type of

compounds, where both members of the compound seem to have equal importance, in the sense
that neither seems to be subordinated to the other, so that they have two semantic heads. They
are called copulative compounds or dvandva compounds.
 172

(27) a. Singer-songwriter
Scientist-explorer
Poet-translator
Hero-martyr
(Plag 2003: 146)
Producer-director
Fighter-bomber
(Lieber 2005: 378)
b. The doctor-patient gap
The nature-nurture debate
A modifier-head structure
The mind-body problem
(Plag 2003: 146)

Copulative compounds fall into two classes. The group in (27a) contains appositional
compounds, which refer to an entity characterized by both members of the compound. A
poet-translator is a person who is both a poet and a translator. The group in (27b) contains
coordinative compounds, which refer to two entities that stand in a particular relationship with
respect to the following noun. The doctor-patient gap is a gap between the patient and the doctor,
the nature-nurture debate is a debate on the relationship between nature and nurture, etc.
If we take into account the phonological factor, we may find rhyming compounds. These
are compounds formed by joining together words that rhyme:

(28) black-Jack
claptrap
night-light (Katamba & Stonham 2006: 305)

Sometimes, these compounds contain bound roots, that is they do not combine words but
things that cannot occur in isolation but are not affixes:

(29) nitwit
 173

titbit
skitwit
helter-skelter
hobnob
namby-pamby (Katamba & Stonham 2006: 305)

We have so far looked at a classification of compounds, including exocentric, endocentric,

copulative (appositional and coordinative) and rhyming compounds. This type of classification
takes into account semantic, that is compositionality of meaning and semantic heads, and
phonological factors (rhymimg words, ablaut). It is now time to look at compounds from a
morpho-syntactic point of view and discuss root and synthetic compounds.

2.2. A classification according to morpho-syntactic factors

From this point of view, there are two forms of compounds that are often distinguished:
synthetic compounds (also called verbal, deverbal, or verbal nexus compounds) and root
compounds (also called primary compounds).
In synthetic compounds, the second element is derived from a verb:

(30) Truck-driver, gift-giving, wind-blown, revenue enhancement, waste disposal (Lieber 2005:
375)

In root compounds, the second element is not derived from a verb:

(31) Dog bowl, file cabinet, red hot, sky blue, blackboard, babysit (Lieber 2005: 375)

3 Root compounds

In English, there are root compounds that represent combinations of the major parts of
speech: (N)nouns, (V)erbs, and (A)djectives / (A)dverbs.
The examples below are taken from Lieber (1992: 80) and Lieber (2005: 378):
 174

(32) Productive combinations of root compounds

NN NA AA AN
File cabinet Sky blue Icy cold Hard hat
Towel rack Leaf green Red hot Bluebird
Catfood Stone cold Green-blue Blackboard
Steelmill Rock hard Wide awake Poorhouse

(33) Less productive combinations of root compounds

AV NV VN VV
Dry farm Handmake Drawbridge Stir-fry
Wet sand Babysit Cutpurse Blow-dry
Double coat Spoonfeed Pickpocket Jump shoot
Sweet talk Machine wash Pull toy Jump start

Root compounds are syntactically right-headed, which, together with the fact that they have
compositional meaning or, at least, transparent meaning, makes them endocentric compounds.
There are, however, a number of exocentric root compounds:

(34) Red head, airhead (Lieber 2005: 378)

‘this unconscious angel-beast division’ (Bauer & Renouf 2001: 113)
‘a very kick-arse attitude’ (Bauer and Renouf 2001: 105)

Since the righthand element is the head in root compounds, it lends its category and
morpho-syntactic features to the compound.
 175

(35) N N A A
2 2 2 2

N N A N N A A A
File cabinet hard hat sky blue blue green

From the point of view of their interpretation of root compound, the same observations we
made before hold: the first element in the compound does not refer to any particular individual
and the compound as a whole denotes a subset of the denotation of the second element, that is a
hyponym of the superordinate. For example, in file cabinet, ‘file’ does not refer to a particular
file, and the whole compound designates a subset of cabinets.

4 Synthetic compounds

Just like root compounds, synthetic compounds are right-headed and quite productive in
English. The examples below are taken from Lieber (1992: 81) and Lieber (2005: 379). We
classify these compounds according to the deverbalizing affix that they take:

(36) Synthetic compounds

-er -ing -en -ion, -ment, -ure,

-al
Truck driver Truck driving Power driven Tax evasion
Meat eater Meat eating Well loved Revenue
enhancement
Cat lover Cat loving Moth eaten School closure
Flower grower Flower growing Home grown Waste disposal
Gift giver Gift giving Law given

Further examples are given below (from Plag 2003: 149)

 176

(37) Synthetic compounds

-er -ing -ance, etc

Beer drinker Pasta-eating Shop clearance
Car driver Window-cleaning
Bookseller Soccer-playing
Churchgoer

Still further examples are given below (from Katamba & Stonham 2006: 321)

(38) Synthetic compounds

-er -ing -en
Moneylender Hay-making Hand-written
Gamekeeper Brick-laying Computer-matched
Shoemaker God-fearing Hand-sewn
Bookseller Self-seeking Time-worn
Anteater Eye-catching Guilt-ridden

The structure of synthetic compounds is represented in (37), where there is a deverbal noun
driver which combines with the noun truck (Lieber 1992: 85, Lieber 2005: 380):

(39) N
2

N N
g 2

truck V -er
g

driver
 177

In principle, there may be another possibility of analyzing the structure of synthetic

compounds. The suffix -er may attach to a compound formed by truck and drive.

(40) N
2
V -er
2
N V
g g
truck drive

However, this possibility is ruled out on the grounds that there are no such compounds in
English as *truckdrive, *booksell, *churchgo, which means that it’s more intuitive to form the
deverbal -er noun first and the compound it with the first element.
As far as the semantic interpretation of synthetic compounds is concerned, it is generally
believed that the first element in a synthetic compound, that is the left-hand constituent is
interpreted as an argument of the verb involved in the synthetic compound (Lieber 1983, Lieber
1992, Lieber 2005, Plag 2003). For example, a cat lover is someone who loves cats, so cat
represents the internal argument of the verb love.
Nevertheless, compounds have a wide range of meanings and, although there have been
many attempts at classifying these meanings, the list of semantic categories is very long and
non-exhaustive. Compounds may refer to location, cause, manner, possessor, material, content,
source, instrument, etc. Moreover, it seems that compounds are ambiguous and can receive many
interpretations depending on the context in which they appear. For example, the following
conversational exchange between a father and his five-year-old daughter illustrates the ambiguity
of the compound cat doctor:

(41) F: What do you want to be when you grow up?

D: A cat doctor.
F: A veterinarian?
D: A doctor who is a cat. (Twitter, July 12, 2017)
 178

Having this ambiguity in mind, we go back to the interpretation of synthetic compounds and
try to see if the first element is generally the internal argument of the verb.

(42) Truck-driver
‘truck’ is the internal argument of ‘drive’, a Theme
Meat-eating
‘meat’ is the internal argument of ‘eat’, a Theme
Flower grower
‘flower’ is the internal argument of ‘grow’, a Theme
Money-lender
‘money is the internal argument of ‘lend’, a Theme
Brick-laying
‘brick’ is the internal argument of ‘lay’, a Theme/Patient
Bear-baiting
‘bear’ is the internal argument of ‘bait’, a Theme/Patient

If the internal argument of the verb appears in the syntax, the first element is interpreted as
an adjunct.

(43) home-schooling of children

‘home’ is an adjunct

If the verb lacks an internal argument, the first element in the compound is again interpreted
as an adjunct.

(44) Lake-swimming
‘lake’ is an adjunct

The adjunct interpretation is usual in case -en synthetic compounds:

(45) Moth eaten - eaten by moths

 179

Handmade - made by hands

Home-grown - grown at home

The first element in a synthetic compound cannot be interpreted as the subject of the verb,
which rules out the existence of compounds like (46):

(46) *Girl-swimming with the meaning ‘swimming done by girls’

*Kid-eating with the meaning ‘eating done by kids’

This distribution of arguments and ajuncts as left-hand elements in a compound is known as

the Argument Linking Principle and it was first proposed by Lieber (1983). according to the
Argument Linking Principle, the left-hand element in a synthetic compound, that is the non-head
is associates (linked) with an argument of of the verb stem in the head.
Sometimes, superficially similar compounds seem to feature a first element that is not linked
to one of the argument of the verb stem in the right-hand constituent, but an adjunct that specifies
the place, time or manner in which the event denoted by the verb takes place:

(47) Chain smoker

Party drinker
Daydreamer
Sunday closing
Spring cleaning
Low-flying

5 Left-headed Compounds

In English, there is a small number of endocentric compounds that have their heads to the
left, therefore contradicting the Right-Hand Head Rule. They add the plural morpheme to the
first element in the compound:

(48) Passer-by passers-by

 180

Notary public notaries public

Grant-in-aid grants-in-aid
Coat-of-mail coats of mail
Mother-in-law mothers-in-law

6 ‘Outbreak’ / ‘breakout’

There are cases in English where it seems that a verb and a particle are involved in deriving
a compound:

(49) a. to download
to outsource
to upgrade
to outbreak
b. the backswing
the input
the upshift

(50) breakdown
break-out
push-up
rip-off

The examples in (49) involve a V + P combination, while those in (50) involve a P + V

combination. In (49a), the result of this combination is a verb but in (49b) the result is a noun,
which cannot be accounted for in terms of compounding. If this were a case of compounding,
feature percolation would turn the resulting compound into a verb.
Examples such as those in (49) are best accounted for in terms of inversion from phrasal
combinations in which the particle follows the base word (Plag 2003: 143):

(51) Load down downloadNOUN/VERB

 181

Come in incomeNOUN/VERB
Put in inputNOUN/VERB
Built in inbuiltADJECTIVE

The examples in (50) are best seen as the result of conversion of a phrasal verb into a noun,
followed by stress shift:

(52) To break dówn VERB a bréakdown NOUN

To push úp VERB a púsh-up NOUN
To rip óff VERB a ríp-off NOUN

To briefly conclude, examples like ‘outbreak’ and ‘breakout’ are not compounds but the
result of two other word-formation mechanism: inversion and conversion.
 182

Exercises

1. Classify the following words as being the products of either inflection, derivation or
compounding. Justify your analysis in the potentially problematic cases (from Plag 2003: 163):

Blackboard eraser
Unacceptability
Flowerpots
Movie monster
Broad-shouldered
Hard-working
Speaking
Developmental

2. Name two general characteristics of English compounds. Use the data below for illustration
(from Plag 2003: 163):

Oak-tree
Drawbridge
Sky-blue
Mind-boggling

3. Classify the following compounds as exocentric, endocentric, possessive, appositional, or

coordinative (from Plag 2003: 163):

Frying pan
Redhead
Maidservant
Author-reader (exchange)
Austria-Hungary
Hardtop
 183

Silkworm
Man-machine (interaction)
Bootblack
German-English
Actor-manager
Gas-light

4. Are underdog, undercoat and overtax, overripe compounds or prefixed derivatives? Discuss.
(from Plag 2003: 164)

5. Rhyme is not the only phonological motivation behind compounding. Determine the
phonological factor that plays a role in the following (from Katamba & Stonham 2006: 306):

Zigzag
Dilly-dally
Sing-song
Tittle-tattle
Tick-tock
Wishy-washy

6. Examine the words below. What is the syntactic and semantic relationship between the two
words that make up each compound? (from Katamba & Stonham 2006: 317)

Schoolboy
Bedroom
Teapot

7. If in synthetic compounds, a noun satisfies the argument structure of the verb in the head of
the compound, show why the words belows should not be treated as synthetic compounds (from
Katamba & Stonham 2006: 323).
 184

Think tank
Call girl
Copycat
Playboy
Punch-line
Showman

8. Match up the definitions on the left with the correct compound on the right:
1. clumsy, awkward all thumbs
2. terrifying pig-headed
3. conceited, boastful bed-ridden
4. stubborn hair-raising
5. too weak to leave one’s bed big-headed
6. a short sleep in a chair a blackleg
7. smb. who attends a party without being invited a nightcap
8. a drink before going to bed a gatecrasher
9. smb. who works when others are on strike a catnap
10. unnecessary bureaucracy a scapegoat
11. smth unpleasant to look at red tape
12. a hopeless, preposterous search an eyesore
13. smb who is blamed for other people’s mistakes a wild goose chase

9. Make up compounds using the word given as the first constituent: fool

a. taking unnecessary risks

b. made in such a way that even a fool can understand and use safely
heart
a. sincere, deeply felt
b. causing deep grief or distress
c. suffering deeply from grief
stop
 185

a. a temporary substitute
b. late news printed in a special column in a newspaper
c. an instrument used for recording the time taken for a race head

a. forward motion, progress

b. self-willed, obstinate
c. words printed in large or heavy type above a newspaper article
d. place from which (e.g. police or army) operations are controlled foot

a. a safe place for the foot, especially when climbing

b. a row of lights along the front of a stage
c. a person who likes wandering across countries

3. Fill in the blanks with compound nouns derived from phrasal verbs from the list below:
break-out lay-off outbreak set-back bypass
uptake look-over outcry upkeep intake offshoot outlook

1. British hopes of a gold medal in the Olympic Games suffered a sharp ….. yesterday, when
Smith failed to qualify during the preliminary heats.
2. Owing to changes in the birth-rate, primary schools have had a smaller…..of new pupils this
year.
3. The British company is a (n) ….. of a much larger American concern.
4. There was a public … when the Post Office proposed higher charges for postal services.
5. Many men have already been made idle by the stoppage, and further ….. will be inevitable
unless agreement is reached soon.
6. Stately homes in Britain need massive sums of money for their…….
7. There was a sudden ….. of violence among students, following a period of relative calm.
8. Only one prisoner remains at large, following yesterday’s ….by six men from Dartmoor
prison.
9. You can avoid going through the town centre by taking the …..
10. We had time only to give the property a quick ……
 186

11. Some people have a very curious ….. on life.

12. I dropped several broad hints, but he seemed to be very slow on the …..
 187

Conclusions
 188

Chapter 1 Words and Morphemes

Reviewing questions and exercises:

Exercise 1.
Explain and illustrate the notions of:

a. Orthographic word

An orthographic word is defined as an uninterrupted string of letters which is preceded by a

blank space and followed by either a blank space or a punctuation mark.
In ‘All my students are amazed by morphology’, there are seven orthographic words.

b. Word-form

A word form is the result of inflectional morphemes attached to a base. Therefore, work,
works, working are all word forms of the same ‘word’.

c. Lexeme

A lexeme is defined as a unit of lexical meaning that exists regardless of the number of
inflectional endings it may have or the number of words it may contain. It is a basic abstract unit
of meaning.
In the examples below, be’, been and being are word forms and the lexeme, that is the
lexical meaning that covers all instance is BE.

Start from the following examples:

a. I have been to Paris twice.
b. I will never be afraid.
c. Any human being would frown upon such a deed.
 189

Exercise 2
Define the following terms and give three examples for each of them:

a. Morpheme

A morpheme is a minimal unit of meaning. Examples would be ‘re-’ in ‘rearrange’, which has
the meaning ‘do X again’, ‘-less’ in ‘homeless’, which has the meaning ‘without’, and ‘boy’,
which is a free morpheme.

b. Prefix

A prefix is a morpheme that attaches to the left of the base: ‘re-’ in ‘rearrange’, ‘a-’ in ‘asleep’,
‘mis-’ in ‘misguide’.

c. Suffix

A suffix is a morpheme that attaches to the right of the base: ‘-less’ in ‘careless’, ‘-ation’ in
‘agitation’, ‘-ment’ in ‘investment’.

d. Affix

An affix is either a prefix or a suffix.

e. Compound

A compound is made up of two bases: ‘greenhouse’, ‘schoolgirl’, ‘bedroom’.

f. Root

A root is a base that cannot be decomposed into further morphemes. For example, in
‘hopelessness’, the root is ‘hope’, arrived at after taking away ‘-bess’ first and then ‘-less’.
 190

g. Truncation

Truncation is a word-formation process that deletes a part of the base, as in ‘disco’, ‘lab’, and
‘demo’, whose bases are ‘discotheque’, ‘laboratory’, and ‘demonstration’.

Exercise 3
Identify the morphemes in the words listed below and specify whether they are free or bound
morphemes, prefixes, suffixes or roots:

a. Unbelievable

The morphemes we can identify are the following: the suffix ‘-able’, the prefix ‘un-’, and the
root ‘believe’. ‘believe’ is a free morpheme, while ‘un-’ and ‘-able’ are bound morphemes.

b. Intersperse
‘intersperse’ can be decomposed into a sequence ‘inter-’, which is a bound prefix, and a
sequence ‘-sperse’, which is a bound root.

c. Teachers

‘teachers’ is made up of the root ‘teach’, the agentive suffix ‘-er’, which is a bound morpheme,
and the inflectional suffix ‘-s’, which marks plural number and is a bound morpheme.

d. Numerous

‘numerous’ is decomposable into a bound root ‘numer-’ and a bound suffix ‘-ous’.
 191

e. Living-room

‘living-room’ is a compound decomposable into the stems ‘living’ and ‘room’, which are free
morphemes. The first term of the compound ‘living’ is, in its turn, decomposable into the
inflectional suffix ‘-ing’ and the root ‘live’.

f. Computerize

‘computerize’ is decomposable into the root ‘compute’, the agentive suffix ‘-er’, and the suffix
‘-ize’. The root is a free morpheme, while the suffixes are bound morphemes.

Exercise 4

a. List all morphemes in the sentence below. How many morphemes have you found?

The morphemes (both derivational, inflectional, and free) are the following: ‘high’, ‘school’,
student’, ‘-s’, ‘some’, ‘time’, ‘-s’, ‘do’, ‘-n’t’, ‘have’, ‘problem’, ‘-s’, ‘with’, ‘friend’, ‘-ly’,
‘advice’.

b. List all complex words and state which type of morphological process (inflection, derivation,
compounding) has affected them.

The complex words in the sentence are the following: ‘high-school’, a result of compounding,
‘students’, a result of inflection, ‘sometimes’, a result of inflection and compounding, ‘don’t’, a
result of inflection, ‘problems’, a result of inflection, and ‘friendly’, a result of derivation.

High-school students sometimes don’t have problems with friendly advice.

 192

Exercise 5
Define and illustrate the main differences between inflectional and derivational affixes.

First of all, inflectional morphemes encode grammatical categories, such as plural, person,
tense or case, while derivational morphemes do not.
Secondly, derivational affixes can occur at either side of the base, while inflectional
morphemes can only be expressed by suffixes.
Thirdly, inflectional affixes do not change the lexical category of the base, while
derivational affixes generally do.
Fourthly, the process of derivation sometimes exhibits semantic opacity, which refers to the
fact that one cannot infer the meaning of the whole based on the sum of the meanings of
constituent parts, in other words, the meaning is opaque or non-transparent.
Finally, inflectional categories tend to be fully productive while derivational categories show
restrictions as to the possible combinations.

Exercise 6 (Plag 2003:19)

Determine whether the adverbial suffix -ly in English is a derivational or an inflectional suffix,
by applying all the criteria discussed in the lecture and mentioned in exercise 5.

Slowly, aggressively, smoothly, hardly, rarely, intelligently, purposefully

The first criterion discussed above realted to the encoding of a specific grammatical category for
inflectional suffixes, such as number, tense, etc. ‘-ly’ does not seem to encode any grammatical
category; however, it does not seem to encode a lexical meaning, either, i.e., it does not
contribute any meaning of itself to the meaning of the derived word. Moreover, ‘-ly’ is
syntactically relevant, which means that syntax requires the derivation of ‘-ly’ formations
whenever an adjective modifies another adjective or a verb. This is an indication of the fact that
‘-ly’ behaves as an inflectional suffix.
 193

Another indication of its inflectional nature is its ordering with respect to derivational adjectival
suffixes. ‘-ly’ is always attached outside derivational suffixes and it is the last one to attach:
attent-ive-ly, care-ful-ly, etc.
The third criterion we mentioned is the change of lexical category induced by derivational
affixes. ‘-ly’ seems to turn adjectives into adverbs, but if we consider that adjectives and adverbs
are actually the same category A, and ‘-ly’ simply appears when adjectives modify adjectives or
verbs, the result would be that ‘-ly’ is not actually category-changing, which is another indication
of its being inflectional, rather than derivational.
If we consider the criterion of semantic opacity, and as stated before, ‘-ly’ makes no meaning
contribution of its own, just like the inflectional morpheme ‘-s’ for the third person singular,
present. There are a number of opaque ‘-ly’ formations, such as ‘hardly’, ‘scarcely’, and ‘barely’,
where the result exhibits a degree of opacity, but these formations are limited in number.
The last criterion has to do with productivity. Inflectional affixes are productive, in that they do
not impose any kind of restrictions on the base they attach to. Any verb will take ‘-s’ in the third
person singular present. Again, ‘-ly’ behaves as an inflectional morpheme, in that most
adjectives take ‘-ly’, with a limited number of exception, such as ‘fast’
The result of this discussion is that ‘-ly’ is rather an inflectional, and not a derivational,
morpheme.

Chapter 2 On Complex Words

Reviewing questions:

1. State the definition of the sign in Saussurian terms.

The linguistic sign is the association between a signifier and a signified; the signifier is the
acoustic image while the signified represents the mental concept associated with the sounds.

2. Saussure characterized the sign as being arbitrary and having relative value. Comment on this.
 194

The arbitrariness of the linguistic sign refers to the fact that there is no motivation in the choice
of a particular signifier for a particular signified. The association between the signifier and the
signified is a matter of conventionalized usage. The linguistic sign has a relative value according
to the contrasts it forms with other signs. The value of a linguistic sign does not come from its
intrinsic signification, nor does it come from its acoustic image. Linguistic meanings are
not independent but dependent on other linguistic signs within their language system.

3. Define the notion of simple and complex linguistic signs, constituent sign.

A simple sign is a sign that cannot be decomposed into two or more constituent signs, such as
‘student’, ‘discuss’, ‘cranberry’. A complex sign is a sign that can be decomposed into two or
more smaller constituent signs, such as ‘rearrange’, ‘misguide’, ‘treeless’. A constituent sign is a
minimal unit of constant form and constant meaning. For example, in ‘rearrange’, there are two
constituent signs - ‘re’ and ‘arrange’, both having a constant form and a constant meaning.

4. Give the definition of morpheme, allomorph, conditioning.

A morpheme is a minimal unit of meaning, such as ‘-er’ with the meaning of agent in ‘teacher’,
‘reader’, etc. An allomorph is a variant of the morpheme, which is conditioned phonologically,
morphologically or lexically. For example, the plural morpheme ‘-s’ has three allomorphs, which
are conditioned phonologically, in that neighboring sounds condition the realization of ‘-s’ as
either /z/, /s/, or /iz/.

Exercises

1. Are the following simple or complex signs?

a. Blackberry
Blueberry
Gooseberry
 195

Butterfly
Withstand
Understand
Undergo
Undertake

The words in a. are simple signs because, although we can identify, by commutation, two
morphemic sequences in their make-up (under + take, with+ stand, butter + fly), these sequences
do not have constant form and constant meaning in all the contexts. We can use ‘butter’ in other
contexts, but the meaning it has in those contexts is not related to its role in ‘butterfly’.

b. Misguide
Misinform
Misplace
Mislead

The words in b. are complex signs. First, we identify by commutation two sequences: ‘mis-’ and
a verbal base. Next, we try to identify is ‘mis-’ has constant form and constant meaning in all the
situations. The meaning identified can be paraphrased as ‘wrongly’.

c. Mistake
Mischief
Mistress

If we apply the commutation test, we identify a sequence ‘mis-’, which seems to be identical
with the one in b. However, a ‘mistake’ is not a wrong take, neither is a ‘mischief’ a wrong chief.
This means that the words in c. are simple signs and ‘mis-’ is not a constituent sign.

d. Discolour
Disclaim
Dismantle
 196

Disagree
Disable

The commutation test uncovers two morphemic sequences: ‘dis-’ and a verbal base. The attempt
at identifying a constant sense associated with a constant meaning for ‘dis-’ proves successful.
‘dis-’ can be paraphrased as ‘the reverse of X’, where X is the meaning of the verb ‘dis-’ attaches
to.

e. Disciple
Discreet
Discuss

The question is now whether the ‘dis-’ identified in the words in e. is the same morpheme ‘dis-’
identified above. The meaning test disproves the hypothesis that ‘dis’- in these words is a
constituent sign. A disciple is not the reverse of a ‘sciple’, nor does being ‘discreet’ entail the
reverse of a ‘creet’ property. This means that the words in e. are simple signs.

f. Retake
Repay
Rearrange
Retry
Rephrase
Reformat
Restate

The commutation test identifies the sequence ‘re-’ and a number of verbal bases. Trying to see if
‘re-’ has constant form and constant meaning, we can paraphrase its meaning as ‘do X again’.
This means that ‘re-’ is a constituent sign and the words in f. qualify as complex signs.
 197

g. Receive
Refer
Resume
Rest
Retain
Remain

The words in g. feature a sequence ‘re-’, as in the examples in f., which is formally similar in all
cases. The question is to determine if it has constant meaning, that is, if it can be paraphrased as
‘do X again’. The answer is obviously no, which means that the words in g. are simple signs.
Upon closer inspection, we notice that they are Latinate words, which are no longer analyzable
as made up of constituent signs.

h. Recover
Refuse
Remove
Report
Refrain
Regard

The words in h. are simple signs, just as the ones in g. The sequence ‘re-’ cannot be understood
as ‘do X again’. We will returns to Latinate words in subsequent chapters.

2. Identify the suffixes in the underlined words. To what word-class (lexical category) do the
words to which the suffixes are added belong, and what word-class results? Decide whether the
suffixes are inflectional or derivational. Justify your decision.

I ducked He was sheepish He ducks

Two ducks Three ducklings
He is humourless You are ducking the issue
 198

I ducked - the inflectional suffix ‘-ed’ for the past tense is added to the lexical category V (verb)
yielding another verb.
Two ducks - the inflectional suffix ‘-s’ for the plural is added to the lexical category N (noun)
yielding another noun.
He is humourless - the derivational suffix ‘-less’, with the lexical meaning ‘without’, is added to
the lexical category N (noun) yielding an adjective (A).
He was sheepish - the derivational suffix ‘-ish’, with the lexical meaning ‘having the
characteristics of’, is added to the lexical category N (noun) yielding an adjective (A).
Three ducklings - the derivational suffix ‘-ling’, with the meaning ‘place of origin or quality’ (as
in ‘earthling’, ‘changeling’, etc.) is added to the lexical category N (noun), the result is also the
category N; then the inflectional suffix ‘-s’ for the plural is added to the lexical category N, the
result being also a noun.
You are ducking the issue - the inflectional suffix ‘-ing’ for the progressive be-ing is added to the
lexical category V (verb), yielding another verb.
He ducks - the inflectional suffix ‘-s’ for the third person singular present is added to the lexical
category V (verb) yielding another verb.

3. What is the plural morpheme in English? What is the allomorph of the plural morpheme that
occurs in the group of words below? Is the choice phonologically, morphologically or lexically
conditioned?

(i) addenda, data, strata, media, desiderata

(ii) stimuli, radii, fungi, alumni
(iii) asses, fishes, badges, leeks, moths, cups, bards, mugs, rooms, keys, shoes

The nouns in (i) come from Latin. They represent the plural of singular neuter nouns, third
conjugation class (singular -um). As such, they are an instatiantion of the plural morpheme /S/,
which is lexically conditioned.
The nouns in (ii) also come from Latin. They represent the plural of masculine nouns, second
conjugation class (singular -us). As such, they are an instantiation of the plural morpheme /S/,
which is lexically conditioned.
 199

The nouns in (iii) exemplify the process of phonological conditioning. The plural morpheme /S/
is pronounced /s/ when it follows a voiceless plosive (cups, leeks); it is pronounced /z/ when it
follows a voiced consonant (mugs); it is pronounced /iz/ when it follows a sibilant (asses, fishes).

4. Study the following examples. Do the underlined morphemic segments below belong to an
invariant morpheme? Explain.

(i) Inexpensive
Impossible
Illogical
Irrational
Irregular
Ingenious
Irresistible
Immobile
Inconsistent
Incomprehensible
Illegal
Illiterate
Incompetent
Irresponsible
Indifferent
Inevitable

The underlined morphemic sequences belong to an invariant morpheme /IN/, with the lexical
meaning ‘the reverse of the property X’. The /n/ sound becomes /m/ when followed by a bilabial
sound (impossible, immobile). It becomes a lateral sound when followed by a lateral sound and a
rhotic sound when followed by a rhotic sound. It is pronounced /n/ elsewhere. As such the
process is one of phonological conditioning (regressive assimilation).
 200

(ii) Baker
Teacher
Driver
Survivor
Beggar
Governor

The underlined morphemic sequences belong to an invariant morpheme /ER/, with the meaning
of Agent. The process at stake is no longer one of phonological conditioning but of lexical
conditioning, the words ‘survivor’, ‘beggar’ and ‘governor’ surviving from Anglo-French with
parts of the original.

5. Multiple affixation: identify the root morpheme and the affixes that make up the complex
words below:

diction, dictate, dictatorial, contradict, contradictory, contradiction, contradictoriness,

benediction

Diction - root ‘dict’ + suffix ‘-ion’

Dictate - root ‘dict’ + suffix ‘-ate’
Dictatorial - root ‘dict’ + suffix ‘-ate’ + suffix ‘-ory’ + suffix ‘-al’
Contradict - prefix ‘contra-’ + root ‘dict’
Contradictory - prefix ‘contra-’ + root ‘dict’ + suffix ‘-ory’
Contradiction - prefix ‘contra’ + root ‘dict’ + suffix ‘-ion’
Contradictoriness - prefix ‘contra-’ + root ‘dict’ + suffix ‘-ory’ + suffix ‘-ness’
Benediction - prefix ‘bene-’ + root ‘dict’ + suffix ‘-ion’

6. Take the free morpheme ‘nation’ and add as many prefixes and suffixes as you can.

National, international, nationality, internationality, nationalize, internationalize, nationalism,

internationalism….
 201

7. Examine carefully the following text:

He had begun to read the novel a few days before. He had put it down because of some urgent
business conferences, opened it again on his way back to the estate by train; he permitted himself
a slowly growing interest in the plot, in the characterisations. That afternoon, after writing a
letter giving his power of attorney and discussing a matter of joint ownership with the manager
of his estate, he returned to the book in the tranquillity of his study which looked out upon the
park with its oaks. (Julio Cortazar, Continuity of Parks)

List five free and three bound morphemes that occur in the sentence.
List three functional morphemes in the sentence.

Five free morphemes: ‘read’, ‘the’, ‘novel’, ‘because’, ‘study’.

Three bound morphemes:‘-s’ in ‘conferences’, ‘-ity’ in ‘tranquility’, ‘-ship’ in ‘ownership’.
Three functional morphemes: ‘-s’ in ‘conferences’, ‘-ed’ in ‘permitted’, ‘-ing’ in ‘writing’.

8. Explain the notion of morphologically-conditioned (base) allomorphy starting from the

following examples (Plag 2003: 42)

Active - activity
Affect - affection
Curious - curiosity

The attachment of suffixes may trigger changes in the stress pattern of the base, along with
changes in pronounciation. This is known as base allomoprhy, conditioned by morphological
factors.
When the suffix ‘-ity’ attaches to the adjective ‘active’, a change in the stress pattern happens.
Stress is shifted from the first syllable of the base to the syllable preceding the suffix: áctive -
actívity.
When the suffix ‘-ion’ is attached to the verb ‘affect’, the sound [t] is palatalized.
 202

When the suffix ‘-ity’ is attached to the adjectival base ‘curious’, a change in the stress pattern
happens. Stress is shifted from the first syllable of the base to the syllable preceding the suffix:
cúrious - curiósity.

Chapter 3 Productivity

Reviewing questions:

1. Make a list of possible problems that arise when thinking about the definition of morphemes
as ‘things’ that perfectly map form and meaning. Illustrate them.

(i) The first phenomenon that poses problems for a definition of morphemes is conversion.
Conversion is a process by means of which words are derived from other words without any
visible marking:

I walk in the park everyday.

Let’s take a walk.

In the case of conversion, we could speak of a zero-morph, which is added to the base to derive a
new word. For example, the verb ‘water’ is form from the noun ‘water’ by adding to the base
noun a zero-morph with the meaning ‘apply X’. (Plag 2003: 22)
(ii) The second phenomenon that poses problems to an analysis of morphemes as perfect
form-meaning mappings is truncation. In this case, it is not clear what the morph is and what
semantic difference there is between truncated and non-truncated forms. We may assume that the
morph is the process of deleting material.

Laboratory - lab
 203

(iii) A third problematic phenomenon is the existence of vowel alternations that add some
meaning to the base:

to fall - to fell
sing - sang
foot - feet

In these cases, it could be argued that the vowel carries causative meanings, the past tense
meanings and plural meanings.
(iv) A fourth problematic phenomenon is the existence of discontinuous morphemes.

the progressive ‘be - ing’

the perfective ‘have - en’

(v) A fifth phenomenon that poses problems to an analysis of morphemes as ‘things’ is the
existence of meaningless morphs:

infer
confer
prefer
refer
transfer

The sequence ‘-fer’ qualifies for the status of morph but it has no meaning for speakers of
Modern English. They are Latinate verbs and the morph ‘-fer’ has a low degree of independence.
In Latin, the root ‘ferro’ carried the meaning ‘to carry’.

2. Define the notion of productivity and give examples of productive and unproductive affixes in
English.
 204

The term productivity refers to the property of an affix to be used in order to coin new complex
words. Productivity varies with different affixes: some affixes are productive, other are less
productive. Productive affixes are: ‘-er’, ‘-able’, ‘-ness’ while non-productive suffixes are: ‘-th’,
‘-ment’, etc.

3. Define the notions of possible and actual words and illustrate them. Give two properties of
possible words.
A possible word is defined as a word whose semantic, morphological structure is in accordance
with the rules and regularities of the language but is not attested in dictionaries, like
‘cannibalizable’. An actual word is a word that is attested in dictionaries and used in actual
speech by the speakers of the language.
One important property of possible words is predictability of meaning. We can predict what
‘cannibalizable’ may mean precisely because it is formed in accordance with the rules of the
language and ‘-able’ means ‘something that can be V-ed’. From this we can extract a second
important property of possible words, namely they cannot be idiosyncratic.

4. Define the notion of mental lexicon. How are lexical items stored (and retrieved from) in the
mental lexicon?

The mental lexicon is a kind of dictionary that is stored in the brain instead of being written on
paper. Lexical items are stored in the lexicon in two ways. The lexicon contains a list of bases
and a list of affixes, which the speakers combine when they speak. However, there is a second
route for speakers: in the case of frequently used derived words, they are stored as such (already
derived) in the lexicon for the speaker to quickly retrieve them.

5. How do we measure productivity?

Productivity is measured by counting the instances of neologisms derived with a specific affix,
which enter the language at a given point in time. We should also take into account the criterion
of frequency, that is how often a neologism formed with a certain affix is used at a given time.
Linguists do that using large text corpora, like the OED.
 205

6. What are the functions of word-formation? Illustrate.

(i) The referential function of word-formation, which refers to those instances when speakers
coin new words to designate new concepts. In such cases, a new word is created in order to give
a name to a new thing or concept.

The Time Patrol also had to unmurder Capistrano’s great-gramdmother, unmarry him from the
pasha’s daughter in 1600, and uncreate those three kids he had fathered. (Kastovsky 1986: 594,
quoted in Plag 2003: 59)

(ii) The syntactic recategorization function of word-formation, which refers to those instances
when speakers use syntactic recategorization to condense the information. Longer phrases or
whole clauses can be replaced by complex words. ‘His clumsiness’ may replace ‘the fact that he
is clumsy’.

Yes, George is extremely slow. But it is not his slowness that I find most irritating.

(iii) The attitudinal function of word-formation, which refers to those instances when speakers
want to convey a particular attitude.

Come here, sweetie.

hat a good doggie.

5. Define pragmatic restrictions on word-formation. Illustrate.

(i) Fashion, a pragmatic restriction which simply refers to certain affixes that are preferred by
speakers at a certain point
mega-
giga-
mini-
 206

super-

(ii) the requirement of existence, which refers to the fact that words are formed to denote only
items, actions or qualities that exist. Existence may cover fictional or mythological domains.
(iii) the nameability requirement. To put it very simply, this restriction requires that the
concepts encoded by derivational categories should be very simple and general:

un- ͇ ‘not X’
-en ͇ ‘make X’
-er ‘the one who does X’

6. Define structural restrictions of word-formation. Illustrate.

(i) Phonological restrictions - refer to individual sounds or to syllable structure or stress. For
example, the nominal suffix ‘-al’ attaches to verbs that end in a stressed syllable, so verbs ending
in an unstressed syllable are excluded as possible bases.
(ii) Morphological restrictions - refer to the fact that affixes can be sensitive to the morphological
structure of their base words. For example, almost every verb derived with ‘-ize’ can be
nominalized by adding ‘-ation’.
(iii) Semantic restrictions, which refer to the fact that, sometimes, affixes attach only to specific
type of semantic categories. For example, ‘-ee’ derivatives must refer to sentient entities.
(iv) Syntactic restrictions, which refer to the fact that word-formation rules are generally
restricted to members of a certain syntactic categories. For example, ‘-un’ has to attach to
adjectival bases (and a limited number of verbs, such as ‘un-earth’, ‘un-fold’, ‘un-dress’),
‘-able’ attaches to verbs, and ‘-al’ attaches to nouns.

11. Define blocking. Illustrate

Blocking refers to the phenomenon where a complex form is not derived because another form
already exists.
(i) Homonymy blocking refers to those case where an already existing form blocks the creation
of a semantically or phonologically identical derived form:
 207

liver ‘inner organ’

?liver ‘someone who lives’

(ii) Synonymy blocking refers to those cases where the creation of a new word is blocked
already existing synonymous lexemes.

*dog-ess
*horse-ess
bitch
mare

Exercises
1. What property of actual words can we infer from the following examples?

Readable
Teachable
Writable
Understandable

Vs.

Knowledgeable
Probable

The examples help us infer one important property of actual words, namely only actual words
can be idiosyncratic. Although ‘knowledgeable’ contains the suffix ‘-able’, it does not have
compositional meaning, in other words it does not mean ‘something that can be knowledged’.
 208

2. Starting from the two ways in which items are stored and retrieved from the mental lexicon
(the whole-word route and the decomposition route), describe the two possible ways in which the
word ‘insane’ is stored in and retrieved from the lexicon. Which route do you think is likely to be
used by most speakers? Why?

The word ‘insane’ is stored and retrieved from the lexicon in two ways. The decomposition route
entails the fact that speakers retrieve the base ‘sane’ and the prefix ‘in-’ and the derive the word
‘insane’ in the morphological component. The whole-word route entails the fact that speakers
retrieve the word ‘insane’ already derived. The route which is likely to be used by most speakers
is the whole-word route, because the word ‘insane’ is frequently used.

3. Take a look at the following table (from Plag 2003: 51). Which are the words for which the
whole-word route is likely to be taken by speakers and why?

Frequencies of ‘-able’ derivatives in the British National Corpus

-able derivative Frequency
Abominable 84
Absorbable 1
Abstractable 2
Abusable 1
Acceptable 3416
Accountable 611
Accruable 1
Achievable 176
Acid-extractable 1
Actable 1
Actionable 87
Actualizable 1
Adaptable 230
Addressable 12
 209

Adjustable 369
Admirable 468
Admissable 2
Adorable 66
Advisable 516

The whole-word route is likely to be taken for words with the frequent occurrences, such as
‘acceptable’, ‘admirable’, ‘advisable’. The decomposition route is likely to be taken for words
with a limited number of occurrences, such as ‘accruable’ and ‘acid-extractable’.

4. Using the discussion on phonological restrictions operating in word-formation, explain the

impossibility of the following formations.

Noun-forming ‘-al’
Arrive - arrival enter - *enteral
Betray - betrayal promise - *promiseal
Construe - construal manage - *manageal
Deny - denial answer - *answeral
Propose - proposal forward - *forwardal

The phonological restriction that operates is stress-related. Nominal ‘-al’ attaches to verbs that
end in a stressed syllable, so verbs ending in an unstressed syllable are excluded as possible
bases. This restriction does not mean that all verbs ending in a stressed syllable can take ‘-al’
(*delayal, *explainal), which means that there are more restrictions operating in word-formation
at the same time.

5. Using the discussion on phonological restrictions operating in word-formation, explain the

impossibility of the following formations.

a. Black - blacken (from Plag 2003: 62)

 210

Fat - fatten
Long/length - lengthen
Loose - loosen
Wide - widen

Suffixation of verbal ‘-en’ is subject to a segmental restriction. The last sound of the base must
be an obstruent or a sonorant.

b. Fine - *finen
Dull - *dullen
Long - *longen
Low - *lowen

The verbal suffix ‘-en’ cannot attach to a base that ends in a sound different from sonorants or
obstruents (liquids, glides, nasals, etc.)

c. Candid - *candiden
Equivalent - *equivalenten
Expensive - *expensiven
Hilarious - *hilariousen
Valid - *validen

The verbal suffix ‘-en’ only attaches to bases that contain one syllable.

6. The nominal suffixes ‘-ation’, ‘-ication’, -ion’, ‘-ance’, ‘-al’, ‘-age’, ‘-y’, and ‘-ment’ are
roughly synonymous. The obvious question is which mechanisms govern their distribution, i.e.
which verb takes which suffix. we will try to answer this question for a subset of verbs, namely
those derived by the suffixation of ‘-ify’, ‘-ize’ and ‘-ate’. The data below exemplify the
nominalization of the pertinent verbs ‘magnify’, ‘verbalize’ and ‘concentrate’. State the
restrictions that constrain the selection of nominalizing suffixes with derived verbs of these types.
(Plag 2003: 70)
 211

Magnification verbalization concentration

*magnify-ation *verbalize-cation *concentrate-ation
*magnify-ion *verbalize-ion *concentrate-cation
*magnify-ance *verbalize-ance *concentrate-ance
*magnify-al *verbalize-al *concentrate-al
*magnify-age *verbalize-age *concentrate-age
*magnify-ment *verbalize-ment *concentrate-ment
Upon close inspection of the verbal bases ‘magnify’, ‘verbalize’, and ‘concentrate’, the result s
that all three have three syllables and their stress pattern is the same, with stress on the first
syllable. Therefore, the restrictions at work in these case cannot be phonological. The next stept
is to investigate the morphological restrictions. All three verbs already contain suffixes: ‘-ify’,
‘-ize’, and ‘-ate’. The suffix ‘-cation’ attaches to all verbs ending in ‘-ify’ (stratify -
stratification); ‘-ation’ attaches to all verbs ending in ‘-ize’ (organize - organization), and ‘-ion’
attaches to all verbs ending in ‘-ate’ (agitate - agitation). this means that the suffixes ‘-cation’,
‘-ation’, and ‘-ion’ are sensitive to the morphological structure of the base words, a phenomenon
known as ‘base-driven restrictions’.

7. The verb-forming suffixes ‘-ify’ and ‘-ize’ impose severe phonological restrictions on their
possible base words. There seem to be three classes of words involved, one class taking
obligatorily ‘-ize’, one class taking obligatorily ‘-ify’, and one minor third class which can take
both suffixes. Try to establish the pertinent phonological restrictions, using the following data,
which are all 20th century neologisms from the OED (Oxford English Dictionary). Hint: consider
the number of syllables and the stress patterns for all derivatives and try to find the appropriate
generalization (Plag 2003: 70-71)

a. -ize derivatives
Academicize accesorize absolutize acronymize adjectivize
Aerosolize anodize anthropologize bacterize Baskonize
Bolshevize Bonderize bovrilize cannibalize capsulize
 212

*artize *massize *bourgeoisize *Japanize *speechize

b. -ify derivatives
Artify bourgeoisify gentrify jazzify Karstify
Massify mucify mythify Nazify negrify
*randomify *federalify *activify *modernify *Germanify

Upon inspection of the data, we notice that ‘-ize’ is always preceded by one or more unstressed
syllables while ‘-ify’ is always preceded by a stressed syllable. Another distinction is that ‘-ify’
can only attach to monosyllabic bases (art, jazz) or to bases with stress on the last syllable
(bourgeoisify), while ‘-ize’ can only attach to words that contain more than one syllable. There
are a few exceptions to these rules: words that end in unstressed ‘i’ (gentry, Nazi) can take ‘-ify’
only of one of the ‘i’s is deleted. From this description, we can infer that ‘-ize’ and ‘-ify’ are in
complementary distribution.

Chapter 4 Structuralist approaches to word structure

Reviewing questions

1. Define ‘distribution’. Give the distribution of bare plurals and bare mass nouns in English and
Romanian.

Distribution refers to the totality of enviroments in which an element appears. In English bare
plurals are allowed in subject position, object position, and the position of objects of preposition,
a distribution which is paralleled by that of bare mass nouns:

Students study.
Students like teachers.

Milk is good for you.

He drank milk.
 213

In Romanian, the distribution of bare plurals and bare mass nouns is restricted to the object
position. The only bare plural subjects in Romanian are post-verbal subjects in episodic
sentences:

Studenții studiază. Laptele e bun pentru sănătate.

Am văzut studenți. Am băut lapte.
Vin studenți mereu aici. Curge apă la baie.

2. Define ‘formal definitions’. Give formal definitions for the direct object in English and for the
category N(oun).

Formal definitions are definitions that exclusively rely on the position of the element in the
sentence. Formal definitions were meant to replace notional definitions, which were thought to
be unreliable. According to a notional definition, the noun is the part of speech that designates
beings, objects, natural phenomena, etc. A formal definition of the noun would look at its formal
features, such as number, gender, case. The formal definition of the noun is the following: the
category that possesses phi-features (gender, number, person) and that may be licensed as an
argument. Similarly, the direct object in English is that category that sits in post-verbal position,
strictly adjacent to the verb [VP__].

3. Complete the following:

(i) IC analysis is a technique of breaking up constituents into smaller units, proceeding in binary
steps.
(ii) Morpheme sequences may display two types of construction: endocentric and exocentric
(iii) A construction is endocentric when the distribution of the whole is the same as the
distribution of the head.
(iv) A construction is exocentric when the distribution of the whole is not the same as the
distribution of the head and there is no attributive relation between the head and the non-head.
 214

Exercises:

1. Identify three strings and three constituents in the following sentence:

Derivational morphology will never be my cup of tea.

Strings: ‘morphology will’, ‘will never’, ‘be my’

Constituents: ‘derivational morphology’, ‘cup of tea’, ‘will be’

2. Identify the ICs and discuss the type of construction evinced by the following morpheme
sequences:

(i) old man - ICs: ‘old’ and ‘man’, endocentric construction

cooking butter - ICs: ‘cooking’ and ‘butter’, ‘cook’ and ‘-ing’, endocentric construction
searchlight - ICs: ‘search’ and ‘light’, endocentric construction
redskin - ICs: ‘red’ and ‘skin’, exocentric construction
bluebeard - ICs: ‘blue’ and ‘beard’, exocentric construction
butterfingers - ICs: ‘butter’ and ‘fingers’, ‘finger’ and ‘-s’, exocentric construction
asleep - ICs: the prefix ‘a-’ and ‘sleep’, exocentric construction
endanger - ICs: the prefix ‘en-’ and ‘danger’, exocentric construction
dethrone - ICs: the prefix ‘de-’ and ‘throne’, exocentric construction
commercialize - ICs: ‘commerce’, ‘-al’, ‘-ize’, exocentric construction
discovery - ICs: ‘discover’ and the suffix ‘-y’, exocentric construction
Unbearable - ICs: ‘un-’ and ‘bearable’, ‘bear’ and ‘-able’, exocentric construction

(ii) sad fiction writer

old car dealer

In these cases, which are examples of syntactic ambiguity, the interpretation of the phrases
depends on the identification of Immediate Constituents. If the identification is [[sad fiction]
 215

writer], the phrase means ‘a writer of sad fiction’. If the identification is [sad[fiction writer]], the
meaning is ‘a fiction writer who is sad’.

3. Examine the words below. What is the syntactic and semantic relationship between the
constituents that make up the compound? Identify the word-class of the two constituent words
and the word-class of the compound; identify the head; -identify the type of compounds

Schoolboy - endocentric compounds, the head is ‘boy’, a schoolboy is a kind of boy, attributive
relation
Bedroom - endocentric compound, the head is ‘room’, a bedroom is a kind of room, attributive
relation
Teapot - endocentric compound, the head is ‘pot’, a teapot is a kind of pot, attributive relation
Water-lily - endocentric compound, the head is ‘lily’, a water-lily is a kind of lily, attributive
relation

Hothouse
Oversight
Greenfly
Skyline
World-wide
User-friendly
Short-lived
Good-natured
Outspoken
Near-sighted
Outstay
Undersell
Passer-by
Notary public
Mother-in-law
Attorney general
 216

All the compounds in this group are endocentric, based on an attributive relation between the
head and the non-head. The distribution of the compound is the same with the distribution of the
head.

Outskirts - exocentric compound, the head is ‘skirts’, outskirts are not a kind of skirts, no
attributive relation
Lazy-bones - exocentric compound, the head is ‘bones’, a person who is referred to as
‘lazybines’ is not a kind of bones. Moreover, the distribution of the compound is not the same
with the distribution of the head.

Dimwit
Butterfingers
Dare-devil
Blockhead
Turncoat
Loudmouth

All the compounds in this group are exocentric compound. There is no attributive or
modifier-head relation between the head and the non-head. The distribution of the compound is
not the same with the distribution of the head. Most exocentric compounds designate types of
people by metaphorical extension.

4. How do you know that ‘head’ is a noun or verb in the following:

The head of the village school has arrived. - ‘head’ is a noun because it appears in the
distribution of the definite article ‘the’ and takes a prepositional ‘of’ genitive.
She will head the village school. - ‘head’ is a verb because it has the distribution of a verb: it is
preceded by a modal and takes a direct object.
The heads of the village schools have arrived. - ‘heads’ is a noun because it is accompanied by
the definite determiner and takes the plural ‘-s’.
 217

She headed that school. - ‘headed’ is a verb because it is inflected with the past tense morpheme
‘-ed’.

5. Mark the syllable that receives main stress in the following pairs. What do you think is the
morpheme that converts verbs into nouns in these cases? (from Katamba and Stonham 2006)

VERB NOUN NOUN VERB

Torment torment pattern pattern
Survey survey advocate advocate
Protest protest patent patent
Digest digest lever lever
Progress progress

In the first group of verbs, stress is on the second syllable. When the lexical items function as
nouns, stress travels to the first syllable. We could talk about a conversion morpheme, which
takes the form of stress.
In the second group of verbs and nouns, stress is always on the first syllable. This time, the
conversion morpheme mentioned above is a kind of zero-morpheme.

6. Consider the data in the following sentences. Which of the four morphemes would you
consider bound and morphemes and which would you consider free morphemes? If you decide
that some of them could be considered free morphemes, what would be the more accurate
treatment of the formations: derived words or compound words?

(i) There was never an error-free text, Cropper said.

(ii) Now the lanes were carless, lawless.
(iii) Arriving on her broomstick at the prison-like school gates, Mildred peered through the
railings into the misty playground.
(iv) She had been a teacher and she had made sure the girl went to a good school: “my granny
had more influence on me education-wise. (Plag 2003: 72)
 218

A bound morpheme can only occur if it is attached to some other morpheme. All italicized
morphemes above can occur on their own. However, only in the case of ‘error-free’ and
‘prison-like’ do the free elements ‘free’ and ‘like’ have the same meaning (error-free = free of
erros, prison-like = like a prison). It follows that ‘free’ and ‘like’ are free morphemes and
‘error-free’ and ‘prison-like’ are compounds.

7. Look at the italicized morphemes below. Are they affixes or bound roots? Are the formations
the results of derivation or compounding?

Biochemistry
Biorythm
Biowarfare

‘Bio-’ is a bound root, with the meaning ‘life’. If it were a prefix and ‘-logy’ a suffix, ‘biology’
would be formed of a suffix and a prefix. The words are compounds, not derived formations.

Photograph
Photoionize
Photoanalysis

‘Photo-’ is a bound root, with the meaning ‘light’. The words it forms are compounds.

Geology
Biology
Neurology

Other examples of compounds with bound roots contain ‘-logy’, with the meaning ‘science of’.
All the words in 7 are examples of neoclassical elements, lexemes borrowed from Latin or Greek,
also known as combining forms, because the roots are obligatorily bound.
 219

8. Look at the following data on gapping. Gapping refers to a phenomenon where it is possible
to coordinate two words by leaving put one of the constitutive elements. Try to find the
appropriate generalization to account for the data.

(i) Possible gapping in compounds

Word and sentence structure
Computer and cooking courses
Word-structure and -meaning
Speech-production and -perception
(ii) Possible gapping with prefixes
De- and recolonialization
Pre- and postwar
Over- and underdetermination
(iii) Possible gapping with suffixes
Curious- and openness
Computer- and internet-wise
Child- and homeless
(iv) Impossible gapping with suffixes
*productiv(e)- and selectivity (productivity and selectivity)
*feder- and local (federal and local)
*computer- and formalize (computerize and formalize)

The close inspection of the data reveals that gapping is only possible with affixes that do not
form one prosodic word with their base.

9. What do the suffixes ‘-ion’ and ‘-ure’ have in common, apart form their being nominalizing
suffixes? Look at the following data and state the generalization. Focus on the
morpho-phonological side of the matter (Plag 2003: 105)

Erode - erosion
Conclude - conclusion
 220

Confuse - confusion
Persuade - persuasion

The bases end in [d], an alveolar voiced plosive. The attachment of the suffix ‘-ion’ triggers
phonological alternations, which generally happens with suffixes of Latin origin and which begin
in a vowel.

Compose - composure
Erase - erasure
Close - closure
Dispose - disposure

The bases end in [z], an alveolar voiced fricative. The attachment of the suffix ‘-ion’ triggers
phonological alternations, which generally happens with suffixes of Latin origin and which begin
in a vowel.
Both suffixes add one more syllable to the base and they incorporate the final consonant of the
base into the onset of the new last syllable: [Link] - [Link].
Finally, it can be noticed that all verbal bases are stressed on the last syllable, which means that
both suffixes are stress neutral.

10. What do the suffixes ‘-ity,’, ‘-ize’, ‘-ify’ and ‘-ism have in common? Look at the following
data and state the generalization. Focus on the morpho-phonological side of the matter (Plag
2003: 105)

Atomic - atomicity
Iambic- iambicity
Historic - historicity
Opaque - opacity
Historic - historicize
Classic - classicize
Erotic - eroticize
 221

Opaque - opacify
Classic - classicism
Romantic - romanticism

Just like the suffixes in exercise 9, these suffixes attach to Latin bases and begin with a vowel. It
can be notice that all bases end in [k] and all derivatives turn that [k] into a [s], a process known
as velar softening. Being vowel-initial, the suffixes incorporate the consonant at the end of the
base into the onset of their syllable: [Link].

Chapter 5 Generativist approaches to morphology

Reviewing questions:

1. Discuss the innateness hypothesis.

The innateness hypothesis refers to the genetic programming of human being that enables them
to learn languages. The native language is acquired (not learned) effortlessly and without specific
instruction, because the brain contains a language acquisition device enabling the acquiring of
the language. In this framework, the child possesses a language faculty, which is part of his
genetic endowment. The child’s mind is not, as Chomsky notes a ‘blank state’, but it possesses a
blueprint for language, a device that allows him to acquire/develop language.
It is important to emphasize the fact that grammars are learnable in the sense that, by the age of
three, any child is able to fully understand the language that is being spoken around him, he will
be able to hear sounds, he will be able to figure out the constituents that sentences are made up of,
he will be able to assign meaning to these words and sentences.

2. What role did morphology have in early Generative Grammar?

With the advent of early Generative Grammar, morphology lost relevance in the general
organization of linguistic theory. If in the framework of structuralism, morphology played a
 222

central part, in generativism syntax is the main focus. The question morphologists have striven to
answer for decades is whether word-formation can be accounted for in terms of syntactic rules or
whether word-formation rules belong to a separate component, such as the lexicon.
In the view of early Generative Grammar, the lexicon contained only simple words, without
including compounds or derived words, which could only be constructed by the transformational
component. All possible variations in form that words and morpheme might show were assigned
to the phonological component.
Generative Grammar assumed no morphological rules. All morphological relations were
accounted for by phrase structure rules, which were allowed to operate on words and
morphemes.
For example, in ‘Aspects of the Theory of Syntax’, Chomsky proposed to use ‘nominalization
transformations’ to account for the relation between word pairs such destroy-destruction, stating
that ‘the phonological rules will determine that nom+destroy becomes destruction’ (Chomsky
1968:11). Therefore, the morphological relation between destroy and destruction was accounted
for by a combination of syntactic and phonological rules.
Inflectional morphology was assumed to operate in a similar fashion. For example, both regular
and irregular inflected verb forms like sang and mended were analyzed as sing+past and
mend+past, ‘where past is a formative with an abstract feature structure introduced by syntactic
rules’ (Chomsky and Halle 1968: 11 quoted in Scalise and Guevara 2005: 149).

3. Define ‘lexicalism’.

Lexicalism refers to the view according to which the processes that form complex words -
derivation and compounding - are accounted for by a set of Lexical Rules, which are independent
of and different from the syntactic rules of the grammar. In the lexicalist framework, Lexical
Rules are assumed to operate in a presyntactic component, the Lexicon.

4. What are the main tenets of Aronoff’s theory?

 223

Aronoff’s model assumes the existence of a separate morphological component (i.e. the lexicon)
which includes:
(a) a list of the atomic and complex words;
(b) a set of word formation rules (Word Formation Rules) that generate potential words on
the pattern of actual words;
(c) a set of obligatory adjustment rules.
The most important feature of Aronoff’s theory is the assumption that word-formation rules
operate over words and not morphemes, i.e. Aronoff adopts a theory of word-based morphology.
A second feature is that Aronoff restricts himself to derivational morphology. All other aspects
of morphology including inflection are regarded as syntactic. In Aronoff’s view, morphemes
cannot serve as the basis of word-formation processes because the very notion of morpheme is
problematic.

5. Discuss Word Formation Rules in Aronoff’s framework.

The main features of word formation rules in Aronoff’s framework are the following:
(i) Word Formation Rules apply only to words; these words must be existing words. Thus,
possible but non-existent words cannot be the base of a Word Formation Rule.
(ii) Word Formation Rules can take as bases only single words; they do not apply to units lower
than or higher than words (i.e. morphemes or phrases)
(iii) both the input and output of a Word Formation Rule must be members of the categories N,
V, A, i.e. the ‘open-ended’ lexical classes.
(iv) Word Formation Rules are intrinsically ordered; that is, they apply each time their structural
description is met in the sense that the condition for the rule application is specified on the base
of the rule.

6. Discuss restrictions on word formation rules.

In order to define the class of possible words of a language, Aronoff provided word
formation rules with a series of restrictions: syntactic, semantic, phonological and
morphological.
 224

(i) Syntactic restrictions - word formation rules have access to the syntactic properties of the base.
The base is one of the major syntactic categories: N, V, A. word formation rules do not apply to
functional categories: articles, pronoun, etc. For example, ‘-able’ attaches to transitive verbs.
Word formation rules also have access to subcategorization frames.
(ii) Semantic restrictions - derivational affixes select the base to which they attach also with
respect to its meaning. For example, ‘-ee’ attaches to [+animate, +sentient] bases.
(iii) Phonological restrictions - the phonology of the base restricts the range of affixes that can
attach to it. For example, ‘-al’ attaches only to verbs with main stress on the last syllable.
Morphological restrictions - some word formation rules are sensitive to the morphological
makeup of the base, crashing the derivation unless the base has the right internal structure. For
example, almost every verb derived with ‘-ize’ can be nominalized by adding ‘-ation’. All other
nominalizing affixes (-ment, -age, -al) are excluded, once ‘-ize’ is attached.

7. Discuss adjustment rules.

The application of the word formation rule to the base triggers different changes: phonological,
morphological and syntactic, which are dealt with in the word-formation component by
adjustment rules.
(i) Phonological adjustment rules - relate to the phonological changes brought about by the
operation of a word formation rule, particularly in terms of the stress pattern of the newly
derived word. Some affixes appear to be sensitive to the stress pattern of the base. As mentioned
before, the suffix -al only attaches to bases that end in a tressed syllable. Moreover, the bases
have a stressed vowel followed (optionally) by a consonant (sonorant/anterior). After -al
attachment, the stress patter of the new word changes:
(ii) Morphological adjustment rules - relate to the allomorphy rules that account for the
allomorphic variation which is lexically or morphologically conditioned by certain specified
morphemes (see Baciu 1998). Allomorphy rules explain the alternation -fy /-fic with the
nominalization suffix -ation as a case of partial suppletion, by writing a rule which states that
when -ation is added to a base ending in -fy, the latter is replaced by the -fic allomorph
(electrify-electrification).
 225

(iii) Syntactic adjustment rules - refer to changes in the subcategorization frame of the base
brought about by the application of a word formation rule. The application of the word formation
rule can affect the subcategorization frame of the base words in three different ways:
➢ Inheritance, where the subcategorization frame of the verb is inherited by the noun, with
minor changes:
The enemy destroyed the city
The enemy’s destruction of the city
➢ Deletion, which deletes elements in the subcatgorization frame of the verb:
They broke the glass into pieces.
The glass is breakable *into pieces
I ate the potato
I overate *the potato
➢ Addition, which adds one element to the subcategorization frame of the base:
He ran
*He outran.
He outran Mary

8. Discuss the Unitary Base Hypothesis.

In Aronoff’s view, an affix selects words of one and only one category. If an affix attaches to
more than one category, we are forced to assume that there are two or more homophonous
affixes. For example, the suffix ‘-able’ attaches to verbs (acceptable) and to nouns (charitable).
According to the Unitary Base Hypothesis, there are two ‘-able’ suffixes, one that attaches to
verbs and one that attaches to nouns.

Exercises:

1. How do the following examples argue against a morpheme-based morphology?

Cranberry, huckleberry, strawberry, blackberry

Refer, defer, prefer, transfer
 226

When we isolate the morpheme berry, we are left with cran and huckle, which do not exist
independently. In this framework, these words are not formed by regular morphological
processes. They are listed in the Lexicon as such. Similarly, we would need to recognize the
stems fe and mit and the prefixes re-, de-, pre-, trans-. However, these units without meaning
cannot be considered minimal linguistic signs. Consequently, Aronoff’s proposal is to look at
morphology as word-based, because words are indeed minimal signs.

2. Taking Scalise & Guevara’s formulation of Word Formation Rules, formulate Word
Formation Rules for the following:

disallow
mislead
rearrange
formulate
careless
boyhood

[WORD]X [[WORD]X + Suf]Y

formula formulate
care less
boy hood

[WORD]X [Pre+[WORD]X ]X
dis allow
mis lead
re arrange

3. Give subcategorization frames for the following:

They always eat their dinner in the kitchen. NP ___ NP

 227

Tom broke the window with a hammer. NP ___ NP

He runs in the park every day. NP ___
They work hard for their families. NP ___
You may depend on me to help you. NP ___ PP
Please refrain from smoking in the auditorium. NP ___ PP
Can you inform me about it right now? NP ___ NP PP
I will fill you in on it immediately. NP ___ NP Prt PP

4. Formulate syntactic restrictions for the attachment of -ness and un-.

‘-ness’ attaches to adjectives to yield nouns; ‘un-’ attaches to adjectives and the result is still an
adjective.

5. Formulate semantic restrictions for the attachment of -er.

‘-er’ attaches to transitive verbs which denote activities under the control of an agent. The
resulting noun is an agentive nominalization.

6. Formulate phonological restrictions for the attachment of -en.

‘-en’ attaches only to one-syllable bases that end in an obstruent or plosive (blacken but not
*lowen or *candiden).

7. Look at the following examples of en- derivation. State the lexical category of the base to
which ‘en-’ is attached. What is the lexical category of the new word? What is the meaning (or
meanings) of ‘en-’ in these words? Is there reason to state that there are two en- homophonous
prefixes? (Katamba & Stonham 2006)

Base New Word Base New word

cage en-cage noble en-noble
large en-large rich en-rich
 228

robe en-robe rage en-rage

danger en-danger able en-able

In the examples in 7, ‘-en’ attaches to nouns (cage, robe, danger) or to adjectives (large, rich,
noble) to derive verbs with the meaning ‘put X in a container / situation’ and ‘cause X to have a
property’, respectively. According to the Unitary Base Hypothesis, if an affix attaches to
different types of bases, we are dealing with two different affixes. Therefore, the examples
illustrate two homophonous suffixes ‘-en’.

Chapter 6 Generativism and lexicalism. The X-Bar Schema. Affixes as Lexical Items

Reviewing questions:

1. Name two important principles of X-bar Theory. What is the general form of the X-bar
schema? In what sense can the X-bar system be extented to word-formation?

Two important principles that articulate X-bar Syntax are the principles of endocentricity and the
principle of binary branching. According to the principle of endocentricity, every phrase is built
round a head and conversely, every head projects up to a maximal projection. Therefore, every
NP is built round an N, every VP is built round a V, a.s.o. According to the principle of binary
branching, the derivation proceeds in binary steps and the representation will always have two
branches stemming from the same mother-node.
The general format of the X-bar schema is the following:
XP (X”) → (Specifier) ˆ X’
X’ → X0 ˆComplements

XP
2
Spec X’
2
X0 Compl
 229

The X-bar system can be extended to word-formation in the sense that complex words ca be
seens as having a hierarchical constituent structure which can be represented by tree-diagrams.

2. Define the notion head in syntax. Can we determine the head of a word in the same way?
Explain by using the following examples:

The head is defined as the nucleus of a phrase. It is a noun for NPs, a verb for VPs, a preposition
for PPs, and an adjective or adverb for APs. The distribution of the phrase is given by the
distribution of the head. In other words, where books may appear, books about the war may also
appear.

Happiness, unhappy, legalize, relegalize

In morphology, it is not as easy to determine the head. If the distribution of the derivative is
given by the properties of the suffix (happiness, legalize), in the case of prefixed derivatives, the
distribution of the whole is given by the base.

Lighthouse, turncoat, drawbridge.

In the case of compound words, it is again less straightforward to determine the head. In the case
of endocentric compounds (lighthouse, drawbridge), the distribution of the whole is given by the
distribution of the head (house, bridge); moreover, there is a modifier-head relation between the
head and the non-head. In the case of exocentric compounds (turncoat), we could still identify a
head, but the relation is no longer that of a modifier-head relation.

3. What is the structure of the lexicon in Selkirk’s model of the morphological component? What
information does it comprise?
 230

In Selkirk’s theory of word-structure the rules of Word-syntax are housed in the lexical
component or lexicon. The lexical component contains a variety of subcomponents.
(i) a list of freely occuring lexical items (assumed to be atomic and complex words) which
represents the dictionary or lexicon in the restricted sense
(ii) a list of the bound morphemes, which together with the dictionary constitutes the extended
dictionary
(iii) the set of word structure rules (WSR) characterizing the possible morphological structure of
the language.
The dictionary also includes those phrases that are usually called idioms and those having
idiosyncratic meaning but not those words that are entirely regular or compositional in form and
meaning.

4. What are the identifying properties of affixes? What kind of information does the lexical entry
for affixes contain? What is the general form of affixation rules in Selkirk’s model?

Since in Selkirk’s model, affixed words are believed to be headed, affixes are treated as lexical
items. They are assigned to a category and have a lexical entry, like any other unbound
morpheme or morphologically complex item, the main distinction between lexical categories
proper and affixes being the fact that affixes are always bound.
Affixes have a lexical entry. The lexical entry of an affix contains three types of information:
syntactic, semantic and phonological.
(i) Syntactic information
A particular affix displays two syntactic properties. The first includes the name (feature bundle)
and type (X-bar level) of the affix’s sister category, and whether the affix is suffixed or prefixed.
For instance, the morpheme ‘-less’ is a suffix and attaches only to a nominal category of the type
N, as in ‘treeless’. The second syntactic property of the affix is the name of the category which
dominates the affix and its sister. The category dominating -less, for example, is always
adjectival.
(ii) Semantic information
The semantic information present in the lexical entry of an affix refers to the semantic analysis of
an affix as a function involving a change in lexical form or to the meaning contribution of the
 231

affix itself. For example, the affix ‘-able’ has associated with it the pair of lexical rules: (i) Obj
→ Subj and (ii) Subj → by Obj/ø. Moreover, the semantic analysis of the suffix ‘-able’ involves
also some characterization of the notion ‘able to be Ved’.
(iii) Phonological information
The phonological attributes of affixes refer to the pronounciation of the affix itself and to
whether they bring about changes in the pronounciation of surrounding morphemes. For example,
when the base ‘magnify’ takes the suffix ‘-ify’, the result will feature the allomorphy ‘fy / fic’ -
magnification.
5. Define neutral/non-neutral affixes. In what sense is the distinction relevant the the distribution
of affixes? What is the AOG? Explain by using the examples below:

monstrosity; proclivitous; fearlessness; tendernessless

*fearlessity; *inselfsufficient; *disbackdate

Neutral affixes are ‘ignored’ or not taken into account by the principles determining the stress
pattern. Neutral affixes are therefore stress-neutral. Non-neutral affixes are not ignored by the
stress pattern principles and enter into the canonical stress patterns of English words.
There is a pattern to the distribution that the non-neutral and neutral affixes have with respect to
each other. Another name that non-neutral and neutral affixes have are the terms Class I and
Class II respectively.

Class I suffixes : +ion, +ity, +y, +al, +ic, +ate, +ous, +ive
Class I prefixes: re+, con+, de+, sub+, pre+, in+, en+, be+
Class II suffixes: #ness, #less, #hood, #ful, #ly, #y, #like
Class II prefixes: re#, sub#, un#, non#, de#, semi#, anti#

Class II affixes may appear outside class I affixes but class I affixes may not appear outside class
II affixes. Selkirk (1982:91) calls this claim the Affix Ordering Generalization (AOG).
For example, the suffixes ‘-ous’ and ‘-ity’ are Class I [non-neutral] affixes, while the suffixes
‘-ness’ and ‘-less’ are Class II [neutral] affixes.
 232

[[monstr-os1]-ity1]
[[procliv-it1]-ous1]
[[fear-less2]-ness2]
[[tender-ness2]-less2]

If the two suffixes in the complex structure of words both belong to either Class I or Class II,
they may appear in any order. However, the order of the two classes themselves is not free.
Suffixes of Class I precede those of Class II:

[[danger-ous1]-ness2], [[activ-ity1]-less2])
*fear-less2-ity1
*tender-ness2-ous1.

Exercises:

1. Pronounce carefully the following pair of words. Place the stress mark ` before the syllable
that receives main stress in each word. Is ‘–ic’ a neutral or a non-neutral affix?

strategy strategic
morpheme morphemic
photograph photographic
democrat democratic

`strategy stra`tegic
`morpheme mor`phemic
`photograph photo`graphic
`democrat demo`cratic
 233

Since the stress is shifted from the first syllable to the second of the third, ‘-ic’ is a non-neutral
suffix.

2. Pronounce carefully the adjectives ‘wide’, ‘long’ and ‘broad’. Pronounce carefully the nouns
‘width’, ‘length’ and ‘breadth’, which are derived from the above adjectives by suffixing ‘–th’.
Pronounce carefully the adverbs ‘widely’ and ‘broadly’, which are derived from the above
adjectives by suffixing ‘–ly’. Determine whether ‘–th’ and ‘–ly’ are neutral or non-neutral
affixes.

Since the stress remains on the first syllable when the adjective are suffixed either with ‘-th’ or
‘-ly’, the result is that the suffixes are neutral.

3. It is generally acknowledged that suffixes trigger changes in the lexical category of the base,
while prefixes do not. Can you think of any category-changing prefixes? How can Selkirk’s
framework accommodate them?

Most prefixes in English are non-category changing affixes. Given the Right hand head rule none
of these prefixes will qualify as heads but will be sisters to a category identical in features with
the mother category. However, a small number of prefixes in English are category-changing.
They are sister to categories whose features are not those of the mother node. These prefixes
include:

a- Ar [a- Vr [sleep ]Vr ]Ar; A r [ a - Nr [kin ]Nr ]Ar

be- V [be- N [ cloud ]N ]V; V [ be- A[calm ]A ]V
en - V [en- A [noble]A ]V; V[en- N[slave]N]V
de- V[de- N[bug]N]V

These affixes are considered marked cases and they will be assigned the category features of
their mother node. For instance, ‘a-’ is Aaff and the others are Vaff.
 234

Chapter 7 Generativism and lexicalism. The permanent lexicon and percolation

Reviewing questions:

1. Discuss the main similarities between Lieber’s framework and Selkirk’s model.

The theory proposed in Lieber (1980, 1983) adopts a framework similar to that of Selkirk (1978,
1982) in that all lexical elements, stems and affixes alike, are given entries in a dictionary and
then undergo insertion into lexical trees. However, the theory differs considerably from that of
Selkirk (1978, 1982) in that the lexicon, called the Permanent Lexicon (Lieber 1980: 62) is
ascribed a different organization and rewriting rules that produce trees are ascribed a different
nature.

2. Discuss the structure of the Permanent Lexicon.

All unanalyzable morphological elements are called lexical terminal elements and are assigned
lexical entries. A lexical entry specifies all information about a terminal element, in the form of:
(i) The category and conjugation or declension class of an item
(ii) Phonological representation
(iii) Semantic representation
(iv) Subcategorization.
Affixes differ from non-affix morphemes in that affixes have frames indicating the category of
items to which they attach and the category of items produced. Subcategorization frames indicate
category (N, V, A) and other diacritic features of the items to which they attach.
un- attaches to adjectives
[A __ [A
-ive attaches to verbs with the diacritic feature [+Latinate]
Preventive, abusive, *understandive, *findive
[+Lat]]V__] A

(v) Diacritics, such as [+/- Latinate] or [+/- Neutral]

 235

(vi) Insertion frames. Lexical terminals are also specified for the syntactic frames into which
they can be inserted. Affixes also have insertion frames.
‘go’ takes a single NP argument, which is the subject
[NP___]
‘throw’ takes two arguments
[NP___ NP]
‘put’ take three arguments
[NP___ NP PP]

‘-ize’ forms verbs with two-place argument structures

The riot factionalized the city.
The lexical entries for affixes are identical to lexical entries for non-affix morphemes except for
the presence of subcategorization information in the entries for affixes. Affixes or bound
morphemes require the presence of some other lexical constituent.
A stem is defined as a morpheme whose lexical entry does not subcategorize another morpheme.
An affix is a defined as a morpheme whose lexical entry specifies another morpheme to which it
can attach.

Exercises:

1. Give the lexical entry for the suffix ‘-ize’ and the prefix ‘in-’

SUFFIX: -ize phonological representation /ayz/

semantic representation: causative
category: ]N__]V
insertion frame: NP__(NP)
diacritics: Level II (Class II)

PREFIX: in- phonological representation /in]/

semantic representation: negative
 236

category / subcategorization: [A __ [A
insertion frame: [___NP; [Link].__]
diacritics: Level I (Class I)

2. State and exemplify the Feature Percolation Conventions.

The framework assumes that, in order to generate complex words, there is a single context-free
rule which generates unlabeled binary branching tree structures.

2
1

3
1

3
1 1

Lexical terminals are inserted into these tree structures subject to their subcategorization
restrictions:

happy]A ness]N

2 ness]N
 237

Grime]N y]A

3
3 s]V

Standard]N ize]V

Trees are labeled by means of four Feature Percolation Conventions.

(i) Convention I
All features of a stem morpheme, including category features, percolate to the first nonbranching
node dominating that morpheme.

N -ize]V
g
[[standard]N

(ii) Convention II (Lieber 1983: 253)

All features of an affix moprheme, including category features, percolate to the first branching
node dominating that morpheme.

V
3
N -ize]V
g
standard]N

(iii) Convention III (Lieber 1983: 253)

 238

If a branching node fails to obtain features by Convention II, features from the next lowest
labeled node automatically percolate up to the unlabeled branching node.

V
3
[counter V
g
[attack]V]V

The prefix ‘counter-’ attaches to nouns (counterweight), verbs (coutersign) and adjectives
(counterintuitive). The resulting word is a verb when the base is a verb, a noun when the base is
a noun, and a adjective when the base is an adjective, i.e. the resulting words seem to have the
features of the bases amd not the features of the outermost morpheme. ‘counter-’ is transparent to
category in the sense that it lacks category features of its own.

(iv) Convention IV
If two stems are sisters (they form a compound), features from the right-hand stem percolate up
to the branching node dominating the two stems.

firetruck
N
3
Fire]N truck]N

3. Show the structure of the complex word ‘breakableness’

Not only category membership passes from a lexical morpheme to a node, but the entire feature
content of a morpheme, including gender features (for languages that have gender, like German
or Romanian) or the diacritic distinction [+/- Latinate] for English. Derived words adopt all
features of their outermost morphemes, along with their category.
 239

N
-Lat
3
A -ness]N
+Lat -Lat
2

V -able]A
g +Lat

break]V
Chapter 8 Argument structure and affixation

Reviewing questions:

1. Correct and/or complete:

b) In the Aspects model of Generative Transformational Grammar the syntactic valency of verbs
is represented by subcategorization frames.
b) In Goverment and Binding Theory the syntactic valency of verbs is represented by argument
structures.
c) There is systematic correspondence between thematic roles and morphologic case and
syntactic function. Refute or support this statement relying on the following examples; identiy
the theta grid of the predicates:

(i) Tom opened the door with a key/ The key opened the door
(ii) Judith loves Tom/ It seems to me that you are tired/ You surprised your teachers
with your theory
(iii) The balloon rose up

There is no systematic correspondence between thematic roles and morphologic case and
syntactic function. In (i), the subject position is associated with an Agent role (Tom) and with an
Instrument role (The key). In (ii), the Experiencer role is discharged on the subject position
 240

(Judith), the prepositional indirect object (to me) and the direct object position (your teachers). In
(iii) yet another theta-role appears in subject position - that of Theme.
Therefore, although it is true that very often the subject position is associated with Agents or
Experiencers and the direct object position is associated with, Themes, the correspondence is far
from systematic.

Exercises:

1. State the subcategorization frames for the predicates below:

Mary is sleeping. NP ___
Tom opened the door NP ___ NP
They put the baby in the crib NP ___ NP PP
They thought about the problem. NP ___ PP

2. State the argument structure of the predicates in the sentences below; identify the external
argument and the internal argument of the predicates; say whether all arguments are obligatory:

Tom opened the door with a key/The key opened the door / The door opened
<Ag, Th, Instr> <Instr, Th> <Th>
The external argument is the Ag, in the first sentence, the Instrument, in the second, and the
Theme, in the third sentence, which we show by underlining. The only obligatory argument is
the Theme argument, which we show as: <(Ag), Th, (Instr)>.

Peter sang a lullaby for the baby.

<Ag, (Th), (Ben)>
The external argument is the Agent, which we show by underlining. It is also the only obligatory
argument.

John broke the window with a hammer/The hammer broke the window/The window broke.
<Ag, Th, Instr> <Instr, Th> <Th>
 241

The external argument is the Ag, in the first sentence, the Instrument, in the second, and the
Theme, in the third sentence, which we show by underlining. The only obligatory argument is
the Theme argument, which we show as: <(Ag), Th, (Instr)>.

A new-born baby weighs between 3 and 4 kilograms.

<Th>
The unique argument of the verb ‘weigh’ is a Theme. If there is one argument in the semtence,
that argument is external.

John positioned his car by the park.

<Ag, Th, (Loc)>
He dropped the stone through the water on to her toe.
<Ag, Th, (Path), (Goal)>

He thought about the problem for a very long time.

<Ag, Th>

Everybody loves linguistics.

<Exp, Th>

3. Study the sentences below and identify the changes that have occurred after affixation (if any):

(i) The City Council developed the city

The City Council’s development of the city
The development of the city by the City Council

After affixation, the noun ‘destruction’ inherits the arguments of the verb ‘destroy’.
<Ag, Th> the enemy, the city

(ii) The grammar generates compounds

The grammar is generative*of compounds
 242

After affixation, the internal argument of the verb is blocked.

<Source, Th> <Source>

(iii) He ran
He outran Mary

After affixation, the intransitive verb gets an internal argument, which is added to its argument
structure.
<Ag> <Ag, Th>
4. Discuss the impact of the affixation of -able, -en, -ed, -ee -ify, on the argument structure of the
lexical item to which they attach in terms of Williams’ hypothesis (externalization,
internalization); represent the two processes schematically:

(i) They read the novel / The novel is readable.

Before affixation, the argument structure of the verb ‘read’ has two arguments: <Ag, Th>. After
affixation, the internal argument of the verb ‘the novel’ is promoted - externalized - to the
subject position: <Th>. This is known as externalization of the internal argument.
<Ag, Th> <Th>

(ii) This shirt is wide/ He has widened the shirt.

Before affixation, the adjective ‘wide’, acting as a predicate, has a unique argument, a Theme,
which surfaces on the subject position: <Th>. After affixation, an external argument is added,
and the former external argument goes to the direct object, on the position of the internal
argument: <Ag, Th>. This is known as internalization of the external argument.
<Th> <Ag, Th>

(iii) The water is pure/ They purified the water.

 243

Before affixation, the adjective ‘pure’, acting as a predicate, has a unique argument, a Theme,
which surfaces on the subject position: <Th>. After affixation, an external argument is added,
and the former external argument goes to the direct object, on the position of the internal
argument: <Ag, Th>. This is known as internalization of the external argument.
<Th> <Ag, Th>

(iv) I teach morphology with pleasure / Morphology is teachable.

Before affixation, the argument structure of the verb ‘teach’ has two arguments: <Ag, Th>. After
affixation, the internal argument of the verb ‘morphology’ is promoted - externalized - to the
subject position: <Th>. This is known as externalization of the internal argument.
<Ag, Th> <Th>

Chapter 9 Argument structure and affixation. Approaches to the thematic

inheritance phenomena

Reviewing questions:

1. Account for the following nominalizations in Williams and Di Sciullo’s framework:

The mayor developed the city.

The development of the city by the mayor.

To account for the cases of pure inheritance, Williams and Di Sciullo assume that the argument
structures of the head and non-head compose (by the mechanism of function composition). The
argument (R) of the affix becomes the external argument of the newly formed word while the
arguments of the base become internal arguments.
develop <Ag, Th>
-ment <R>
 244

develop ment ----> development (R, Ag, Th)

<Ag,Th> <R>

The barbarians destroyed the city.

The destruction of the city by the mayor.
destroy <Ag, Th>
-tion <R>
destroy -tion ----> destruction (R, Ag, Th)
<Ag,Th> <R>
The professor examined the students.
The examination of the students by the professor.
examine <Ag, Th>
-ation <R>
Examine ation----> development (R, Ag, Th)
<Ag,Th> <R>

2. Explain the notion of control in the following pairs:

Employer vs. Employee

Detainer vs. Detainee
Cheater vs. Cheatee

To account fo such cases, Di Sciullo and Williams introduce the mechanism of control. The
argument (R) of the affix, after composing with the argument structures of bases, controls (or
binds) one of the arguments of the base, which is no longer expressed in syntax. These affixes
attach to verbs that have in their argument structure the arguments they control, external or
internal, respectively. In each case the control relation satisfies the controlled argument so that it
cannot be expressed in syntax independent of its controller. The relation of control is expressed
by placing the same index on the controller and the controllee:
employ er R controls external argument
<Agi,Th> (Ri)
 245

employ ee R controls internal argument

<Ag, Thi> (Ri)

detain er R controls external argument

<Agi,Th> (Ri)
detain ee R controls internal argument
<Ag, Thi> (Ri)

cheat er R controls external argument

<Agi,Th> (Ri)
cheater ee R controls internal argument
<Ag, Thi> (Ri)

3. Explain the notion of implicit argument starting from the following examples.

The light was turned off.

In this sentence, the presence of the passive voice indicates the presence of an implicit agent,
which is not overt (by someone).

The light was turned off [PRO to scare the children].

The implicit agent is able to control the PRO (non-overt) subject of the infinitival clause. The
same person who turns off the light scares the children.

The light was turned off by the clown [PRO to scare the children].
In this sentences, the agent is made explicit with the help of a ‘by-phrase’.

4. Can implicit agents control the null subject of an adverbial of purpose with -ing compounds,
nominalizations, and -able adjectives?

meat-eating [PRO to gain weight]

 246

The non-expressed agent of the infinitival clause (the one who gains weight) is controlled, that is
co-referential with the implicit agent in ‘meat-eating’, that is, the person who eats meat is the
same with the person who gains weight.

the eating of meat [PRO to gain weight]

The non-expressed agent of the infinitival clause (the one who gains weight) is controlled, that is
co-referential with the implicit agent in ‘meat-eating’, that is, the person who eats meat is the
same with the person who gains weight.

Goods are exportable [PRO to improve profits]

Whenever suffixed with ‘-able’, the resulting adjective contains an implicit agent. If goods are
exportable, then someone exports them. This implicit agent controls the non-overt PRO subject
of the infinitival clause.

5. Explain the notion of inferred cognitive agency starting form the contrast below:

*the thief of the bank /*the taxman of hidden assets

the robber of the bank / the taxer of hidden assets

There is a distinction between the notion of implicit agent - which is thematic - and the notion of
inferred or ‘cognitive agency’. Implicit arguments (agents) pressupose the existence of a theta
grid, while inferred thematic roles - the mere notion of agency - do not. ‘thief’ and ‘taxman’
entail a cognitive notion of agency but do not have a thematic grid and that is why the ‘of’-
phrase, which refers to the theme argument makes *the thief of the bank / *the taxman of hidden
assets ungrammatical. The verbs ‘tax’ and ‘rob’ have a thematic grid that is maintained by the
suffix’ -er’. ‘Robber’ and ‘taxer’ have complements and it is to these complements that the
‘of’-phrases refer.

Exercises:

1. Explain the ill/well-formedness of the following ‘-er’ nominalizations in Roeper’s approach:

 247

*a trier
*an intender
*a hoper
*a disappearer
teacher
runner
swimmer
writer
The suffix ‘-er’, which creates nouns referring to an Agent, only attaches to verb bases which
have the Agent role. ‘-er’ has the theta-grid <Ag,Th> and attaches to verbs that have an identical
theta grid: teacher, runner, swimmer, writer.
*a trier, *an intender, *a hoper are ungrammatical because the thematic grid of the verb does not
match the thematic grid of the affix, that is they have an Experiencer theta-role in their argument
structure.

2. Explain the ill/well-formedness of the following ‘-able’ adjectives in terms of Roeper’s

approach:

*tryable
*comeable
*dieable
*fallable
*decidable
readable
understandable
loveable

The suffix ‘-able’ has the theta-grid <Ag,Th>, and attaches only to transitive verbs. ‘Learnable’,
‘defendable’, ‘pushable’ are well-formed since the theta-grid carried by ‘-able’ matches the
theta-grid on the verb-base.
 248

The adjective *comeable is ill-formed because ‘-able’ is a transitive affix but it attaches to
an intransitive verb, so there is a mismatch between their theta-grids. ‘-able’ has its own thematic
grid <Ag,Th> which must match the grid of the verb-base. *tryable, *decidable are ill-formed
because the verbs ‘try’ and ‘decide’ take sentential complements.

3. -er’ nominalizations are called ‘agentive nouns’ because ‘-er’ controls the agent argument of
the base verb. Analyze the following and determine whether the analysis of thematic affixes can
account for these cases.

a. a lover of the road / a lover of life / a lover of good wine

b. The freezer is in the kitchen
c. Salicylic acid is an inducer of pathonogenesis-related proteins
d. owners of intellectual property

The verbs ‘love’ and ‘own’ do not have Agents in their argument structures; therefore, in
principle, they could not pe suffixed with the agentive ‘-er’. On the other hand, ‘inducer’ and
‘freezer’ do not refer to Agents, but to objects. One possible explanation is the fact that ‘-er’
helps in the personification of objects. A more technical explanation is that ‘-er’ is sometimes
able to bind the Instrument role instead of that of the Agent. As for ‘lovers’, it may be that, in the
context of usage, it is no longer understood as an Experiencer but as an Agent: a lover of the road
is the one who travels a lot.

Chapter 10 Nominalizations

Reviewing questions:

1. Explain which are the possible interpretations of nominalizations.

(i) Complex event nominalizations:

The barbarians destroyed the city.

 249

The destruction of the city the barbarians.

The barbarians’ destruction of the city.

The verb ‘destroy’, which has the argument structure <Ag, Th>, passes these arguments on the
nominalization ‘destruction’. ‘the city’, which is the Theme argument in the argument structure
of the verb ‘destroy’ preserves its Theme status with ‘destruction’. ‘the barbarians’, which is the
Agent argument in the argument structure of ‘destroy’, preserves its Agent status. This status is
obvious when looking at the preposition ‘by’, which introduces Agents in the passive. In the
third example, ‘the barbarians’ appears in the Genitive case, which is usually associated with the
Agent. Complex event nominalizations describe an event, in the same way as the verb from
which they are derived does. An important feature of complex event nominalizations is that they
take internal arguments, with the help of the preposition ‘of’.
(ii) Result nominalizations. Sometimes, the same nominalization may get a different
interpretation, depending on the presence of different grammatical elements, such as the plural
marker ‘-s’. Result nominalizations no longer have an argument structure. They describe a
variety of things related to the verb, such as the result state of the event denoted by the verb (5)
or a related concrete object (6).

The destructions (*of the city) were painful to look at.

The examination was on the table.

(iii) Simple event nominalizations. They describe an event which is associated with the verb
from which they are derived, but do not have an internal argument.

The destruction took a long time.

2. Is there a nominalization which is non-ambiguous? Explain.

While the vast majority of nominalizations are ambiguous between an event reading and a result
reading, there is one type of nominalization which always designates complex events: ‘-ing’
 250

nominalizations. Therefore, by investigating the behavior of ‘-ing’ nominalizations, we can

extract features that will help disambiguate between event and result nominalizations.
(i) ‘-ing’ nominalizations require obligatory internal arguments, just like their verbal
counterparts:

They felled trees.

*They felled.
The felling of the trees.
*The felling.
They destroyed the city.
*They destroyed.
The destroying of the city.
*The destroying.

(ii) ‘-ing’ nominalizations can control the PRO subject of an adverbial of purpose clause. ‘-ing’
nominalizations can also co-occur with a ‘by-phrase’. This means that they contain an Agent in
their argument structure:

The fracturing of a leg [PRO to realign the bones]

The running of the store by Bill [PRO to get better profits]

(iii) ‘-ing’ nominalizations only allow the definite article

The shooting of rabbits is illegal

*A/*one/*that shooting of rabbits is illegal.

(iv) ‘-ing’ nominalizations do not take the plural marker:

*The mowings of the lawn were interminable.

 251

(v) ‘-ing’ nominalizations never occur in predicative position, which means that they cannot
occupy the position across the copula:

*That was the / an eating of apples.

(vi) ‘-ing’ nominalizations co-occur with the same aspectual modifiers as their verbal
counterparts. As you may remember, accomplishments take ‘in’-phrases while activities take
‘for’-phrases.

a. John built the house in two years.

b. The reading of the book in two days.
a. They ate apples for two hours.
b. The eating of apples for two hours.

Therefore, ‘-ing’ deverbal nominalizations always designate complex events.

3. Give examples of the tests that help us disambiguate between different readings of
nominalizations.

(i) One way of disambiguating between event and result interpretations of nominalizations is by
using aspectual modifiers like ‘constant’ and ‘frequent’. If the nominalization accepts these
aspectual modifiers, the presence of the internal argument is obligatory and the result reading is
excluded.

a. This assignment is to be avoided

b. *The frequent / constant assignment is to be avoided
c. The frequent / constant assignment of unsolvable problems is to be avoided

(ii) A second major way of disambiguating between the event and result reading is by using a
Genitive with an Agent role, accompanied by Agent-oriented modifiers like ‘deliberate’ or
‘intentional’, which determine a genuine Agentive reading of the possessive subject.
 252

a. The examination took a long time

b. The doctor’s examination took a long time
c. *The doctor’s deliberate / intentional examination took a long time
d. The doctor’s examination of the patients took a long time

(iii) A third important way of disambiguating between the event and the result interpretations of
nominalizations is looking at the different determiners they take. Result nominalizations are free
in the use of determiners. They co-occur with the definite article ‘the’, the indefinite article ‘a /
an’, with the numeral ‘one’, or with demonstratives like ‘this / that’.

She graded the / an / one / that examination.

Another feature of result nominals is that they freely take the plural morpheme.

She had graded so many examinations that she was blue in the face.

Event nominalizations do not pluralize.

*The constant examinations of the students made her faint.

Event nominalizations admit only the definite article, or it is possible to have no determiner at
all.

(The) assignment of difficult problems has caused problems.

When we compare the distribution of result and event nominalizations, we can notice that result
nominalizations pattern like countable nouns, while event nominalizations pattern like mass
nouns. Very intuitively speaking, this indicates the fully nominal nature of result nominals but
the hybrid nature of event nominalizations, which are verbs incorporated into a nominal layer.
 253

Another observation related to the distribution of nominalizations is that complex event nominals
do not occur predicatively while result nominals do:

a. That was the / an assignment.

b. *That was the / an assignment of the problem.

(iv) A fourth major way of disambiguating between complex event nominalizations and result
nominalizations is by looking at the presence of postnominal genitives and temporal possessives.
These elements always indicate result nominals:

This assignment of John’s drove him nuts.

This week’s assignment drove all my students nuts.

*This semester’s constant assigment of difficult problems drove all my students nuts.
The constant assignment of unsolvable problems this semester led to disaster

(v) A fifth way of disambiguating between complex event ad result nominalizations is by

looking at the acceptability of infinitival clauses functioning as adverbials of purpose. The
existence of infinitival clauses with nominalizations is explained by the presence of the argument
structure with event nominals, an argument structure which contains an Agent. Therefore,
complex event nominalizations can control the PRO subject of the infinitive clause, while result
nominals cannot.

The assignment of difficult problems [in order PRO to drive my students nuts.]
*The assignments of difficult problems [in order PRO to drive my students nuts].

(vi) complex event nominals admit the same aspectual modifiers as their verbal counterparts.

The bombing destroyed the city in two days.

The total destruction of the city in two days.
They observed the patient for a few days.
 254

The observation of the patient for a few days.

4. Is there evidence showing that there is a verbal head inside nominalizations?

There is syntactic evidence in favor of the view that there is a real VP under the -tion node,
coming from adverbs (a) and VP-ellipsis (b):

(a) While the removal of evidence purposefully is a crime, the removal of evidence
unintentionally is not.
His resignation so suddenly gave rise to wild speculation.
(b) John’s destruction of the city and Bill’s doing so too.
Under the view that do so refers to VPs and adverbs also modify VPs, it is clear that there is a
real VP under the nominalization node. This explains the inheritance of the argument structure of
the verb by the corresponding nominalization.

Exercises:

Discuss the following sentences. Identify the interpretation of the nominalizations and point to
the elements that disambiguate the interpretation.
(i)
The assignment of difficult problems bothers the students.
The presence of the internal argument ‘of difficult problems’ entails a complex event reading.

The transformation of the catterpillar was complete.

The presence of the internal argument ‘of the catterpillar’ entails a complex event reading.

*the change of the catterpillar was complete.

The noun change is a simple result nominal, so the presence of the internal argument is barred.

The frequent assignment of difficult problems bothers the students.

 255

The presence of the internal argument and the presence of the modifier ‘frequent’ entail a
complex event reading.

*The frequent problem bothers the students.

‘Problem’ is a result nominal so it cannot accept either the presence of ‘frequent’ or of the
internal argument.

Their performance of the play in only two hours surprised the critics.
The presence of the internal argument and the presence of the aspectual modifier ‘in two hours’
entail a complex event readig.

*Their dance in only two hours surprised the critics.

‘Dance’ is a simple event nominal and, therefore, it cannot appear with aspectual modifiers like
‘in two hours’.

Transformation of the catterpillar into a butterfly as rapidly as possible is essential for the
survival of the insect.
The presence of the internal argument, the lack of any determiner, and the modification by an
adverb (rapidly) lead us to conclude that we are dealing with a complex event nominalization.

*The catterpillar’s change into a butterfly as rapidly as possible is essential for the survival of the
insect.
Because ‘change’ is a simple event nominal, it cannot take a Genitive subject and cannot be
modified by adverbs.

(ii)
The construction of the building lasted a long time.
The presence of the internal argument and the occurrence in the frame ‘lasted X time’ indiate a
complex event reading.

The construction was frequent and lasted a long time.

 256

The lack of the internal argument and the occurrence of ‘frequent’ across the copula indicate a
simple event reading.

The new construction he built was tall.

Modification by adjectives indicates a result reading.

The destruction of the building in an hour was hard to watch.

The presence of the internal argument and the presence of the aspectual modifier ‘in an hour’
indicate a complex event reading.

Lisa’s explanation of the problem in two minutes flat impressed me.

The presence of the internal argument, the presence of a genitive subject and the presence of the
aspectual modifier ‘in two minutes’ indicate a complex event reading.

*Lisa’s explanation that there was a problem in two minutes flat impressed me.
Complex event nominalizations with genitive subjects cannot take sentential complements.

John’s suggestion of that possibility for so many years got tiresome.

The presence of the internal argument, the presence of the external argument as a genitive
subject and the presence of the aspectual modifier ‘for so many years’ indicate a complex event
reading.

*John’s suggestion that he father a child for so many year got tiresome.
Complex event nominalizations with genitive subjects cannot take sentential complements.

(iii)
Bill’s cultivation of the tomatoes
The bomb’s destruction of the town
The volcano’s fortuitous burial of Herculaneum
The emperor’s restoration of monarchy
The proposal’s creation of controversy
 257

The sun’s illumination of the room.

Bill’s suspension of the activity

In all the examples in (iii), there are both external arguments as genitive subjects and internal
arguments, so all examples contain complex event nominalizations.

Chapter 11 Compounds

Exercises

1. Classify the following words as being the products of either inflection, derivation or
compounding. Justify your analysis in the potentially problematic cases:

Blackboard eraser
‘eraser’ is the product of derivation; ‘blackboard’ is the product of compounding; ‘blackboard
eraser’ is the product of compounding.

Unacceptability
‘unacceptability’ is result of derivation with the prefix ‘un-’ and the suffixes ‘-able’ and ‘-ity’.

Flowerpots
‘flowerpots’ is the result of compounding, followed by the addition of the inflectional plural
marker ‘-s’.

Broad-shouldered
Although ‘broad-shouldered’ may look like a compound, ‘shouldered’ consists of the root
‘shoulder’ and the suffix ‘-ed’, which indicates that ‘broad-shoudered’ is a derivative with the
phrase [broad shoulder] as its base. The idea is reinforced by the fact that ‘broad-shouldered’
means ‘having broad shoulders’ and the ornative suffix ‘-ed’ in English produces this meaning.
 258

Hard-working
‘hard-working’ is an adjectival compound with a participle as right-hand element, which is
modified by ‘hard’.

Speaking
If ‘speaking’ is the present participle of ‘speak’, it is the result of inflection. If it is a deverbal
noun that denotes a process, it is the result of derivation. If it is used as an adjective, it is also the
result of derivation.

Developmental
‘developmental’ is the result of derivation. The suffixes ‘-ment’ and ‘-al’ are added to the root
‘develop’.

2. Name two general characteristics of English compounds. Use the data below for illustration:

Oak-tree
Drawbridge
Sky-blue
Mind-boggling

First, compounds are binary structures that consist of roots, words, or phrases. Our examples are
compounds that consist of words. Second, compounds are right-headed. All our examples can be
interpreted in a way in which the non-head modifies the head (an oak-tree is a kind of tree). in
the case of ‘mind-boggling’, the left-hand constituent is an argument of the head, which means
that the compound is synthetic.

3. Classify the following compounds as exocentric, endocentric, possessive, appositional, or

coordinative:

Frying pan - endocentric compound

 259

Redhead - exocentric, possessive compound = ‘a person possessing red hair’

Maidservant - appositional compound = ‘a person who is both a maid and a servant’
Author-reader (exchange) - coordinative compound = ‘a particular relation between reader and
author’
Austria-Hungary - appositional compound = ‘country that consists of both Austria and Hungary’
Hardtop - exocentric, possessive compound
Silkworm - endocentric compound
Man-machine (interaction) - coordinative compound = ‘a particular relation between man and
machine’
Bootblack - exocentric compound = a person who blackens boots’
German-English - appositional compound in ‘German-English family’ but coordinative in
‘German-English contact’
Actor-manager - appositional compound
Gas-light - endocentric compound = a kind of light

4. Are underdog, undercoat and overtax, overripe compounds or prefixed derivatives? Discuss.
(from Plag 2003: 164)

The discussion starts from the fact that the elements ‘under’ and ‘over’, which in these case
appear as bound morphemes, that is prefixes, may also appear as free morphemes, which would
makes these formations compounds. If they are compounds, ‘under’ and ‘over’ would have to
have the same meanings as in their free occurrences. In other words, ‘over’ in ‘overtax’ should
have the same meaning as in ‘over the rainbow’. The meaning of ‘under’ in ‘underdog’ means
‘in a state of inferiority’, which is the same as in ‘under the regulations’. In ‘undercoat’, ‘under’
means ‘beneath’, a meaning that ‘under’ has in ‘under the table. We can conclude that
‘undercoat’ and ‘underdog’ are preposition-noun compounds.
‘Over’ in ‘overripe’ means ‘too much’ and so does in ‘overtax’, which means ‘taxing too
heavily’. it follows that ‘overtax’ and ‘overripe’ are also preposition-noun compounds.

5. Rhyme is not the only phonological motivation behind compounding. Determine the
phonological factor that plays a role in the following (from Katamba & Stonham 2006: 306):
 260

Zigzag
Dilly-dally
Sing-song
Tittle-tattle
Tick-tock
Wishy-washy

In these examples, an important role is played by ablaut. The two bases have identical
consonants but different vowels.
Conclusions
 261

Bibliography

Alexiadou et. al. 2011. Scaling the variation in Romance and Germanic nominalizations. In
Sleeman, P. and H. Peridon (eds.) The Noun Phrase in Romance and Germanic, 25-40.
Amsterdam: John Benjamins

Aronoff, M. 1976. Word-Formation in Generative Grammar. MIT Press, Cambridge: Mass.

Aronoff, M. 1982. Potential Words, Actual Words, Productivity and Frequency. Preprints from
the 13th International Congress of Linguists, Tokyo.

Baciu, I. 1998. English Morphology: Word Formation. A Generative Perspective. Bucharest

University Press: Bucharest

Baker, M. 1985. The Mirror Principle and Morphosyntactic Explanations. Linguistic Inquiry, 16,
373-416.

Baker, M. 1988. Incorporation: A Theory of Grammatical Function Changing. University of

Chicago Press, Chicago.
 262

Barker, C. 1998. Episodic -ee in English: a thematic role constraint on a new word formation.
Language 74: 695-727

Bauer, L. 1978. The Grammar of Nominal Compounding. Odense University Press: Odense.

Bauer, L. 1983. English Word-Formation. Cambridge University Press: Cambridge.

Bauer, L. 1988. Introducing Linguistic Morphology. Edinburgh University Press: Edinburgh.

Bauer, L. 2001. Morphological Productivity. Cambridge University Press: Cambridge

Bauer, L. 2005. Productivity: Theories. In Handbook of Word-Formation, Štekauer, P. and R.

Lieber (eds.), 315-335.

Bauer, L. and A. Renouf. 2001. A Corpus-based study of compounding in English. Journal of

English Linguistics 29: 101-123.

Bobalijk, J. D. 2002. Auspicious compounds. McGill Working Papers in Linguistics vol 17.2,
65-71.

Bolinger, D. 1948. On defining the morpheme. Word 4, 18-23.

Bolinger. D. 1975. Aspects of Language. New York: Harcourt, Brace and Jovanovich, 87-286

Borer, H. 1990. V+ing: it walks as an adjective, it talks as an adjective. Linguistic Inquiry 21,
95-103.

Borer, H. 2003. Exo-skeletal vs. Endo-skeletal explanations: syntactic projections and the
lexicon. In The Nature of explanation in Linguistic Theory, Moore, J. and M. Polinsky (eds.),
31-67.
 263

Chomsky, N. 1957. Syntactic Structures. Mouton: The Hague

Chomsky, N. 1965. Aspects of the Theory of Syntax. MIT Press, Cambridge: Mass.

Chomsky, N. 1970. Remarks on Nominalizations. In Readings in English Transformational

Grammar, Jacobs. R and Rosenbaum, S (eds), Blaisdell, Waltham: Mass

Chomsky, N. and M. Halle. 1968. The Sound Pattern of English. Harper and Row: New York.

Cornilescu, A. 1995. Concepts of Modern Grammar. Bucharest University Press: Bucharest.

Crystal, D. 1995. The Cambridge Encyclopedia of the English Language. Cambridge:

Cambridge University Press.

Di Sciullo, A.-M. 1990. Formal Relations and Argument Structure. In Contemporary

Morphology, Dressler, W. et al. (eds.), Mouton de Gruyter: Berlin.

Di Sciullo, A.-M. 1992. Deverbal Compounds and the External Argument. In Thematic Structure:
Its Role in Grammar, Rocca, I. (ed.) Mouton de Gruyter: Berlin, 65-78

Di Sciullo, A.-M. and E. Williams. 1987. On the Definition of Word. MIT Press, Cambridge:
Mass.

Grimshaw, J. 1990. Argument Structure. MIT Press, Cambridge: Mass.

Harley, H. 2008. The morphology of nominalizations and the syntax of vP. In Quantification,
Definiteness and Nominalization, Rathert, M and A. Giannankidou, Oxford: Oxford University
Press, 320-342.

Hjelmslev, L. 1943. Prolegomena to a Theory of Language. Indiana University Press,

Bloomington.
 264

Katamba, F. 1993. Morphology. Basingstoke: Macmillan

Katamba, F. and J. Stonham. 2006. Morphology. Palgrave: Macmillan.

Kornfilt, J. and J. Whitman. 2011. Afterword: Nominalizations in syntactic theory. Lingua 121,
1297-1313.

Lieber, R. 1980. On the Organization of the Lexicon. Ph.D Diss: MIT

Lieber, R. 1983. Argument Linking and Compounds. Linguistic Inquiry 14, 251-285.

Lieber, R. 1992. Deconstructing Morphology. Word Formation in Syntactic Theory. University

of Chicago Press: Chicago.

Lieber, R. 2009. Introducing Morphology. Cambridge: Cambridge University Press

Lyons, J. 1968. Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press

Lieber, R. 2005. English Word-Formation Processes. In Handbook of Word-Formation, Štekauer,

P. and R. Lieber (eds.), 375-429.

Magionos, G. 2016. Nominalizations, functional structure, and argument structure. MA Diss:

Patras University

Marchand, H. 1969. The Categories and Types of Present-Day English Word-Formation. C. H.

Beck Verlagsbuchhandlung: Munich.

Matthews, P. H. 1974. Morphology. Cambridge University Press: Cambridge

Matthews, P. H. 1991. Morphology. 2nd Edition. Cambridge University Press: Cambridge

 265

Moulton, K. 2013. Simple Event Nominalizations: roots and their interpretation. In

Cross-Linguistic Investigations of Nominalization Patterns, I. Paul (ed.), 119-144.

Pesetsky, D. 1995. Zero Syntax: Experiencers and Cascades. MIT Press

Plag, I. 2003. Word-Formation in English. Cambridge University Press: Cambridge

Roeper, T. 1987. Implicit Arguments and the Head-Complement Relation. Linguistic Inquiry 18,
267-310.

Roeper, T. 2005. Chomsky’s Remarks and the Transformationalist Hypothesis. In Hanbook of

Word-Formation, Štekauer, P. and R. Lieber (eds.), 125-147

Rose, J. H. 1973. Principled limitations on productivity in denominal verbs. Foundations of

Language 10, 509-526.

Saussure, F. de. 1983. Course in General Linguistics. London: Duckworth

Săvescu, O. 2014. Word-Formation. Lecture notes. Bucharest University

Scalise. S. 1984. Generative Morphology. Dordrecht: Foris.

Scalise, S. and E. Guevara. 2005. The Lexicalist Approach to Word-Formation and the Notion of
the Lexicon. In Handbook of Word-Formation, Štekauer, P. and R. Lieber (eds.), 147-189.

Selkirk, E. 1982. The Syntax of Words. Linguistic Inquiry Monographs 7, MIT Press,
Cambridge: Mass.

Siegel, D. 1974. Non-sources of Un-passives. Syntax and Semantics II, J. Kimball (ed.), Seminar
Press: New York.
 266

Spencer, A. 1991. Morphological Theory. An Introduction to Word Structure in Generative

Grammar. Basil Blackwell: Oxford.

Spencer, A. and A. M. Zwicky (eds.) 1998. The Handbook of Morphology. Oxford: Blackwell

Spencer, A. 2005. Word-Formation and Syntax. In Handbook of Word-Formation, Štekauer, P.

and R. Lieber (eds.), 73-99.

Štekauer, P. and R. Lieber (eds.). 2005. Handbook of Word-Formation. Dordrecht: Springer

Williams, E. 1981. On the Notions ‘Lexically related’ and ‘Head of a word’. Linguistic Inquiry
12, 245-274