Negation in Modern Standard Arabic: an Lfg Approach

Modern Standard Arabic (MSA) uses five different particles to express sentential negation: the invariant particle maa, the particle laa and its tensed counterparts lam (PAST) and lan (FUT), and laysa which is marked only for SUBJ agreement. Partial analyses of these elements are offered in other frameworks, notably Minimalism (Shlonsky, 1997; Benmamoun, 2000), but have not to date received an analysis within LFG. We propose an approach to four of these particles: the fifth one, namely maa, raises a number of additional issues and we leave it to one side for reasons of space. laa, lam, lan show distinctions of TENSE, occur only with imperfective forms of the verb (excluding the perfective) and must immediately precede the verb itself. They are limited to occurrence in verbal sentences. We propose that the adjacency requirement follows from the fact that these negative particles are non-projecting words adjoined to the (imperfective) V. On the other hand, laysa is a fully verbal element, and is thus a negative verb, occurring only with present tense interpretation. 1 Data 1.1 Negative Particles In Modern Standard Arabic (henceforth MSA) five different particles are used to express sentential negation: the (invariant) particle maa, the item laa and its (temporally) inflected counterparts lam and lan and (variously inflected) forms of laysa. Amongst these elements, laysa is unique in inflecting for SUBJ agreement. In the present paper, we will have nothing to say here about maa and concentrate uniquely on the forms of laa and laysa. 1.2 Laa, Lan, Lam There are good grounds for distinguishing between laysa on the one hand, and laa, lam and lan on the other. For laa, lam and lan the basic facts are as follows.1 Firstly, all these negative forms occur in sentences which have a verbal element as the main predicate. There is a basic morphological opposition in Arabic between imperfective and perfective verbforms, and laa, lam, lan all co-occur only with imperfective forms of the verb: substituting perfective verbforms in all of the following examples would lead to ungrammaticality. The pairs in (1) (3) exemplify the particle laa negating an imperfective indicative (with a present tense reading); (1) and (2) additionally illustrate SV(O) order and (3) shows VSO word order. Note that irrespective of word order, the negative particle laa immediately precedes the imperfective verb in all of these examples. (1) a. t ̇ -t ̇ ullaab-u the-students-NOM ya-drus-uu-n 3M-study.IPFV-3MP-IND The students study/are studying. We are grateful to Tracy Holloway King and the audience at LFG09 for comments and suggestions (in particular Ash Asudeh and Ron Kaplan) and to members of the Essex Arabic Syntax Workshop for discussion of contemporary work on MSA and the Arabic vernaculars. Note: glossing is morphological, reflecting the standard morphosyntactic desrciption of MSA. Where examples have been taken from sources, transliterations have been standardized to the DIN31635 format (and some randomly omitted case marking has been reinserted in some examples from Benmamoun (2000)). b. t ̇ -t ̇ ullaab-u the-students laa NEG ya-drus-uu-n 3M-study.IPFV-3MP-IND The students do not study/are not studying. (Benmamoun, 2000, 95) (2) a. Zayd-un Zayd-NOM y-aktub-u 3M-write.IPFV-3MS.IND al-yawm-a the-day-ACC al-risalat-a the-letter-ACC Zayd is writing the letter today. b. Zayd-un Zayd-NOM laa NEG y-aktub-u 3M-write.IPFV-3MS.IND al-yawm-a the-day-ACC al-risalat-a the-letter-ACC Zayd is not writing the letter today. (3) a. Y-aktub-u 3M-write.IPFV-3MS.IND Zayd-un Zayd-NOM al-yawm-a the-day-ACC al-risalat-a the-letter-ACC Zayd is writing the letter today. b. Laa NEG y-aktub-u 3M-write.IPFV-3MS.IND Zayd-un Zayd-NOM al-yawm-a the-day-ACC al-risalat-a the-letter-ACC Zayd is not writing the letter today. The following set of data illustrate the basic facts with respect to the tensed forms of laa, namely lam and lan. (4) and (5) show that the future may be expressed by means of an imperfective (indicative) verb with the prefix sa-, and additionally that the future form verb is negated by using the particle lan in combination with a subjunctive mood imperfective (without the prefix sa-): again, adjacency is required between the particle and the main verb irrespective of sentential word order. (4) a. t ̇ -t ̇ ullaab-u the-students-NOM sa-ya-d ̄ hab-uu-n FUT-3M-go.IPFV-MP-IND The students will go. b. t ̇ -t ̇ ullaab-u the-students-NOM lan NEG.FUT ya-d ̄ hab-u 3M-go.IPFV-MP.SBJV The students will not go. (Benmamoun, 2000, 95) (5) a. sa-ya-d ̄ hab-u FUT-3M-go.IPFV-MSG-IND t ̇ -t ̇ ullaab-u the-students-NOM The students will go. b. lan NEG.FUT ya-d ̄ hab-a 3M-go.IPFV-MSG.SBJV t ̇ -t ̇ ullaab-u the-students-NOM The students will not go. Finally (6) shows that the combination of the particle lam with an imperfective verb in jussive mood corresponds to an (affirmative) perfective verb. It should be noted that in the Arabic vernaculars, the basic constrast is between the marked form (IPFV.IND) in the affirmative and the unmarked form in the context of the tensed negative particle (that is, the JUSS/SBJV distinction in neutralised in the vernaculars). (6) a. t ̇ -t ̇ ullaab-u the-students-NOM d ̄ ahab-uu go.PFV-3MP The students left. b. t ̇ -t ̇ ullaab-u the-students-NOM lam NEG.PAST ya-d ̄ hab-uu 3M-go.IPFV-MP.JUSS The students did not go. (Benmamoun, 2000, 95) c. *lam NEG.PAST t ̇ -t ̇ ullaab-u the-students-NOM ya-d ̄ hab-uu 3M-go.IPFV-MP.JUSS The students did not go. To summarise, laa, lam and lan occur with verbal forms in the imperfective but not with perfective forms of the verb. In all cases, the negative particle must be adjacent to this form, see (6c). laa occurs with the indicative imperfective and cannot be used for sentences in the future or past. lam occurs with the jussive imperfective expressing negation in the past, and lan with the subjunctive imperfective, expressing negation in the future: thus lam and lan appear to be negative particles which carry temporal information. (7) TENSE AFFIRM FORM NEG FORM PRES IPFV.IND laa + IPFV.IND PAST PFV lam + IPFV.JUSS FUT sa-IPFV.IND lan + IPFV.SBJV 1.3 Future Negation: A Further Data Point It is generally claimed that laa canot co-occur with tensed verbs (Benmamoun, 2000; Bahloul, 1994). In fact, however, things are slightly more complicated. It is certainly true that ‘double’ expression of FUT is impossible (shown by (9) and (8)), but it is not completely accurate to state that laa cannot combine with a future marker. This is because there is an alternative analytic realization of future, namely the use of the particle sawfa with an (unprefixed) imperfective indicative form. As the data shows, laa can combine with safwa but not with prefixal future forms in sa(hence the contrast between (11) and (12)). (8) *sawfa FUT lan NEG-FUT y-ah ̇ dur-a. 3SM-come-SBJV He will not come. (9) *t ̇ -t ̇ ullaab-u the-students-NOM lan NEG.FUT sa-ya-d ̄ hab-uun/-uu FUT-3M-go.IPFV-MP.IND/-MP.SBJV The students will not go. (10) lan NEG-FUT y-ahdur-a 3M-come.IPFV-SM.SBJV He will not come. (11) *t ̇ -t ̇ ullab-u the-students-NOM laa NEG sa-ya-d ̄ hab-uu-n FUT-3M-go.IPFV-3M-IND The students will not go.(Benmamoun, 2000, 101) (12) Sawfa FUT laa NEG y-ah ̇ dur-u 3M-present.IPFV-3MS.IND He will not come. (Fassi-Fehri, 1993, 82) 1.4 Laysa laysa differs in several respects from the invariant forms laa, lan, lam. It realizes (SUBJ) agreement and is not required to be adjacent to the verb. (13) SG DU PL 1 lastu lasnaa 2M lasta lastumaa lastum 2F lasti lastumaa lastunna 3M laysa laysaa laysuu 3F laysat laysataa lasna (14) a. laysa NEG.3MS h ˇ ālid-un Khalid-NOM ya-ktub-u 3M-write.IPFV-3MS š-šiQr-a the-poetry-ACC Khalid does not write/is not writing poetry. b. laa NEG ya-ktubu 3M-write.IPFV-3SM h ˇ ālid-un Khalid-NOM š-šiQr-a the-poetry-ACC Khalid does not write/is not writing poetry. (Benmamoun, 2000, 103) A third difference is that it occurs in both verbal and verbless sentences (unlike laa, lan, lam), that is, sentences with nominal and adjectival predicates. (15) a. laysa NEG.3MS Pah ˇ ii brother.my muQalliman-an. teacher-ACC My brother is not a teacher. b. laysa NEG.3MS muQalliman-an. teacher-ACC He is not a teacher. (Benmamoun, 2000, 53) laysa shows the typical behaviour of a verb in that number agreement is defective when it precedes the SUBJ: (16) a. al-awlad-u the-boys-NOM lays-uu NEG-3MP ya-ktub-uun. 3M-write.IPFV-3MP-IND The boys do not write. b. lays-a NEG-3MS al-awlad-u the-boys-NOM ya-ktub-uun. 3M-write.IPFV-3MP-IND The boys do not write. laysa is compatible only with IPFV.IND verbs and receives a present interpretation. (17) a. *laysa NEG.3SM r-rağul-u the-man-NOM ijakala eat.PERF.3SM The man did not eat (Benmamoun, 2000, 105) b. *laysa NEG.3SM r-rağul-u the-man-NOM sa-ya-ijkulu FUT-eat.IPFV.3SM ġadan tomorrow The man will not eat tomorrow (Benmamoun, 2000, 105) 1.5 Compound Tenses We use purely morphosyntactic glossing throughout. Verbs show a morphological distinction between PFV and IPFV forms: such forms are used to express both temporal and aspectual distinctions: the opposition between them in sentences containing a single analytic form broadly encodes a PAST/NONPAST temporal distinction. (See Fassi-Fehri (2004) for some discussion.) The INDIC imperfective further inflects for FUT (or combines with the particle sawfa). The imperfective stem also shows what are traditionally called distinctions of MOOD: INDIC, JUSS, SBJV. Compound tenses involve the combination of a finite auxiliary with the perfective and imperfective indicative (finite) forms. They are not required to be adjacent. The table below illustrates various compound tenses. (18) FORM REALIZATION MEANING PFV katab-tu t-taqrı̄r-a PAST I wrote the report. IPFV ijaktub-u t-taqrı̄r-a PRES I write/am writing the report. FUT-IPFV sa-ijaktub-u t-taqrı̄r-a FUT I will write the report. PFV + PFV kun-tu qad katab-tu t-taqrı̄r-a PAST PRFT I had written the report. PFV + IPFV kun-tu ijaktub-u t-taqrı̄r-a PAST PROG I was writing/used to write the report. PFV + FUT-IPFV kun-tu sa-ijaktub-u t-taqrı̄r-a PAST FUT I was going to write the report. IPFV.IND + PFV ijakūnu qad katab-tu t-taqrı̄r PRES PRFT I (always) have written the report. FUT-IPFV + PFV sa-ijakūnu qad katab-tu t-taqrı̄r FUT PRFT I will have written the report. FUT-IPFV + IPFV sa-ijakūnu ijaktub-u t-taqrı̄r-a FUT PROG I will be writing the report. 2 Minimalist Approaches Negation in MSA (and in the Arabic vernaculars) has received a reasonable amount of theoretical attention within Minimalism (and its precursors), the major references being Benmamoun (2000); Ouhalla (2002) and Shlonsky (1997). Of these, the most extensive discussion is Benmamoun (2000), and for this reason we briefly present his approach here. The basic structural assumptions made in this account (which discusses negation in the vernaculars (concentrating on Moroccan Arabic (MA)) and MSA, involves a NegP projection situationed between TP and VP, as in (19).2 (19) TP XP T’ T NegP Neg VP XP V’ V The crucial points of this analysis concern the assumptions about what features are inherent to each node. First, sentential negation (the Neg node), is taken to be specified for the categorial feature [+D] (Benmamoun, 2000, 69). The elements laa, lam and lan are generated in Neg. Second, Tenses are associated with different bundles of features generated on the T node, as follows (Benmamoun, 2000, 50): The ordering of functional heads is critical to Benmamoun’s proposal, but Shlonsky (1997) takes Neg to be higher than T in the hierarchical structure in Arabic (Shlonsky, 1997, 103-4). (20) T → [+D] (Present) T → [+D, +V] (Past, Future) T → [+V] (Imperative) Suppose the node T is generated with the feaure bundle [+Past, +D, +V] or [+Fut, +D, +V] (“the V feature must be checked by verbal heads, while the D feature can be checked by nominal heads or by verbs that carries (sic) agreement” (Benmamoun, 2000, 99)). By assumption, the Neg node is also specified for [+D]. In order for both the +V and the +D features of the T node, to be appropriately “checked”, it is necessary that both the V and the Neg move to the T node. A derivation such as the following will ensue, in which V raises to Neg and then Neg and V together raise to T. The spell out of the resultant T node is the combination of lam + verb, likewise if +Fut is generated on the T node, then the spell out will be lan + verb. As for Neg and V “they are both in tense supporting the tense feature and checking the categorial [+V] feature” (Benmamoun, 2000, 100). (21) TP XP T’ T[+past, +D, +V] [Neg+Vi]j T NegP Neg [+D] tj VP XP V’ V ti The alternative might be to try to move the verb directly to deal with the +V feature (and spell out the tense): presumably such a verb could also check the D feature of the T node, as it carries subject agreement, but this violates Minimality, or take the Neg also but spell out the features on the verb, not the negation. This is ruled out by the assumption that tesne must be spelled out on the head of the complex, which is Neg (Benmamoun, 2000, 102). Suppose now that the T node is generated with the feature bundle [+Pres, +D]. The +D feature can be checked by a nominal. Because there is no +V feature on T, neither the verb (nor the Neg) is required to raise to T. However given that laa and the V are required to be adjacent, something must require this: “merger between laa and verb must be due to some property of laa itself. The property in question is the categorial feature [+D] feature of laa. The merger betwen laa and the verb, carrying subject agreement, allows the latter to check the categorial [+D] feature on the negative” (Benmamoun, 2000, 100). (22) TP XP T’ T[+pres, +D] NegP Neg [+D] laa +Vi VP XP V’ V ti In contrast to traditional accounts, which view laysa as a verbal element, Benmanoun takes it also to be a Neg particle (specified for [+D]). The idea is that since laysa itself inflects for SUBJ agreement, then this feature is checked by the SUBJ and so Neg (i.e. laysa) does not raise to T for purposes of feature checking. This means that in principle, it is free to be non-adjacent to the inflected verb (unlike laa). (23) TP XP Subj T’ T[+D] NegP Neg [+D] laysa VP XP Subj V’ V yaktubu Although it would take us too far afield to attempt here any substantial critique of this (or other Minimalist) proposals, we will make a number of brief observations about the account. The first is that it is far from complete in its present form. It does not explain how (by which mechanism) different negatives select different forms (moods) of the verb, and given that that there are no lexical differences postulated between laa, lam, lan (they result from the spell out of different sets of features in different tree locations, as far as we understand it), it is not obvious how this will be treated. Second, the account is radically incomplete in that there is no attempt to extend it to the more complicated facts of negation with compound tenses. Third, the assumption that Neg is categorially specified as +D plays a crucial role in terms of ensuring that forms of laa and the verb are strictly adjacent: the subject agreement features of V are required to check the +D specification of Neg heads. While this diacritic approach does indeed appear to produce the desired result, it is unclear what it actually represents (other than a diacritic). Moreover there is perhaps some unwelcome asymmetry in the treatment of the laa+V adjacency (which involves only this +D checking requirement) and that of the lam/lan + V adjacency, which additionally involves the verb checking the +V feature of T (and thus raising alongside Neg to T). Fourth, it is unclear what checks the +D feature of the T[+Pres, +D] node, in the case where laa + V occurs in Neg and in the case where laysa occurs in Neg.3 Fifth, there is no discussion or analysis The issue here is perhaps only one of unclarity of presentation, making the resultant analysis opaque to those less than totally familiar with the assumptions of the framework. of the multiple agreements on the negative laysa and the following verb, while most of the previous approaches within this framework have postulated multiple functional (Agr) projections to account for this data. 3 Analysis of Laa, Lam and Lan 3.1 Adjacency and Selection In short, we argue that adjacency follows because the negative particle and the verb form a small construction, that is, the particle is a non-projecting word in the sense of Toivonen (2003). Neg and V do not constitute a single morphological word. Unlike laysa, laa, lam and lan are non-projecting elements which occur as sister to I, and therefore occur with verbal elements. The behaviour of the negative particles laa, lam and lan is strongly reminiscent of the particles discussed in Toivonen (2003). (24) Property laa, lam, lan Swedish Verbal Particles Take complements No No Can be modified No No Bear stress Yes Yes Adjoined to verb Yes, left Yes, right Separable No Yes, but not by object (25) I −→ Î ↑ = ↓ I ↑ = ↓ (26) IP NP Zaydun I’ I Î laa I y-aktub-u S VP NP al-risalat-a Each particle places certain co-occurrence requirements on its sister, and thus a question arises as to whether these are cor f-structure constraints. We turn to this in the following subsection. 3.2 Selection In order to discuss the matter of selection we will need to say much more about the tense and aspect system. There is some literature on this question, but accounts often appear to be driven more by theory-internal requirements than by the empirical facts. For the moment, we simply make the following analytic assumptions. Firstly, although some researchers argue that MSA is a tenseless language (largely based on very theory-internal reasoning rather than data), we take it that MSA has tense as well as aspect and that TENSE involves distinctions of PAST/NON-PAST and FUT/NON-FUT. Additionally, as we have seen, the Arabic verb makes a morphosyntactic distinction between three moods, JUSS, SBJV and INDIC. Only the last of these, the INDIC, encodes distinctions of TENSE. JUSS and SBJV forms only occur when selected for. In principle, selection might be in terms of a MOOD feature or directly on c-structure form, and we return to this question. With this in place we can formulate the lexical entries to capture the basic facts. The basic agreement information for 3MPL forms is provided in the template (27). Illustrative lexical entries for indicative verb forms (perfective, imperfective and future-imperfective) are in (28)-(30), and for the other moods in (32)-(33).4 (27) 3MPL ≡ (↑ SUBJ NUM) = PL (↑ SUBJ PERS) = 3 (↑ SUBJ GEND) = MASC (28) d ̄ ahab-uu I (↑ PRED) = go < SUBJ > Perfective Form (↑ TENSE PAST) = + @3MPL (29) ya-drus-uu-n I (↑ PRED) = study < SUBJ > Imperfective Form (↑ TENSE PAST) = − @3MPL (30) sa-ya-d ̄ hab-uun I (↑ PRED) = go < SUBJ > Imperfective Form (↑ TENSE PAST) = − (↑ TENSE FUT) = + (↑ POL) = POS @3MPL (31) sawfa Î (↑ TENSE FUT) = + (32) ya-d ̄ hab-uu I (↑ PRED) = go < SUBJ > Imperfective Jussive Form (↑ MOOD) = JUSS @3MPL (33) ya-d ̄ hab-uu I (↑ PRED) = go < SUBJ > Imperfective Subjunctive Form (↑ MOOD) = SBJV @3MPL Treating the value of the FUT feature as instantiated would prevent (30) co-occurring with (??) (thanks to Tracy Holloway King for pointing this out). However it is not yet completely clear to us what co-occurrence restrictions should be treated at f-structure and which ones are more properly considered to be part of c-structure or even morphological restrictions, so we have not used instantiated features here. To recap, the behaviour we need to capture is summarised in (34). (34) laa coccurs with an imperfective indicative verbform lam expresses PAST= + and selects the jussive verbform lan expresses FUT=+ and selects the subjunctive verbform sa(and sawfa) express POL = + Consider first the treatment of saand sawfa. (30) limits the saform to occurrence in a positive clause, whereas sawfa does not place this restriction. This will be used in accounting for (11) and (9) permitting laa to co-occur with sawfa (12).5 The entries for the particles are as follows. The TENSE specification in the entry for laa means it cannot combine with Perfectives, the POL specification prevents it combining with the saImperfective. If it were to combine with JUSS or SBJV then there would overall be no TENSE which would be a problem. So the f-structure for (1b) is shown in (36) (35) laa Î (↑ TENSE PAST) 6= + (↑ POL) = NEG (36) 