Marie-Thérèse Vinet, Driss Seghrouchni, Abderrazzak Tourabi, for helpul remarks and ...... quantificational categories (Dressler (1968), Xrakovskij (1989)).
Abdelkader Fassi Fehri Mohammed V University, IERA, Rabat
Causativity, Transitivity, and Iterativity as Pluralities
*
In this paper, two essential problems are addressed: (a) a descriptive one, consisting in the appropriate characterization of the variable behaviour of the morpho-syntactic Arabic conjugation forms (or Semitic binyanim), which has until now resisted any systematic (or unifying) treatment, and (b) a theoretical one, namely the appropriate theory of transitivity or causativity (or more generally complexity of lexico-syntactic diathesis). I argue that these two problems can be dealt with adequately only through the introduction of Number (= Nb) as a feature category of «verbs», and the configurations in which they are found, in parallel to nominal Nb. Various singular (= Sg) or plural (= Pl) Nb configurations would then lead to various syntactic distributions
and
alternations.
Furthermore,
the
collective/distributive
dichotomy of Nb is shown to play an important role in determining configurational and semantico-aspectual differences associated with complex predicates.
Arabic (as well as Semitic) provides clear morphological evidence for general nominal/verbal Nb marking (the morphology used in nominal forms to express Pl being essentially the same as that used in verbal forms). If Nb is generalized (and is orthogonal) to categories, then Nb theory can serve as the
* Part of this work has been presented at the UQAM Asymmetry Project, Montreal, December 16, 1999, the 5th Conference of Afroasiatic languages, Univ. of Paris VII, June 28-30, 2000, and during my visit to the Department of Linguistics at MIT. It would like to thank the audiences there, and especially Anna Maria Di Sciullo, Ken Hale, Alec Marantz, Noam Chomsky, Richie Kayne, Marie-Thérèse Vinet, Driss Seghrouchni, Abderrazzak Tourabi, for helpul remarks and discussions. An early version of part of this work appeared in Fassi Fehri (2000).
Causativity, Transitivity, and Iterativity as Pluralities
10
basis for providing a unifying treatment of (a) the “intensive“ (or repetitive) and the “causative“ (or transitive) alternations (of e.g. FII), and (b) the verbalization (or “inchoativization“) and the transitivization (or “causativization”) properties (of e.g. FIV). Further properties of other forms (e.g. FIII, FVI, and FX) are analyzed along similar lines.
The general Nb theory of event/argument complexities receives strong support from alternative uses and complex polysemies of causative reduplicating
morphemes
across
various
languages.
Significant
consequences and predictions of language variation are then naturally accounted for.
Nb theory is implemented in the Minimalist framework (as defined in Chomsky 1995, 1998), coupled with Distributed Morphology (cf. Halle & Marantz 1993, and subsequent work). It builds on significant ideas advanced in the literature with regard to verbal Nb marking and reduplication uses (typically Kulikov 1993, 1999, Greenberg 1991, Dolinina 1999, and Moravcsik 1978a &b, and the many references cited there), as well as ideas proposed in the context of VP shell structure (Larson 1988), L-syntax composition (Hale & Keyser 1993, 1998, 1999), little v (Chomsky, ibid), and recent work on word and category formation (Marantz 1997, 2000, and other work along similar lines). As far as implementation is concerned, it is proposed that Nb is a (potential) functional projection of “verbs“ and that its realization involves either a Sg Merge (collective), or a Pl Merge (distributive).
1. Issues
Abdelkader Fassi Fehri
11
1.1. Problem 1: Semitic morpho-syntax
The correspondence between morphologically complex Arabic forms (or Semitic binyanim) and their syntactico-semantic properties, alternations, and uses has until now resisted any systematic and/or convincing treatment. For example, how can we concile two of the traditionally acknowledged properties of e.g. so-called Form II (= FII) and Form IV (= FIV)? (a) on the one hand, FIV (which involves glottal prefixation) in (1) and FII (which involves second consonant gemination) in (3) appear as the transitivized forms of Form I (= FI) in (2) and (4), respectively: (1) ?-axraja
r-rajul-u
caus-went.out the-man-nom
l-walad-a the-child-acc
The man made the child go out. (2) xaraja
l-walad-u
went.out the-child-nom The child went out. c
(3) addama
l-?amr-a
made-important the-matter-acc He made (took) the matter (as) important. c
l-?amr-u
became-important
the-matter-nom
(4) aduma
The matter became important. (b) On the other hand, FII and FIV are used in (5) and (6), without inducing any adicity increase (or transitivization): (5) a. kassara broke.intens.
l-walad-u
l-ka?s-a
the-child-nom the-glass-acc
The child broke the glass into pieces.
b. jawwala
r-rajul-u
Causativity, Transitivity, and Iterativity as Pluralities
12
took.walks.intens. the-man-nom The man took a lot of walks. (6) a. kasara broke
l-walad-u
l-ka?s-a
the-child-nom the-glass-acc
The child broke the glass. b. jaala
r-rajul-u
took.walks the-man-nom The man took walks. (7) a. ?-aflata
r-rajul-u
come.escaped the-man-nom The man came to escape. b. falata
r-rajul-u
escaped the-man-nom The man escaped. In this use, morphology is rather associated with "Aspect", be it "inchoative", as in (5), or "intensive", as in (6). Furthermore, the aspectual and transitive meanings (and/or alternations) are not taken to be relatable, and hence no unifying treatment of these uses is proposed.
1
1.2. Problem 2: Transitivity theory
What is the source of transitivity? Does it arise from a VP shell structure (Larson 1988), the projection of little v (Chomsky 1995), or the projection of Voice (Krazer 1996)? Chomsky (1995, 1998), relying on work by Larson (ibid) and Hale & Keyser 1993 and passim) has proposed that transitivity arises
1
In addition to these two characteristics, variable meanings and uses of these forms are documented (for details on the matter, see e.g. Fassi Fehri 1996 a & b).
Abdelkader Fassi Fehri
13
through a v functional head, which is external to the projection of VP, and which:
2
(a) licenses a second Merge, and introduces an external argument (in its Spec) (b) forms a case relation («Agree») with an object Further work has shown that
3
(c) v has a verbalizing property (combines with a root R to form a «verb») (d) v has various semantic flavours (agentive, causative, process, stative). Consider (b) and (d). Unergative v’s have the property (b) only if they are 4
transitive; but not if they are intransitive. As for (d), it is not a (semantic) definition of v, since nothing unifies these various semantic flavours. Let us then turn to (a) and (c). If vp’s and their np nominalizing counterparts have the same internal functional structure, then (c) cannot be of strictly categorial (morpho-syntactic) nature. It has to be replaced by some
2
These two properties form the so-called Burzio’s (1986) generalization. Chomky (1995, pp. 315-6) claims that «… if a verb has several internal arguments, then we have to postulate a Larsonian shell […], where v is a light verb to which V overtly raises. […] The internal arguments occupy the positions of specifier and complement of V. Accordingly, the external argument cannot be lower than [Spec, v]. If it is [Spec, v], as I will assume, then the v-VP configuration can be taken to express the agentive or the causative role of the external argument. It would be natural to extend the same reasoning to transitive verb constructions generally, assigning them a double VP structure […]». He then adds that «If intransitive (unergative) verbs are hidden transitives, as Hale and Keyser suggest, then only unaccusatives lacking agents would be simple VP structures. […] The external role is a property of the v-VP configuration, and a specifier bearing this role is therefore a necessary part of the configuration ; a transitive verb assigns an external theta-role by definition». Nut note HK’s analysis is located at the L-syntax, a level which deals with lexical complexity and diathesis, but not necessarily with Case. Thus HK appear to deal with argument complexity in the lexicon, rather than transitivity as a functional syntactic notion, in which Acccusative case plays a defining role. 3 See in particular Marantz (1997, 1999, 2000), Alexiadou (2000), Harley (1999), Arad (1999), and Embick (2000), among others. 4 Following HK (1993), Chomsky (1995) analyzes unergatives as kinds of transitives (see footnote 2), a position that I will not be adopt here, assuming that lexical transitivity (or diadicity) can be kept distinct from syntactic transitivity, typically associated with accusativity.
14
Causativity, Transitivity, and Iterativity as Pluralities
“lexico-aspectual“ type feature, let it be ev (for eventuality), to be able to take into account the parallel functional structure of nominalizations, as in (8): (8) ntiqaad-u
r-rajul-i
5
c
l-masruu -a
criticizing-nom the man-gen the project-acc The man’s criticizing the project.
Assume that ev licenses an arg in its Spec. Arg is then «internal» in unaccusatives, and «external» in transitives and unergatives. The internal or external positioning of argument is semantically motivated (as shown by Krazer 1996). Suppose then that ev is merged with an internal argument (let it be ev1), and assigns it an internal theta role (= i-th). In the case of transitives, a second ev is merged with the external argument (let it be ev2), and assigns it an external theta role (= e-th). If this view is correct, then transitivity is derived through the occurrence of two ev’s, one of each licensing one of the two arguments of the transitive verb. Another way to put it is to say that the bivalent (or complex) argument structure of transitives is (bidirectionally) correlated with the bivalent/complex structure of ev’s.
The bidirectionally of complexity between args/Specs and evs/heads is questionable, however. For example, unergatives might be taken to be 6
headed by diadic evs, although they merge only with one e-Spec. Likewise, some causative and anti-causative predicates might be analyzed as merging only an internal Spec, although they are headed by diadic evs. If this is so (as
5
I take here a neutral position with respect to the category label of these nominalizations. Another plausible option is an early introduction of v, which is later nominalized in the structure (a position which is adopted by Marantz 1997, among others, and criticized by Alexiadou 2000). It might be that v is a canonical realization of ev, as suggested in work by Hale & Keyser (passim), and Fassi Fehri (1993), but this picture has to be refined to distinguish event nominalizations from «pure» verbs, presumably through structure heights.
Abdelkader Fassi Fehri
15
I will argue), then the projection of an extra syntactic functional layer (or ev2) does not force transitivity. Furthermore, if v has the morpho-syntactic property (c), i.e. that of category forming, then it is not clear how (a) and the latter property combine. In other words, we are in need of a connection between the category forming property and the complexity property. I take these two problems (the lack of direct correlation between ev complexity and arg complexity and the lack of correlation between category forming and complexity forming) to undermine seriously the external v theory of transitivity. The objections raised against (the external) little v (or vVP) as a source of transitivity carry over to the VP shell theory (proposed by Larson 1988) and the VoiceP (advanced by Krazer 1996) if they are construed as structural sources of transitives. Voice (applied to L-syntax, to determine the «lexical» internal/external diathesis, or to S-syntax in Hale & Keyser’s (1993, 1997) sense can be viewed as the source of hierarchical organization of arguments (= HOA), which reflects argument complexity only indirectly, at different heights of functional structure (cf. infra subsection 4.3 for further discussion).
2. Number Theory
2.1. Ingredients of NbT
Any theory of transitivity should be able to take into account the following properties:
6
I interpret Krazer’s proposal in this sense, since the source of e-th for the external arg necessarily involves internal binding into the lower V (via what she calls «event identification»).
Causativity, Transitivity, and Iterativity as Pluralities
16
(i) the complex nature of arg/Spec structure in transitives (ii) the (potential) complex nature of ev’s/H’s (iii) possible relations between complex Specs (= c-Spec) and complex heads (= c-H), with no necessary bidirectionality (iv) relations between the use of ev/v as «category» forming, and its use as complex predicate forming. I claim that complexity relations described in (i)-(iv) can be captured by Number Theory (= NbT). NbT is based on the following hypotheses: (a) Nb (with Pl value) is the source of complexity (b) Nb is a property of H/ev or Spec/arg (or both, hence “distributed plurality“, with potential multiple Specs) (c) Pl Nb when applied to H or Spec can be realized as single or multiple Hs or Specs, depending on whether it is non-distributive (collective) or distributive. (d) various cases of complexity are instances of various Nb configurations (e) Voice may affect complexity, but is not the crucial factor in determinig complexity.
2.2. Verbal plurality and distributed Nb
Arabic FII is acknowledged to have the characteristics of event plurality. Medieval Arabic grammarians associate it with the essential meanings of taktiir “multiplicity“ and mubaalagah “exaggeration“ (see e.g. Astarabadii). Western philologists translate these two meanings as “intensive“ and «extensive», respectively (see e.g. Caspari/Wright 1898). Greenberg (1991)
Abdelkader Fassi Fehri
17
has shown that this form (as well as the Semitic Piel) has the basic characteristics of verbal plurality, as originally stated by e.g. Swadesh (1946). The latter consist in:
7
(a) reduplication (although partial, instantiated e.g. by gemination of the second consonant) (b) temporal repetition (or repetitive action or event) (c) plural action on/by many (d) “natural ergativity“ (i.e. action on many with transitives, and action by many with intransitives). The two first characteristics are easily observed in the so-called «intensive» repetitive FII. They are illustrated in (5a) and (5b) above.
Action on plural objects is illustrated by the following constrasts: (9) a. jarraha
l-junuud-a
wounded.intens the-soldiers-acc He wounded many (of) the soldiers. b. jarraha
r-rajul-a
wounded.intens the-man-acc He inflicted many wounds (to) the man.
(10) a. gallaqa-t
l-?abwaab-a
closed.intens-f the-doors-acc She closed (locked) many (of) the doors. b. gallaqa-t
l-baab-a
closed.intens-f the-door-acc She closed (locked) the door firmly.
7
Greenberg (1991) takes (c) and (d), which I have given as distinct, to be the same property. Also, I do not discuss here another characteristic of plurality, namely «spatial dispersion».
18
Causativity, Transitivity, and Iterativity as Pluralities
The difference between these constructions is that examples (a) have, in addition to the repetitive/intensive reading, an interpretation which examples (b) cannot have, namely the interpretation of distributed action on many objects. This interpretation of the so-called intensive is sufficiently documented for Standard Arabic, Arabic dialects, and other Afro-Asiatic languages (as reported e.g. in Greenberg 1991).
Action by many (or plurality of subjects) is not so documented. Greenberg (ibid) provides, after Caspari/Wright (1859/1973), the contrast in (11): (11) a. baraka
l-jamal-u
kneeled the-camel-nom The camel kneeled. b. barraka
c
n-na am-u
kneeled.intens the-drove-nom The (whole) drove (of camels) kneeled.
Let us take the property (b), i.e. temporal repetition (or repeated action), to be an instance of Pl Nb on verbs (or evs, following in essentials the Arabic tradition and Greenberg), marked here by reduplication. The question then is how this property relates to property (c), i.e. to plurality of action on (or by) many? A natural way to relate these two properties is to take advantage of their configurational positioning: temporal repetition arises from Nb operating on ev/H, and action on/by many from Nb operating on args/Specs. In other words, Pl Nb applies to the H/Spec domain (a minimal domain formed by Merge), and can be a property either of H/ev or its Spec/arg. I will use the expression distributed Number to designate the placement of Nb in the second instance, and non-distributed in the first instance.
2.3. Distributed plurality
Abdelkader Fassi Fehri
19
Greenberg takes the two properties (b) and (c) (as well as the ergative character of verbal Nb, stated in (d)) to follow from the distributive nature of Nb, marked prominently by gemination. «[If …] the expression of distributive plurality in the verb has as its original sound symbolic meaning «temporal repetition», then the transition to the meaning «plural object» is not so difficult» If one does something repeatedly, he adds, then «… it is highly likely to be action on separate objects, hence to involve a distributive plural». (p. 584).
Greenberg then goes on explaining why (d) holds: «If one or more agents act on a plurality of things, the action on plural objects is almost certain to be viewed distributively as separate action on a number of distinct objects. On the other hand, separate agents can only carry out the same act on a particular plurality of objects by a concerted action, thus collectively. For an intransitive verb, there is no object toward which action is being directed, each subject will in the usual case be acting independently» (pp. 583-4).
8
I will leave at the moment aside the natural ergativity question, taking it to be a different question from that of the distributed question of Nb, as instantiated by properties (b) and (c). My implementation of the instances of distributed Nb will take into account this separation. On the other hand, while I agree that the distributive/collective dichotomy is relevant for accounting for the various configurations of Pl Nb, I will be using it differently, though, to distinguish between various Pl H’s and Spec’s, which are realized either as multiple discrete segments (giving rise to distributivity), or as single segments (which are non-distributive or collective). In repetitive events, I take H Pl to be
8
It is worth noting that he keeps the causative meaning as separate from that of the intensive, thus postulating no connection.
Causativity, Transitivity, and Iterativity as Pluralities
20
«collective» or non-distributive, in the sense that the Pl ev is realized by a single H. I also take action on many or by many in the examples above to be an instance of weak distributivity applied to Specs (or args), since I assume that the plural interpretation obtains in a single Spec. In this view, the distributed Pl in H or Spec is not taken as (strongly) distributive in both cases. I then have to look for cases where strong distributivity applies to H and/or Spec.
2.4. Causative complexity, verbalization, and distributivity
As I have said above, the introduction of glottal prefixation (FIV) may or may not result in transitvization. This variation depends on the nature of the root to which this form applies. If the root is non-verbal, then the form «verbalizes» it (without adicity increase), but if the root is already «verbal», then the outcome is transitivized. We then have to look for how these two properties of the form can be related, although they do not appear to be so at first glance.
The verbalizing property of the glottal prefix /?/ is supported by its ability to create «verbs» from non-verbal roots, thus forming denominal intransitives, as in (12): (12) ?a-nbata
l-haql-u
inch-planted the-field-nom The field became with plants (started having plants). Likewise, the same prefix verbalizes roots which can be thought as adjectival, or prepositional, as illustrated by (13) and (14), respectively: (13) ?a-xbat-a-t tough-ed-f
l-mar?at-u the-woman-nom
The woman became tough.
Abdelkader Fassi Fehri c
(14) ?arba -a
21
r-rajul-u
spring-ed the-man-nom The man came to be in spring. The transitivizing property is illustrated by the construction (1) above, compared to (2), repeated here as (15a) and (15b), for convenience: (15) a. ?-axraja
r-rajul-u
l-walad-a
caus-went.out the-man-nom the-child-acc The man made the child go out. b. xaraja
l-walad-u
went.out the-child-acc The child went out. In this case, the outcome of the prefixation is necessarily transitive. Why is that so, and how is transitivization forced?
It is reasonable to think that transitives like (15a) have both double or complex heads (= c-Hs) and double or complex Specs (= c-Specs). The c-H is morphologically realized as/?-V/, where/?/ can be taken as another discrete instance of ev (or V/v). In other words, /?/ can be taken as reduplicating ev/V, and hence pluralizing it. The Spec complexity is instantiated by the existence of two args, an indication of another form of «reduplication» or plurality. Suppose then that the complexity arises from the reduplicating plurality in both cases. Since the c-H is not composed of identical evs, and the c-Spec is not either, I take plurality to be of «distributive» character in both cases (with distributive c-Hs and distributive c-Specs). Thus the application of this prefix morphology to already formed «verbs» results in creating a reduplicated «verb» configuration, which is read as distributive (by virtue of having two discrete morphemes for the two «verbs», although they form a single c-H. Likewise, the realization of discrete Specs in transitives is an indication of the distributive character of Spec Nb.
Causativity, Transitivity, and Iterativity as Pluralities
22
If this is true, then /?/ prefixation is read off as reduplication of ev/V, although it is not a straightforward mark of reduplication. However, despite the fact that /?/ prefixation results in plurality, it cannot be interpreted as «intensive», i.e. as denoting temporal repetition, for example. This follows automatically if /?/ if taken to be (necessarily) interpreted as distributive (due to its «discrete» nature), whereas the gemination morphology of FII is not (necessarily) so interpreted. Hence FII is normally «intensive» because reduplication here is collective. Consequently, we derive the core «meanings» of FIV and FII, acknowledged in the traditional literature, namely that FIV has a transitivizing property, and FII an «intensive» property (see Fassi Fehri, passim, on these core properties). But as also acknowledged in the literature, matters are more complex, because FII can also be a transitivizer. I turn to this matter in the next section.
2.5. Two sources of transitivity
I have shown in the last section that transitive FIV arises from plural distributivity applied to both (verbal) Hs and Specs. On the other hand, we have seen that FII does not necessarily yield to transitives when it applies to verbal roots, as in (5b) above, compared to (6b). The latter constructions are repeated here as (16a) and (16b), for convenience: (16) a. jawwala
r-rajul-u
took.walks.intens. the-man-nom The man took a lot of walks. b. jaala
r-rajul-u
took.walks the-man-nom The man took walks. The reason (16a) has no extra argument, compared to (16b), can be attributed to the fact that reduplication marks collective plurality of the event. Collective
Abdelkader Fassi Fehri
23
plurality of evs is interpreted as c-H (read collectively, hence «intensitivity»), without necessarily inducing c-Spec. But there are cases in which FII results in transitivization when applied to verbal roots, as exemplified by (3) above, compared to (4). Likewise, there is a version of FII in (16), which is transitivized, as illustrated in (17): (17) jawwala
r-rajul-a
made.took.walks the-man-acc He made the man take walks. If reduplication is a mark of verbal plurality, how can it be the source of both (16), the «intensive», and (17), the «causative»? In other words, can the intensive and the causative properties be related? I would like to suggest that the answer is positive. In order to establish this relation, we need to establish (a) how FII can behave like transitivized FIV, and (b) how FII can be intensive intransitive, whereas FIV cannot be.
I have shown that FII is a plural form. When Pl is read collectively on H, intensive results. Suppose now that Pl on FII can also be read distributively (on Spec), just like what happens with Form IV. Then transitivity is yielded in the same way. In other words, Pl can be read ambiguously with FII, thus giving rise either to (16) or to (17). The distributive/collective ambiguity of Pl of FII recalls here the ambiguity found in the system of universal quantification described in Fassi Fehri (1999). There it is shown that under specific structural conditions the universal quantifier kull can be ambiguously interpreted as distributive or non-distributive. The ambiguity of kull is exemplified in (18) and (19): (18) ?akala kull-u ate
r-rijaal-i
dajaajat-an
all-nom the-men-gen chicken-acc
All the men ate a chicken. (19) ?akala kull-u ate
rajul-ayn-i
dajaajat-an
all-nom man-dual-gen chicken-acc
24
Causativity, Transitivity, and Iterativity as Pluralities Every two men ate a chicken.
The quantifier in (19) is strongly distributive in the sense that there should more than one chicken, and each two men ate one. It is collective in (18), in that there could be only one chicken eaten by all the men, or only weakly distributive, in the sense that that the existence of more than one chicken is not required by the interpretation. The transitivizing property of FII is hardly questionable, and Arabic dialects have simplified their conjugation form system by doing without Form IV, and transfering its functions to Form II (which then becomes the unique form for expressing both intensive and causative meanings). If this is true, both the 9
ambiguity of FII and its transitivizing property come as no surprise. We have then answered question (a), and half of the question (b). To complete the
9
In fact, this picture has to be refined by taking into account the considerations brought up in Fassi Fehri (1996 a &b), in particular the fact that transitive FII selects a stative (adjective or noun) as an internal complement, whereas FIV selects non-stative verbal complements (see doublets there and their analysis). For example,?a-taala-hu «he lengthened it» means «he made it lengthen» (which can presuppose that it was long, but was made longer), whereas tawwala-hu means «he made it long» (presupposing that it was short). In other words, transitive FII induces a change of state of the complement, but FIV does not. But the picture is more complex, since transitive FII can also select non-statives, in addition to statives. When this happens, it is understood as «coercive» or implying the use of energy or force by the causer, against the will of the internal agent. As a matter of fact, the subject of the internal verb controls the action with FIV, but that with FII it does not. Compare the following pairs: (i) ?a-ktaba-hu «to make someone write», kattaba-hu «to make someone write by using force, or against will» (ii) ?arqasa-hu/raqqasa-hu «to make someone dance (forcefully)» (iii) ?axraja-hu/xarraja-hu «to make someone go out (forcefully)» On the other hand, intensive FII, when applied to intransitive or transitive verbs, can only select non-stative verbs, as exemplified by (5a) and (5b) above. The geminate counterpart carrafa-hu (bi-hi) of carafa-hu «he knew it», for example, cannot mean «he knew him intensively», but rather «he made him know (it)», i.e. the causative of the transitive. There is no intensive of statives. The latter restriction follows presumably from the semantics of verbal plurality. The detailed treatment of these questions is beyond the aims of this paper, however, and further research on the matter is needed (but see section 4 below for additional discussion).
Abdelkader Fassi Fehri
25
picture, we have to answer the other half of question (b), namely the question why FIV cannot be intensive. In order to do so, we have to look into more details of verbal plurality morphology.
2.6. Parallel plural morphology
Up till now, I have assumed that (consonantal) gemination/reduplication, on the one hand, and glottal prefixation, on the other hand, are marks of verbal plurality, and they also mark the distributive/non-distributive distinction of plurals. Since I have established the parallel with nominal quantification for the latter property (or feature), it would be a welcome step to establish parallels for the other features/properties of plurality. Such parallels can be established (and are in fact partly acknowledged in the literature) by carefully examining the morphology of plurality in the nominal, the adjectival, and the verbal systems. I will point to some (presumably) non-accidental similarities here, leaving aside the details for further research. Typically, glottal prefixation as well as gemination (both consonantal and vocalic) play a role in the nominal and the adjectival systems which parallel its role in the verbal system. Consider first glottal prefixation. There is a class of the so-called broken plurals which is taken to denote small sets (or number of individuals, from c
three to ten normally), named jumuu l-qillah «paucity plurals» by traditional grammarians, and which have the essential common property of being headed by the glottal prefix /?/. These paucity plurals are illustrated in (20): c
c
(20) a. ayn «eye»?a yun; di?b «wolf»?ad?ub; b. yawm «day» ?ayyam; zaman «time» ?azmaan; c
c
inab «raisin»?a naab; tiql «heaviness»?atqaal; c
c
burj «tower» ?abraaj; unuq «neck»?a naaq c
c
c
c
c. imaad «support»?a midah; ta aam «food»?at imah; ragiif «flat loaf of bread, roll»?argifah; burj «tower» ?abrijah; c
c
amuud «column»?a midah
Causativity, Transitivity, and Iterativity as Pluralities
26
I take /?/ to be the essential morphological segment for marking plurality in those nominal forms. As can be easily observed, vowels of the singular are not transfered to the plural in any significant sense. For example, vocalic change is not regularly «apophonic».
10
Whenever vocality is relevant, it has to
do with lenghtening/geminating the vowel (and opening it in the case of ?aCCaal), a property to which I return. It is clear, however, that the unifying feature of these forms is glottal prefixation, and that the latter morphology parallels that of the «verbal» plural causative of FIV.
11
Consider now gemination and/or lengthening. Consonantal gemination is involved in the so-called «forms of exaggeration» (siyag al-mubaalagah) by Arabic grammarians. These forms designate normally the subject of the event which repeats «exaggeratedly» the action, as in the following examples (21) ?akkaal «a lot eater», sarraab «a lot drinker», massaa? «a lot walker», qarraa? «a lot reader», kaddaab «a lot lyer», nawwam «a lot sleeper», sikkiir «a lot drinker (of alcohol)». The translation with a quantity adverb (or adjective) is a mere approximation. I take the gemination as expressing an «intensitivity» applied to an adjectival subject.
10
12
For various relevant versions of apophony paths and their motivation, see Kurylowicz (1973), Seghrouchni (1989), and Guerssel & Lowenstamm (1996), among others. Clearly, apophony paths are involved in the derivation of various broken plural forms, the vocalism of which is presumably morpho(no)logical, compared to that of single forms. 11 Further questions might arise with regard to how far parallelism goes, and how glottal prefixation can be related to vowel lengthening. I return to these questions below, some of which are presumably too premature to answer at this stage, although they are worth pursuing. 12 It is worth noting that these subject forms cannot be «derived» from the verb, in the traditional sense of derivation, since the corresponding intensive verbs do not exist to my knowledge (cf. e.g. * sakkara r-rajul-u «the man drank a lot», *?akkala r-rajul-u «the man ate a lot»). Consonant gemination does occur in broken nominal plurals, which are normally of adjectival forms (e.g. saa ?im «fasting» suwwam/suyyam, saajid «prostrating oneself» sujjad, but also kaatib «writer» kuttaab «writers», daabit «officer» dubbaat «officers», where the plural is also
Abdelkader Fassi Fehri
27
Other intensive forms involve vocalic gemination (or lenghtening), as in the following examples: (22) sakuur «very grateful», waduud «very friendly», kasuul «very lazy, jahuul «very ignorant», waluud «giving birth to many children, fertile», c
saruub «big drinker», ?akuul «big eater», samii «big listener», c
qadiir «very apt», aliim «big knower», sariib «big drinker». As far as I can tell, these forms do not designate active subjects, and they behave more like stative adjectives, denoting properties (and indicating high degree or gradation).
13
Vowel gemination (or lengthening) appears also as an essential building segment of a class of broken plurals: (23) a. di?b «wolf» di?aab; jabal “mountain” jibaal; rumh “lance” rimaah; rajul rijaal b. qayd “constraint” quyuud; jisr “bridge” jusuur;
induced via vowel lengthening). It seems then that verbs are not to be distinguished from nouns via consonant vs. vowel lengtheing, especially when further contrasts are taken into account (cf. infra). Moreover, since traditional grammarians also take exaggeration forms to be subject forms, the question arises as to how their derivation proceeds. I return to the question of the derivational source below. 13 Questions arise with respect to the difference in interpretation of doublets or triplets found in these lists, as well as to why there are no intensive adjectival objects. The modern lexicon tries to determine distinct uses of the object and the subject functions, as the following constrasts indicate: (i) rajul sarraab «a big drinker man» (ii) * maa? sarraab «a big drinking water, i.e. a drinkable water» (iii) maa? saruub «a big drinking water, i.e. a drinkable water» (iv) ?? rajul saruub (v) rajul jahuul «a very ingnorant man» (vi) * rajul jahhal «a very ingnorant man» These contrasts suggest that exaggeration forms in (21) apply to unergative subjects, whereas those in (22) normally applies to ergative subjects (or understood objects). I return to some of these questions below.
28
Causativity, Transitivity, and Iterativity as Pluralities namir “tiger” numuur; burj “tower” buruuj
These plurals are taken to be less marked, compared to paucity plurals exemplified in (20). They are interpreted as «plurals of abundance» only when contrasted with the latter, otherwise they are just plurals, with no mere specification (cf. Caspari/Wright (1859/ 1973)). They can also be contrasted c
with «plurals of plurals» (muntahaa l-jumuu ), as exemplified in (24):
14
(24) a. qawl «saying» ?aqwaal ?aqawiil b. jamal «camel» jimaal jamaa?il c. kuttaab «Koranic school» kataatiib; jumhuur «public» jamaahiir; diiwaan «cabinet» dawaawiin
The vowel gemination has a verbal counterpart in FIII, which according to traditional grammarians has the meaning of «participation» (musaarakah). The latter form is exemplified in (25): (25) a. maasaa-hu walked.part.-him He walked with him (or: he took the same path as him). b. saabaqa-hu speeded.part-him He raced him. c. saaraba-hu drank.part.-him He drank with him. What «participation» means in this case is the fact that both participants are interpreted as actors of the action (or the event). But although the event is the same (e.g. «the race»), it has plural participants. The plurality of participants
14
These plurals involve double gemination of the vowel, or a double pluralization process (with various forms, arguably related to a much simpler plural form). But the important idea (originally
Abdelkader Fassi Fehri
29
(the «nominal» part of the verb meaning) appears then to be distributed on Pl Specs. FVI, the reciprocal form, plays the role of unifying these two subject participants: (26) ta-saabaqa zayd-un
c
wa- amr-un
recip.-raced Zayd-nom and-Amr-nom Zayd and Amr raced each other. But consider the following pair:
(27) saabaqa zayd-un raced
c
amr-an
Zayd-nom Amr-acc
Zayd raced Amr. c
(28) saabaqa amr-un raced
Zayd-an
Amr-nom Zayd-acc
Zayd raced Amr. What is the difference between the two? Although the interpretation of both (27) and (28) appears to involve reciprocal action, this is not to be confused with the reciprocal. In FIII, the subject appears to be more active than the object, and the two sentences are not equivalent to the reciprocal FVI in (26). With the latter form, both participants are «subjects», while they are not so with FIII. Since both participants are understood as «subjects» of the «same» event, I will take these subjects to have «fissioned» into subject and object positions, been licensed by Pl Spec. If vowel gemination is a functional Pl head, then it is satisfied through Pl Specs in this case.
found in traditional grammar) is that these forms are semantically «intensive» plurals, because they involve «double» plurality.
30
Causativity, Transitivity, and Iterativity as Pluralities Suppose this is so. Then the question arises as to why this plural cannot be
read as intensive. This situation recalls that found with FIV, which cannot be interpreted as intensive either, being only distributive. Now surprisingly and interestingly, FIII is also only distributive.
Interestingly enough also is the fact that FIV and FIII can be, and in fact have been, morphologically related. McCarthy and Prince (1990), adopting Levy’s (1971) view, propose that the two forms are related through a local metathesis of the open vowel in the first syllable. Inspired by this relatedness view, I will assume, however, a more morpho-syntactic approach of the matter. Suppose the two forms are related by virtue of being both marked as Pl, realized by the long vowel [aa], and that this long vowel fissions in FIV, in the sense of Fassi Fehri (1996c), i.e. it has two segment head (s), instead of one. In other words, FIV has more structure than FIII, i.e. one extra head up. The two segments of the Pl head in FIV then license «discrete» evs, whereas the Pl head in FIII does not; it licenses only one ev head, hence the «participation» meaning. By contrast, FIV has no such a participation meaning, since the two participants are subjects of different events. Put simply, we take FIII to have only distributive Spec, but no distributive H, whereas IV has both.
Let us turn now to the difference between FII and FIII plurals. We have seen that FII can be intensive or transitive, and we have interpreted its transitivity as distributivity of Spec, and its intensitivity as collective Pl on H. The question then is why FIII cannot be intensive (and intransitive), just like FII is. One possibility is to take its distributivity to be encoded «lexically», much like what happens with quantifiers each and all in English.
15
But there is no
evidence for such lexical/vocabulary encoding. Another possibility is to
Abdelkader Fassi Fehri
31
explore a difference in marking Pl in FIII (and FIV) on the one hand, and FII, on the other hand.
FII implicates the second syllable of the word in encoding Pl as gemination. FIII (and its derived «daughter» FIV), on the other hand, encodes Pl gemination on the first syllable. In the nominal system, Pl is only encoded in the second syllable. Suppose then that Pl is normally encoded on the second syllable only, and as suggested earlier, «verbal» plurality in FII is encoded on consonants, whereas «nominal» plurality is encoded on vowels.
16
When
plurality is encoded on a single segment H in the second syllable, we get the collective/intensive interpretation. Now the fact that the Pl is found in the first syllable in FIII is somehow unexpected. This «displacement» might be in fact an indication that movement has occurred internally, thus creating a two-segment head, which then licenses two Specs (or distributive Spec), hence the transitivity. The absence of the intensive reading appears to be correlated with this displacement, much like what happens with transitive FII and FIV.
17
The difference with FIV is that the latter has no participation/interaction meaning. I have interpreted the participation configuration as arising from split/distributive Specs, but not distributive Hs. By contrast, transitive FIV has both distributive Hs and distributive Specs. If FII, FIII, and FIV are
15
See e.g. Belleghi & Stowell (1997) for such a lexical approach. As I explained earlier in footnote 11, the consonant/vowel gemination contrast cannot be taken as a distinctive category feature. For example, finite verbs of FII and their participles have geminated consonants (as in haddada «to threaten», muhaddid, muhaddad), but in action nominalizations the gemination is vocalic (tahdiid «threatening»). 17 The intransitive couterpart of FIII appears to be FXI, which denotes in fact intensivity/plurality. Thus we have hmaarra, which means «to become very red», compared to FI hamira «to be red», or FIX hmarra «to become red». It is usually thought that FXI is derived from FXI by vowel 16
32
Causativity, Transitivity, and Iterativity as Pluralities
morpho-syntactically related, then if FIII «derives» from FII via one step movement «inside» the consonantal stem, then FIV is «derived» from FIII by «peripherical» movement, outside the consonantal stem. This further step and splitting movement which occurs in FIV, coupled with presumably more structure (and height), is what is behind double distributivity found in FIV, which excludes both the intensive/collective meaning (by virtue of fissioning like FIII in a first step), and the participation meaning (by virtue of fissioning to the periphery).
18
2.7. Summary
I have shown that consonant gemination in FII expresses multiple or repetitive action, interpreted as event plurality. The plurality of action can also be interpreted as operating on multiple individuals, hence the argument multiplicity. Thus the plurality/multiplicity, which is encoded on the verb, can (ambiguously) apply to the event or to the arguments. Vocalic gemination in FIII expresses plurality of participants (or arguments). But the participants are not treated as “equal”, hence the transitivity. The latter is a reflection of a partitioned plurality of participants. But in the reciprocal FV, the participants are assembled, hence the intransitivity, even though the event is plural. Glottal prefixation in FIV, analyzed as a fissioned vocalic gemination, also expresses partitioned plurality of participants, hence the transitivizition. FII may also
lengthening. Since FIX expresses inchoativity, then FXI expresses both the latter meaning and the intensive meaning. 18 It is possible to view the reciprocal participation meaning in FIII coupled with its obligatory transitivity as arising from a sort of (a fissioned) cumulative Pl, which licenses two Specs, but not two events (e.g. in the race case, the race has to be one and the same). On the other hand, it is worth pursuing the matter of parallelisms much further. For example, normal plurals and plurals of plurals exhibit internal stem processes, thus paralleling FII and FIII, but paucal plurals exhibit an external stem process, thus paralleling FIV. On the other hand, only verbal forms seem to exhibit internal fissioning, as in FIII. The counterparts of FIII and FIV are also found in the reflexive system through FV and FVII, respectively (cf. infra subsection 3.2).
Abdelkader Fassi Fehri
33
express plurality of participants, and become transitive through partitioned plurality, which alternates with the assembled plurality in the repetitive. FII then contrasts with FIV in this respect, since the latter cannot be an assembled plural. If the inchoative FIV is conceived as a distributive plural over events, and the causative as a distributive over arguments, then the alternating semantics of FIV is limited to distributivty. Thus Pl (encoded via consonant or vowel gemination) is behind intensive and transitive meanings, via fissioning movement and (various kinds of) distributivity interpretation. Complexity of Hs can be separated from complexity of Specs, and H distributivity separated from Spec distributivity. Event/argument complexity issue can be kept separate from HOA, a point to which I return in subsection 4.3.
19
3. Cross-linguistic evidence Up till now, I have provided reasons to think that the Arabic (or Semitic) verb system makes significant uses of plural quantification, which not only parallels that of nominal quantification, but interestingly enough, enables us to derive the «intensitivity» of verbs, the complexity of event/argument structure (e.g. in transitives or causatives), and also accounts for alternating (or
19
There are no doubt other sources of transitivity (or intransitivity) which I do not address here, but which can be probably subsumed under predicate complexity, dictating the occurrence of distributive Specs (see Fassi Fehri 1996 a & b for details). For example, apophonic [a] is used to form denominal transitive verbs (compare hazina «to be sad» with hazana «to put sadeness into someone»), iambic reduplication to create «attenuation» forms of verbs like samsama «to sniff something», compared with samma «to smell something», xanna «to speak through the nose or to nazalize», xanxana «to speak through the nose or to nazalize repeatedly», xasxasa «to make rustle or rattle», xarxara «to snore», gargara «to gargle», etc. In these reduplicating forms, each part can be taken as a small part of event, and the event can be seen as somehow «fissioned» or partitioned over the two pieces of the root. Intransitivity or transitivity does not depend, however, on this process, hence confirming our view that event complexity should be kept separate from adicity complexity. Note that the fissioning event can be thought as a sort of paucity plural, in that the whole event is made of «small» or «paucal» (parts of) events.
34
Causativity, Transitivity, and Iterativity as Pluralities
multiple) behaviours of verbal Nb morphology (intensive and causative, causative and “verbalized” inchoative, etc.). It would be a welcome step to see how such a view holds for other languages.
As a matter of fact, a number of convincing descriptive studies have already established the correspondence between reduplication morphology and (a) causative or transitive formation, (b) verbal plurality, or more generally event quantification, and (c) variable behaviours and meanings of the same form. For example, Moravcsik (1978a) has focused typically on the (b) relation, investigating various forms of reduplication and «… the meanings of increased quantity, intensity, diminution, and attenuation which are concepts capable of pulling together many superficially disparate uses of reduplication constructions…». Similarly, Greenberg (1991) has especially highlighted the relevance of the (b) correlation with respect to Semitic intensive forms. The two studies have in common that they do not have an answer to the question of how the properties (a) and (b) can be connected. But other compelling work, however, provides rather strong empirical evidence for correlating the (a) to (c) properties (see in particular Kulikov 1993, 1999, and Dolinina 1999, and the many references cited there), typically outlining the role of reduplication (or other morphologies) in causative and transitive formation, and the extensional use of this morphology to express various (b) meanings. In 3.1, I first examine some of the relevant data and analyses which provide motivation for this direction of research, and in 3.2. I discuss Moravscik’s resistant cases to her view of reduplication, which in fact add further support to the correlation of the (a) to (c) properties.
3.1. Causatives, transitives, and event quantification
Abdelkader Fassi Fehri
35
Crosslingustically, morphological reduplication (in particular) is used to form not only plural verbs (interpreted as repetitive, expressing multiple actions, or multiple participants, distributivity, etc.), but it can also be used for forming single causatives or transitives, double causatives, or exhibiting multiple behaviour meanings.
3.1.1. Causativization and transitivization
Causative or double causative formation through reduplication is well-documented in the literature. Consider the following constructions brought up by Kulikov (1993) from various sources: (29)
a. wanu “to die” b. wanu-chi “to kill” c. wanu-chi-chi “to cause to kill”
(30)
a. rupit(a) “to work” b. rupita-pt(u) “to cause to work” c. rupita-ptu-pt(u) “to ask to work”
(31)
a. ut’ “to sleep” b. ut’-k’ “to make to sleep” c. ut’-k’-ek’ “to cause to make to sleep”
In Huallaga Quechua (29) (taken from Weber 1989), the causative morpheme “chi” forms a transitive causative in (29b) and, when reduplicated, a double causative in (29c). In Mansi Vogul (30) (quoted from Robandeeva (1973)), the morpheme “pt” marks the causative in (30b), and is reduplicated to mark double causation in (30c), with a “requestive” meaning (see below subsection 4.3.4). Finally, in Hunsib Daghestan (31) (taken from Isakov (1986)), the
Causativity, Transitivity, and Iterativity as Pluralities
36
causative morpheme “k” in (31b) reduplicates to form the double causative in (31c).
20
This data provides clear evidence that reduplication is involved in
causative formation. When a morpheme is used to causative a verb, then its reduplication forms a double causative. Reduplication is not of the root, but of the external morpheme. It is worth noting that in these cases, the causative morpheme (as described in these studies) does not originate from an intensitive or other quantificational feature source, and that reduplication of this the morpheme provides the double causative meaning.
Moravcsik (1978a) lists (partial) reduplication as a source of transitive formation, a case that she takes as resistant to her view of reduplication (see the next subsection). She brought up the following verbal pair from Sundanese, after Robins (1959), to illustrate the case: (32) a. narerewas “to frighten” b. rewas “to be afraid” Here, the reduplication is root-based, as in the Arabic cases. Both root-based and morpheme-based reduplication appear then to play the role of transitivization or causativization.
3.1.2. Multiple behaviour Intimate relationships between causative and “aspectual” meanings of causative morphemes have been repeatedly noticed in typological studies (see in particular Nedjalkov 1966 and many other references cited in Kulikov 1999). Various types of polysemies of these morphemes have been observed.
First, in Turkish, either causative or intensive/iterative interpretation of a given form can be acceptable in precisely the same context. Thus double
20
See Kulikov (1993) for these references.
Abdelkader Fassi Fehri
37
causative morphemes may refer either to double causation, or to intensive/iterative causation (Zimmer 1976, after Kulikov, ibid): (33) Müdür-e
mektub-u ac-tïr-t-tï-m.
director-dat letter-acc open-caus-caus-Past-1sg a. I had someone make the director open the letter. b. I made the director open the letter forcefully. Second, in Yukaghir, the verbal suffix –t- expresses cumulatively both causative and distributive meanings (Maslova 1993): (34) a. sel’ge-t “to break several distinct things” b. joyê-t “to open (tr.) several times” Similarly in Aleut, the causative –dgu- instantiates the cumulation of causative and distributive meanings (Golovko 1993; see (4”) below). Third, in Bouma Fijian, the prefix va’a forms causatives with some verbs and intensives with others (Dixon 1988; after Kulikov ibid): (35) a. vuli(-ca) “learn, study” b. va’a-vuli(-ca) “teach” (36) a. mate “die” b.va’a-. mate-a “kill” This distribution is strictly parallel to that of Arabic, in which FII marks causatives with some verbal roots, and intensives with others: c
c
(37) a. aduma “be great” addama “take as great, make great” c
c
b alima “learn smth” allama “teach sb ” (38) a. kasara “break sth” kassara “break sth into pieces” c
c
b. qata a “cut sth” qatta a “cut sth into pieces”
Further polysemies are provided by Kulikov (1993), who observes that in Tuvan, the double causative can function as iterative of the first causative: (39) asak
Bajyr-ga
inek-ti
dile-t-ken
Causativity, Transitivity, and Iterativity as Pluralities
38
old-man Bajyr-DAT cow-ACC look for-CAUS-PAST An old man caused Bajyr to look for the cow (several times). (40) asak
Bajyr-ga
inek-ti
dile-t-tir-ken
old-man Bajyr-DAT cow-ACC look for-CAUS-CAUS-PAST An old man caused Bajyr to look for the cow (several times). The causative morpheme can also express the plurality of certain participants, as in Carib, where a double causative morpheme serves inter alia for expressing the plurality of objects of causation, or causees (cf. Hoff 1981): (41) kaiku:si ?wa kisi:wopoi «Do not let him be killed by the jaguar» (42) kisi:wopo:poi kaiku:si ?wa «Do not let him be killed by all these jaguars» According to Hoff, the double causative po:poi is used because of the plurality of the causees.
Kulikov further observes that in same cases “… the boundary between iteratives to causatives and plural causatives is rather vague”. In Aleut, there are two constrasting causative morphemes, -t-, which denotes causation, as in (43b), and –dgu-, which is used mainly with plural causees, as in (43-c): (43) a. igluqa-r hide-SG
qaka-ku-r dry-NON.FUT-3SG
The hide is dry. b. ayagar
igluqa-r qaka-t-i-ku-r
woman-SG hide-SG dry-CAUS-Epenth-NON.FUT-3SG The woman is making/makes the hide dry. c. ayagar
igluqa-s qaka-dgu-ku-r
woman-SG hide-PL dry-CAUS-NON.FUT-3SG The woman is making/makes the hides dry.
Abdelkader Fassi Fehri
39
But Golovko (1993) argues that –dgu- “… renders not the plurality of the causee but rather the plurality of the caused events, thus expessing causativity and distributivity at the same time”.
3.1.3. Event quantification
Reduplication is used to form distributive verbs, as documented in various languages. Dolinina (1999) illustrates this by (44) from Classical Nahuatl (quoted after Suarez (1989)), (45) from Cree (after Horden (1934)), and (46) from Moses Columbian (after Kinkade (1977)): (44) (te t.tla)-maka
(te t.tla)-ma?-maka
to give sb.sth (45) makiw
to give sth to each person mu-makiv
he gives
he gives distributively
(46) leq-laq-
lx
lx
sit.Sg/Redupl/Dist-Activity Non-Sg/Agent Each has a position in which to sit.
In Russian, Ukrainian, and Polish, verbal prefixes are used for Subject, Object, and (so-called) Diversative distributivity (expressing motion from plurality of locations to single place). These meanings are illustrated in (47a) to (47c), respectively: (47) a. Butylk-i
po-pada-li
s
plok-i
bottle-PL:NOM DIST-fall-PAST:PL from shelf-GEN The bottles fell from the shell. b. Vona
pere-bil-a
she:NOM DIST-break-PAST:3SG:FEM She broke all the dishes.
uves’ posud all
dish.COLL.ACC
Causativity, Transitivity, and Iterativity as Pluralities
40 c. Ptak-i
z-leciat-y
Bird-PL.NOM
sie
do ogrod-u
DIST-fly-PAST.PL REFL in garden-SG.ACC
The birds flew (from different directions) to the garden. In Evenki (48), instead, a suffix encodes the multiplicity of locations (see Nedjalkov (1989), quoted in Dolinina, ibid): loko-t
(48) loko hang (smth)
hang (separately (and in different places))
Similar phenomena occur in Aleut, Eskimo, Turkish, Armenian, Japanese, Indonesian, etc. (Dolinina, ibid, p. 191 ).
Moltman (1997) observes that the same morphology may be used for iterative (collective) or distributive verbs. Thus in Copanaila Zoque, the idea of repetition is marked by a non-differentiated iterative-distributive prefix, which can have both readings (see Suarez 1983; Dolinina, ibid): (49) min-ge?t-u He came again/ he also came. But in Classical Nahuatl (Suarez, ibid, Dolinina, ibid), event plurality, marked by reduplication, expresses (only) iteration: (50) wi.tek wi?-wi.tek to hit
to hit repeatedly
Likewise, in Tepetotula Chinantec (Westley, 1991, Dolinina, ibid), a specialized affix marks repetition in time: (51) ki-ka-?nau?
za
ha?
REP-PAST-seek.TR.ANIM.3 3SG animal She/he repeatedly hunted the animal.
Summarizing this subsection, we have shown that causativity can be associated with distributivity or iterativity, or vice-versa. Furthermore,
Abdelkader Fassi Fehri distributivity, multiplicity of actions, or agents
41 is also associated with
iterativity.
3.2. Moravcsik’s resistant cases
Moravcsik (1978a) examines what she calls derivational uses of reduplication, and she observes that there is a «… large and indeed disparate set of examples of derivational meanings that cannot be seen related [our underlining; FF] to any of the non-derivational meaning categories surveyed…». Consequently, she concludes that «given that reduplication is neither the exclusive expression of any one meaning category in languages, nor are the meanings that it is an expression of all subsumable under general classes, no explanatory or predictive generalization about the meanings of reduplicative constructions can be proposed». But the list of resistant cases she establishes is precisely the sort of data which can be used to support our Nb theory of adicity (i.e. they come as «good news»).
Transitivization and intrantivization uses are typically among the cases which Moravcsik takes as resistant to her more or less quantificational picture (see the transitization case in Sundanese via reduplication she quotes from Robins 1959, i.e. example () above). But if transitivity is a form of abstract reduplication, as I have shown, these morpho-syntactic alternations represent exactly the kind of verbal behaviour that we expect. As a matter of fact, the Arabic counterparts of these pairs are related through glottal prefixation, which I take to be a more abstract form of the reduplicating verbal plurality, as argued earlier: c
(52)?a-fza a «to frighten someone» c
(53) fazi a «to be afraid»
42
Causativity, Transitivity, and Iterativity as Pluralities In the same direction, and also of great interest, the author mentions, after
Harrisson (1973, p. 415) and Christaller (1875, 64), respectively, that in Mokilese and in Tiwi, intransitives are derived from transitives, as the following contrasting pairs indicate: (54) a. koskos «cut (intr)» b. kos «cut (tr.) » (55) a. didi «eat (intr.)» b. di «eat (tr.)» Such a direction of derivation is also found in Arabic.
21
Furthermore, the more
general phenomenon of morphologically marking the intransitive with respect
21
This is what happens, for example, when FIV is used as a resultative (the so-called mutaawacah in traditional grammars), as in the following contrast, in which the more complex form is the intransitive: (i) basar-tu-hu announced-I-him I announced good news to him (ii) ?absara He rejoiced (at good news) Note that (i) is a more classical use of FII in (iii), whereas (ii) is an equivalent of FX in (iv): (iii) bassar-tu-hu announced-I-him I announced good news to him (iv) stabsara He rejoiced (at good news). Likewise, both members of the FIV pair in (v) and (vi) appear to be related to the simple intransitive in (vii): (v) ?a-bta?a r-rajul-a caus.slowed.down the-man-acc He made the man slow down. (vi) ?a-bta?a r-rajul-u caus.slowed.down the-man-nom The man was made (came) to slow down. (vii) batu?a r-rajul-u slowed.down the-man-nom The man slowed down.
Abdelkader Fassi Fehri
43
to the transitive is not that surprising cross-linguistically. In these examples, it is the intransitive which is reduplicated, a matter that could appear counterintuitive. However, more has to be said about the meaning of these intransitives. For example, these verbs may denote small partitioned events (or attenuation), as in Arabic. At any rate, what is important here is that we have substantial evidence from other languages that reduplication can be used for marking intransitive/transitive alternations, whatever the direction of the derivation is, and this is exactly the kind of data that we look for, to corroborate the view that transitives (or complex intransitives) are plurals.
22
With respect to derivational reduplication (our «verbalization»), Moravcsik mentions as related (to the general meanings of reduplication) denominal verbs (which she takes to have distributive meaning), deverbal agent or action nominalizations (with habitualness or continuity «connotations»), and denominal adjectives (with connotation of fullness of something). I will examine only the verbal case here, since nothing hinges for the point on the other cases. Quoting Watson (1966, 99), the author takes the Pacoh’s examples in (35) to instantiate distributive meaning: (56) pampe «to divide by three» (pe «three»)
Unfortunately, the translation is incomplete. Thus we do not know whether the verb is transitive or intransitive, and no explanation is given as to the
This variation is expected once complexity of event structure and that of argument structure are taken to be not directly correlated, and also when the various plurality effects are treated appropriately. 22 Among the «unrelated» derivational cases, Moravcsik enumerates adverbs which are derived from verbs or adjectives (as in Thai), and she also enumerates as unrelated the non-derivational use of reduplication as expressing perfectivity in Sanskrit, Greek, Latin, Gottic, and Germanic. These cases have to be analyzed in detail, but I do not see why they should be treated as «negative» cases, given the relation between Nb and Aspect, on the one hand, and adjectives (or
44
Causativity, Transitivity, and Iterativity as Pluralities
source of the «divide» meaning. For example, the Standard Arabic tallata (literally «to three» or «to triple») is normally transitive, and it can mean «to make two people three, by joining them», a cumulative meaning, or «to create three parts from a unique thing», or «to do something three times», or «to take off a third», etc. In Moroccan Arabic, tellet is used transitively and intransitively with the same meanings as in Standard Arabic, plus other extra meaning. Thus it can further mean, as intransitive, «to join two people to become three», or it can mean «to take three parts», or «do something three times», etc. None of these meanings are in fact distributive.
Moravcsik does not give any example of denominal verb reduplication with no «related» meaning. Had she done, she would have considered them «unrelated». The question is then whether such unrelated reduplicated verbs exist. As far as I can tell from the Arabic data, not only they do, but they are quite productive. Now, it might be that their existence without «aspectual connotation», as she puts it, is only a matter of neutralization. If that were the case, then maybe «Aspect» is behind verbalization. Reduplication and Aspect would then be correlated. But if that were so, transitivization would remain «unrelated». However, if transitivization is a form of abstract reduplication, then there is a way to relate verbalization to transitivization, on the assumption that the transitive morpheme is a reduplicating verb.
4. Conceptual motivations and competing analyses
As observed above, Nb is a cross-categorial quantificational semantic category (or feature), which holds for events and objects (or things). Being orthogonal to categories (or types), it enables us to account for
verbs) and adverbs, on the other. Related derivational uses include denominal verbs (with distributive meaning), and deverbal agent or agent nominalizations.
Abdelkader Fassi Fehri
45
event/argument complexity in causatives and transitives, multiplicity of events or objects, well-acknowledged interactions, as well as alternative (or cumulative) uses of argument and event quantification morphemes (seen traditionally as aspectual). Consequently, it eliminates potential competitors which cannot account for these properties.
4.1. Little v: verbalizer or transitivizer?
Consider again little v, taken as a source of transitivity (in e.g. Chomsky (1995)). The latter is construed as a transitive head only by stipulation; by the stipulation that there is a bidirectional relation between event complexity and argument complexity. Given this bidirectionality, it is not clear how to capture structural
differences
between
e.g.
unergatives
and
transitives,
or
anti-causatives (or inchoatives) and causatives. Furthermore, differences between objective cases (accusative and partitive) are not naturally captured. Likewise, if v is a transitive head, then it is not clear how it can be equally taken as a forming category (a verbalizer). Nothing in the v theory of transitivity predicts this correlation.
In Hale and Keyser (1999; = HK), transitivization and verbalization are distinct operations. Transitives arise from two different structural sources: (a) a complement source (e.g. calve, from have a calf, cake, from make a cake, laugh, from do a laugh), which allows no intransitive alternant, and (b) a specifier source, which is optionally internalized, hence the (in)transitive alternation. According to HK, unaccusatives force the projection of a Spec, and it is this property which permits transitivization (HK, 1999, p.55). Unergatives or transitives, by contrast, do not project (internal) Spec. Differences between the two constructions with respect to transitivization are then related to differences in specifier or complement branching. Transitivization results from embedding a verbal projection as the
Causativity, Transitivity, and Iterativity as Pluralities
46
complement of another verb (HK, 1999, p. 53). Unaccusative verbs alternate because both the inner head and the outer head are verbs, and the intransitive is simply the inner projection unmerged with another verb (HK, 1999, p. 57). Furthermore, V (or v) is category forming of both unergatives and unaccusatives. In Harley (1995), eventiveness in inchoatives and causatives is argued to be introduced by v (Become and Cause, respectively), and not the root. The complement of the vp, the rootP, is a predicative structure denoting a state, the end result of the change of state introduced by the v head (interpreted as spontaneous with inchoative, or caused, with the causative, because of the agent argument projected). But Harley (1999), building on HK’s work, adopts the view that roots may also denote things or events, in addition to properties. In the case of cough, cake (The mud caked), dance, it would be wrong to posit a predicative denotation for the root (and a “resultant state” meaning for the complement). If that were the case, i.e. that these verbs were predicative, then they should have internal arguments (of which the underived form would denote a property), and they would have been able to undergo the causativization process. It must then be the case that the root R is simply a bare nominal (denoting a thing or an event), and the v head is Do or Make, as in the following structure (his (5)): (57)
vp /
\
DP
v’ /
\
v
Rp
/ \
I
v Rdancei R I ti
Abdelkader Fassi Fehri
47
Harley then proposes an underspecified view of v, in which v is determined contextually (58)
via a taxonomy of verbal environments:
a. v + State, no Spec
= Become
b.v + State, Spec
= Cause
c. v + Thing, Spec
= Make
d. v + Event, Spec
= Do
Embick (2000) argues that v is a verbalizer (with e.g. -ize, -en spell outs), which forces an eventive interpretation (the latter being incompatible with the environment requiring stative interpretation). When there is no attachment of v, there is no eventive interpretation, hence the structure of pure statives (which are simple states, with no implication of prior event). Asp attaches either to Root, or above v. But height of attachment of Asp (as well as the contribution of v) is not sufficient to differentiate Stative Passive from Eventive Passive. In stative passives (which describe a state resulting from event in complement), the Result Asp selects a v [- AG], while the Perf Asp in eventive passives (with completive or perfective aspect) requires v [+ AG], with no external argument. Thus it is the selectional relationship between Asp and its complement which distinguishes the two passive v’s. With respect to height of attachment, Embick assumes that normal verbs have the pass feature attached to v, and hence exhibit passive syntax (no external argument, no Case). Deponent verbs, however, have the pass feature attached to the root, and hence active syntax is possible.
We can then see that most authors posit the verbalizing property of v, but as already observed in subsection 1.2. above, there has been no natural proposal to relate this property to transtiivization (via e.g. reduplication or other morphology). Harley and Embick “flavoured” views are essentially diacritical, and do not differ radically from Marantz’s previously analyzed proposals. In our view, the first v is a verbalizer, and the upper v is a
Causativity, Transitivity, and Iterativity as Pluralities
48
reduplicator, which can then transitivize the structure under appropriate conditions (but not necessarily; cf. e.g. iteratives).
4.2. Aspect
Traditionally, noun quantification and event quantification have been regarded as belonging to different domains of grammar: Nb (atemporal quantification) is associated with the former, and Aspect (temporal quantification) is associated with the latter. But as correctly observed by Dolinina (1999), in the case of event plurality, quantification is more abstract than Aspect, because event plurality can be triggered by repetition on the axis of time (and thus be related to Aspect), but also by plurality of activities of the participants, and hence lies in other dimensions than Aspect. For example, distributive constructions obligatorily contain a component of quantification, but only in a few of them is it due to repetition in time (and hence Aspect). A distributive construction can be insensitive to the axis of time, but never to Nb, a clear indication of the dominance of the quantificational meaning over the aspectual one. Distributivity then belongs to the (general) cluster of quantifier categories, and not to the temporal cluster. Semantics of distributivity, its typological encoding, and the oppositions it deals with are broader than Aspect. Within quantification, distributivity is a category with a dual nominal/event nature, combining the meanings of nominal Pl and event Pl.
23
23
Dolinia observes that there have been renewed attempts to unify the area of affiliation of quantification, with some linguists regarding nominal quantification as aspectual (Rijjkoff (1991), Verkuyl (1993)), and others arguing that event quantification belongs to the cluster of quantificational categories (Dressler (1968), Xrakovskij (1989)). But as she puts it, the question is
Abdelkader Fassi Fehri
49
Similarly, Moltman (1997) argues that the event argument in distributives is a group event, rather than a single event, just as what is required in repetitive events. In the latter case, the event is a group, but unlike the distributive, it consists of single events that all have the same participant. If the members of an event group stand in the relation denoted by the verb to parts of a participant group, the result is not a repetitive, but rather a distributive reading of the verb. Cross-linguistic evidence supports the assimilation of repetitive and distributive readings of verbs: a verb whose event argument is a group event is marked with a special distributive morphology, regardless of whether this group event constitutes a repetitive or distributive interpretation (pp. 56-57).
In
observing
causative/iterative
or
intensive
polysemies
and
interdependencies, many authors have identified the latter meaning as “aspectual” (see e.g. Nejdjalkov (1966), Comrie (1985), Li (1991), Kulikov (1993), among others). The latter author traces the polysemy (or what he calls “causative split”) back to intensivity, which he takes as the essential component of causation, because, as he puts it “Causing someone to do something implies channeling extra force from outside into the situation. The meaning “more forcefully”, “more effectively” may be thus the common semantic denominator shared by causativity […] and intensitivity, iterativity, etc. […] It is for that reason that these aspectual meanings can become associated with causativity and, in a sense, appear as its side effects” (p.26). But the identification of iterativity or intensivity with Aspect (instead of e.g. Nb and Manner or Force), and the mechanical association of causation with Force are questionable. For example, there are Arabic causative doublets,
not to choose only Nb or only Aspect, but to see which of these features is the “dominant” component of meaning.
Causativity, Transitivity, and Iterativity as Pluralities
50
which are precisely distinguished by implying (or cumulating) or not implying Force. Thus when contrasted, FII (when causative) cumulates intensity, while FIV does not: (59) ?a-xraj-tu
l-walad-a
caus-go.out-past.I the-child-acc I made the child go out. (60) xarraj-tu
l-walad-a
go.out-gem-past.I the-child-acc I made the child go out (by force; against his will). Note, however, that the distinction is made only when FII is cumulative. Most instances of FII are not, and they express either intensive or causative, but not both (see examples (37) and (38) above). Note also that in the causative/intensive complementary distribution, the choice between FIV and FII is fairly fixed, FIV being necessarily causative, and FII necessarily intensive/iterative/distributive. The following contrasts illustrates this fixed meaning opposition: c-
(61) qatta a
l-walad-u
l-lahm-a
cut.gem-past the-child-nom the-meat-acc The child cut the meat into pieces. c-
(62) ?a-qta a
r-rajul-u
l-walad-a
l-lahm-a
caus-cut-past the-man-nom the-child-acc the-meat-acc The man made the child cut the meat. (63) kassar-a
l-walad-u
l-ka?s-a
broke.gem-past the-child-nom the-glass-acc The child broke the glass into pieces. (64) ?a-ksar-a
r-rajul-u
l-walad-a
l-ka?s-a
caus-broke-past the-man-nom the-child-acc the-glass-acc The man made the child brake the glass.
Abdelkader Fassi Fehri
51
Clearly then, FII is essentially intensive, and FIV causative. FII can also be causative, or cumulative, but FIV is not cumulative, at least with repetitive/intensive meaning. Consequently, no unified source/account of the two causative forms can be provided through intensitivity (seen as Force/Aspect).
24
Hopper and Thomson (1980) takes as an essential evidence for the correlation between transitivity and tense/aspect well-known oppositions found in the Finnish case system, between accusative or partitive objects. The latter
contrast
correlates
with
the
aspectual
opposition
perfective/imperfective), respectively. Clauses with partitive objects are shown to be less transitive than those with accusative ones. Likewise, in “split ergative” systems (such as
Hindi/Urdu, some Australian and Ameridian
languages), the ergative construction is limited to perfective and preterite environments, whereas its absolutive counterpart is confined to imperfective or non-preterite (Dixon 1979). In a similar vein, Kulikov (1999) examines how in Vedic (and Ancient Greek), the transitive-causative constructions are mostly employed with the present, whereas perfect constructions are typically intransitive.
But this connection has to do with another ingredient of
transitivity, that is Case, and typically accusative or ergative Case.
25
While we
agree that Tense/Aspect (typically Perfective) and Accusative Case, for example, are interdependent, I deny a direct connection between this Aspect and the argument complexity component involved in transitives or causatives.
24
Form IV might be conceived as cumulative with the inchoative, in that it triggers change of state.On the other hand, intensivity and iterativity are far from being one and the same notion. That suggests that in addition to Nb, Force (seen as a sort of Manner) is involved in deriving FII. Recall that our unifying feature of the two forms is only Nb. 25 We follow here Hopper and Thompson (ibid) in taking transitivity to be a complex set of features, rather a simple binary opposition, but these ingredients distribute over heights, and are not “a continuum”, as they put it.
52
Causativity, Transitivity, and Iterativity as Pluralities
More crucially, I deny any clear direct connection between intensity and iterativity (seen as aspectual) and transitivity.
In sum, the fact that verbal Nb encodes plurality of actions, or plurality of arguments operating these actions, is not accidental. It is rather motivated by semantic considerations. The semantic duality of repetitive/ distributive morpheme requires a much broader category than Aspect, and transivity and causativity require complex functional ingredients in which Nb is playing a major role, to determine common derivational sources and intricate, although principled, polysemies.
26
26
The connection between Nb and Aspect has been repeatedly noted in the literature. A classical test distinction is that of telicity (to eat an apple vs to eat apples). But as far as I can tell, no precise analysis of the connection has been proposed. Holisky (1985), for example, argues that it is the verb Aspect, rather than Nb (plurality), which is relevant for determing the (im)perfectivity of the root in the Aorist in Tsova-Tush, contra Guagua (1962) and Schiefner (1859), who go for the opposite, that is Nb, and the fact that « … the imperfective root is used in the Aorist group when the object (of a transitive) or subject (of an intransitive) is in the plural. » (p. 454). To this conclusion, Holisky objects that he «… found both perfective and imperfective roots with plurals in the Aorist» (ibid). However, as he pointed out, the difference between the two is that with the perfective, the plural objects are brought all at once, while they are brought at separate occasions with the imperfective. In fact, what Holisky shows is that what makes the difference is not the single/plural dichotomy, but rather the distributive/non-distributive distinction, in our sense. Presumably more has to be said about the Nb/Aspect connection, but the counterdata provided is certainly useful for clarifying how plurality can be (un)ambiguously used, depending on (under)specification of other relevant features of plurality. Likewise, it is conceivable to relate intransitivization to reduplication (plurality) and at the same time imperfective Aspect. Radckliffe (p.c.) informed me that as as far he knows, the more wide spread case is that reduplication (or plurality) be related to intransitives, or to imperfective Aspect. In the same vein, Lea Nash pointed out to me that in Georgian, there is an identical
Abdelkader Fassi Fehri
53
4.3. Voice Voice distinguishes (a) active (agentive or unergative) predicates from non-active (ergative or middle) ones, or (b) passive (eventive or stative) verbs from non-passive ones. Voice can thus be seen as operating hierarchical organization of arguments (= HOA) with respect to predicates, rather than taking care of argument complexity.
27
Other theories of Voice have treated
argument complexity, HOA, and even specific properties of external arguments under Voice.
28
In our theory, complexity, stated in terms of Nb, is
the source of (in)transitivity. H complexity and Spec complexity being dissociated, distributive complex Specs turn out to be the typical configuration for transitivity. On the other hand, there are cases of HOA, which are best accounted via Voice, but not via Nb.
morpheme which marks both plural nouns and imperfective verbs. Such Nb/Aspect alternative uses and interconnections are more than expected under our general theory of Nb. 27 For various recent views of voice, and how it connects to transitivity, see, among others, Krazer (1996), Sells (1997), Jelinek (1998), Borer (1998), Embick (1998, 2000), Fassi Fehri (1988), and also Benreniste (1950). 28 According to McGinnis (1999), voice morphology distinguishes “active” transitives and unergatives from “nonactive” passives, unaccusatives, middles, and/or reflexives, and is usually assumed to reflect the presence or absence of a syntactic external argument. Furthermore, voice is also sensitive to certain properties of the external argument (Embick (1998)). However, in Lidz (1999), the reflexive/nonactive morphology reflects the absence of specifier of vP. If that were the case, then we would expect active morphology in e.g. “La lettre est écrite lundi”, where vP has an overt Spec. In many languages, reflexive clitic constructions use nonactive voice, although there is evidence that such constructions do have a syntactic external argument If so, then voice does not simply reflect presence or absence of external argument, “… voice morphology makes reference to a particular specifier of vP (the external argument), as well as the properties of that specifier (Embick (1998)). Additional observations from Greek provide evidence that morphology inserted in the v node can refer both to the (intrinsic) features of v and to the (contextual) features of the external argument in spec-vP” (p. 153). Active morphology must then be able to make specific reference, not only to the presence or absence of phi-features in Spec of vP, but also to the presence or absence of such features in a particular Spec of vP, namely the external argument (p. 155). Case on v is not necessary for the insertion of active voice morphology. V has a fully specified external argument, and the insertion of active voice is predicted to be contextual (p. 157).
54
Causativity, Transitivity, and Iterativity as Pluralities Functional heads expressing Voice (diathesis or HOA) occur at two
different heights: (a) a low position, which can be found lower than morpho-syntactic categorization (v, n, a, p), and which enables us to distinguish arguments of internal ev or external ev basically. Let us call the internal positioning ergative diathesis, and the external unergative diathesis. (b) a high position, which is projected higher than morpho-syntactic category formation, and which organizes arguments with respect to Case, in particular Accusative Case in the Accusative/Nominative Case systems, and Ergative in the Absolutive/Ergative languages . Low voice or diathesis applies at L-syntax, while high voice applies at S-syntax, and distinguishes, for example, active accusatives from passive non-accusatives. The double distinction in Voice systems is not knew. For example, Benveniste (1950) proposes two dichotomies: (a) active/middle and (b) active/passive, which parallel broadly the distinctions made here. Likewise, Burzio’s (1981) generalization (associating unergativity and accusativity) may be seen as a reflection of this system. Furthermore, Moravcsik (1978b) has convincingly established the mixed nature of languages with respect to ergative/accusative patterns. There are various instances of event/argument structure organization which can be treated as low voice distinctions, and which can be kept separate from argument complexity, treated as arising from Nb specification.
4.3.1. Anti-transitive reflexives
FI transitives may have two anti-transitive (or anti-causative) counterparts (or sorts of reflexives) which correspond to different interpretations, and which should be taken as doubly-headed, or exhibiting complex heads, although they are monadic with respect to their args/Specs. This is the case of the verb
Abdelkader Fassi Fehri
55
gasal «to wash someone/something» with its two anti-transitive counterparts in (65) and (66): (65) n-gasala
l-makaan-u
ref-washed the-place-nom The place has been washed. (66) g-t-asala
r-rajul-u
ref-washed the-man-nom The man washed himself.
As well-known, the two morphologies ([n] and [t]) apply only to transitive roots. They indicate the «directionality» of derivation, i.e. that the intransitive ismorphologically marked, compared to the transitive. This formal complexity is coupled with complexity of interpretation, yielding a reflexive interpretation in (66), and an «affected» anti-causative in (65).Thus one can show by various tests (control structures, agent-oriented adverbs, passivization) that (66) is agentive, whereas (65) is not. These properties suggest that the two constructions have Pl distributive Hs, as is normally the case with transitives (or causatives), although they have only a Sg arg. What is then the difference between the two constructions, if the Nb of their Hs and Specs is the same? I think that the difference has to do with HOA (or diathesis). I take the subject of (65) to be its internal object (or the ergative subject), whereas the subject in (66) is the external (or unergative subject).
29
Suppose this is so. Then how this difference is implemented? One might explore the vocabulary difference between [n] and [t] as well as their placement in the word (prefixation vs. infixation through metathesis). This vocabulary approach may or may not be right. But I would like to introduce a
Causativity, Transitivity, and Iterativity as Pluralities
56 more
syntactic
(and
lexeme-based)
approach.
Suppose
that
the
detransitivizing morpheme is expressing an (internal) ev, and it can be placed either internal to the stem, or external to it. When it is internal to the stem, the subject of the entire stem immediately c-commands a segment of the stem which includes (or c-commands) the affix [t], which is then internalized. Consequently, the external unergative subject is a subject of the complex ev, dominated by the external (unergative or active) ev. The active reflexive interpretation then results from the fact that there is a unique subject of both evs, in which the higher ev and its subject are active. On the other hand, when the (internal) affix is placed externally, the «internal» subject has to be «externalized», presumably to «match» the ergative [n], and the unergative subject is not realized (presumably because it can only be external), hence the «passive» anti-causative interpretation. Note that this process of derivation recalls in part the internal/external distinction between FII and FIV.
30
A similar case of distributed ev/arg complexity can be raised with benefactive FVIII, compared to the transitive FI. Let us compare the following pair: c
(67) qata a l-lahm-a cut
the-meat-acc
He cut the meat. c
(68) q-t-ata a l-lahm-a
29
Many other doublets point to the same analysis: n-fataha «to open (by itself)» f-t-ataha «to open something to his benefit», n-sawaa «to get roasted», stawaa «to roast something to his benefit», n-sagala «to get occupied», stagala «to occupy oneself », etc. 30 It is striking that the two morphological forms are unified into one in most Arabic dialects, which is uniformly prefixal, rather than infixal. Likewise, only prefixation is available to detransitivize FII, FIII and FIV, thus yielding FV, FVI and FX, respectively. With FV, we get anti-causative/reflexive, with FVI reciprocal, and with FX anti-causative/reflexive also.
Abdelkader Fassi Fehri cut.ref
57
the-meat-acc
He cut the meat to his benefit. In (68), the number of arguments is not increased syntactically, although it is understood that the verb has an extra theta slot. I will assume that the verbal head there is three-way complex, although the Spec configuration is only double. Note that the affix here cannot be [n], since the benefactive is an active reflexive.
4.3.2. Reflexive causatives
Reflexive causative FX provides an interesting case of complex morphological derivations. It is usually agreed that FX can be the reflexive form of causative FIV. This reflexivization comes in two forms: (a) pure reflexives, as in (69), and benefactives, as in (70): (69) a. salima «to be sane» ?aslama-hu «to deliver it (sane)» sta-slama «to deliver himself; to surrender» b. ?a-yqada-hu «to wake him up»
sta-yqada «to wake (one’s self)
up (70) a. xaraja «to go out» ?a-xraja-hu «to make him go out» sta-xraja-hu «to make it go out for one’s self; to extract» b. hadara «to be present» ?a-hdara-hu «to make it present» sta-hdara-hu «to make it present for one’s self» c. kataba «to write» IV ?a-ktaba-hu «to make him write» sta-ktaba-hu «to make him write for his benefit» I take the difference between the structure associated with those forms and those studied in (65)-(68) above to be that FX is not only morphologically more complex, but
it
can
be
shown
to
be
structurally
more
complex.
For
example, unergatives transitivize through FIV, but not through FI, and thus have FX as causative intransitives.
Causativity, Transitivity, and Iterativity as Pluralities
58
4.3.3. Agentive and «expositive» causatives
Consider the following pair of FIV causatives: c
(71) ?a-ba -tu
zayd-an
l-faras-a
caus-sold-I Zayd-acc the-horse-acc I made Zayd sell the horse. c
(72) ?a-ba -tu
l-faras-a
caus-sold-I the-horse-acc I made the horse to sell (to be sold). The two sentences differ in interpretation. In (71), the horse has been sold, but in (72), it may or may not have been. It is only «exposed» for selling, traditional grammarians say. This is so because in (71), the agent of selling is realized, and hence the object is an object of an «unergative/active» internal verb, whereas in (72), the object might be an «ergative» subject of a middle/passive verb. This difference in interpretation and structure complementation appears to follow from complexity captured through Nb, but this is not necessarily so, due two reasons. First, there is no ergative (active) verb «sell» in Arabic, i.e. in the high Voice system, and the only way to express the middle is via Passive morphology. Second, the agent of «sell» can be omitted, but not its theme, hence the ungrammaticality of (73): c
(73) * ?a-ba -tu
r-rajul-a
caus-sold-I the-man-acc I made the man sell (his horse).
Abdelkader Fassi Fehri
59
It is possible to think that the verbal root is diadic in both cases, and that (43), when interpreted as «expositive» instantiates a sort of low passive Voice.
31
4.3.4. «Requestive» causatives
In FX causatives analyzed above, we have seen that the reflexive or benefactive is also the causer. But this interpretation does not obtain in so-called requestive causatives (talabiyyah), which are instantiated in the following pair (74) sta-fham-tu-hu caus.ref-understood-I-him I asked him to explain to me (make me understand). (75) sta-xbar-tu-hu caus.ref-informed-I-him I asked him to inform me. In these constructions, the interpretation is (unexpectedly) not: «I made him understand/inform to my benefit», but rather something like «I made him make me understand/ be informed». How do we get this interpretation? One possibility is to take these constructions to be instances of double causatives. In this case, the subject (or causer) of the first (higher) cause is object of the second (embedded) cause, and the object of the first cause is subject of the second cause. By contrast, in (70) above, the subject of the (unique) cause is the beneficiary, and the object of the cause the subject of the event. How can we then obtain this reversal of ordering? It seems to us that this reversal recalls in fact the double use of FIV, as expressing both agentive causativization, or the result or effect of the caused event on the object c
(mutaawa aah). Then FX here appears as a combination of both structures.
31
Another possibility is that the causative morpheme selects an ergative verb in (72), which then provides the middle meaning. Even under such an approach, one may want to argue that it is
Causativity, Transitivity, and Iterativity as Pluralities
60
But note that the difference between the two structures of FIV is a difference in Voice, rather than a difference in number of arguments.
32
4.2.5. Ergative Number and intensive forms
Consider again the intensive Nb on verbal and non-verbal forms, as well as the restrictions on its distribution and interpretation. The forms to consider here are: verbal forms, participle forms, and exaggeration forms. The first two forms can be taken as verbal, whereas the last one is adjectival. Furthermore, Pl Nb is distributed according to the ergative parameter in that it applies to objects of transitives and subjects of intransitives (as the examples (76) and (77) show), and to their participles in the same way: (76) zayd-un
c
muqatti -un
l-lahm-a
Zayd-nom cutting.intens.-nom the-meat-acc Zayd is cutting (cutter of) the meat into pieces. (77) l-lahm-u
c
muqatta -un
the-meat-nom cutting-nom The meat is cut into pieces. It cannot apply to subjects of transitives. For example, (78) cannot mean that many people or various groups of people cut the meat: c
(78) qatta a
n-naas-u
l-lahm-a
cut.intens. the-people-nom the-meat-acc The people cut the meat into pieces. Likewise, the intensive is formed from non-stative intransitives which can be taken as ergative, but not from the unergative intransitives:
(79) a. jawwala
r-rajul-u
diathesis, rather than Nb, which is playing the essential role. 32 For a similar observation, though in a different framework, see Larcher (1996).
Abdelkader Fassi Fehri
61
took.walks.intens. the-man-nom The man took a lot of walks. b.* sakkara
r-rajul-u
drank.intens. the-man-nom The man took a lot of drinks. c. * raqqasa
r-rajul-u
danced.intens. the-man-nom The man danced intensively. These observations suggest that verbal Nb (when intensive or collective) is in fact ergative.
Consider now exaggeration forms, which can be taken as naming subjects, following traditional grammarians. Contrary to what happens with subject or object participles, it is striking that there are unergative intensive adjectives, as in?akkaal, raqqaas, sikkiir, etc. How are these adjectives formed? And how come that they refer to the unergative subjects of say ?akal, rather than to the ergative? In order to answer these questions, we need to assume that exaggeration forms are derived from «basic» transitive or intransitive roots alike (to name subjects), but they cannot be derived from derived roots. Evidence for this hypothesis comes from the non-existence of exaggerations of augmented or derived forms. Only participles can derive from these forms. The deeply rooted derivation of these adjectives may also provide a clue to the absence of relevance of the subject/object dichotomy, assuming that adjectives have subjects, but no objects.
33
33
According to traditional grammarians, adjectival intensives of e.g. FII and FIV take the form of the instrumental, which is also used for FI. But the instrumental does not seem to carry the meaning of the augmented (or derived) form. Thus mizwaaj “who marries a lot” appears to be related to tazawwaja “to marry” (rather to the transitive zawwaja “to marry someone”), miqtaac «who or which cuts a lot» is related to FI, rather than FII, mictaa? «who gives a lot» is not
62
Causativity, Transitivity, and Iterativity as Pluralities
4.4. Further empirical motivations
4.4.1. Ergative and non-ergative Number in event plurality and transitivity
Natural ergativity of Number has been observed with respect to the distribution of Nb in intensive verbs, i.e. if the action applies to arguments, then it would be the object if the verb is transitive, and the subject if it is intransitive (or ergative; see above subsection 2.2). Such natural ergativity is also observed for transitive/intransitive alternations (see e.g. English and Arabic “laugh”/dahika, which do not transitivize, compared to Engish “open”, or Arabic hamara “pour out”, which do). If transitives arise from realization of Pl Nb, then the effects of natural ergativity in the two instances (intensives and transitives) are unified.
On the other hand, English does not have a morpheme for repetitive plurality of events, while Arabic does. The prediction then is that in English,
necessarily related to FIV. These observations corroborate the low derivational status of these forms. It is important to note that active subject adjectives, stative adjectives and exaggerations forms may be only parts of the same system of adjectival forms. This is why they are found in complementary distribution in some cases. For example jawwaal appears to be the subject participle of the intransitive jawwala, baliid «stupid» is the adjectival subject of baluda «be stupid», etc. The count/non-count (or mass) dichotomy can be used to distinguish active subject participles from subject adjectives. Thus the complement of a participle can be countable, but not that of an adjective, hence the following contrast: (i) r-rajul-u ?aakil-un l-lahm-a the-man-nom eating-nom the-meat-acc The man is eating the meat. (ii) r-rajul-u ?akkaal-un li-l-lahm-i the-man-nom big.eater of-the-meat-gen The man is big eater of meat.
Abdelkader Fassi Fehri
63
Ergative Nb can only be distributive (on the assumption that transitives use distributive Nb), an option provided by UG syntax in terms of Pl Merge (of Specs). In other words, we assume that syntactic or analytic Nb (which is made use of English) is universal, whereas synthetic Nb (used in the Arabic intensive or repetitive) is language specific. Furthermore, analytic Nb when used in L-syntax is subject to natural ergativity, but it is free in S-syntax, hence the contrast (80) * I laughed John (81) I made John laugh
Morphological causatives do not observe the Ergativity condition, as amply documented in the literature (see e.g. Japanese and Arabic), thus paralleling S-syntax, not L-syntax. This does not necessarily imply that they are biclausal (because other tests can be used to decide whether they are monoclausal or biclausal; see Gelbuth). This difference can be captured in terms of structure heights, typically height of Nb placement (see nect subsection).
Structure heights may also be used in distinguishing morphological forms. For example, Arabic FIII and FIV do not observe the Ergativity Condition, and they express only distributive Nb. This suggests that they are morphologically complex, with one functional level up, compared to collective FII.
4.4.2. Direction of derivation: conceptual and functional transitivity
The direction of derivation of morphological objects is normally determined by complexity. Thus in the cases of (in)transitivity, the direction of derivation is clearly determined: the morphologically less complex member of the pair is the source of derivation, be it intransitive or transitive. It is normally the case
64
Causativity, Transitivity, and Iterativity as Pluralities
that causatives derive from inchoatives, and this direction of derivation correlates with extra morphology. Likewise, anti-causatives derive from causatives via additional morphology. But the direction of derivation of so-called lexical transitives (or causatives) is not so determined. Thus there is crosslinguistic variation in the direction of derivation, from intransitive to transitives or vice versa (compare English intransitive “open” with Arabic transitive fataha). This variation might be attributed to conceptual/cultural considerations. Let us call the transitivity formed at the lexical conceptual level (or L-syntax) conceptual transitivity, and the one formed through morphology or syntax functional transitivity. We can then say that functional transitivity has a clear unidirectional source, whereas conceptual transitivity does not; it is subject to conceptual variation.
Functional transitivity clearly increases argument complexity, whereas functional intransitivity decreases argument complexity. I take the latter functional operations to be manifestations of pluralities. The transitive is distributive/partitioning
plural,
whereas
the
intransitive
is
a
collective/assembling plural.
4.4.3. Adicity, (in)transitive alternations, and multiple uses
If a dyadic predicate is treated as Pl (distributive), then it is expected to have two different realizations, depending on whether Nb is analytic, or synthetic. It is transitive when Nb is analytic, and intransitive when synthetic. Reflexives and reciprocals instantiate the analytic/synthetic variation. Reflexives may be syntactically dyadic in English, but they are syntactically dyadic or monadic in Arabic, depending on roots: (82) The man washed himself (83) a. g-ta-sala r-rajul-u
Abdelkader Fassi Fehri
65
b.* gasala r-rajul-u nafsa-hu (84) a. wabbaxa r-rajul-u nafsa-hu The man blamed himself b. * ta-wabbaxa r-rajul-u The man blamed himself Reflexives can be seen as arising from a distributive (plural) event, with one (and the same) argument, which is partitioned in syntax, but unified in morphology. With reciprocals, similar (although not identical) phenomena and contrasts arise: c
(85) a. haddada n-naas-u
c
ba d-u-hum ba d-an
menaced the-people-nom some-nom some-acc The people menaced each other. b. * ta-haddada n-naas-u (86) a. ta-haabba
n-naas-u
rec.-loved the-people-nom The people menaced each other. b.* haabba n-naas-u
c
c
ba d-u-hum ba d-an
loved the-people-nom some-nom some-acc c
(87) a. saara -a n-naas-u fought
c
c
ba d-u-hum ba d-an
the-people-nom some-nom some-acc
The people fought each other. c
b. ta-saara a n-naas-u rec.-fought the-people-nom The people fought each other. Here both the event and the argument are plural (distributive), but the latter can be partitioned in syntax, or assembled through morphology.
Causativity, Transitivity, and Iterativity as Pluralities
66
The morphology often unifies various constructions which may have different structures, but share adicity building properties. For example, Morrocan Arabic has the same morphology for reflexives, reciprocals, and passives. The same is true in Romance. This is expected if the morphology is associated with an assembling plural. Likewise, causatives alternate with passives in Japanese/Korean, and with inchoatives or (resultative) anti-causatives in Arabic, again a manifestation of partitioning or assembling .plurality.
5. Nb theory and Nb heights
5.1. Sg and Pl Merge
Up till now, I have shown that verbal plurality is realized in various morphological or syntactic ways. Suppose that the source of verbal plurality is a functional head Nb, which is an extended projection of verbs (or eventive roots). Then plural Nb can be morphologically realized on the head, and it is interpreted as repetitive event or action denoted by the verb. It can also be interpreted as the same action operated on many objects (when plurality is distributed over arg/Spec, rather than H/ev). Second, Pl Nb can be morphologically realized on the head, but the plural interpretation is induced through the multitude of Specs. The plurality is then read off somehow cumulatively, through additioning Sg Specs. This is how I view the situation found, for example, with transitive FII or FIV. Third, there are cases where the head is not overtly realized, although the transitivity is still visible through multiple Specs. This is the case of alternant doublets discussed in Fassi Fehri (1987) and (1996), exemplified by the following contrasting pair: (88) a. hamara
c
d-dam -u
poured.out the-tears-nom
Abdelkader Fassi Fehri
67
Tears poured out. b. hamara
c
d-dam -a
poured.out the-tears-acc He poured out tears. Finally, there are cases of (in) transitivization, which appear to be «lexicalized», in the sense that no alternation appears to be possible. Leaving the morphology aside, I will take the four instances of plurality to be subsumable under two subcases of Merge: (a) Sg Merge (with «intensive» Nb), and (b) Pl Merge (with transitives). We then have to see how languages make use of the two options. 5.2. Language variation Arabic, English, and Japanese (and presumably quite many – if not all languages) make use of the (b) option for forming transitives. They differ, however, in that Arabic and Japanese can transitivize unergatives, but English cannot. This difference can be attributed to the “richness” of morphology. Compare the following constructions:
34
(89)* John laughed Mary (90)* dahaka r-rajul-u
l-walad-a
laughed the-man-nom the-child-acc The man laughed the child. (91) ?a-dhaka
r-rajul-u
caus.laughed
l-walad-a
the-man-nom the-child-acc
The man made the child laugh. I take the rich morphology to allow more structure than does the poor one, hence the difference between English and Arabic, and in fact between Arabic (90) and (91). In fact, I take Nb to be Ergative in the English alternations and
34
Richie Kayne (p.c.) informed me that The doctor coughed John is not that bad, in which case there is a more complex story to be said about English.
Causativity, Transitivity, and Iterativity as Pluralities
68
the Arabic (88). And I will assume that Enb is projected at the root level.
35
The
“rich” morphology, which allows the transitives/causatives of unergatives is nothing
else
but
Unergative
Nb
(= UeNb). I take UeNb to be projected higher in the structure than that root. The two structures might be as follows (I take the predicate here to be v for simplicity sake, although this is not so): (92)
vp / \ John v+pl / \ the door v | opened
(93)
vp / \ rajul vp +pl / \ walad v+pl / \ dahik v
In (93), I am adapting a HK’s view of unergatives as transitives, and hence plurals in my view. The analysis is tentative, and is not crucial for the point. In this view, it is the v+pl which licenses the occurrence of another verb, and (distributively) its specifier. Note that it is the rich plural morphology which allows also the occurrence of transitive statives in Arabic, but not English. This is illustrated by (3) above, repeated here as (94), for convenience: c
(94) addama
l-?amr-a
made-important the-matter-acc He made (took) the matter (as) important.
35
For a more theta-based view, see Pylkk? nen (1999, 2000).
Abdelkader Fassi Fehri
69
Consider now Sg Merge of Pl, which yields the intensive or the repetitive, as exemplified in (5) above, repeated here as (95), for convenience: (95) a. kassara
l-walad-u
l-ka?s-a
broke.intens. the-child-nom the-glass-acc The child broke the glass into pieces. b. jawwala
r-rajul-u
took.walks.intens. the-man-nom The man took a lot of walks. English does not have a couterpart of the intensive, i.e. it does not make a direct use of this subcase of Merge. Yet it can form a complex predicate through the use of aspectual particles or adverbs, to convey a similar (although not identical) meaning. We have taken the repetitive Pl as synthetic and assembling, because it does extra arguments, nor does it yield separate discrete verbs. The representation of (95) might then be as follows: (96)
vp / \ rajul v+pl | jawwal
In this case, the plurality is confined to the “ergative” verb, by ENb. To recapitulate what we have been saying so far, the picture in (98) emerges, based on the characteristics in (97): (97) a. Pl: +/- Part (-Part: assembling or collective; + Part: partitioning or distributive) b. Pl: Erg or Unerg c. Spec/Arg: Pl or Sg d.H/ev/v: Pl or Sg (98) a. Repetitive: - Part H, -Part Spec (or assembling H and assembling Spec)
Causativity, Transitivity, and Iterativity as Pluralities
70
b. Causative: + Part H, + Part Spec c. Transitive: - Part H, + Part Spec (“lexical transitives”) d. Analytic causative: + Part H, + Part Spec e. Synthetic reflexive (reciprocal, passive): +Part H, -Part Spec Moreover, recall how we view the relation between verbalization and causativization. The former operates on a single v, and reduplication of the same (or similar) v induces its plurality (or duality), a feature property of causatives.
6. Summary and Conclusion
In this paper, I have argued that verbal Nb plays a central role in characterizing transitive or causative constructions, seen as instances of (distributive) plural configurations, generated through Pl Merge. Thus when a Pl morpheme occurs on verbal heads, it can be interpreted collectively (as an iterative or intensive event), or it can be interpreted distributively (as a distributive action operated on distributive roles or arguments). This ambiguity of verbal plurality recalls that found in the nominal plural quantification, which suggests that Nb is orthogonal to verbal and nominal categories. If Nb is generalized to verbs, and extended to transitivization, then there is no need to appeal to a conceptually distinct category (such as Aspect, Voice, or v) to account for verbal complexity and quantification, an option which turns out to be both empirically and conceptually deficient. SA is an instance of languages in which the morphology of plurality in verbs and nouns is essentially uniform: it is based on vocalic or consonantic gemination (of the second syllable, or partial reduplication). Verbal reduplication is used to express both the iterative (or intensive) plural and the causative. Fissioning or partitioning reduplication is a typical mechanism which derives only causatives, or in the case of iambic reduplicating roots, “attenuative” or “paucal” partitioned event forms. That
Abdelkader Fassi Fehri
71
reduplication is used to mark verbal plurality is amply documented in various languages. Likewise, its exclusive or ambiguous uses as repetitive, distributive, or causative is a matter of crosslinguistic variaion. The correlation of these uses is expected under our general Nb theory, but not under Aspect, v, or Voice theories. The latter categories play various specific roles in functional structures, but their roles (although interacting with transitivity and Nb) have to be kept distinct. Indeed, Voice, Aspect, and Tense have been argued in the literature to interact with transitivity. But Voice, as a functional head encoding hierarchical organization of arguments, has to be kept separate from Nb, which accounts for event/argument complexities. On the other hand, Aspect and Tense exhibit dependencies with Case, but they interact with event/argument complexities only indirectly.
References
Causativity, Transitivity, and Iterativity as Pluralities
72
Alexiadou, A.: 1999, On the Syntax of Nominalization and Possesssion: Remarks on Patterns of Ergativity. Habilitation Thesis. Universität of Postdam. Alexiadou, A.: 2000, Deriving Words and Categories. Paper delivered at GLOW 23. Vitoria Alsina, A., J. Bresnan, & P. Sells, eds. 1997, Complex Predicates. Stanford: CSLI. California. Arad, M.: 1999, On “Little v”. M I T W P L 33, 1-25. Astarabaadii, R.: 12th century, Šarh aš-Šaafiya. Beyruth: daar al-kutub c
l- ilmiyyah.1975. Beghelli, F. & T. Stowell: 1997, Distributivity and Negation: The Syntax of each and every. 71-107. Szabolcsi, A. ed. Benveniste, E. 1950, Actif et moyen dans le verbe. Problèmes de linguistique générale. 169-175. Paris: Gallimard. Bittner, M. & Hale, K.: 1996, Ergativity: Toward a Theory of a Heterogeneous Class. Linguistic Inquiry. 27. 531-604. Borer, H.: 1998, Deriving Passive without Theta Roles. Morphology and its relation to phonology and syntax. 60-110. S. Lapointe ed. Stanford: CSLI. Bresnan, J.: 1982, The Passive in Lexical Theory. The Mental Representation of Grammatical Relations. J. Bresnan ed. Cambridge: MIT Press. Caspari/Wright, W.: 1859/1973, A Grammar of the Arabic Language. Cambridge: Cambridge Univ. Press. Chomsky, N.: 1995, The Minimalist Program. Cambridge, Mass: MIT Press. Chomsky, N.: 1998, Minimalist Inquiries. M I T W L. 15. Cambridge: M I T. Comrie, B.: 1985, Causative verb formation and other verb-deriving morphology. Language Typology and Syntactic Description. Vol. III. 309-348, T. Shopen ed. Cambridge: Cambridge Univ. Press. Cinque, G.: 1995, Adverbs and the Universal Hierarchy of Functional Projections. GLOW Newsletter. 34. 14-15.
Abdelkader Fassi Fehri
73
Cinque, G.: 1997, Adverbs and the Universal Hierarchy of Functional Projections. Oxford: Oxford Univ. Press. Cuvalay-Haak, M.:1997, The Verb in Literary and Colloquial Arabic. Berlin: Mouton de Gruyter Di Sciullo, A. M.: 1999, Verbal Structure and Variation. Semantic Issues in Romance Syntax. E Trevino & J. Léma eds. Amsterdam: John Benjamins. Di Sciullo, A. M.: (to appear). Local Asymmetry Connection and Lexical knowledge. MITWPL. Cambridge: MIT. Dixon, R. M. W.: 1994, Ergativity. Cambridge/ Cambridge University Press. Dobrovie-Sorin, C.: 2000, Eventive and Generic Predicational Readings of SE-Passives and the theory of «little v». GLOW Newsletter. 44, 20-21 Dolinina, I. B.: 1999, Distributivity, More than Aspect. Tense-Aspect, Transitivity, and Causativity. Essays in honour of Vladimir Nedjalkov. W. Abraham & L. Kukikov eds. Amsterdam, John Benjamins. Dressler, W. 1968, Studien zur verbalen Pluralitât. Wien: Böhlau in Kommission. Durie, M.: 1986. The grammaticalization of number as a verbal category. BLS.12. Berkley Linguistics Society, University of California. Embick, D.: 1998, Voice systems and the syntax/morphology interface. MITWPL. 32. H. Harley ed. Embick, D.: 2000, Participle Structures and Participles Asymmetrics. ms. MIT & UTA. Fassi Fehri, A.:1985, al-mucjam al-carabii. Casablanca: Toubqal Publishers Fassi-Fehri, A.: 1987, Anti-causatives in Arabic, causativity and affectedness. LPWP. 15. Cambridge: MIT. Fassi Fehri, A.: 1988, Arabic Passives as aspectual predicates. Ms. MIT & Faculty of Letters, Rabat. Fassi Fehri, A.: 1991/1993, Issues in the Structure of Arabic Clauses and Words. Dordrecht: Kluwer Academic Publishers.
Causativity, Transitivity, and Iterativity as Pluralities
74
Fassi Fehri, A.: 1996a. Configurations and Transitivity Splits in the Arabic Lexicon. Configurations. 53-78. A. M. DiSciullo ed. Somerville: Cascadilla Press. Fassi Fehri, A.: 1996b. Selective Diathesis and Arabic Forms. Situated Languages, Technology, and communication. 156-184. A. Fassi Fehri et al eds. Rabat: IERA Publications. Fassi Fehri, A. : 1996c, Distributing Features and Affixes in Arabic Subject Verb Agreement Paradigms. Linguistic Research. 1.2, 1-30. Rabat: IERA. Fassi-Fehri, A.: 1999a, cani t-tawaaruti fii l-huduudi wa-bacdi xasaa?isi t-taswiir-I l-kullii (On Inheritance in determiners and some properties of universal quantification). Al-murakkabaat-u l-ismiyatu wa-l-haddiyyah fii llisaaniyyat-il-muqaaranah, 9-46. A. Fassi Fehri et als. eds. Fassi Fehri, A.: 1999b. How Asymmetric (and Modular) is Lexical Diathesis? Asymmetry Project Lecture. UQAM. Montréal. Fassi Fehri, A.: 2000 Transitivity as Plural Number. Linguistic Research. 5.1. Rabat: IERA Frajzyngler, Z.: 1985. Ergativity, Number, and Agreement. BLS.11. Berkley Linguistics Society, University of California. Gil, D.: 1996, Universal Quantification in Hebrew and Arabic. Studies in Afroasiatic Grammar.1. J. Lecarme, J. Lowenstamm, & U. Shlonsky. Eds. The Hague: Holland Academic Graphics. Golovko, E.V.: 1993, On non-causative effects of causativity in Aleut. Causatives and Transitivity. B. Comrie & M. Polinsky eds. Amsterdam, John Benjamins. Greenberg, J.: 1991, The Semitic intensive as verbal plurality. A. Kaye ed. Semitic studies in honor of Wolf Leslau. II, Wiesbaden: Harrassowitz. Guerssel, M.: 1996, On Berber verbs of change. LPWP. 12.Cambridge: MIT. Guerssel, M. & Lowenstamm, J.: 1996, Ablaut in Classical Arabic Measure 1 verbal forms. J. Lecarme, J. Lowenstamm, & U. Shlonsky eds.
Abdelkader Fassi Fehri
75
Haiman, J.: 1980, The iconicity of grammar: Isomorphism and Motivation. Language. 56.3. Hale, K. & Keyser, J.: 1993, On argument structure and the lexical expression of of syntactic relations. The view from building 20. K. Hale, K. & J. Keyser, eds .111-176. Cambridge, Mass: MIT Press. Hale, K. & Keyser, J.: 1997, On the complex nature of simple predicators. Alsina et als. eds., 29-65. Hale, K. & Keyser, J.: 1998, The Basic Elements of Argument Structure. MITWPL.32. 73-116. H. Harley ed. Cambridge: MIT Hale, K. & Keyser, J.: 1999, Bound features, Merge, and Transitivity Alternations. MITWPL.35. 49-72. L. Pylkkänen, A. van Hout & H. Harley eds. Cambridge: MIT Halle, M. & A. Marantz.: 1993, Distributed Morphology and the Pieces of Inflection. The view from building 20. K. Hale, K. & J. Keyser, eds. 111-176. Cambridge, Mass: MIT Press. Harley, H.: 1995, Subjects, Events, and Licensing. Ph.D. Cambridge, Mass, MIT. Harley, H.: 1999, Denominal verbs and Aktionsart. MITWPL.35. 73-86. L. Pylkkänen, A. van Hout & H. Harley eds. Cambridge: MIT Harley, H. & Noyer, R.: 1998, Licensing in the Non-Lexicalist Lexicon: Nominalizations, Vocabulary Items, and the Encyclopedia. H. Harley ed. Higginbotham, J.: 1985, On Semantics. Linguistic Inquiry 16, 547-593. Holisky, D.A.: 1985, A Stone’s Throw From Aspect to Number in Tsova-Tush. International Journal of American Linguistics. 53, 453-4. Hooper, P. J. & S.A. Thomson: 1980, Transitivity in Grammar and Discourse. Language. 56.2. Van Hout, A.: 1994, Projection based on Event Structure. Lexical specification and Lexical insertion. P. Coopmans Everaert & J. Grimshaw eds. Hillsdale:Lawrence Erlbaum Associates Ibn Aqiil, B.: 14 th century, Šarh alfiyyat Ibn Maalik. Cairo: daar al-fikr. 1979
Causativity, Transitivity, and Iterativity as Pluralities
76
Jackendoff, R.S.: 1995, Parts and boundaries. Lexical & Conceptual Semantics. B. Levin and S. Pinker eds. Cambridge, B. Blackwell. Jelinek, E.: 1998, Voice and Transitivity as Functional Projections in Yaqui. The Projections of Arguments. M. Butt & W. Geuder eds. Stanford, CSLI Publications. Kayne, R.: 1994, The Antisymmetry of Syntax. Cambridge, Mass: MIT Press. Krazer, A.: 1996, Severing the external argument from its verb. Phrase Structure and the Lexicon. 109-137.
J. Rooryck & L. Zaring eds.
Dordrecht: Kluwer Academic Press. Kulikov, L.: 1993. The “second causative: a typological sketch. Causatives and Transitivity. B. Comrie & M. Polinsky eds. Kulikov, L.: 1999. Split Causativity. Tense-Aspect, Transitivity, and Causativity. Essays in honour of Vladimir Nedjalkov. W. Abraham & L. Kukikov eds Kurylowicz, J.: 1973 , Studies in Semitic Grammar and Metrics. London: Curzon Press. Landmann, F.: 1996, Plurality. The Handbook of Contemporary Semantic Theory. S. Lappin ed. Oxford: Blackwell Larcher, P.: 1994, Un phénomène de surdérivation en arabe classique: à propos de la Xe forme verbale istafcala. Annales Islamologiques. XXVIII, 215-230. Larcher, P.: 1996, Sur la valeur «expositive» de la forme?afcala de l’arabe classique. ZFAL..31, 7-26. Larcher, P.: 1998, La Forme IV?afcala de l’arabe classique: faire faire et laisser faire. ZFAL..35, 14-28. Larson, R.: 1988, On the double object construction. Linguistic Inquiry. 19, 335-391.
Abdelkader Fassi Fehri
77
Li, F.: 1991, An examination of causative morphology from a cross-linguistic and diachronic perspective. CLS. 27. 344-359. Dobrin, L. M et al eds.Chicago: Chicago Linguistic Society. Lidz, J.: 1999, Causativity, Late Insertion, and the Projection of vP. MITWPL.35. 117-136.
L. Pylkkänen, A. van Hout & H. Harley eds.
Cambridge: MIT Marantz, A. 1997. No Escape from Syntax. Ms. Cambridge: MIT. Marantz, A. 1999. Creating words above and below little v. ms. MIT. Marantz, A. 2000. Reconstructing the lexical Domain with a single Engine. ms. Cambridge: MIT. Marantz, A. 2001. The Universality of Root and Pattern Morphology. ms. Cambridge: MIT. Maslova, E.S.:1993. The causative in Yukaghir. Causatives and Transitivity. B. Comrie & M. Polinsky eds. McCarthy, J. & Prince, A.: 1990, Foot and word in Prosodic Morphology. Natural Language and Linguistic Theory. 8, 209-282. McCarthy, J.: 1993, Template Form in Prosodic Morphology. FLSM III. McGinnis, M.: 1995, Fission as Feature Movement. MITWPL. 27, 165-187. R. Pensalfini & H. Ura eds. Cambridge, Mass: MIT. McGinnis, M.: 1999, Reflexive clitics and the specifiers of vP. MITWPL.35. 137-160. L. Pylkkänen, A. van Hout & H. Harley eds. Cambridge: MIT Miyagawa, S.: 1997, Light Verb Make and the Notion of Cause. ms. Cambridge: MIT. Moltmann, F.: 1997, Parts and Wholes in Semantics. Oxford: Oxford Univ. Press. Moravcsik, E.A.: 1978a, Reduplication Constructions. Universals of Human Language. III, 297-334 J. Greenberg ed.. Stanford: Stanford Univ. Press. Moravcsik, E.A.: 1978b, On the distribution of Ergative and Accusative Patterns. Lingua. 45, 233-279.
Causativity, Transitivity, and Iterativity as Pluralities
78
Nash, L. & A. Rouveret: 1997, Proxy Categories in Phrase Structure Theory. NELS 27. Nedjalkov, V. P.: 1966, Ob arel’nyxuniversalijax (na materiale kausativnyx glagolv. Konferencia po problemam izucenijauniversal’nyx i areal’nyx svostv jazyka. 55-58. Moskva: Nauka. Pesetsky, D.: 1995, Zero Syntax. Cambridge: MIT Press. Pylkkänen, L.: 1999, Causation and External Arguments. M I T W P L. 35. 161-183. Pylkkänen, L.: 2000, The Syntax of Internal causation. ms. Cambridge: MIT. Ratcliffe, R.: 1997, Prosodic Templates in a Word-based Morphological Analysis of Arabic. Perspectives on Arabic linguistics,.X, 148-171. M. Eid & R Ratcliffe eds. Amsterdam: John Benjamins. Rijkhoff, J.: 1991, Nominal Aspect. Journal of Semantics. 8. 291-309. Seghrouchni, D.: 1989, Les schèmes en arabe. ms. Rabat: Faculty of Letters. Sells, P.: 1997, The Functions of Voice Markers in the Philippine Languages. Morphology and Its Relation to Phonology and Syntax. S. G. Lapointe, D. K. Brentai, 1 P. M. Farell eds. Stanford, CSLI Publications. c
Sibawayhi, A., 8th century, al-Kitaab. ed. A. Haaruun. Cairo: aalam al-kutub. 1974. Szabolcsi, A.: 1987, Functional categories in the noun phrase. Approaches to Hungarian. I. Kenesei. ed. Vol. 2, 167-189. Jate Szeged. Szabolcsi, A.: 1997. ed. Ways of Scope Taking. Dordrecht: KUP. Verkuyl, H. J.: 1993, A Theory of Aspectuality.The Interaction between temporal and atemporal structure. Cambridge: Cambridge Univ. Press. Webelhuth, G.:1998, Causatives and the nature of Argument Structure. Syntax and Semantics. 30. Complex Predicates in nonderivational syntax . New York: Academic Press.
Abdelkader Fassi Fehri
79
Xrakokovskiij, V.S.: 1989, Semanticeskie tipy mnozestva situacij I ix estestvennaja klassifikacija. Tipologija iterativnyx konstrukcij. 5-53. Xrakokovskiij, V.S.:ed. Leningrad: Nauka.