You are on page 1of 29

Morphology (2020) 30:61–89

https://doi.org/10.1007/s11525-020-09350-w

Non-sound’ verb Inflection in Arabic: Allomorphic


variation and paradigmatic uniformity

Samira Farwaneh1

Received: 16 February 2015 / Accepted: 27 January 2020 / Published online: 11 February 2020
© Springer Nature B.V. 2020

Abstract Focusing on the Levantine Arabic variety, this paper investigates the
paradigmatic asymmetry observed in the inflection of sound verbs whose stems con-
tain three or four consonants and non-sound verbs whose stems contain only two
consonantal realizations. It provides a unified account of verb inflection within Opti-
mality Theory and the theory of paradigms. The target of investigation is the allomor-
phic variation in the non-third person markers which appear in their basic allomorph
in inflected sound verbs, but appear in their augmented éeC form in the paradigms of
non-sound (weak and geminate) verbs. Previous analyses have viewed this paradig-
matic inconsistency as arbitrary exceptions requiring highly specific rules or allomor-
phic postulates, thus treating the sound and non-sound verb systems as two distinct
types. This paper shows that the interaction of independently-motivated markedness
constraints with paradigmatic uniformity constraints is capable of producing these
allomorphic effects without recourse to ad-hoc rules or arbitrary allomorphic state-
ments. The two verbal types are therefore treated as a unified system.
The Optimality-theoretic account of affix allomorphy accords a prime role to the
markedness constraint optimizing binarity of prosodic constituents, and rhythmic and
prosodic uniformity within and across inflectional subparadigms the interaction of
which explains the emergence of the augmented allomorph of the affix. The paper ad-
vocates an expanded definition of paradigms allowing grammatical categories such as
gender, person, case, etc. to form subparadigms subject to what I refer to as paradigm-
to-paradigm faithfulness requiring paradigmatic identity within a grammatical cate-
gory regardless of verb type. Hollow verbs which present an interesting challenge
to the analysis are discussed and accounted for by highlighting the role of Anchor
in differentiating between hollow verbs and other non-sound verbs. The paper ends
with a comparison with the contrast analysis of Broselow (2008) demonstrating on
empirical grounds the superiority of the uniformity against the contrast account.

B S. Farwaneh
farwaneh@email.arizona.edu
1 The University of Arizona, Tucson, AZ, USA
62 S. Farwaneh

Keywords Arabic verbs · Optimality · Allomorphy · Inflection · Paradigm


uniformity

1 Verb types and inflectional asymmetries in Levantine


The verb morphology of Classical Arabic received a great deal of attention in the
last four decades and was the impetus behind the development of several theoretical
models of morphology. On the other hand, little attention was given to verb deriva-
tion and inflection in Spoken Arabic, until the last decade when dialectal data were
employed to put to the test the theories developed for Classical Arabic often with
intriguing results (cf. Benmamoun 1999, 2003). The vast literature on the verbal sys-
tem of Arabic, often focusing on Classical or Modern Standard Arabic, usually be-
gins with an analysis of the basic Measure I of the sound verb, which constitutes the
model to which all other non-sound verbs are later compared. A sound verb is that
whose stem phonetically realizes three or four autonomous consonants, that is, with
no doubling/gemination. Example (1) gives examples of sound verbs from Levantine
focusing on triconsonantal stems. All outputs exhibit the zero-inflected third person
masculine singular form.1
(1) Sound Verbs
Verb Gloss Verb Gloss
katab ‘wrote’ nizil ‘left’
èamal ‘carried’ rikib ‘rode’
rasam ‘drew’ misik ‘held’
Other verbs exist whose stems are limited to two consonantal realizations. I will refer
to these verbs collectively as non-sound verbs for ease of reference. The non-sound
verb category includes geminate, weak and hollow verbs. Geminate verbs are char-
acterized by final consonant gemination, while weak verbs are vowel final; hollow
verbs are monosyllabic with a medial long vowel. Representative examples of gemi-
nate, weak and hollow verbs are shown in (2), (3) and (4) respectively.2
(2) Geminate verbs
Verb Gloss Verb Gloss
èabb ‘loved’ radd ‘answered’
marr ‘passed by’ laff ‘wrapped’
(3) Weak Verbs
Verb Gloss Verb Gloss
rama ‘threw’ rid.i ‘agreed’
baka ‘cried’ nisi ‘forgot’

1 The following phonetic symbols are used throughout this paper:


è= voiceless pharyngeal fricative; Q = voiced pharyngeal fricative; P = glottal stop; š = voiceless
alveopalatal fricative; ž = voiced alveopalatal fricative; X = voiceless uvular fricative; ġ = voiced uvular
fricative; t./d./s. = emphatic coronals.
2 In the more familiar root-based analysis informed by the descriptive contributions of the Arabic linguistics
tradition, geminate, weak and hollow verbs are hypothesized to be derivatives of biliteral, glide-final and
glide-medial roots respectively.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 63

(4) Hollow Verbs


Verb Gloss Verb Gloss
zaar ‘visited’ žaab ‘brought’
raaè ‘went’ saal ‘leaked’
In addition to the number of consonant realizations distinguishing the input of sound
and non-sound stems, the two types are also distinct inflectionally. The inflectional
paradigm of the perfective form of sound verbs is systematic, preserving the identity
of both the stem and the affix, as the paradigm of the verb [katab] ‘write’ in (5)
illustrates.3

(5) Strong verb paradigm


1P 2P 3P
-F -PL katábt katábt kátab
+F -PL katábt katábti kátabat
+PL katábna katábtu kátabu
The inflectional paradigm of weak verbs on the other hand selects an augmented
allomorph of the first and second person affixes, as shown in the inflection of [rama]
‘throw’.

(6) Weak Verb Paradigm


1P 2P 3P
-F -PL raméet raméet ráma
+F -PL raméet raméeti rámat
+PL raméena raméetu rámu
Here we find an augmented allomorph of the first and second person marker with
an initial long stressed mid front vowel. Geminate verbs likewise show the same
augmented affix allomorph. The paradigm of the verb [radd] ‘answer’ is an example.4

(7) Geminate Verb Paradigm


1P 2P 3P
-F -PL raddéet raddéet rádd
+F -PL raddéet raddéeti ráddat
+PL raddéena raddéetu ráddu
The anomalous behavior of weak and geminate verbs in Levantine and other spoken
dialects received little attention compared to the plethora of studies on stem and affix
allomorphy in Classical Arabic and Modern Standard Arabic (Brame 1970 within
Standard Generative Phonology, McCarthy 1979, 1981 in Non-Linear Phonology,
McCarthy and Prince 1986 in Prosodic Morphology, McCarthy (2002) within the
Optimal Paradigm model of OT to name a few examples). This paper contributes to

3 The following inflectional category abbreviations are used: 1p, 2p, 3p = first, second and third person;
f = feminine; pl = plural.
4 Interestingly, all Arabic varieties inflect geminate verbs with the –VVC extension. I know of no Arabic
dialect preserving the Classical Arabic pattern of splitting the geminate before consonantal affixes as in
[radad-tu] instead of [raddeet] ‘I responded’.
64 S. Farwaneh

filling this gap by highlighting the paradigmatic difference between the inflection of
sound and non-sound verbs in dialectal Arabic.
The inflectional paradigm of the perfective sound verbs consists of outputs faithful
to their base components (stem+affix), while first and second person affixes appended
to geminate and weak verbs appear in an augmented form with an initial long stressed
mid (or high) vowel. Thus, while third person markers are invariable, non-third mark-
ers have two allomorphs, a consonant-initial allomorph specific to sound verbs, and
a vowel-initial allomorph appended to non-sound verbs.
Previous analyses have viewed this paradigmatic inconsistency as arbitrary ex-
ceptions requiring special allomorphy statements (McCarthy 1986), thus treating
the two systems as two distinct types. In this paper I provide an explanatory ac-
count of affix allomorphy in Arabic within the framework of Optimality Theory
(henceforth OT) (Prince and Smolensky 1993; McCarthy and Prince 1993a, 1995),
and paradigm-based morphology (Gafos 2003; Broselow 2008; Kenstowicz 2005;
McCarthy 2002 and works in Downing et al. 2005), all of which adopt a stem-based
approach, I show that affix allomorphy is not arbitrary, but is governed by the prosodic
well-formedness of the output on the one hand, and the symmetry of the paradigm on
the other. The verbal system is therefore treated uniformly despite the nonuniformity
of the base. That is, the analysis presented in this paper applies to all verbs in PA
regardless of type, sound, hollow, geminate, or weak. Section 2 gives an analysis of
the sound and non-sound stems that serve as a base for faithfulness evaluation, while
Sect. 3 offers an analysis of the observed allomorphy within OT, showing how such
an approach can account for all cases of allomorphy through the proper ranking and
interaction of faithfulness, markedness, and paradigm uniformity constraints. Sec-
tion 4 focuses on the hollow verb class which seems to defy the analysis presented
for weak and geminate verbs. Section 5 compares the uniformity-based account pro-
posed here to the contrast-based account of Broselow (2008) demonstrating, based
on empirical evidence, the superiority of a paradigmatic uniformity-based analysis
over a contrast-based analysis.

2 Deriving the uninflected stem


Before addressing the source of the augmented non-third person suffix, this sec-
tion will focus on the derivation of the uninflected stem of the four verb types:
sound, geminate, weak and hollow verbs. The proposal assumes that the Spoken
Arabic lexicon consists of fully specified stems not roots. this assumption is based
on empirical evidence from Arabic and Hebrew including broken plurals (Mc-
Carthy and Prince 1990; Ratcliffe 1998), diminutives (McCarthy and Prince 1990;
Ratcliffe 1997), hypocoristics (Farwaneh 2006, 2007), vocalic patterns (Benmamoun
2003), loan words (Bat-El 1994), denominal verbs (Bat-El 1994; Ussishkin 1999;
Farwaneh 2006), verb morphology (McCarthy 1993; Benmamoun 1999, 2003; Gafos
2003; Ussishkin 2005). In addition, the stem-based assumption adopted in this pa-
per is congruent with paradigm-oriented approaches which by default assume fully-
specified stems.5 A survey of the Levantine verb stems show that verbs are subject

5 Psycholinguistic experimental works present robust evidence of the psychological reality of the conso-
nantal root (Frost et al. 2000; Boudelaa and Marslen-Wilson 2011; Ussishkin et al. 2015). However, these
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 65

to two major requirements: Foot Binarity which limits the size of the stem discussed
in Sect. 2.1, and Final consonantality which demarcates its right edge discussed in
Sect. 2.2.

2.1 Constraints on stem size: Foot binarity

The stem of a sound verb in Levantine like [katab] displays the same prosodic and
alignment characteristics observed in Classical Arabic and Modern Standard Ara-
bic, namely that it is disyllabic ending in a consonant (McCarthy and Prince 1986;
McCarthy 1993). In the approach pursued in this paper, these requirements do
not stem from a prescribed morphemic template, as in McCarthy’s earlier work
(McCarthy 1979, 1981), or from a templatic prosodic constraint specific to verbs
as in McCarthy (1993), but rather from independently-motivated constraints appli-
cable to lexical categories in general. Deriving templatic effects through constraint
interaction without specific reference to a verb template figured in the Optimality-
Theoretic literature as Generalized Templatic Theory (McCarthy and Prince 1994;
Ussishkin 2005). The disyllabicity requirement is translatable as a markedness con-
straint governing the size of the stem, be it a verb or otherwise, defined in (8)6 :

(8) Constraints on Arabic stems


Ft-Bin: Feet are binary branching.
(McCarthy and Prince 1986; McCarthy 1993)7

Requiring feet to be binary-branching entails that a foot may not be monomoraic or


trisyllabic. The Ft-Bin constraint thus imposes a lower and upper bound on stem size,
limiting it to two syllables or two moras. Since according to the Prosodic Morphol-
ogy Hypothesis (McCarthy and Prince 1986, 1990, 1993a, 1993b) a lexical category
is co-extensive with a prosodic category, Ft-Bin derives the notion of minimal word:
A word is minimally a bimoraic or disyllabic foot. Surveying all verb types in Levan-
tine shows conformity to the foot binarity constraint, either in its moraic or syllabic
interpretation as shown in (9).

works focus on strong or sound triconsonantal forms, and do not include all classes of defective or non-
sound forms as stimuli. Twist (2006) work on Maltese shows that her experiments elicited faster reaction
times for strong verbs as opposed to weak verbs, and strong verbs also elicited fewer errors. Hence, the
primacy of the root in non-sound forms is still a debatable subject.
6 Caballero (2006) argues for template-specific constraints to account for cases of abbreviated reduplication
in Guarijio where the base truncates to match the truncated reduplicant. Verb templates in Arabic are
more straight-forward and consistent and are derivable from independently-motivated constraints applying
elsewhere in the language; hence postulating specific templates would be an unnecessary overloading of
the lexicon with predictable information. Thanks to an anonymous reviewer for bringing Cavallero’s work
to my attention.
7 Ussishkin (2000) proposes the Foot branching constraint (FtBranch) to avoid the ambiguity of FT-BIN,
since foot binarity could be interpreted quantitatively to mean bimoraic or disyllabic foot. In a quantity-
sensitive language like Arabic, adoption of this constraint requires limiting its effect to one layer of the
prosodic hierarchy, for allowing both syllables and moras to branch would result in an impermissibly large
syllables or feet. To avoid this, Ussishkin proposes an Alignment constraint as a “size-restricter constraint”
penalizing feet containing mor than two moras.
66 S. Farwaneh

(9) Type shape size example


sound CVCVC disyllabic katab
geminate CVCC bimoraic madd
weak CVCV disyllabic rama
hollow CVVC bimoraic zaar

Ft-Bin is inviolable since no verb stem in Arabic is smaller than two moras or larger
than two syllables.

2.2 Constraint on alignment: Final-consonantality

The second requirement, final consonantality, is an alignment constraint requiring


that stems end in a consonant. This constraint is stated in (10).

(10) Final consonantality


Final-C: stems are consonant final. (McCarthy 1993)

All verb stems except weak verbs are consonant final. Weak verbs of the shape CVCV
as in [rama] are analyzed in a root-based account as a derivative of a glide-final root.
The glide is subsequently vocalized stem-finally via series of specifically-ordered
phonological rules (Brame 1970). In the word-based approach adopted here, the in-
put of weak verbs is vowel-final with no underlying glide representation.8 With an
input lacking a final consonant, weak verbs like [rama] are the only verb type that
violates Final-C. Two strategies are available to remedy the stem to align it with a
final consonant in accord with Final-C, both of which run into conflict with other
constraints: The first strategy is to delete the final vowel and reduce the stem to a
single syllable *[ram], but this strategy reduces the stem to subminimality, thus vi-
olating Ft-Bin. The second strategy is to copy the final consonant as in *[ramam],
which violates integrity defined in (11).

(11) Integrity (McCarthy and Prince 1995)


No element in the base has multiple correspondents in the output.

Integrity is inviolable in the verbal system of Levantine since copying of the initial
or final consonant is disallowed. In other Semitic languages, e.g., Classical Arabic or
Modern Hebrew (McCarthy 1979 et seq.; Ussishkin 1999), a well-known asymmetry
first observed in McCarthy (1979) involves the spreading of final but not initial stem
consonant; thus, an allomorph like [samam] is observed but *[sasam] is unattested.
Neither the CVCi VCi nor Ci VCi VC configurations are permitted in LA. For a stem
to contain identical elements strict contiguity must be observed; as in the geminate

8 The argument for positing an underlying glide in weak verb roots comes from the observed ablaut pattern
in CA where the perfective surface vowel [a[ corresponds to either the imperfective [u[ and nominal [w],
or imperfective [i] and nominal [y], e.g., [ramaa yarmii ramyun] ‘throw perf., imperf, nom.’ versus [SaHa
yaSHuu SaHwun] ‘awaken perf., imperf., nom.’. in Levantine Arabic, all weak forms contain a high front
vowel or glide, indicating the disappearance of phonemic short /u/, cf. [ramaa/yirmi rami] versus [SiHi
yiSHaa SaHayaan]. Thus, the evidence for an underlying glide no longer holds.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 67

stem CVCi Ci .9 Integrity outranks both FT-Bin and Final-C. The ordering of Ft-Bin
over Final-C has been justified on the basis of the ban on monosyllabic consonant-
final but subminimal CVC in favor of the minimal but vowel-final CVCV output. The
ordering of Integrity over Ft-Bin is motivated by the outputs of geminate verbs where
integrity-compliant but monosyllabic CVCi Ci are favored over disyllabic forms with
non-adjacent identical consonants CVCi VCi , a point to which we shall return later.
With the three constraints ranked in (12), the evaluation of output candidates from
the input [rama] is illustrated in the tableau below.

(12) Integrity » Ft-Bin » Final-C


Candidate Integrity Ft-Bin Final-C
a. ram *!
b. ramam *!
c.  rama *
In addition to the active effect of the three constraints Ft-Bin, integrity and
Final-C, faithfulness constraints encompassing the now familiar Max(imality) and
Dep(endency) (McCarthy and Prince 1995) which guard against deletion and
epenthesis also participate in monitoring identity relations between input and out-
put. The uninflected stems of all verb types do not pose a challenge to Max and
Dep, but we add them to the hierarchy in anticipation of the discussion to follow.
The faithfulness constraints Max and Dep are further subcategorized to reflect the
unequal status of consonants and vowels in Arabic; a remnant of the old division
of labor between the consonantal root serving as a content morpheme and the vo-
calic melody serving as a functional morpheme. The subconstraint Dep-C prohibits
epenthetic onsets, while Dep-V prohibits epenthetic nuclei. Although Arabic is an
epenthetic language permitting epenthetic nuclei liberally, restriction on consonantal
epenthesis is stringent. Epenthetic consonants are limited to word-initial position.
This fact supports a hierarchy where Dep-C occupies a higher rank than Dep-V. Dep-
C ranks lower than Ft-Bin, mandating consonantal epenthesis word-initially. Dep-V
is subordinate to all constraints, while the markedness constraint Onset, requiring
all syllables to have a consonantal onset, is undominated. This derives the partial
hierarchy in (13):

(13) Onset » Ft-Bin » Dep-C » Dep-V

This hierarchy is motivated by the mandatory consonantal epenthesis in the impera-


tive form of verbs whose stem begins with a consonant cluster; e.g., /ktub/ > [Puktub]
‘write!’. The ranking in (13) causes the candidate with consonantal epenthesis to
emerge despite Dep-C and Dep-V violations, as shown in (14).

9 Noncontiguous identical elements are allowed in non-verbal stems, as in verbal noun (masdar) [malal]
‘bordam’ from [mall] ‘to be bored’, or [Haabib] ‘like-act.part.’ from [Habb] ‘to like’. Integrity violations
in verbal morphology is observed only in derived causatives and intensives compelled by the necessity to
realize the causative or intensive morpheme, e.g., [sammam] ‘poison-causative’ from [samm] ‘poison’.
68 S. Farwaneh

(14) Onset » Ft-Bin » Dep-C » Dep-V


Candidate Onset Ft-Bin Dep-C Dep-V
a. ktub *!
b. uktub *! *
b.  Puktub * *
Likewise, Max-C outranks Max-V, as preservation of input consonants takes prece-
dence over the retention of input vowels. This retention discrepancy arises from the
necessity to realize inflectional categories, most of which are expressed vocalically,
ultimately leading to the low faithfulness of stem vowels (Gafos 2003:349). The rank-
ing of Dep-C over Max-V is motivated by vowel hiatus (*VV) resolution. Whenever
two vowels come into juxtaposition, as a result of morpheme concatenation, two
strategies become available: Deletion of one of the vowels, a Max-V violation, or
epenthesis of a consonantal onset, a Dep-C violation. Arabic favors vowel deletion
over consonantal epenthesis in these cases, thus *VV > V not *VCV. The inflection
of vowel-final verbs with vocalic affixes discussed in Sect. 3.2.2 illustrates the inter-
action of Dep-C and Max-V. The following ranking in (15) is now supported.
(15) Partial Hierarchy
Onset » Integrity » Ft-Bin » Max-C » Dep-C » Max-V » Dep-V
The basic stem of all verb types, be it CVCVC, CVCC, CVVC or CVCV can be
derived by the hierarchy in (15) without reference to a specific template. Complexities
arise when consonantal and vocalic affixes are appended which is the focus of the
following section.

3 Inflectional paradigms and paradigmatic identity


This section will address the main objective of this paper, namely the affix allomorphy
in non-third person inflection. The inflectional paradigms of sound vs. geminate/weak
verbs illustrated in (1), (2), and (3) clearly show that while third person allomorphs
are invariable, non-third person affixes have two allomorphs, a basic and augmented
one. The following table isolates the allomorphs of person markers and shows their
distribution.
(16) Person markers
1P 2P 3P
Sound Non-Sound Sound Non-Sound Sound Non-Sound
-F –PL -t -éet -t -éet -0 -0
+F -PL -t -éet -ti -éeti -at -at
+PL -na éena -tu -éetu -u -u

The source of the long /ee/10 in the augmented affixes is puzzling. It appears at first
glance as an epenthetic vowel separating consonant clusters, /èabb-t/ > [èabbeet]. But

10 Diachronically, the extension /ee/ evolved from the coalescence of the diphthong /ay/ attested in earlier
varieties of Arabic, including Standard, and attested in some modern dialects, e.g., Lebanese, Maltese,
and Yemeni. But since Palestinian Arabic has no diphthongs, I assume that the mid /e/ vowel has been
phonemicized.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 69

as McCarthy (1986) points out, the final stressed vowel in the augmented allomorphs
of the non-third affix cannot be an epenthetic vowel inserted to preempt consonant
hiatus, since no language documented so far has long stressed epenthetic vowels. The
augmented affixes then seem to be the result of allomorphy and not the phonology.
Pre-OT derivational analyses have viewed allomorphic variation responsible for this
and similar paradigmatic inconsistency as class-specific stipulative statements speci-
fying the form of the exceptional allomorph and its distribution. In the case at hand,
we can postulate an allomorphy rule adapted from McCarthy (1986) in (17) specify-
ing the form and environment of the two stem allomorphs.
(17) Allomorphy Rule for Weak and Geminate Verbs
affix[#eeC(V)] ----> /Ci {V, Ci }
affix[#C(V)] ----> /
This rule stipulates that the augmented affix eeC is inserted if the stem ends in an open
CV syllable or a geminate Ci Ci , otherwise, non-augmented affix is inserted. Thus, the
rule is applicable only in the derivation of weak and geminate verbs. There are two
outcomes of such a rule, both are negative. First, the rule is ad-hoc and serves no pur-
pose in the grammar of the language beyond the input specified in the rule. Second,
it treats the sound and geminate/weak verbal classes as two distinct systems. Alter-
natively, the choice of an allomorph may be determined by characteristics internal to
the stem. Recall that the two constraints governing verb stems are foot-binarity and
final-consonantality. Verbs which select the augmented -éeC allomorphs are Gemi-
nate stems, which are monosyllabic but consonant-final, and weak stems which are
conversely disyllabic but vowel-final. Thus, neither stem size nor right-edge align-
ment is responsible for the selection of the affix allomorph.
I propose that an OT approach recognizing both intra- and interparadigmatic corre-
spondence is superior to earlier derivational or classic OT approaches. I will demon-
strate that affix augmentation in the perfective verb paradigms is governed by the
prosodic well-formedness of the output on the one hand and the uniformity of the in-
flectional paradigm on the other. The OT-oriented proposal derives affix allomorphy
through the interaction of markedness constraints which shape the optimal form of
the output with identity constraints governing members of inflectional subparadigms
(intraparadigmatic identity) as well as identity across paradigms (interparadigmatic
identity). The unified account to be developed in this section applies to all verb stems
regardless of size or quality. I will first give a brief overview of paradigm-oriented
OT in Sect. 3.1, while the allomorphy problem is discussed in Sect. 3.2.

3.1 Paradigms and paradigmatic identity

In addition to input-output and output-output correspondence, the effect of paradig-


matic correspondence or identity relations among members of the same or similar
paradigms has been recognized in recent works (Kenstowicz 1996, 2005; McCarthy
2002; Gafos 2003). Members within similar paradigms may influence the phonolog-
ical structure of one another sometimes creating exceptions to an otherwise regular
morphological process (Downing et al. 2005 and articles therein). Recognizing the
role of paradigm leveling or analogy as a driving force explaining extensions of a
70 S. Farwaneh

pattern or exceptions to a prescribed rule resurrects the methodological notion of


‘analogy’ or qiyas used extensively in the descriptive Arabic grammars of Medieval
grammarians and linguists
For the purposes of this paper, which focuses on inflection not derivation, a
paradigm is defined as in McCarthy (2002): “A set of inflected forms based on a
common lexeme or stem”. We assume further following Gafos (2003) and as cited in
McCarthy (2002 footnote 3) that inflectional paradigms have internal structure.11 The
perfective verb paradigm in Arabic draws a distinction between the 3rd and non-3rd
subparadigms, also observed in Broselow (2008).12 This third vs. non-third person
distinction is forced by the structure of person affixes, where non-third person affixes
are consonant-initial, while third person affixes are null or vowel-initial.13
Having defined the notion ‘paradigm’, we now address the issue of paradigmatic
identity and how it is assessed. Within different models of OT, candidates to be eval-
uated by a constraint hierarchy are fully-specified output forms, but in an Optimal
Paradigm (OP) model (McCarthy 2002), each candidate subject to evaluation is an
entire paradigm or possible paradigm. Every member in a paradigm stands in a cor-
respondence relation with other members in the same paradigm and the relationship
is symmetric giving no precedence to one member over others (McCarthy 2002).
In a third person paradigm, for example, third masculine singular inflection must
bear segmental and suprasegmental resemblance to the third person feminine singu-
lar and plural forms. Intra-paradigmatic output-output faithfulness constraints guard
against any changes within the paradigm, unless mandated by higher-ranked marked-
ness constraints. The optimal paradigm among the possible paradigm candidates is
the one which incurs the fewest violations, i.e., shows fewer or zero alternations.
In addition to intra-paradigmatic identity which recognizes identity relations be-
tween members of the same paradigm, the present proposal recognizes a non-local
type of identity relation holding across related paradigms as well. As I show below,
there are types of resemblance that cannot be explained locally without reference to
a more global paradigmatic correspondence. An example of such resemblance which
requires looking beyond the local paradigm is the pluralization of the word [mudiir]
‘director’ in Levantine dialects to [mudaraaP]. The form has crept into the news me-
dia and received acceptance as a Modern Standard Arabic derivative despite its vi-
olation of the active participle plural formation rule. The prescriptive rule stipulates
that active participles formed by prefixing /mu-/ may not take broken plural forma-
tion, but must be pluralized by suffixation of /-iin, -uun/. Nothing in the inflectional

11 Gafos and Ralli (2001) have similarly argued for an internal structure of inflectional paradigms based
on evidence from two Greek dialects of Lesbos where intra-paradigmatic identity constraints hold along
the morphosyntactic dimensions of Person and Number.
12 Arabic textbooks have recently recognized the pedagogical validity of teaching third and non-third in-
flections as separate subsystems (Brustad et al. 1997), a striking contrast from the pre-90s textbooks which
adopted the animacy hierarchy of the Germanic/Romance languages (1sg/2sg/3sg/1pl/2pl/3pl) thus intro-
ducing a non-uniform paradigm difficult to internalize.
13 McCarthy’s assumption of a flat paradigm makes sense for CA and MSA where the 3p feminine plu-
ral affix /-na/ is the only consonant-initial 3p affix and thus patterns more with non-third person affixes.
Compare [the prosodic structure and stress placement in [katábna] ‘they-f wrote’ which patterns with
[katábtu] ‘I wrote’ and not [kátabat] ‘she wrote’. Neutralization of plural affixes in many dialects in favor
of the masculine marker /-u/ renders the third person paradigm more symmetric, hence a person paradigm
decomposition is justified in regional and not Literary Arabic.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 71

paradigm of [mudiir] or even in the active participial paradigm in general can ex-
plain this anomalous plural. Yet it is evident that the analogue of [mudaraaP] is the
plural form [wuzaraaP] ‘ministers’ (sg. waziir) and [PudabaaP] ‘writers’ (sg. Padiib).
Speakers somehow perceived that [mudiir] and [waziir] belong to the same paradigm
and thus are pluralized analogously despite the distinct first vowel, /a/ in [waziir]
vs. /u/ in [mudiir], which indicates that they belong to different lexical categories.
Yet the semantic resemblance, both indicating a high-status profession, and prosodic
resemblance, both being iambic feet, justified the paradigmatic relation. Segmental
identity is then not a requirement in inter-paradigmatic relation, but suprasegmental
and prosodic identity are. The four layers of identity relations are exemplified in (18).
(18) Levels of Faithfulness (Identity relations)
IO-Faith: monitors identity relationships between input (stem) and output.
OO-Faith: Monitors identity between two morphologically-related outputs.
OP-Faith: monitors identity relationships among members within an inflec-
tional paradigm intraparadigmatic correspondence).
PP-Faith: monitors identity relationships across related paradigms (inter-
paradigmatic or paradigm-to-paradigm correspondence).
I use the abbreviation PP-Faith to distinguish inter-paradigmatic correspondence
from OP-Faith which entails correspondence within optimal paradigms.

3.2 Resolving the affix allomorphy problem

Having introduced the partial constraint hierarchy in (15) and the different identity
relations in (18), and with Markedness constraints introduced as needed, we now
examine the inflectional paradigms of various verb types.

3.2.1 Analysis of geminate verbs

Geminate alternation is observed in Modern Standard Arabic, but not in the Regional
Arabic dialects, where the geminate is retained throughout the paradigm as shown in
(19) using the stem [madd] ‘stretch’ as an example.
(19) Geminate Verbs
MSA RA
1/2P 3P 1/2P 3P
madádtu mádda maddéet mádd
madádta máddat maddéet máddat
madádna máddu maddéena máddu
Geminate verbs in Literary Arabic show surface alternation between a strong disyl-
labic CVCi VCi allomorph preconsonantally, and a monosyllabic geminate CVCi Ci
prevocalically. Within derivational approaches, the strong variant is assumed to be the
underlying form from which the geminate output is derived by syncope (Brame 1970;
McCarthy 1979 and others). The Arabic dialects I have examined have dispensed with
such alternation in favor of the geminate monosyllabic form which appears through-
out the inflectional paradigms of both the perfect and the imperfect. Given the ab-
sence of a surface strong allomorph such as *[madad-] and *[samam] which we have
72 S. Farwaneh

attributed to the dominance of the Integrity constraint, there is no reason to postulate


it as the input of derivation. I assume, then, following Gafos (2003) in his analysis
of Literary Arabic, that the monosyllabic form CVCi Ci is the base for the geminate
verb.
Beginning the discussion with third person vocalic affixes using the stem [madd]
for illustration, affixing vocalic suffixes to the stem [madd], result in an output faith-
ful to its base: /madd-at/ > [maddat], /madd-u/ > [maddu]. In contrast, consonantal
affixes added to the stem result in outputs violating one or more markedness and/or
faithfulness constraints. A geminate stem like [madd] when inflected with a conso-
nantal suffix may yield the following possible outputs: *[máddt] a faithful candidate
with gemination retained, *[mádt] with degemination, *[madádt] with geminate split,
or *[máddit] with epenthesis to rectify the consonant cluster; all of which are unat-
tested.
The first candidate *[maddt] violates a markedness constraint motivated by the
distribution of geminates. Geminates show a pattern of distribution akin to that of
consonant clusters: both geminates and clusters may occur word-finally, intervocal-
ically, or before a syllable; both are restricted before extrasyllabic consonants. The
distribution in (20) is based on the stems [madd] ‘stretch’ and [katab] ‘write’.

(20) Distribution of Geminates and Clusters in Levantine


Position Geminates Gloss Clusters Gloss
CC# madd# ‘3m.sg.’ katabt# ‘1sg.’
VCCV maddu ‘3pl’ katabtu ‘2pl’
VCC.CV madd.ha ‘loved-3m-3f’ ka.tabt.ha ‘wrote-1sg-3f’
VCC.C# *mad.t# ‘1sg’ *ka.tabt.š# ‘1-neg’
The similar distribution of and restriction on geminates and clusters suggests an upper
bound on the size of a mora limiting it to two segments. Trisegmental moras of the
structure *VCC]μ or *CCC]μ are banned by the constraint in (21).

(21) Constraint on Moras


*xxx]μ -trisegmental moras are prohibited.

The constraint against triple-branching moras in (21) rules out the first candidate
*maddt]. The second and non-optimal form *[madt] with degemination can be elim-
inated by a constraint on input-output length faithfulness proposed by Davis and Za-
waydeh (1999) for their treatment of hypocoristics. The constraint formulated in (22)
requires that length in the input is preserved in the output.

(22) Length Faithfulness


IO-Max-μ: Moraic structure of the input is preserved in the output; no short-
ening or degemination.

Max-μ is subordinate to *xxx]μ (21) to capture the generalization that in hollow


verbs long vowels shorten persistently before tautosyllabic clusters. Thus, when a
hollow verb like [xtaar] ‘choose’ is inflected with the first person marker /-t/, the
vowel shortens; e.g., /xtaar-t/ > *[xtaart] > [xtart]. Thus, length faithfulness is sacri-
ficed to satisfy the ban on trisegmental moras. The following tableau shows the effect
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 73

of Max-μ and *xxx]μ interaction in selecting the correct form of inflected hollow
verbs.

(23) Inflection of hollow verbs


*xxx]μ » Max-μ
Candidate *xxx]μ Max-μ
a. xtaart *!
b. $  xtart *
On the other hand, *xxx]μ is subordinate to Max-C, since coda deletion is never em-
ployed to rectify triple mora violation. Thus, the following subhierarchy is supported:
Max-C » *xxx]μ » Max-v.
As for the third candidate with geminate split *[madadt], it is rejected by Max-μ;
because spreading compromises input length as well as the integrity of the CVCC
stem. The fourth epenthetic output *[maddit] seems to fare well on all candidates
and is expected to incorrectly emerge as the winning candidate as the comparison
tableau in (24) illustrates. The tableau shows the relevant constraints only, but the
ranked constraints appear above the tableau to help the reader follow the constraint
hierarchy as it develops. The broken vertical lines in the hierarchy indicate lack of
interaction as in the case of *xxx]μ and Max-μ with respect to Dep-C. Since Dep-C
involves a non-moraic consonantal onset, while the constraint on branching moras or
mora deletion apply to codas, no conflict between Dep-C and the two constraints on
moras is expected to arise.

(24) Geminate Stems with Consonantal Affixes Integrity » Max-C » *xxx]μ »


Max-μ | Dep-C » Max-V » Dep-V
Candidate Integrity *xxx]μ IO-Max-μ IO-Dep-V
a. maddt *!
b. madt *
c. madadt *! *
d.  madd(i)t *
To explain the failure of a seemingly optimal form like [èáddit] requires departure
from the familiar input or base-output relationship into a global view of verbs as stems
in paradigms (Gafos 2003; McCarthy 2002) and as the proposed layout of identity re-
lations in (18). The falsely optimal candidate *[maddit] which incurs an IO-Dep-V
violation yields a mismatched non-third paradigm like *{maddit, maddti, maddtu,
maddna, . . . 14 } where epenthesis is needed before consonantal but not syllabic af-
fixes, thus allowing intraparadigmatic vowel/zero alternation in the paradigm. The
OP constraint OP-Dep-V proposed in McCarthy (2002) to account for verb inflection
in Classical Arabic is invoked to resist hybrid paradigms where epenthetic vowel oc-
curs in one member but does not carry over to other members in the paradigm. The
constraint is given in (25).

(25) OP-Dep-V: No vowel/zero alternation is allowed in the same paradigm.

14 Left and right braces are a notational convention used to mark members of same paradigm.
74 S. Farwaneh

There is a crucial conflict between OP-Dep-V which requires uniform vowel inser-
tion within the paradigm and Dep-V which rejects insertion. With OP-Dep-V ranked
over IO-Dep-V, paradigm uniformity takes precedence over faithfulness to the in-
put, and the uniformly augmented paradigm {maddéet, maddéeti, . . . } emerges at
the expense of the IO-faithful but OP-nonfaithful paradigm {mádd(i)t, maddti, . . . }.
McCarthy’s analysis of Classical Arabic inflection demonstrates the inviolability of
OP-Dep-V, as vowel-zero alternations never occur in the inflectional paradigms of
CA. However, McCarthy states that “through ranking permutation, there are lan-
guages where these same constraints (OP-Dep-V) are crucially dominated and not
visibly active” (McCarthy 2002) citing modern Arabic dialects as example. I believe
the implicit reference is to the alternation seen in the sound verb inflection in some
varieties of Spoken Arabic where epenthesis sometimes breaks up final clusters, e.g.,
{katábit, katábti, katábtu, . . . }. I maintain that the ranking of OP-Faith and IO-Faith
is the same in Classical Arabic and Levantine, and that the vowel/zero alternations
observed in the inflection of sound verbs like [katabit] in some dialects (e.g., Iraqi) is
forced by higher-ranked sonority-based markedness constraints.
More important, the unattested epenthetic form *[máddit] further creates an asym-
metric stress pattern within inflectional paradigms. The stress rule of Palestinian Ara-
bic is stated roughly as follows: Stress the last heavy syllable in the word, if none
exists, stress the initial one. Consonantal affixes of the 1st and 2nd person inflection
create ultimate or penultimate syllables targeted by stress; vocalic affixes of the 3rd
person paradigm, on the other hand, yields series of open syllable forcing stress to
fall on the initial syllable, as the [katab] paradigm repeated here shows:

(26) Inflection of [katab] and stress placement


1P 2P 3P
-F -PL katábt katábt kátab
+F -PL katábt katábti kátabat
+PL katábna katábtu kátabu
In the epenthetic form *[maddit], epenthesis creates a final light syllable, forcing
stress to fall on the initial stem syllable, resulting in a non-uniform stress pattern
across the disyllabic first/second conjugation. Disyllabic sound verbs receive final
stress [katábt] while disyllabic geminate verbs display initial stress *[máddit]. By
augmenting consonantal affixes uniformly with a long /ee/ extension, all paradigmatic
mismatches discussed earlier are regularized: the same extension vowel appears in
all inflected forms of the stem in satisfaction to OP-Dep-V, and stress lands on the
penultimate syllable on a par with other first/second subparadigms as the comparing
table illustrates:

(27)
[katab] [madd]
1 -PL katábt maddéet
2 -F -PL katábt maddéet
2 +F -PL katábti maddéeti
2 +PL katábtu maddéetu
1 +PL katábna maddéena
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 75

The success of the augmented paradigm is attributed to an inter-paradigmatic con-


straint enforcing stress uniformity across paradigms of the same inflectional category;
that is, second person inflection for example must be metrically similar regardless of
the base. The constraint is proposed in (28).

(28) Interparadigmatic Constraint


PP-IdentAccent: Stress position is invariant across similar paradigms.

IdentAccent is a constraint on paradigms rather than individual output. Furthermore,


it assesses similarities across paradigms rather than within a paradigm. That is, all
first person conjugations of sound and defective verbs are evaluated for stress uni-
formity. IdentAccent assesses each possible paradigm of a defective verb against that
of a sound verb for metrical similarity. In (29), PP-IdentAccent compares the two
possible paradigms under examination {máddit, máddti, máddtu, . . .} with epenthe-
sis applying as needed and the second paradigm {maddéet, maddéeti, maddéetu,
. . .] with consistent augmentation against the invariable paradigm {katábt, katábti,
katábtu, . . . }. By appending a long vowel to consonantal affixes, stress lands on the
final or penultimate syllable on par with other first/second subparadigms in satisfac-
tion of PP-IdentAccent.
(29) Paradigm comparison
Paradigm IdentAccent
a. {katábt, katábti, katábtu, . . . } & {máddit, máddti, máddtu, . . . } *!
b. $  {katábt, katábti, katábtu, . . . } & {maddéet, maddéeti, maddéetu, . . . }

The function of the augmenting extension /-ee-/ is, then, to ensure a uniform stress
placement and avoid vowel/zero alternation across all same person paradigm in the
inflection of geminate verbs. The following section examines the regularizing role of
the vocalic extension in the inflection of weak verbs.

3.2.2 Weak verbs

Weak verb stems are those ending in a vowel (traditionally analyzed as derivatives of
a glide final root). Appending consonantal affixes faithfully to the stem [rama], used
to exemplify the weak verb type, yields outputs with initial stress: *{rámat, rámati,
rámatu, . . . }. Once again, IDENTACCENT would reject these forms despite their di-
syllabicity and faithfulness to their base in favor of the augmented paradigm }raméet,
raméeti, raméetu, . . . } to parallel the sound paradigm }katábt, katábti, katábtu, . . . }.

(30) Weak verb paradigm rama + Consonantal Affixes


Paradigm Comments
a. *}rámat, rámati, . . . } Stress falls on initial instead of second
vs. {katábt, katábti, . . . } syllable; violates PP-Ident-Accent
b. $  {raméet, raméeti, . . . } Stress uniform but missing final stem vowel;
vs. {katábt, katábti, . . . } violate Max-V.
76 S. Farwaneh

As stated earlier and illustrated by the hierarchy in (15), Max-V is low on the con-
straint hierarchy because stem vowels are more susceptible to deletion under the pres-
sure of markedness constraints than stem consonants. Therefore, a conflict between
PP-IdentAccent and Max-V is resolved in favor of paradigm uniformity over faith-
fulness to the input. The tableau in (31) shows the evaluation of the augmented but
uniform paradigm versus the faithful but non-uniform one.
(31) Candidate PP-IdentAccent IO-Max-V
{rámat, rámati, . . . } *!
$  {raméet, raméeti, . . . } *
Now we shall consider the consequences of appending vocalic affixes to the stem of
weak verbs which yields a number of possibilities, four of which will be examined
shown below in (32).
(32) rama + u
a..ra.ma.u. Faithful with onsetless syllable
b..ra.mau. Faithful with vowel hiatus
c. ra.ma.Cu. Consonantal epenthesis
d. ra.mu. Vowel deletion (melodic overwriting)
The worst candidates are the ones most faithful to their input, candidate (32a) and
(32b). The vowel hiatus resulting from appending a vocalic affix to a vowel-final
stem, if heterosyllabic as in candidate (32a), runs afoul of ONSET, whose undom-
inated status has been demonstrated repeatedly; if the two vowels are tautosyllabic,
as in candidate (32b), then the resulting output violates *VV, which bans hiatus (pro-
posed in Davis and Zawaydeh (1999)). Two possible resolutions of vowel hiatus can
be considered; epenthesizing a consonantal onset, or deleting the stem-final vowel.
Epenthetic onsets are banned except in word-initial position, giving Dep-C a domi-
nant position over *VV. The winning candidate D [ramu] satisfies all size-controlling
constraints; its only problem is the loss of the second stem vowel; a minor violation
of the constraint Max-V.15 Tableau (33) shows the outcome of the evaluation.
(33) Candidate Evaluation for /rama-u/
Candidate Onset Dep-C *VV Max-V
a. ra.ma.u. *!
b. ra.mau. *!
c. ra.ma.Cu. *!
d. $  .ra.mu. *
The same evaluation result may be achieved if vowel hiatus is resolved by deleting
the affixal vowel, /rama-u > *[rama]. But as stated in Ussishkin (1999), Affix-Faith
outranks Root-Faith, contrary to McCarthy and Prince’s (1995) Root±afix faithful-
ness metaconstraint. The precedence of Affix-Faith over Root-Faith preserves the
morphological information embodied in the affix; on the other hand, semantic infor-

15 The low effect of MAX-V is also observed in Standard Arabic, where vowel-final imperfective forms
lose the final vowel before a vocalic affix; e.g., /ya-rmii-uun/ > [yarmuun] ‘they throw’.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 77

mation of the stem is still recoverable despite the deletion of a stem vowel. If hiatus
can be avoided without sacrificing the stem vowel, that is, without violating Max-V,
the stem vowel is preserved. This is the case with perfective weak verb variants with
high front vowel16 where the plural affix and the ensuing hiatus forces hardening of
the stem vowel into a glide onset.

(34) Weak Verbs with High Vowels


3-SG 3-PL Gloss
rimi rimyu *rimu ‘threw’
liPi liPyu *liPu ‘found’
riDi riDyu *riDu ‘agreed’
biki bikyu *biku ‘cried’
Following the pattern in (33), the tableau in (35) evaluates the possible outputs from
the input /rimi-u/ but with the addition of the fifth candidate [rimyu].

(35) Candidate evaluation for /rimi-u/


Candidate Onset Dep-C *VV Max-V
a..ri.mi.u. *!
b..ri.miu. *!
c..ri.mi.Cu. *!
d..ri.mu. *
e.  .rim.yu
While violation of Max-V in (33) did not prevent the output [ramu] from emerging as
the winning candidate, violation of the same constraint rules out the parallel candidate
[rimu] since another candidate [rimyu] fares better on Max-V. Thus, violation of the
same constraint may eliminate a surface form while admitting another, depending on
the quality of outputs available for evaluation.

3.3 Allomorphs of the augmented affix

One of the advantages of Optimal Paradigm Theory within OT is that it can ex-
plain, without recourse to special constraints or strategies, the different outputs de-
rived from different allomorphs of the base. The earlier discussion focused on the
third plural conjugation of the two base variants, [rama] and [rimi], showing that
the parallel forms [ramu] and *[rimu] are not equally optimal despite their struc-
tural overlap. The blocking of *[rimu] follow straightforwardly without recourse
to different rules or rule ordering statements. This leads us to test the analysis
against the different surface manifestations of the first conjugation of the Ci Ci and
CaCa variants. The table in (36) compares the 1/2 conjugations of CaCa and Ci Ci
verbs.
16 The low and high vowel variants (baka/biki) are used interchangeably within the same dialects but the
high vowel variant is more prominent in Lebanese, Syrian and northern Palestinian.
78 S. Farwaneh

(36) CaCa Ci Ci Gloss


raméet rmíit ‘threw’
laPéet lPíit ‘found’
bakéet bkíit ‘cried’
maDéet mDíit ‘signed’
Paréet Príit ‘read’
There are two noticeable differences between the two variants [rameet] and [rmiit]:
The first difference involves the distribution of the two affix allomorphs /-eet/ and
/iit/, and the second is the presence versus absence of the initial syllable vowel. The
distribution of the affix allomorph is determined by the height of the base vowel: A
base with a low vowel CaCa selects /-eet/ with a non-high vowel while a base with
a high vowel Ci Ci selects /-iit/. We propose that affix selection is dictated by an IO-
faithfulness constraint requiring input and output vowels to be of the same height,
formalized in (37).

(37) IO-Ident-High: input and output vowels must not alternate in height.

This constraint is not specific to inflected weak verbs, but regulates allomorph selec-
tion elsewhere in the language. The third person feminine singular affix, for example,
is realized as /-at/ or /-it/ depending on the height of the stem vowel; [katabat] versus
[nizlit] ‘she descended’.
The second difference, the vowel/zero alternation observed in the first syllable
of the two weak stem variants, is attributed to a markedness constraint against the
realization of weak nuclei given in (38).

(38) *Ci]w

The constraint against weak syllables in (38) rejects all metrically weak open syl-
lables with high nuclei, unless blocked by high-ranking markedness constraints. We
observe the effect of the weak syllable constraint in many Arabic dialects, where
medial open syllables with unstressed /i/ are reduced by syncope; e.g., /nizil-u/ > [ni-
zlu] ‘they descended’. Some dialects characterized as Onset dialects (Broselow 1992)
restrict the application of syncope to medial position but blocks it where it would oth-
erwise create a consonant cluster. To capture the underapplication of syncope in these
dialects, a markedness constraint banning clusters such as *Complex must dominate
*Ci]w . In Levantine, the restriction on weak nuclei is extended to initial syllables
creating a word-initial cluster, /nizil-t/ > [nzilt] ‘I descended’ indicating a reverse
constraint order *Ci]w > *Complex. The following tableau provides an evaluation of
the possible output forms from/rimi-VVt/.

(39) /rimi-VVt/
Candidate IO-Ident-High *Ci]w *Complex
a. rimeet *! *!
b. rmeet *! *
c. rimiit *!
d.  rmiit *
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 79

Candidates (39a) and (39b) are rejected for their violation of the IO-faith constraint
that requires height identity among input and output vowels. Candidate (b) incurs an
additional violation of *Ci]w which bans high vowels in open syllables. Candidate
(39c) fares a little better but it is still ruled out because of the retention of the ini-
tial weak syllable, leaving the fourth candidate (39d) as the optimal output despite a
minor violation of the cluster-blocking constraint *Complex, not a very active con-
straint in many Levantine dialects. Appending the basic unaugmented affix /-t/ to a
Ci Ci stem follows the same pattern of evaluation shown with respect to CaCa stem;
the reader may refer to the discussion at the beginning of this section and tableaus
(30) and (31). This concludes the discussion of weak verbs and their inflection. Be-
fore dealing with the final type, hollow verbs, the following section gives an overall
summary with a list of active constraints and their effect.

3.4 Summary

Non-sound verb stems differ structurally from sound stems, being either monosyl-
labic consonant-final or disyllabic vowel-final, compared to the disyllabic consonant-
final shape of sound stems. This structural difference has numerous inflectional con-
sequences, most important of which is the utilization of three allomorphs of the non-
third person affix: the default consonant-initial affix C(V) appearing consistently in
sound paradigms, and two augmented allomorphs eeC(V) and iiC(V). Explaining
these complex alternations require access to multiple levels of faithfulness. An in-
flected output establishes multiple correspondence or identity relations with a num-
ber of targets; its input and other members in the same paradigm. The paradigm itself
bears an identity relation with morphologically-related paradigms that do not neces-
sarily share the same lexeme but share a morphological relation, as in second person,
gender, number, etc. The following list contains the constraints identified so far and
their functional categories.
(40) Markedness constraints optimizing the structure of the output
• Onset: Each syllable must have a consonantal onset.
• Ft-Bin: A stem is equal to a binary branching foot (two syllables or two
moras).
• Final-C: Stems must end in a consonant.
*xxx]μ : Moras must not contain three segments.
• *VV: No vowel hiatus.
• *Ci]w : No high unstressed open syllables.
• *Complex: No consonant clusters.
IO-Faithfulness constraints: monitor input-output correspondence.
• Max-C: Every consonant in the input must have a correspondent in the
output; no deletion.
• Dep-C: Every consonant in the output must have a correspondent in the
input; no epenthesis.
• Max-V: Every vowel in the input must have a correspondent in the output;
no deletion.
• Dep-V: Every vowel in the output must have a correspondent in the input;
no epenthesis.
80 S. Farwaneh

• Integrity: Every input segment must have one correspondent in the output;
no copying.
• Max-μ: Length in the input must be preserved in the output; no shortening
or lengthening.
• IDENT-HIGH: Input and output vowels may not alternate in height; no
raising or lowering.
OP-Faithfulness constraints: monitor identity among members of the same
paradigm.
• OP-Dep-V: No vowel/zero alternation in a paradigm; if epenthesis applies
in one member, it must carry over to other members.
PP-Faithfulness constraints: monitor similarity across morphologically-
related paradigms
• PP-IDENTACCENT: stress must be uniform across related paradigms; no
stress alternation across paradigms.
The ranking of the constraints presented in (40) is given in (41).
(41) Onset » Integrity » Ft-Bin » Max-C » *xxx]μ » Max-μ | Dep-C | PP-
IDENTACCENT » *VV » IDENT-HIGH » *Ci]w » Max-V » Final-C | OP-
Dep-V » Dep-V » *Complex
Ass stated earlier, the broken bar indicates non-interaction, usually between con-
straints targeting different domains. For instance, Dep-C which involves non-moraic
onsets does not conflict with constraints on moras. Also, Dep-c does not interact with
IdentAccent which monitors stress uniformity.

4 Hollow verbs: The role of alignment

The fourth verb type to be investigated here, that of the so-called hollow verbs,
present an interesting challenge to the analysis developed so far. In traditional anal-
yses, hollow verbs are derived from glide-medial roots. They are dubbed “hollow”
because the medial glide /w/ or /y/ manifests itself as a long vowel in the output
form; thus, in a stem-based analysis, a hollow verb stem consists of a closed mono-
syllable with a long peak. Hollow verbs exhibit a more interesting pattern that does
not parallel that of weak and geminate verbs. The augmented allomorph expected be-
fore consonantal affixes is substituted instead by an unexpectedly reduced allomorph.
The basic allomorph appears before vocalic affixes as usual, as the paradigm in (42).
(42) Hollow Verb Paradigm
1P 2P 3P
-F -PL zurt zurt zaar
+F -PL zurt zurti zaarat
+Pl zurna zurtu zaaru
Assuming that the third person singular form is the base stem, the input to inflectional
derivation is then [zaar]. Consider first the adjunction of vocalic affixes to the bare
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 81

stem. This yields the third person paradigm {zaar, zaarat, zaaru}, all members of
which satisfy all constraints.
Augmenting the stem [zaar] with consonantal affixes yields a number of possibil-
ities among which are a faithful [zaart], a shortened form [zart], and an augmented
form [zaareet], evaluated in the tableau (43).

(43) Candidate *XXX]μ Max-μ Dep-V *Complex


a. zaart *! *
b. zart * *
c. $  zaareet *
Candidate (43a) *[zaart] is ruled out by the stringent ban on trisegmental moras, while
candidate (43b) appears to violate Max-μ since input vocalic length has not been re-
tained; both candidates incur a mild violation of *Complex as well. The unattested
candidate (43c) fares well on all constraints, except Dep-V. How can we explain the
optimal yet unattested *[zaareet]? Such forms are not only foot-binary, they are also
syllabically well-formed in terms of the presence of syllable onsets and the absence
of complex edges, and are faithful to the structure and quantity of the input. Addi-
tionally, augmenting the output with the long /ee/ in *[zaaréet] satisfies the paradigm
uniformity IdentAccent as it places stress on the final syllable on par with [katábt]
and [èabbéet].
The answer lies in the notion of edge alignment McCarthy and Prince (1993b) later
reintroduced as Anchor which governs the alignment of elements defined in (44).

(44) Anchor (McCarthy and Prince 1995)


Any element at the designated periphery of the input or base has a corre-
spondent at the designated periphery of the output.

In Farwaneh (2007), a more specific interpretation of Anchor referencing syllabic


position of edged segments was introduced to account for the invariant onset position
in partial reduplication, e.g., [faraè] > [farfaè] and not *[fafraè]. The constraint is
given below in (45).

(45) Relativized Anchor


Anchor-Onset (IO): All output correspondents of the left edge of the input
must occupy onset position in the output.
Anchor-Coda (IO): All output correspondents of the right edge of the input
must occupy coda position in the output.

The preservation of the syllabic position of edged segments required by Anchor17 in


(45) rules out an output like *[zaareet] since the final stem consonant [zaar] occu-
pies a coda position but shifts to an onset position after augmentation of the affix.
The Anchor constraint in (45) stipulates that the right edge segment occupying coda
position be situated at the right edge of the syllable, i.e., the stem-final consonant

17 A similar relativized alignment constraint is introduced in Kenstowicz (2005) to account for the distri-
bution of diminutive affix allomorphs /-it/ and /-cit/ in Spanish, where preserving the syllabic position of
base segments determine the choice of the affix allomorph, [casa] > [casita] vs. [limon] > [limoncito].
82 S. Farwaneh

must also be syllable-final. Thus, stems with augmented affixes where the vowel /ee/
follows the stem-final consonant, as in *[zaareet] would constitute an Anchor viola-
tion.
To determine the ranking of Anchor relative to other constraints, we survey
all measures of hollow verbs to find that the medial vowel always shortens be-
fore consonant-initial affixes; /xtaar-t/ > [xtart] ‘I chose’, /stafaad-t/ > [stafatt] ‘I
benefited’.18 This persistent shortening is the response to the constraint on triple-
branching moras *XXX]μ proposed in (21) which can also be satisfied by appending
an augmented affix *[xtaareet] which satisfies Max-μ at the expense of Anchor. This
preference suggests that Anchor outranks Max-μ. With this ordering, tableau (46)
reevaluates the same candidates considered earlier in (43).

(46) zaar + t
Candidate *XXX]μ Anchor-Coda Max-μ *Complex
a. zaart *! *
b. $  zart * *
c. zaareet *
Again, the hierarchy established so far continues to incorrectly select the unattested
form *[zart] which appears to possess the correct shape but wrong vowel quality since
the attested allomorph is [zurt] as shown in (42) above. The reduced allomorph [zur]
emerges consequence to the conflict between IO and OO correspondence constraints
forcing perfective outputs to be identical to their inputs on the one hand and to parallel
the corresponding imperfective outputs on the other (Gafos 2003). (47) illustrates the
segmental correspondence between perfective and imperfective forms of hollow verb
Measure I.
(47) Segmental Correspondence: Perfective & Imperfective
Perfective Imperfective Gloss
zurt Pazuur ‘visit’
jibt Pajiib ‘bring’
nimt (*namt) Panaam ‘sleep’
To account for the similar vowel quality in the perfective and imperfective counter-
parts we propose the following output-output faithfulness constraint.

(48) OO-Ident-V
Vowel quality is preserved across morphologically-related paradigms.

When the imperfective vowel is high, OO-Ident-V is invoked triggering a similar


vowel in the perfective form. However, when the imperfective vowel is low, seg-
mental identity is blocked and the default high front vowel emerges as in the per-

18 Kathem Al-Saher, the renowned Iraqi singer and composer, functions as one of the jurors on the two
popular programs The Voice and The Voice Kids. Observing his utterances shows many instances of aug-
mented hollow verb conjugation, e.g., [xtaareet] ‘I selected’ and [stafaadeet] ‘I benefited’ (cf. PA [xtart]
and [stafatt]). This may indicate that in some Iraqi dialects, Anchor is subordinate to MAX-μ, thus al-
lowing augmentation to apply to all nonsound verbs without exception. More research on nonsound verb
paradigms in various Arabic dialects is needed for a full typological analysis of Arabic verb morphology.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 83

fect/imperfect alternation /nim- -naam/ ‘sleep’ and /xif- -xaaf/ ‘be afraid’. By virtue
of OO-Ident-V, The imperfective form /-zuur/ imposes its segmental identity on the
perfective allomorph selecting [zurt] over *[zart]. The competition between the two
outputs [zurt] and *[zart] exemplifying the conflict between OO-Faith and IO-Faith is
resolved to the benefit of OO-Faith indicating its dominance over IO-Faith. Tableau
(49) shows a reevaluation of the outputs from /zaar-t/: the faithful [zaart] the segmen-
tally faithful but suprasegmentally unfaithful [zart], the augmented [zaareet] and the
reduced and OO-faithful [zurt].
(49) /zaar-t/
Candidate *XXX]μ Anchor OO-Ident-V IO-Ident-V
a. zaart *! *
b. zart *
c. zaareet *! *
d. $  zurt *
It seems superficially that the alignment analysis runs counter to the analysis pre-
sented in Sect. 3 where the augmented allomorphs of the non-third affix are selected
by weak and geminate verbs for stress uniformity purposes. We expect the Anchor
constraint to rule out forms like [rameet] and [èabbeet], but these forms somehow
escape Anchor. Weak, geminate and hollow stems have one thing in common: their
surface representations contain two consonants only. What sets them apart is that in
the case of weak stems, both stem consonants are syllable onsets in the input and
retain their syllabic position in the augmented outputs, e.g., [ra.ma] and [ra.meet]
therefore no Anchor violation ensues. Geminate stems likewise may safely be aug-
mented since the final stem segment surfaces as a geminate which violates alignment
requirements anyway given the positional duality of geminates. A geminate conso-
nant is an underlyingly moraic segment (Hayes 1989) which surfaces as a coda and
onset of two successive syllables illustrated in the representation in (50).
(50) Geminates
Input Output
σ
/
μ μ/
| |/
C C
A geminate occupies two edges, the right edge of one syllable and the left edge of the
following one and therefore a geminate is not uniquely anchored and would violate
any alignment constraint.
Returning to the hollow verb paradigm, the Anchor analysis is challenged by the
3P feminine and 3P plural forms [zaarat] and [zaaru] where the stem consonant is syl-
lable final in the input but syllable initial in the output. Anchor violation is inevitable
in this case because preserving the syllabic position of the stem-final consonant leads
to either onset violation *[zaar.at] or Dep-C violation *[zaar.Cat]. Stem allomorphy
in hollow verbs presents yet another question concerning the selection of stem al-
lomorphs. I have argued that the reduced output [zur] is the optimal candidate for
84 S. Farwaneh

consonantal affixes to append to, given its proper alignment and its faithfulness to the
imperfect allomorph. One may envisage a paradigm where the reduced allomorph
also combines with vocalic affixes yielding such forms as *[zur-at] and [zur-u].19
Since OO-Ident-V outranks IO-Ident-V, *[zurat] is expected to win over [zaarat] as
the tableau (51) shows.
(51) /zaar-at/
Candidate Onset Dep-C Anchor OO-Ident-V IO-Ident-V
a. zaarat * *
b. zurat * *
Both candidates (51a) and (51b) satisfy Onset and Dep-C while violating Anchor. The
Anchor violation stems from the misalignment of a final stem segment occupying
coda position in the input as a syllable onset in the output. The constraint next in
the hierarchy OO-Ident-V decides the competition in favor of [zurat] over [zaarat].
The reduced outputs *[zurat] and *[zuru] should also be favored for their small size
relative to [zaarat] and [zaaru]. Kager (1996) proposes a universal constraint favoring
the phonologically minimal shortest morpheme to mark a morphological category.
Hargus (1997) expands upon this notion and proposes the principle Brevity given
below:
(52) Brevity
The phonologically shortest allomorph is preferred.
The two principles dictate how allomorphs should be ranked against each other, all
things being equal. That is, if EVAL fails to discern the winning candidate solely on
the basis of a given constraint hierarchy, it may resort to an allomorph hierarchy fa-
voring some allomorph over others based on a certain inherent characteristic of that
allomorph. In this case, Kager and Hargus propose the length of the allomorph to be
that distinguishing feature. There seems to be one difference between the Kager and
Hargus principles, however. Kager’s principle seems more specific and applies only to
morphemes which mark morphological categories, i.e., functional morphemes. Har-
gus’s Brevity principle, on the other hand, seems more general and could be appli-
cable to stem allomorphs as well. This being the case, the allomorph [zur] would be
favored over the longer [zaar], and the zero-derived third person masculine singular
form is expected to be [zur]. This allomorph is however subminimal and runs afoul of
Ft-Bin. The third person feminine and plural suffixes can safely append to either allo-
morph [zur] or [zaar], producing the forms [zurat] or [zaarat], [zuru] or [zaaru], all of
which are admitted by minimality. If the Brevity constraint prefers the shorter forms
[zurat] and [zuru], it will result in deriving an incongruent third person paradigm such
as the following hypothetical one:
(53) Third Person Paradigm
M-Sg F-Sg Pl
zaar zurat zuru

19 I would like to thank Sharon Hargus for bringing to my attention the Brevity principle, and for discussing
its implication on stem allomorphy.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 85

Here we have another incongruent paradigm where one form [zaar] is faithful to its
input while the other forms are faithful to the corresponding imperfective output.
A paradigm-regularizing constraint OP-IDENT-V resists vowel quality alternation
within a subparadigm and rejects the short allomorph in favor of a more congruent
paradigm {zaar, zaarat, zaaru}. We deduce from these facts that OP-IDENT-V out-
ranks OO-IDENT-V which in turn outranks IO-IDENT-V.
To summarize, the reduced allomorph of the hollow verb stem emerges as a result
of the competition of three forces, faithfulness to the perfective input, faithfulness
to the imperfective output, and identity within the person subparadigm. The possible
augmented output *[zaareet] is ruled out by a relativized interpretation of Anchor
which resists alternation in syllabic position in inputs and outputs.

5 Uniformity versus contrast

The terms ‘paradigmatic uniformity’ and ‘paradigmatic contrast’ are sometimes used
interchangeably (Downing et al. 2005), but in this section, I argue that the two notions
lead to different outcomes, and that uniformity accounts for many empirical observa-
tions that contrast fails to capture due to its locality. Broselow (2008) discusses the
same array of facts presented here with special reference to Iraqi Arabic. We concur
with Broselow that the different realizations of inflected stems do not lend themselves
straightforwardly to a phonological solution, and that morphological constraints must
be invoked to explain the apparent anomaly. However, Broselow attributes the pres-
ence of the /-ee/ stem extension to ‘morphological distinctness’ or contrast. Her ex-
planation goes as follows: There is a distinction in Arabic verb inflecting between
first/second person on the one hand, and third person on the other. Since third person
suffixes are either vowel-initial or phonetically empty, stress falls on the initial sylla-
ble; e.g., [nísa, nísat, nísaw] ‘forgot’. By extending the stem of the non-third person
forms, the extender /-ee/, with its bimoraic nucleus, pulls stress away from the initial
syllable onto the second syllable, thus establishing the desired contrast between third
and non-third person paradigms. In this paper, the linker /ee-/ is attributed to a unifor-
mity requirement that monitors stress placement and prosodic structure across tokens
of the same sub-paradigm: the disyllabicity of and final stress in [katábt] optimizes
the augmented weak and geminate forms [raméet] and [èabbéet] over monosyllabic
alternatives with or without epenthesis, *[rám(i)t] and *[èább(i)t]. A number of shifts
in Arabic dialects confirm the uniformity over the contrast explanation.
The first piece of evidence comes from the conjugation of glottal-initial verbs
like [Paxað] and [Pakal] in Kuwaiti and Egyptian. In Kuwaiti, the first, second and
third conjugations drop the initial syllable [Pa] reducing the stem to one syllable,
rendering it subminimal. This subminimality triggers affix augmentation, [xaðéet] ‘I
took’, [xaðéeti] ‘you-f took’ [xaðéetaw] ‘you-pl took’, [xaða] ‘he took’ and [xaðat]
‘she took’.20 The contrast analysis fails to account for the extension in xaðéet since
the old form Paxáðt satisfies the three versus 1/2 person contrast. The uniformity

20 These forms were extracted from data collected for my dissertation. The consultant was a Kuwaiti stu-
dent working toward his PhD at the same institution. Further data collected from various Kuwaiti serials
broadcast on MBC TV corroborate the pattern.
86 S. Farwaneh

account, on the other hand, explains this shift as an analogy to open (weak) verbs like
raméet and [èabbéet]. In Egyptian, the initial syllable is dropped reducing the stem
in all person conjugation without augmentation as in the list below.
(54) Egyptian
P 2P 3P
-F –PL kalt xatt kalt xatt kal xad
+F –PLkalt xatt kalti xatti kalit xadit
+PL kalna xadna kaltu xattu kalu xadu
The 3p and 1/2p stems of xad and kal are identical. But the contrast analysis requires
some contrast between the two forms. Uniformity has nothing to say about these
forms since third person and first/second person inflection of these two verbs match
that of other verbs as shown below. Syllables relevant for comparison are underlined.
(55) 3P 1/2P
ká.ta.(b) ka.táb.(t)
xá.0.(d) .0.xát.(t)
Thus, since the stem is monosyllabic, the syllable may serve as the initial syllable in
third person forms and final syllable in first/second person forms, and no uniformity
violation ensues.
Another alternation attributed to contrast effect in Broselow’s analysis is the vo-
calic alternation in the third versus non-third person allomorphs of hollow verbs citing
Iraqi šaaf vs. šift ‘see’. Since the monosyllabic stem does not allow for stress alterna-
tion, contrast manifests itself instead as contrast in vowel melody. This account works
successfully in Iraqi and similar dialects which utilize one vocalic melody to mark
perfective verbs. Other dialects like Egyptian and Levantine show /i/-/u/ alternation,
the choice of which is determined by the quality of the imperfective vowel. Following
Gafos (2003) assumed here, this is an OO-faith effect; the quality of the perfective
vowel is governed by a correspondence relation that binds the perfective form with
its imperfective counterpart; thus zurt { Pazuur, and jibt { Pajiib. Since both vowels
satisfy the contrast constraint, the contrast analysis cannot trigger the correct choice
between the two high vowels.
Another challenge to the contrast analysis comes from the stress of subject and
object-inflected perfective verbs in Iraqi (as well as Levantine and Egyptian). Ac-
cording to Broselow’s analysis, contrast not only accounts for the vocalic extension
in niséet ‘I forgot’ but also for the opaque stress-epenthesis interaction in kitábit
(from kitabt). Stress rules in Iraqi require initial stress in a triple open syllable forms;
yet in kitábit stress falls on the penultimate syllable. Departing from earlier analyses
attributing the unexpected penultimate stress to the invisibility of epenthetic vowels
to stress, Broselow’s analysis attributes the disruption of stress to the morphological
constraint Contrast rather than the presence of the epenthetic vowel. Although the
contrast analysis is appealing as it unifies sound and non-sound verb inflection and
stress placement under the rubrics of morphology, it cannot be extended to object-
inflected verbs where stress in third and non-third forms is neutralized on the penul-
timate syllable. Consider the forms in (56) which show the stress difference in third
person forms when an object suffix is added.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 87

(56) Subject Subject+Object


3p kítab kitábha
1/2p kitábit kitábitha
As the inflected forms of kitab in (2) show, stress contrast is observed in the subjective
forms, but once an object affix is added, the prosodic structure forces stress to land
uniformly on the penultimate syllable. For the contrast analysis to hold, it would
be necessary to restrict the effect of the contrast constraint to the earliest level or
stratum by demoting its ranking in later levels, perhaps using a modified version
of Kiparsky’s (2003) Stratal OT. The uniformity-based account does not face this
problem because it only requires that inflectional paradigms across verb types be
uniform. This is indeed the case as the 3 versus non-3 subject and subject_object
inflection of sound and nonsound verbs is shown in (57).

(57) 3p+subj 1/2p+subj 3p+subj+obj 1/2p+subj+obj


Sound kátab katábt katábha katábtha
Weak ráma raméet ramáaha raméetha
Geminate mádd maddéet máddha maddéetha
Hollow záar zúrt záarha zúrtha
Stress falls uniformly on the initial syllables in 3p-inflected verbs of all verb types but
shifts to the second stem syllable when a consonant-initial objective suffix is added;
this uniformity is observed in all verb types as each column in (57) illustrates. The
monosyllabic stems of geminate and hollow verbs pose no problem to the uniformity
account since the two moras of the trochaic foot match both the initial and second
syllable of a disyllabic trochee.

6 Conclusion

In conclusion, I have provided an account of some of the complex facts concern-


ing the inflection of sound and non-sound verbs, which involve affix and stem allo-
morphy. The account recognizes four levels of correspondence, input-output, output-
output, within and across paradigms. The interaction of correspondence constraints
characterizing each of these four levels with markedness and alignment constraints
needed elsewhere in the language, derives the desired results without recourse to tem-
platic or form-specific markedness constraints. The effect of paradigm uniformity or
leveling on Output optimization is a promising approach to elucidating the anoma-
lous behavior of non-sound verbs; thus, eliminating the need for the sound/non-sound
verb dichotomy in earlier analyses of Arabic verb morphology.
Crucial to the analysis is the inter-paradigmatic constraint PP-IDENTACCENT
which regularizes stress placement across same person paradigms, thus requiring
augmentation of non-third person conjugation of weak and geminate verbs. Aug-
mentation is blocked in hollow verb conjugation due to the constraint Anchor which
preserves the alignment of the input’s peripheral segments. The paper also demon-
strates that paradigm uniformity with its global scope is better suited to account for
such apparent anomalies than the locally-restricted contrast.
88 S. Farwaneh

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

References
Bat-El, O. (1994). Stem modification and cluster transfer in Modern Hebrew. Natural Language and Lin-
guistic Theory, 12, 571–596.
Benmamoun, E. (1999). Arabic morphology: The central role of the imperfective. Lingua, 108, 175–201.
Benmamoun, E. (2003). The role of the imperfective template in Arabic morphology. In J. Shimron (Ed.),
Language processing and acquisition in languages of semitic root-based morphology, Amsterdam:
Benjamins.
Boudelaa, S., & Marslen-Wilson, W. D. (2011). Productivity and priming: Morphemic decomposition in
Arabic. Language and Cognitive Processes, 26, 624–652.
Brame, M. (1970). Arabic phonology: Implications for phonological theory and historical semitic. Ph.D.
dissertation, MIT.
Broselow, E. (1992). Parametric variation in Arabic dialect phonology. In E. Broselow, M. Eid, &
J. McCarthy (Eds.), Perspectives on Arabic linguistics IV (pp. 7–46). Amsterdam: Benjamins.
Broselow, E. (2008). Stress-Epenthesis Interactions B. Vaux & A. Nevins (Eds.), Rules, constraints, and
phonological phenomena (pp. 121–149). Oxford: Oxford University Press.
Brustad, K., Al-Batal, M., & Al-Tonsi, A. (Eds.) (1997). Al-kitaab fii tac allum al-c arabiyya: A textbook
for Arabic. Georgetown: Georgetown University Press. 1997.
Caballero, G. (2006). “Templatic back-copying” in Guarijío abbreviated reduplication. Morphology, 16(2),
273–289.
Davis, S., & Zawaydeh, B. (1999). Hypocoristic formation in Ammani-Jordanian Arabic In E. Benmamoun
(Ed.), Perspectives on Arabic linguistics XII (pp. 113–139). Philadelphia: Benjamins.
Downing, L., Hall, T. A., & Raffelsiefen, R. (Eds.) (2005). Paradigms in phonological theory. Oxford:
Oxford University Press.
Gafos, A., & Ralli, A. (2001). Morphosyntactic features and paradigmatic uniformity in two dialectal
varieties of the island of Lesvos. Journal of Greek Linguistics, 2, 41–73.
Gafos, A. I. (2003). Greenberg’s asymmetry in Arabic: A consequence of stems in paradigms. Language,
79(2), 317–355.
Farwaneh, S. (2007). Hypocoristics revisited: challenging the centrality of the consonantal root. In Per-
spectives on Arabic linguistics (Vol. 20, pp. 25–48). Amsterdam: Benjamins.
Farwaneh, S. (2006). Surface effects in Arabic morphology. International Journal of Languages and Lin-
guistics, 17, 35–46.
Frost, R., Deutsch, A., & Forster, K. (2000). Decomposing morphologically complex words in a nonlinear
morphology. Journal of Experimental Psychology, 26(3), 751–765.
Hargus, S. (1997). Prosodically conditioned allomorphy in optimality theory: Brevity in Witsuwit’en. Un-
published manuscript, University of Washington
Hayes, B. (1989). Compensatory lengthening in Moraic phonology. Linguistic Inquiry, 20(2), 255–306.
Kager, R. (1996). On affix allomorphy and syllable counting. In Ursula Kleinhenz (Ed.), Studia Grammat-
ica: Vol. 41. Interfaces in Phonology (pp. 155–171). Berlin: Akadamie Verlag.
Kenstowicz, M. (1996). Base identity and uniform exponence: Alternatives to cyclicity. In J. Durand & B.
Laks (Eds.), Current trends in phonology: Models and methods (pp. 363–393). Paris-X and Salford:
University of Salford Publications.
Kenstowicz, M. (2005). Paradigmatic uniformity and contrast. In L. Downing (Ed.), Paradigm in phono-
logical theory.
Kiparsky, P. (2003). Syllables and moras in Arabic. In R. van de Vijver & C. Féry (Ed.), The Syllable in
Optimality Theory (pp. 147–182). Cambridge: Cambridge University Press.
McCarthy, J. (1979). Formal problems in semitic phonology and morphology. Doctoral dissertation, MIT.
McCarthy, J. (1981). A prosodic theory of nonconcatenative morphology. Linguistic Inquiry, 12, 373–418.
McCarthy, J. (1986). OCP effects: Gemination and antigemination. Linguistic Inquiry, 17, 207–263.
McCarthy, J. (1993). Template form in prosodic morphology. In L. Smith-Stvan (Ed.), Papers from the
third annual formal linguistics society of MidAmerica conference (pp. 187–218).
McCarthy, J. (2002). Optimal paradigms. Rutgers optimality archives, 485.
McCarthy, J., & Prince, A. (1986). Prosodic morphology. Ms, University of Massachusetts-Amherst and
Brandeis University.
Non-sound’ verb inflection in Arabic: Allomorphic variation and. . . 89

McCarthy, J., & Prince, A. (1990). Prosodic morphology and templatic morphology. In M. Eid &
J. McCarthy (Eds.), Perspectives on Arabic linguistics, II: Papers from the second symposium on
Arabic linguistics (pp. 1–54). Amsterdam: Benjamins.
McCarthy, J., & Prince, A. (1993a). Prosodic morphology, I: Constraint interaction and satisfaction. Rut-
gers optimality archive, 482.
McCarthy, J., & Prince, A. (1993b). Generalized alignment. In G. Booij & J. van Marle (Eds.), Yearbook
of morphology 1993 (pp. 79–153). Dordrecht: Kluwer.
McCarthy, J., & Prince, A. (1994). The emergence of the unmarked: Optimality in prosodic morphology.
In Proceedings of NELS (Vol. 24, pp. 333–379). Amherst: GSLA.
McCarthy, J., & Prince, A. (1995). Faithfulness and reduplicative identity. In J. Beckman, L. W. Dickie,
& S. Urbanczyk (Eds.), University of Massachusetts occasional papers in linguistics (Vol. 18, pp.
249–384).
Prince, A., & Smolensky, P. (1993). Optimality theory. Rutgers Optimality Archive, 537.
Ratcliffe, R. (1997). Prosodic templates in a word-based morphological analysis of Arabic. In M. Eid &
R. Ratcliffe (Eds.), Perspectives on Arabic linguistics (pp. 147–172). Amsterdam: Benjamins.
Ratcliffe, R. (1998). The ‘broken’ plural problem in Arabic and comparative semitic: Allomorphy and
analogy in Arabic and comparative semitic. Amsterdam: Benjamins.
Twist, A. E. (2006). A psycholinguistic investigation of the verbal morphology of Maltese. PhD thesis,
University of Arizona, Tucson, Arizona.
Ussishkin, A. (1999). The inadequacy of the consonantal root: Modern Hebrew denominal verbs and
output-output correspondence. Phonology, 16, 401–442.
Ussishkin, A. (2000). The emergence of fixed prosody. PhD dissertation, UC Santa Cruz.
Ussishkin, A. (2005). A fixed prosodic theory of nonconcatenative templatic morphology. Natural Lan-
guage and Linguistic Theory, 23, 169–218.
Ussishkin, A., Reimer Dawson, C., Wedel, A., & Schluter, K. (2015). Auditory masked priming in Maltese
spoken word recognition. Language, Cognition and Neuroscience, 30, 1096–1115

You might also like