Chapter 1: Introduction
1.3 Aspects of Tunisian morphology
This section describes some selected aspects of Tunisian morpho-syntax with the intent to help the reader which is not familiar with Arabic to have a better understanding of the examples and the concepts discussed in this dissertation.
Personal pronouns in Tunisian come in two forms: tonic pronouns and clitic ones. They present gender, number and person agreement features as the table in (15) illustrates.
(15) Personal pronouns in Tunisian (Talmoudi 1980, 144–45)
tonic forms direct object obj. of preposition & possessive
1st ena -ni -ī/-ya
2nd inti -k/-ik -k/-ik
3rd Masc. huwwa -h/-u -h/-u
3rd Fem. hiyyǝ -he -he
1st ahna -ne -ne
2nd ntum -kum -kum
3rd hum -hum -hum
The data presented in this subsection is taken from Talmoudi’s (1980) grammar of the dialect of Soussa, a coastal city in the north west of the country.
The second feminine and masculine singular person of the tonic pronouns are conflated into one form: inti. For the sake of completeness, I must mention that this conflation occurs in the varieties of Tunisian spoken in the bigger cities in the north of the country (e.g. Tunis, Sousse, Sfax, Monastir) while rural varieties even within the same northern region retain the feminine masculine dichotomy ent/inti. Tonic pronouns can occur as subjects or in direct object position under stressed prosodic conditions:
(16) hiyyǝ ukht-ī she sister-mine
‘She is my sister.’
(17) šuft hiyyǝ fī-l-car see.perf her in-the-car
‘I saw HER on the bus (not, her sister…)’
Given their dependent status, clitic pronouns occur cliticized onto a lexical item whose category may vary. As Table (15) illustrates, with the exclusion of the 1st singular forms, the same clitics serve a range of uses. They may perform the syntactic function of direct object when the pronoun
cliticizes onto a transitive predicate, (18) and (19); they can function as possessive pronouns if cliticized onto a nominal head, (20) and (21); and finally they substitute the argument of a preposition as in (22) and (23):
(18) ẓarb-u hit.perf-him
‘He hit him.’
(Talmoudi, 1980: 144)
(19) ra-h see.perf-him
‘He saw him.’
(Talmoudi, 1980: 144)
(20) umm-he mother-her
(21) ktēb-u book-his
(Talmoudi, 1980: 145)
(22) maʕ-he with-her
(23) fī-h in-it
(Talmoudi, 1980: 145)
As the pair of examples (18) and (19) illustrate, the two 3rd person singular variants -h/-u are allomorphs and according to the informant who helped me to unify the transliteration of the examples, their distribution depends on the individual preference of the speaker. The two 1st person singular clitic pronouns present a form, the pronoun, namely -ni, that only replaces a noun phrase occurring in direct object position (24), and a pair of allomorphs, the forms -ī/-yya, that occur as either a possessive pronoun (25) and (26) or in the complement of a preposition (27) and (28).
(24) šāf-ni see.perf-me
‘He saw me.’
(25) umm-ī mother-mine
(26) khu-yya brother-mine
(Talmoudi, 1980: 146)
‘In front of me’
(28) bi-ya with-me
(Talmoudi, 1980: 146)
The distinction between the accusative and the oblique form of the first-person clitic pronoun, i.e.
-nī and -ī/-yya, will become essential in chapter 3 since these forms display the only case distinctions overtly realized in Tunisian.
Tunisian is a variety of Arabic and, consequently, is a Semitic language. Thus, Tunisian presents the distinctive non-concatenative morphology of this language family. I will not discuss template formation in detail but, because I will refer to this notion in the course of chapter 4, I will introduce some basic aspects of this derivational mechanism. Non-concatenative morphology in very simple terms means that word formation operates not only via the addition of affixes to a root, but it also operates through its internal modification. Semitic morphology, in fact, is said to be non-concatenative because the consonants forming a root, generally three, are not “stringed” together but allow the insertion of different vocalic patterns and affixes within the root boundaries. These patterns of root modification are called templates. Templates may derive nouns, adjectives or verbal forms. The basic meaning of a verbal root, for instance, in Standard Arabic corresponds to a template consisting of a sequence of three syllables CVCVCV in which the consonant positions are filled by the phonemes forming the root, for instance [q][t][l], while the vowel slots are filled with a short middle vowel1 deriving the form qatala (29)a. Some other examples of derivation via templates taken from Ratcliffe’s (1998) explanation of Standard Arabic morphology, are listed in (29)b-d:
(29) a. qatala “he killed” CaCaCa Basic meaning b. qutila “he was killed” CuCiCa Passive voice c. qattala “he massacred” CaCCala Intensive meaning
d. qaatil “one who kills” CaaCiC Active Participle of a.
or “killer” (Ratcliffe 1998:24 (2))
Tunisian templates are essentially the same as Standard Arabic. The relations between the root and the basic form of a verb, for instance ‘to write’, or the root and its active participle ‘writer’
are expressed respectively by the templates in (29)a and (29)b, resulting in the forms ktēb and kētib respectively.
Tunisian does not present periphrastic causative forms like the ones found in Romance languages or in English. In this language as in the other Arabic varieties, causation is expressed by putting a root in a specific template, i.e. the form in (29)c resulting in the form kattāb ‘to make someone read’. Modern Arabic grammars generally respect the convention of labelling the templates that derive verbal forms with a number that goes from I to X. The basic form of a verb, here (29)a, is called Form I, while the causative template in (29)c is called Form II. I will not adopt this convention and I will simply refer to verb forms like kattāb as causative forms.
1.3.3 Inflected verbal paradigms
Derivation in Arabic does not only operate via root internal modification, but it also operates via prefixation and suffixation, just as it happens in the Romance language family. This means that φ features on verbs, nouns and adjectives are realized by means of affixes. Tunisian like all other Arabic varieties has two inflected verbal paradigms. Verbs agree with the subject for person, gender and number features; gender and number are not marked on all the forms of the paradigms, whereas person is and the difference between the two paradigms is precisely in the position where person agreement is realized. Generative works on Arabic languages customarily refer to the two paradigms as “perfective” and “imperfective” (for instance, Benmamoun 2000; Aoun, Benmamoun, and Choueiri 2009; Hallman 2015b), I will follow the same convention clarifying that these labels do not precisely reflect the feature endowment of the forms.
Perfective verbs display only perfective morphology. Table (30), adapted from Saddour (2010, 119: table 9), exemplifies the perfective paradigm by presenting the full conjugation of the verb ktib ‘to write’:
(30) Perfective paradigm Suffixes
1st -t ktib-t
2nd Masc. -t ktib-t
2nd Fem. -t(i) ktib-t(i)
3rd Masc. - ktib
3rd Fem. -it kitb-it
1st -nā ktib-nā
2nd -tū ktib-tū
3rd -ū kitb-ū
In the other inflected paradigms, namely the morphologically imperfective one, person agreement is expressed by means of prefixes, conversely gender and number features are perfective. Table
(31) below, adapted also from Saddour (ibid.), exemplifies the imperfective paradigm by presenting again the full conjugation of the verb ktib ‘to write’:
(31) Imperfective paradigm Affixes
1st n- n-kteb
2nd Masc. t- t-kteb
2nd Fem. t- -i t-kteb-i
3rd Masc. y- y-kteb
3rd Fem. t- t-ketb
1st n- -ū n-kteb-ū
2nd t- -ū t-kteb-ū
3rd y- -ū y-ktub-ū
Naming the two sets of forms, in fact, is a non-trivial issue that remains a controversial point in the field of Arabic studies. In the literature the two paradigms have more traditionally been labelled on the basis of the interpretation with which they are attributed. Benmamoun (2000) adopts the labels “perfective/imperfective” but does not associate any aspectual or temporal feature to their morphology. Ryding (2005) in her grammar adopts the opposite temporal approach, and refers to the two paradigms respectively as “past tense” and “present tense”. Mixed approaches have also been proposed, as in the case of Comrie (1976) who suggests that the Arabic verbal system combines tense and aspectual properties.
As I will show in chapter 5, the Tunisian available data supports the approach proposed by Hallman (2015b), meaning that perfective verbs express perfective aspect, whereas tense is not necessarily expressed by the verbs that occur in this form. As for the imperfective paradigms, Hallman treats them as default verb forms which are underspecified for both independent temporal features and aspectual ones. This is the view that I adopt as well.
To conclude this subsection a brief explanation of my glossing choices is mandatory. In the remainder of this work I will keep the glossing system as simple as possible in order to focus on the aspects that are relevant for our discussion. For instance, a gloss associated to an inflected verb form will not specify person, gender and number features unless required by the discussion.
In all other contexts I will simply indicate whether a verb occurs in the perfective or in the imperfective form, as done so far and illustrated by examples (32) and (33):
(32) semi dhin zuz dīār fī-žanfie Semi paint.perf two houses fi-January
‘Semi painted two houses in January.’
(33) semi ydhin d-dīār Semi paint.imp the-houses
‘Semi paints houses.’
Tables (30) and (31) respectively show that perfective and imperfective verb forms always convey subject agreement features, in the glosses, however, I will systematically omit this information
which represents a superfluous level of information in most aspects of the discussion I intend to carry out here. Most sentences will present an overt subject which will allow us to recover the information about subject verb agreement. The reader should nonetheless keep in mind that Tunisian is a pro-drop language and that rich verb agreement allows us to retrieve the φ features associated to pro. Finally, for ease of comparison I will use my own glossing system also for the examples taken from other authors.
Finally, I use the third singular masculine person of the perfective paradigm as citation form as customary. This means that I will cite the Tunisian form ktib as ‘to read’, even though in no context ktib is appropriately translated with the infinitival form of the verb.
1.3.4 Participles and participial sentences
Participles in Tunisian have two forms, active and passive. Active and passive participles are similar to past participles in Romance languages in the sense that their morphology patterns with that of adjectives. Like adjectives, in fact, they present gender and number but not person agreement. See the following table:
(34) Active and passive participles2
Singular ‘to write’ Plural
Active Passive Active Passive
F kētib-a maktūb-a kētb-īn maktūb-īn
M kētib maktūb
Tunisian participles differ from Romance passive participles in that they appear to be the only verbal predicate of a root sentences. In this respect they recall the behavior of ‘benoni’ in Hebrew (Shlonsky 1997) which is a finite verb form with full rights although from the viewpoint of its morphology it is a nominal element.
Participial sentences of this kind receive a variable temporal and aspectual interpretation depending on the aspectual class of the underlying predicate. For instance, participles derived from an activity predicate in Vendler’s (1957) terms receives a present perfect interpretation while directed motion predicates are interpreted in the progressive:
(35) semi šērib kuka barka Semi drink.prtcpl coke only
‘Semi has drunk only coke.’
(36) semi sēig l-adžmal li-l-bir Semi lead.prtcpl the-camels to-the-well
‘Semi is leading the camels to the well.’
The reason for this alternating interpretation has been the object of previous study (Brustad 2000;
Boneh 2010; Hallman 2015a); participial sentences, in fact, do not occur only in Tunisian but represent a well-known construction which to my knowledge spreads across all Arabic varieties.
Chapter 5 and 6 will deal, among other things, with the syntactic properties of participial sentences in Tunisian.
Going back to the morphology of participles, table (34) presents the full paradigm of a participle derived from Form I of a verb. However, since a root may occur in different forms (i.e.
templates) this same root may derive more than one active and one passive participle. I will not discuss these alternative forms here since it would bring additional complication to a matter that is already complex. Thus, in the remainder of this work I keep the issue simple and limit my examples as much as possible to the participles whose form complies with the pattern presented in (34). Nonetheless, the reader who is interested in Arabic participial morphology can refer to Hallman’s (2016) work; in this work Hallman presents an exhaustive list of all the possible forms in which participles can occur.
1.3.5 The auxiliary kēn/ykūn
Certain temporal, modal and aspectual distinctions are expressed by means of verbal periphrases formed by placing the auxiliary kēn/ykūn ‘to be’, before the lexical verb. The auxiliary, on a par with lexical verbs, presents two inflected forms: perfective and imperfective, see table (37) below:
(37) The auxiliary kēn/ykūn3
Perfective paradigm Imperfective paradigm
1st kun-t 1st n-kūn
2nd Masc. kun-t 2nd Masc. t-kūn
2nd Fem. kun-ti 2nd Fem. t-kūn
3rd Masc. kēn 3rd Masc. y-kūn
3rd Fem. kēn 3rd Fem. t-kūn
1st kun-na 1st n-kūn-u
2nd kun-tum 2nd t-kūn-u
3rd kēn-u 3rd y-kūn-u
Referring to kēn/ykūn solely as “auxiliary” is not completely accurate since this verb also functions as copula. The perfective forms of this auxiliary occur in copular constructions referring to the past.
(38) semi kēn ġanī Semi aux.perf rich
‘Semi was rich.’
The imperfective forms of the auxiliary, i.e. ykūn and the rest of the forms in the paradigm, do not occur in copular contexts referring to the present since Tunisian like other Arabic varieties require a null copula when the reference is to the present time.
(39) semi ġanī Semi Ø rich
‘Semi is rich.’
Rather the imperfective form ykūn is used as copula in embedded contexts. For instance, it occurs between the subject and the nominal predicate whenever a copular sentence is embedded under a desiderative verb such as ḥebb ‘to want’:
(40) semi yḥebb ykūn ġanī Semi want.imp aux.imp rich
‘Semi wants to be rich.’
A question that I will touch on in the course of the dissertation is whether kēn is the only auxiliary verb in Tunisian or not. There are two additional verbal elements in Tunisian that contribute to the temporal interpretation of the sentence but not its lexical content: the future marker beš and progressive qāʕed. The two elements are considered to be active participles, though future beš is morphologically invariable and, consequently, it does not display subject agreement.
Under standard analysis auxiliaries merge in the same clause as the main verb and express the main temporal relation. This is also what future beš and progressive qāʕed seem to be doing, hence I treat them as auxiliary although in the course of chapter 5 we will see that they display different properties.
A sentence in Tunisian is negated by two co-occurring elements: the proclitic element ma- and the enclitic element -š. This complex negation is by no means distinctive of this variety, but it is also attested in other Arabic dialects such as Moroccan, Libyan, Algerian, Palestinian, Yemeni etc. The syntax of the complex negation ma- and -š is very similar in the varieties in which it appears although certain small variations occur in each language.
This conclusive section focuses on the syntax of sentential negation in Tunisian Arabic only.
Negative construction will be used throughout the dissertation as the main diagnostic for verb movement, therefore an introduction to their syntax is mandatory in this introduction.
The complex nature of the negation ma- and -š raises several issues on its syntactic, semantic and morphologic status. A major problem discussed in the literature is how this configuration is derived. As in other languages in which negation is syntactically represented by two superficially distinct elements, the question is whether the two clitics ma- and -š occur: (a) in the specifier and the head of the same syntactic projection, (b) both in the head of the same syntactic projection or (c) in two different negative heads.
In light of the different licensing properties displayed by ma- and -š, I adopt Soltan’s (2014) proposal and assume that negation entails the presence of two distinct projections whose hierarchical order reflects the linear order of the two elements ma- and -š and that the two negative head are each hosted in the head of their own projection. Unlike Soltan, however, I will claim in chapter 6 that the Tunisian data supports an analysis whereby the negative heads occur at different
“heights” in the syntactic structure, i.e. above the TP, between TP and AspP and, once again, between AspP and the vP.
Tunisian presents two different strategies to negate a sentence. I follow Halila (1992) and call the two alternative strategies “affixal negation” and “independent negation”. According to Halila, affixal negation is derived via verbal movement to a negative projection, landing in an intermediate position between ma- and -š. This movement derives the sequence: ma-V°-š found in the following example (41). The same assumption is also adopted in the present work.
(41) semi ma-qre-š
‘Semi did not study.’
If the direct object occurs in the form of a clitic pronoun, the object must move to the NegP along with the verb on which it cliticizes. This means that the pronominal object ends up in-between the two negative elements together with the verbal head, see (42):
(42) semi ma-qar-ū-š
‘Semi did not study it.’
The same is true for an indirect object when it is realized in the form of a clitic pronoun; example (43) illustrates that the indirect object clitic moves along with the verb and ending up between ma- and -š:
(43) semi ma-kteb-lhe-š
‘Semi did not write to her.’
When the direct object and the indirect object are not clitic pronouns but full DPs they remain in situ, meaning that they follow the complex formed by the negative particles and the verbal head as illustrated by (44) and (45):
(44) semi ma-qre-š le-ktēb Semi neg-study.perf-neg the-book
‘Semi did not study the book.’
(45) semi ma-kteb-eš l-marwa Semi neg-write.perf-neg to-Marwa
‘Semi did not write to Marwa.’
Not all types of predicates can undergo movement to negation. While perfective predicates always merge with the negative clitic; if the predicate is a participle movement to negation is not allowed, see (46).
(46) a. *semi ma-mēši-š d-dār Semi neg-go.prtcpl-neg the-house
b. semi ma-hu-š mēši d-dār Semi neg-he-neg go.prtcpl the-house
‘Semi is not going home.’
As example (46) illustrates, a participial predicate is incompatible with affixal negation; therefore, participial sentences must be negated via an alternative strategy that Halila refers to as independent negation. We are not interested in the reasons why such movement is banned, but see Halila (1992) for a possible analysis of the Tunisian facts and Benmamoun (2000) for an approach which includes other Arabic varieties.
Whenever a predicate does not undergo movement to the negative projection, and irrespective of the reason why this movement does not occur, the two negative particles ma- and -š cliticize onto a personal pronoun or onto one another.
The clitic pronoun involved in this construction and the predicate of the sentence agrees with the same subject, sharing person, number and gender features:
(47) marwa ma-he-š tekl fī-l-kosksi Marwa neg-she-neg eat.imp fi-the-couscous
‘Marwa is not eating the couscous.’
‘Marwa is not eating the couscous.’