Nicholas LaCara · University of Toronto lin333 · 1 April 2019

Root-and-pattern morphology
For our final lecture, we look at the complexities that arise for the kind of non-concatenative root-and-pattern morphology
found in the Semitic languages. This presents a similar problem to the one infixation did for our morphosyntactic theory:
While it appears that words in these languages are composed of various subparts, they are not linearized one after the other,
but interspersed amongst each other. We will build on our approach to infixation and reduplication to solve this problem.

1 Overview

• In the last two lectures, we focused on the problems presented by infixation and
reduplication.

• On the accounts presented, both processes have a significant phonological com-

ponent, referring to non-morphological and non-syntactic structures.

– Infixation places infixes between segments and at various prosodic bound-

aries rather than at pre-existing morphological boundaries.
– Reduplication utilizes morphemes whose exponents are prosodic tem-
plates, morphemes with prosodic structure that lack underlying segmen-
tal material.

• Semitic root-and-pattern morphology requires us to build on both of these ap-

proaches.

– Like infixation, this morphology is non-concatenative, but to an even Such morphemes, as you may
more extreme degree. Two morphemes appear to become interspersed recall, are sometimes called
transfixes.
with each other.
– The patterns of these interspersions varies depending on derivational and
inflectional characteristics of these words, suggesting a templatic account
similar to the one used for reduplication.
√
(1) Arabic verbs formed from the root ktb: [Link]
wiki/K-T-B
a. kattaba b. kutiba
‘he made someone write’ ‘it is written’

• The discussion today centres on the properties of this kind of morphology, with
an emphasis on verbs.

– We’ll first look at the properties of words derived in this fashion in Hebrew,
Arabic, and Amharic.
– We’ll then look at the syntax underlying the verbs in these languages.
– Finally, we’ll look at an autosegmental phonological analysis that attempts
to account for the patterns by the used of prosodic templates.

1
 Nicholas LaCara · Root-and-pattern morphology 2

2 The data

• In this section we will go over what the morphology of verbs and nouns looks
like in Arabic, Hebrew, and Amharic.
• In these languages, verbs can be broken down into (at least) three different mor- The divisions here are based on
phemes. the discussion in Arad (2005);
this is slightly different from
what Katamba (1993: 166)
i. A consonantal root, which contributes core, idiosyncratic lexical meaning. states; vowels have different
ii. A vocalic melody (i.e., vowels) that are related to the voice of the clause. contextual allophones in
different tense/aspects. As is
iii. A template that determines the order of consonants and vowels in the re- customary, I'm setting aside
sulting word and which is associated argument structure. agreement affixes in much of
this discussion, concentrating
• Despite our focus on verbs, noun and adjective formation relies on similar mech- on verb stem formation.
anisms.

2.1 Root-and-pattern morphology in Hebrew

• Arad (2003, 2005) takes roots to be atomic elements devoid of syntactic or func-
tional material, which matches the traditional definition of root most morphol-
ogists assume.
• These are arranged into patterns called binyanim (sg. binyan).
• In some cases, the meanings are transparent. Many words derived from the root
√
šmn have a meaning related to fat:
√
(2) Words derived from šmn (Arad 2003: 742): Though some patterns display
geminate consonants, there
Category Binyan Form Gloss are no geminates in Modern
Hebrew (though there were in
Verb a. hiCCeC hišmin ‘grow fat/fatten’ Biblical Hebrew). However,
Noun b. CeCeC šemen ‘oil/grease’ there are four-consonant roots
that use these binyan, hence
c. CuCaC šuman ‘fat’ the double-Cs.
d. CuCaC šimen ‘grease’
e. CaCCeC šamenet ‘cream’
Adjective f. CaCeC šamen ‘fat’

• But the meanings are not always transparent. Though it may be possible to see
some semantic connection, what the connection is could be quite vague:
√
(3) Words derived from bxn (Arad 2003: 743): There is a phonological
alternation between [b] and
Category Binyan Form Gloss [v] in these examples.

Verb a. CaCaC baxan ‘test, examine’

b. hiCCiC hivxin ‘discern’
Noun c. miCCaC mivxan ‘an exam’
d. CoCaC boxan ‘a quiz’
e. maCCeCa mavxena ‘a test-tube’
f. aCCaCa avxana ‘a diagnosis’
 Nicholas LaCara · Root-and-pattern morphology 3

√
• In (3) it seems as though bxn acquires two different meanings as verbs, and
Hebrew speakers must acquire the two interpretations that this root has.

• It has a number of apparently related but specialized meanings when it is a noun.

• So, to some extent, roots must be underspecified in their meanings.

– If roots occur only in the contexts of the different words derived from them, This is related to the idea from
it is difficult to access the semantic core of the root. Marantz (1997) that elements
like roots only receive their
– When we put together the words made from the same root, we might be meanings in context. In some
able to extract some semantic core shared by them sense, under DM, all words are
idioms.
• There are a couple of important things to see here, related to issues we’ve seen
in this course.

– The same root can occur in words of different categories. For instance,
√
šmn can occur in adjectives, nouns, and verbs.
– The thing that determines the category of a word is the binyan. Consonan-
tal roots are not restricted to any particular category.
– The idiosyncrasy of the meanings is similar to other derivational morpho-
logical processes we’ve seen (cf., reality, realist, realize, really).

• Indeed, as Arad (2003) discusses at length, words formed by this process can
serve as the input to further morphological derivation.
√
– For instance, it is possible to use sgr to derive both nouns and verbs:
√
(4) Words derived from sgr (Arad 2003: 746):
Category Binyan Form Gloss
Verb a. CaCaC sagar ‘close’
b. hiCCiC hisgir ‘extradite’
c. hitCaCCeC histager ‘cocoon oneself ’ Check out the metathesis in
this form; cf. McCarthy 1979:
Noun d. CeCeC seger ‘closure’ 228.
e. CoCCayim sograyim ‘parenthesis’
f. miCCeCet misgeret ‘frame’

• It is possible to take the noun (4f) misgeret ‘frame’, and derive a verb from it,
using a binyan with four consonants:

(5) Verb derived from noun misgeret (Arad 2003: 746):

Category Binyan Form Gloss
Noun miCCeCet misgeret ‘frame’
Verb CiCCeC misger ‘to frame’
CuCCaC musgar ‘to be framed’
 Nicholas LaCara · Root-and-pattern morphology 4

√
• The resulting verb has the same root as the noun ( sgr), but it isn’t being derived
directly from the root; it has several properties of the noun:

– The verb has the meaning of ‘frame’ derived in the original noun.
– The morphology contains the initial [m] segment from the prefix in the Arad notes there is no
miCCeCet binyan, treating it as if it were another root consonant. explanation for why the suffix
-et is lost.
– The vowels of the the vocalic melody are those of the the verbal paradigm. This is most evident in the
passive form musgar ‘be
• So this has properties of a derivational process – namely, category changing, and framed’.
idiosyncratic meaning.

• However, this system plays a role in determining verbal paradigms as well.

√ √
(6) Hebrew binyan with roots Qmd and qpl (Arad 2003: 742):
Binyan Root Verb Gloss
√
1 CaCaC Qmd Qamad ‘be standing’
√
2 niCCaC Qmd neQamad ‘stand up’
√
3 CiCCeC qpl qipel ‘fold’ (trans.)
√
4 CuCCaC qpl qupal Passive of 3
√
5 hiCCiC Qmd heQemid ‘make stand up’
√
6 huCCaC Qmd huQamad Passive of 5
√
7 hitCaCCeC qpl hitqapel ‘fold’ (intr.)

– Different binyanim result in different forms of the ‘same’ verb.

– Notice, too, that the primary difference between binyanim 3 and 4, on the
one hand, and 5 and 6 on the other is the vowels.

• Here is one more (unfortunately incomplete) paradigm, with only one root:
√
(7) Paradigm with gdl (Ussishkin 2003):
Binyan Verb Gloss
1 CaCaC gadal ‘he grew’ (intr)
3 CiCCeC gidel ‘he raised’
4 CuCCaC gudal ‘he was raised’
5 hiCCiC higdil ‘he enlarged’
6 huCCaC hugdal ‘he was enlarged’

– Again, we see a characteristic (though somewhat opaque) semantic rela-

tion between elements derived from the same root.
– However, these are all verbs, and information like active vs. passive is en-
coded in these different forms.
 Nicholas LaCara · Root-and-pattern morphology 5

2.2 Root-and-pattern morphology in Arabic

• As can be seen in the chart below, Arabic words are also based on consonantal
roots arranged into different templates.
√
(8) Arabic words derived from ktb (Tucker 2011: 178): This is very similar to the list
given by Katamba (1993: 164).
Category Template Form Meaning Much of the from Tucker and
Katamba (1993) ultimately
Verb a. CaCaCa kataba ‘he wrote’ come from McCarthy (1979,
b. CaCCaCa kattaba ‘he made someone write’ 1981).
c. nCaCaCa nkataba ‘he subscribed’
d. CtaCaCa ktataba ‘he copied’
Noun e. CiCaaC kitaab ‘book’
f. CuCCaaC kuttaab ‘Koranic school’
g. CuCayyiC kutayyib ‘booklet’
h. maCCaCa maktaba ‘library, bookstore’
g. muCaaCiC mukaatib ‘correspondent, reporter’

– As in Hebrew, the same root may appear in both nouns and verbs.
– There is some degree of idiosyncratic meaning here (cf. kuttaab ‘Koranic
school’ and mukaatib ‘reporter’).

• Also, as in the Hebrew examples, the difference between active and passive verbs
in many paradigms is the vowels (though there are more patterns here).

(9) Perfective active and passive paradigms of fQl, “doing, action”: Cited in Tucker (2011: 180)

Template Form Template Form

I C1 aC2 aC3 faQal I C1 uC2 iC3 fuQil
II C1 aC2 C2 aC3 faQQal II C1 uC2 C2 iC3 fuQQil
III C1 aaC2 aC3 faaQal III C1 uuC2 iC3 fuuQil
IV PaC1 C2 aC3 PafQal IV PuC1 C2 iC3 PufQil
V taC1 aC2 C2 aC3 tafaQQal V tuC1 uC2 C2 iC3 tufuQQil
VI taC1 aaC2 aC3 tafaaQal VI tuC1 uuC2 iC3 tufuuQil
VII nC1 aC2 aC3 nfaQal VII n/a
VIII C1 taC2 aC3 ftaQal VIII C1 tiC2 iC3 ftiQil
IX C1 C2 aC3 C3 fQall IX n/a
X staC1 C2 aC3 stafQal X stuC1 C2 iC3 stufQil
Q1 C1 aC2 aC3 aC4 faQalal
Q2 taC1 aC2 C3 aC4 tafaQlal

– So again, just as in Hebrew, part of the inflectional system is based in root-

and-pattern morphology as well, not just derivational morphology.
– Notice here, though, that the distinction between voice (i.e., the choice of
vowels) is much more transparent.
 Nicholas LaCara · Root-and-pattern morphology 6

• It is worth pointing out, too (though we will not talk about it as much) that
Arabic verbs take suffixes to express agreement

(10) Inflection of Perfective Verbs Exemplified in Form I: Suffixes are the same across all
perfective forms. Circumfixes
Person Gender Singular Plural are used in the imperfective
forms.
1 — faQal-tu faQal-naa
2 Masc faQal-ta faQal-tum
Fem faQal-ti faQal-tunna
3 Masc faQal-a faQal-uu Most of the verbs from
Fem faQal-at faQal-na Katamba, discussed below, are
3rd person singular masculine
and bear the -a suffix.
• These can be treated just like suffixes in any other language; there won’t be much
interesting to say about them.

2.3 Other Semitic languages

• There are many other Semitic languages (including Amharic, Tigrinya, Maltese,
and Aramaic) whose root-and-pattern systems have received far less attention
in the theoretical literature.
• These languages, too, display similar behaviour to Hebrew and Arabic. This can
be seen in Amharic (Ethiopia):
√ √
(11) Amharic words formed from brk’ and lbs: Data from Ayalew 2013: 8 and
Kramer 2009: 157.
Root Template Form Gloss
√
brk’ C1 äC2 C2 äC3 ä bärräk’ä ‘lighten; thunder’ (v.)
C1 C2 äC3 mä-bräk’ ‘lightening, thunder’ (n.)
C1 aC2 C2 äC3 ä barräk’a ‘go off accidentally (gun)’
C1 C2 C3 brk’ ‘scarce, rare, precious’
C1 C2 äC3 äC3 C2 C2 äC3 ä t-bräk’ärräk’ä ‘glitter’ This sure looks like internal
√ reduplication for
lbs C1 äC2 C2 äC3 ä läbbäsä ‘wear’ pluractionality.
C1 1C2 C3 l1bs ‘clothing’
C1 1C2 aC3 l1bas ‘cover’

• There is some work on the morphosyntax of other Semitic languages (see Kramer
2009 on Amharic and Tucker 2013 on Maltese), but I couldn’t find much on the
root-and-pattern morphology outside Arabic and Hebrew.

2.4 Theoretical issues

• If we want to account for this sort of morphological phenomenon under our
current understanding, we need to address several issues.

– If morphemes are heads, as we have been assuming, what heads corre-

spond to which parts of the Semitic word?
– Since the morphology is non-concatenative, how do get explain how each
morpheme comes to be interspersed among the others?
 Nicholas LaCara · Root-and-pattern morphology 7

3 The syntactic derivation of Semitic verbs

• Most discussions in the literature pay the most attention to how verbs are de-
rived, so we’ll concentrate on that here.

• We’ll start by looking at the syntactic structure of the verbs before turning to
issues having to do with the morphophonology.

– We can map each of the subcomponents of the Semitic verb onto a different
head in the syntax.
– Prosodic morphology of the type we used to explain reduplication will
allow us to explain how these pieces are put together.

3.1 Deriving verbs in the syntax

• We saw above that we can break Hebrew and Arabic verbs up into three subparts:

i. A consonantal root that contributes idiosyncratic semantic information.

ii. A vocalic melody that represents information about voice.
iii. A pattern of consonants and vowels that is related to argument structure
and syntactic category.
√
• We can see each of these at work in Hebrew verbs formed from qpl.
√
(12) Hebrew verbs derived from qpl:
Binyan Verb Gloss
CiCCeC qipel ‘fold’ (transitive)
CuCCaC qupal ‘be folded’ (passive)
√
– The root qpl contributes the meaning ‘fold’.
– The template CVCCVC is used for agentive verbs.
– The vowels /i. . . e/ is used for actives, while /u. . . a/ are used for passives.

• Similar observations come from the Arabic data we saw in (8) and (9):
√
(13) Arabic verbs derived from f Ql:
Template Verb Gloss
CaCaC faQala ‘he does’
CuCiC fuQila ‘it is done’

– Here the template and root remain the same; only the vocalic melody changes,
affecting the voice of the verb.
– In fact, it’s clearer in Arabic that the consonants, vowels, and template act
independently of one another, since the templates aren’t tied to the vowels
in the same way as they are in Hebrew.
 Nicholas LaCara · Root-and-pattern morphology 8

• Arad (2005: 45) uses these correlations to argue for the following correlations
which Tucker (2011) adopts for Arabic as well.

– The Semitic root corresponds to the DM notion of root.

– The vocalic melody corresponds to Voice0 .
– The pattern is a template corresponding to the verbalizing head v0 .

• Thus, the underlying structure for a verb like Hebrew qipel ‘fold’ or Arabic faQal
‘did’ (ignoring tense and agreement morphology) might look as follows:

(14) a. VoiceP b. VoiceP

Voice′ Voice′

Voice0Act vP Voice0Act vP
/i. . . e/ /a. . . a/
√ √
v0 P v0 P
CVCCVC CVCVC
√ √
qpl f Ql

• It is generally thought that verbs move to T0 in Hebrew. This dates back at least to
Doron 1990.
– This means that these elements will all undergo head movement to T0
where they will be placed adjacent to tense and agreement morphology.

(15) TP

DPSub
i T′

T0 VoiceP

Voice0 T0 DPi Voice′

v0 Voice0Act Voice0 vP
/i. . . e/
√ √
qpl v0 v0 P
CVCCVC
√
qpl

• Because these elements are all grouped together under T0 , they will be treated
together as a word by the phonology (they are all in a single M-word).

• Arad (2005: 43) explains that the various changes in the vocalic melody changes
in active verbs with different templates can be explained by analyzing them as
exponents of contextually specified Vocabulary Items:
 Nicholas LaCara · Root-and-pattern morphology 9

(16) Vocabulary of active Voice0 :

a. Voice[act] ↔ /a. . . a/ / CVCVC
b. Voice[act] ↔ /i. . . a/ / nVCCVC
c. Voice[act] ↔ /i. . . e/ / CVCCVC
d. Voice[act] ↔ /i. . . i/ / hVCCVC
e. Voice[act] ↔ /a. . . e/ / hitCVCCVC
(17) Vocabulary of passive Voice0 :
a. Voice[pass] ↔ /u. . . a/ / CVCCVC, hVCCVC
b. Voice[pass] ↔ niCCaC / CVCVC

• The pattern inserted in v0 is presumably determined by features on v0 along with Unfortunately, Arad does not
what root it co-occurs with. say what these features are or
would be, though to her credit
this is not what the central goal
• Verb movement in Arabic is a bit more complex due to the interaction of auxil-
of her work is. Given what we
iaries with vso and svo word orders (see Tucker 2011: 184–191). have seen so far, the relevant
features may be transitivity
– For now, we can limit ourselves to assuming that the verb moves to T0 . and causativity, though
pinning down the exact
– vso orders occur when the subject remains in SpecVoiceP. meaning of each binyan
remains difficult.
(18) TP

T0 VoiceP

Voice0 T0 DPSub Voice′

v0 Voice0Act Voice0 vP
/a. . . a/
√ √
f Ql v0 v0 P
CVCVC
√
qpl

3.2 Brief aside on deriving verbs from nouns

• Recall, too, that verbs can be derived from nouns:
√
• The root sgr has a number of possible interpretations:
√
(19) Words derived from sgr:
Category Binyan Form Gloss
Noun miCCeCet misgeret ‘frame’
Verb CiCCeC misger ‘to frame’
CuCCaC musgar ‘to be framed’
 Nicholas LaCara · Root-and-pattern morphology 10

• We can use the same sort of derivational processes that we have assumed all
along to create an appropriate structure.
√
– The noun misgeret ‘frame’ can be formed by combining the root sgr with
an appropriate n0 .
– The verbs can be derived by combining this n0 with verbal material.

(20) a. Noun formation: b. Verb formation from a noun:

n0 → [misgeret] [musgar] ← Voice0

√
sgr n0 v0 Voice0Pass
miCCeCet /u. . . a/
[misgeret] ← n0 v
CVCCVC
√
sgr n0
miCCeCet

• However, as far as I know, there is no analysis of the morphophonology of these

sorts of verbalizations. Things to wonder about in the coming discussion:

– Why does the /m/ segment get preserved rather than the /t/?
– What happens to the vowels in the noun? Do they get overwritten? How? This might get a partial answer
in some of the OT-based work
– Why doesn’t the template act more like a reduplicant morpheme? Why on Semitic morphology; see
should the consonants in the noun get remapped to the new template? Ussishkin (2000, 2003).

4 The morphophonology of Semitic verbs

• Under the syntactic derivation above, the root, template, and vowels are each
separate morphemes.

• Under our current assumptions about morpheme linearization, we end up de-

riving nonsensical linearization statements:
√ ⌢
(21) a. qpl CVCCVC b. CVCCVC⌢/i. . . e/

• Notice how similar this is to the linearization statements we derive when we try See the example of O'odham on
to derive reduplication in our syntactic system. page 8 of the lecure notes on
reduplication .
• However, the same system that let us understand reduplication gives us a way of
accounting for root-and-pattern morphology as well.

• Work in Autosegmental Phonology, starting with McCarthy (1979), proposed

that each morpheme could be associated with a different tier:
 Nicholas LaCara · Root-and-pattern morphology 11

(22) The morphology of qipel ‘folded’ (Hebrew), following McCarthy (1979):

√
q p l Root tier (= )

C V C C V C Skeletal tier (=v0 )

i e Vocalic melody tier (=Voice0 )

– v0 introduces a template, just like reduplicant morphemes do.

– The slots in this template must come to be associated with material in the
root and the vocalic melody.

• So now let’s look at how this happens.

4.1 Association
• Let us go through the example of Arabic kataba ‘he wrote’, discussed by Katamba
(1993: 165–167).
• In a simple case like this, the underlying template is CVCVC. The final vowel in the form is
actually an agreement affix.
– Consonants from the root tier associate with each C slot from left to right. We'll come to those below.

– Under Autosegmental Phonology, it is widely assumed that two adjacent

identical segments are often disallowed (under the Obligatory Contour
Principle).
– It follows that there is only one underlying vowel /a/ that spreads to each It is worth noting that this
vowel position. spreading rule needs to be
language-specific. Some
languages with similar
(23) The morphology of perfective stem katab- ‘wrote’ (Arabic): morphology do not show this
spreading, and as we saw with
k t b reduplication, a single vowel
does not fill every available
C V C V C spot in most languages.

a
• In cases like (22) above, we must also assume that consonants have the ability to
spread to fill empty C slots:

(24) The morphology of causative perfective kattab- ‘caused to write’ (Arabic):

k t b

C V C C V C It is worth contemplating why

this spreading occurs. One
a could imagine other forms, like
*kaktab or *katbab, or *zazrara,
– Here, it must be possible for /t/ to be associated with the middle two C- as McCarthy (1979: 249–250)
slots in the template. does. He ultimately (p. 255–257)
states that *katbab is formed
but a rule goes back and erases
• In fact, such spreading appears even in templates without geminate consonants. the association line between
There are roots with only two consonants, like zr ‘pull’, and in these the second /b/ and the third consonant
consonant spreads to all remaining C slots: slot.
 Nicholas LaCara · Root-and-pattern morphology 12

(25) a. zarar- ‘pulled’: b. zarrar- ‘caused to pull’:

z r z r

C V C V C C V C C V C

a a

• It is also worth pointing out that quadrilateral roots (those with four consonants)
also exist, and they use b

(26) dahraj- ‘caused to roll’:

d h r j

C V C C V C

4.2 Infixation in Arabic

• As you may have noticed, Arabic template VIII contains an infixed /t/:

(27) Templates I and VIII:

Template Active Passive

I C1 aC2 aC3 faQal fuQil
VIII C1 taC2 aC3 ftaQal ftiQil

• This is apparently some sort of reflexive marker, as can be seen in the following
Syrian Arabic examples:

(28) Reflexive infixation in Arabic:

Root Form Gloss
smQ samiQa ‘hear something’
stamaQa ‘hear oneself (listen)’
rf Q rafaQa ‘lift something’
rtafaQa ‘lift oneself ’

• The question is how to incorporate this element into the templatic root-and-
paradigm approach used here.

• Following McCarthy (1981), we can further incorporate morpheme tiers into our
representations, showing which
 Nicholas LaCara · Root-and-pattern morphology 13

(29) Structure of kutiba ‘it was written’:

µ√

k t b
C V C V C V
u i a
µVoi µ Agr

– Notice that this gives us a way to add the requisite agreement suffix to the
representation!

• If we assume that the reflexive morpheme is a morpheme of its own, then it /t/
will be totally independent of the root, melody, and template.

– This means it will be added to the representation with its own µ node and
consonant slot.
– When this affix is next to the verb, it undergoes infixation. There is no need
for association lines to cross.

(30) Derivation of rtafaQa ‘he lifted himself ’:

µ√ µ√

r f Q r f Q
a. C + C V C V C V → b. C C V C V C V
t a a t a a
µRe f l µVoi µ Agr µRe f l µVoi µ Agr

• We need only assume an infixation rule of the sort we’ve used previously:

(31) /t/ → [Wd C-t-V . . . / + [Wd CV. . . cf. McCarthy 1979: 252.

4.3 Beyond the template

• Saying that there is a single morpheme that introduces the template has come
under criticism.

– A look over most of the templates we’ve seen reveals is they tend toward
having CV(C) syllables.
– The Hebrew forms, in fact, are largely bi-syllabic.

• This has led some people working in theories like OT, specifically Ussishkin
(2000, 2003), to suggest that the template isn’t real.

• The reason for this is that the patterns imposed by the templates can be made
to fall out from general principles about how phonology likes to organize con-
sonants and vowels, and what kinds of prosodic structures are permissible.
 Nicholas LaCara · Root-and-pattern morphology 14

• Tucker (2010, 2011) takes a very interesting take on this approach.

– Basically, if you take a bunch of consonants and a bunch of vowels, the only
way the phonology would want to arrange them is by interspersing them.
– Thus, if you give the phonology C1 C2 C3 + V1 V2 , it’s gonna wanna spit out
C1 V1 C2 V2 C3 .

• For those who can appreciate an OT analysis, he proposes:

(32) Faith: A cover constraint for: (33) *Complex: A cover constraint for:
a. Max: a. *Complexons :
No deletion. No complex onsets.
b. Dep: b. *Complexcod :
No epenthesis. No complex codas.

(34) Integ(rity):
A segment in the output has a single correspondent in the input.
(35) Contig(uity): McCarthy and Prince 1995
The portion of the input and output strings standing in correspondence
forms a continuous string.

• These can derive Arabic Template I: But see also Tucker 2010 for
other templates.

(36) Verb stems in template I:

/fQl/ + /a/ Faith *Complex Integ. Contig.
☞ a. faQal * **
b. faQl *! *
c. fQal *! *
d. Qal *! *

– We have to split up the root, violating Contiguity.

– We also must duplicate the input vowel, violating Integrity.
– But this is better than having complex onsets or codas, or deleting input
material.
Nicholas LaCara · Root-and-pattern morphology 15

Terms
nonconcatenative morphology A form of word formation that (Katamba and Stonham 2006). The core meaning of the word
does not involve stringing morphemes together sequentially. is usually associated with this element.
root The core of a (lexical) word. template A series of CV-slots on a skeletal tier introduced by
root In traditional morphology, the irreducible core of a word, some morphemes.
with absolutely no other morphological elements attached to it transfix A discontinuous affix interspersed in its base.

References
Arad, Maya. 2003. Locality constraints on the interpretation of and morphology. Doctoral Dissertation, Massachusetts Insti-
roots. Natural Language & Linguistic Theory 21:737–778. tute of Technology.
Arad, Maya. 2005. Roots and patterns: Hebrew morpho-syntax. McCarthy, John J. 1981. A prosodic theory of nonconcatenative
Dordrecht: Springer. morphology. Linguistic Inquiry 12:373–418.
Ayalew, Bezza Tesfaw. 2013. The submorphemic structure of McCarthy, John J., and Alan Prince. 1995. Faithfulness and redu-
Amharic: Toward a phonosemantic analysis. Doctoral Disser- plicative identity. In Massachusetts occasional papers in linguis-
tation, University of Illinois at Urbana-Champaign. tics, ed. J. Beckman, L. W. Dickey, and S. Urbanczyk, 249–384.
Doron, Edit. 1990. V movement and VP-ellipsis. Ms., The He- GLSA Publications.
brew University of Jerusalem. Published as Doron 1999. Tucker, Matthew A. 2010. Roots and prosody: The Iraqi Ara-
Doron, Edit. 1999. V movement and VP-ellipsis. In Frag- bic derivational verb. Recherches linguistiques de Vincennes
ments: Studies in ellipsis and gapping, ed. S. Lappin and E. Ben- 39:31–68.
mamoun, 124–140. Oxford: Oxford University Press. Tucker, Matthew A. 2011. The morphosyntax of the Arabic verb:
Katamba, Francis. 1993. Morphology. New York: St. Martin’s Toward a unified syntax-prosody. In Morphology at Santa
Press. Cruz: Papers in honor of Jorge Hankamer, ed. Nicholas LaCara,
Katamba, Francis, and John Stonham. 2006. Morphology. Mod- Anie Thompson, and Matthew A. Tucker, 177–211. Santa Cruz:
ern Linguistics. Basingstoke: Palgrave Macmillan, 2nd edition. SlugPubs.
Kramer, Ruth. 2009. Definite markers, phi features and agree- Tucker, Matthew Alan. 2013. Building verbs in Maltese. Doctoral
ment: A morphosyntactic investigation of the Amharic DP. Dissertation, University of California, Santa Cruz.
Doctoral Dissertation, University of California, Santa Cruz. Ussishkin, Adam. 2000. The emergence of fixed prosody. Doc-
Marantz, Alec. 1997. No escape from syntax: Don’t try morpho- toral Dissertation, University of California, Santa Cruz.
logical analysis in the privacy of your own lexicon. In Proceed- Ussishkin, Adam. 2003. Templatic effects as fixed prosody: The
ings of the 21st Annual Penn Linguistics Colloquium, ed. Alexis verbal system in Semitic. In Research in Afroasiatic Grammar
Dimitriadis, volume 4 of Penn Working Papers in Linguistics, II: Selected papers from the fifth conference on Afroasiatic lan-
201–225. guages, Paris, 2000, ed. Jacqueline Lecarme, 511–530. Amster-
McCarthy, John J. 1979. Formal problems in Semitic phonology dam: John Benjamins.