Typically, the causative agency-head g does not embed a structure which contains both v and a root with an argument. Presumably, this is so since structural accusative Case can only be assigned once in Hebrew, whereas a sentence like (28) would have two arguments, x and y, which require Case, in addition to the nominative z.
(28) z hisˇbir et y et x
z break-caus accy accx ‘z made y break x’
But there are cases where transitive verbs can nonetheless be embedded under the causative head g. For transitive simple verbs, there are two patterns of causativization, described by Cole (1976) and Cole and Shridhar (1977) but left unexplained since: the ‘causee’ (originally the subject of the simple verb) is oblique for some causative verbs derived from transitive verbs, but accusative for others.
It turns out that causative verbs where the causee is oblique are derived from simple verbs which have locative/ experiencer subjects (e.g. ahav ‘love’, sana ‘hate’, ra’a ‘see’, sˇama ‘hear’, kalal ‘include’, lavasˇ ‘wear’, katav ‘write’, nasa ‘take bride’). I assume that what characterizes these verbs is that the root may assign inherent Case to their locative/ experiencer argument. Inherent case does not show up with the simple verb, where this argument is assigned structural nominative Case, as in (29a), but it does show up with causative verbs (29b) and with adjectival passives (29c), where this argument is marked by the preposition al ‘on’.
(29) a. ha-talmid sana et ha-miqcoa’ the student hate-simpl accthe-subject ‘The student hated the subject.’
b. ha-sefer hisni al ha-talmid et ha-miqcoa’ the subject hate-caus on the-student accthe subject ‘The book made the student hate the subject.’
c. ha-miqcoa’ sanu al ha-talmid
the subject hate-simpl-pass-part on the-student ‘The subject is hateful to the student.’
The fact that structurally, locative/experiencer subjects have a diVerence position than agentive/causative subjects has already been argued for by
Landau (1999; 2002), on the basis of independent sets of data. Landau (1999) investigates the distribution of Possessive Datives (Wrst discussed by Borer and Grodzinsky (1986)). According to Landau, Possessive Datives are generated in the speciWer of the possessee and raised to spec-VP. This position is available as a landing site for raising in structures with agentive verbs such as qilqel ‘damage’, since their own subject does not occupy spec-vp, but spec-vp, as shown in (30a). Landau shows that Possessive Datives never co-occur with locative/ experiencer subjects, as exempliWed in (30b) and (30c).
(30) a. [ [spec-vp rina] [[Vþv qilqela] [ [spec-vp le-gili]
Rina damaged to-Gil
[tV et [dp[spec-dp ti] ] ha-sˇa’on] ]
acc the watch
‘Rina damaged Gil’s watch.’
b. gil sana le-rina et ha-tisroqet Gil hated to-Rina accthe hairstyle ‘Gil hated Rina’s hairstyle.’
c. ha-xesˇbon kalal le-rina et ha-aruxa the bill included to-Rina accthe meal ‘The bill included Rina’s meal.’
Landau’s account is that locative/experiencer subjects are mapped to SPEC- VP, unlike causative/agentive subjects, thereby Wlling up the Possessor Dative’s landing site and blocking its extraction in (30b, c). My own account will follow Landau’s in this respect. I assume that the subject argument of locative/ experiencer verbs is not an argument of v (but of the root).
(31)
y sana et x
y hate-SIMPL ACC x ‘y hated x’
R λe [hate (e,x) & Loc (e,y)] y λy λe [hate (e,x) & Loc (e,y)]
x [R sn’] R
(32) y kalal et x
y include-SIMPL ACC x ‘y included x’
R λe [include (e,x) & Loc (e,y)]
y R
x [R kl]
λy λe [include (e,x) & Loc (e,y)]
λx λy λe [include (e,x) & Loc (e,y)] When nominative Case is assigned to a higher argument, e.g. the subject of the causative verb, then the locative/experiencer argument is assigned inherent Case by the root, al in example (33) and be-in example (34):
(33) z hisni et x al y z hate-CAUS ACC x
λe [hate (e,x) & Loc (e,y) & Cause (e,z)] z λy λe [hate (e,x) & Loc (e,y) & Cause (e,z)]
R λe [hate (e,x) & Loc (e,y)]
y R x on y ‘z made y hate x’ g g g
λz λe [Cause (e,z)]
λy λe [hate (e,x) & Loc (e,y)]
[R sn’] λx λy λe [hate (e,x) & Loc (e,y)] (34) z hixlil et x be-y
z include-CAUS ACC x in y ‘z included x in y’
λe [include (e,x) & Loc (e,y) & Cause (e,z)] z
y x R
R
λe [include (e,x) & Loc (e,y)]
g
g
g
λy λe [include (e,x) & Loc (e,y) & Cause (e,z)] λy λe [Cause (e,z)]
λy λe [include (e,x) & Loc (e,y)] [R kl] λx λy λe [include (e,x) & Loc (e,y)]
There is a second type of verb which describes emotional states, sometimes called psych-verbs (e.g. paxad ‘fear’, da’ag ‘worry’, ka’as ‘be annoyed’), where, at least according to the analysis of Belletti and Rizzi (1988), the subject of emotion is a direct argument of the verb. This agrees with the intuition that these verbs classify the subject of emotion as aVected. It also is the case that adjectival passives can be predicated of the subjects of psych-verbs (ka’us ‘annoyed’, da’ug ‘worried’). I think it would be correct to consider the oblique argument of psych-verbs, the source of the emotion, a Cause. Indeed this oblique argument is usually marked by the preposition me-‘from’, which is also used to express the causative relation (as is common cross-linguistically). It turns out that when the causative template embeds a psych-verb, it does not introduce a new argument. Rather, it is the source of emotion which surfaces as the subject of the causative verb. These verbs therefore require a correction to the generalization of section 7.2 whereby the causative template always adds an argument to the simple verb. In examples based on psych-verbs, g and the preposition ‘from’ replace each other. ‘From’ is the expression of inherent Case assigned by the root to its Cause argument.
(35) a. dani paxad me-ha-kelev Dani fear-simpl from-the-dog ‘Dani feared the dog.’
b. ha-kelev hifxid et Dani the dog fear-caus accDani ‘The dog scared Dani.’
Crucially, the relevant thematic role is expressed once but not twice (simi- larly to Pesetky’s (1995) T/SM restriction). Since g assigns the same thematic role as the root, an additional argument of g and the Cause argument of the root cannot co-occur in a single event.
(35) c. ha-nevixot hifxidu et Dani me-ha-kelev the-barking fear-caus accDani from-the-dog ‘The barking caused Dani to fear the dog.’
The same pattern is found for some location verbs. Just as the thematic role of an experiencer subject is the extension of a locative role in the locative/ experiencer subject verbs, the thematic role of the subject of a psych-verb is the extension of an aVected locative role. Cause arguments of roots are also found with verbs which describe such locative relations (e.g. nadaf ‘emanate’, nazal ‘drip’, yarasˇ ‘inherit’, saxar ‘rent’, lava ‘borrow’):
(36) a. re’ax ra’ nadaf me-ha-kelev smell bad emanated-simpl from-the-dog ‘Bad smell emanated from the dog.’
b. ha-kelev hidif re’ax ra’
the dog emanated-caus smell bad ‘The dog emanated bad smell.’
c. ha-haznaxa hidifa re’ax ra’ me-ha-kelev the-neglect emanated-caus smell bad from-the-dog ‘Neglect caused bad smell to emanate from the dog.’
The fact that the subject of psych/locative verbs is a direct argument of the root is further demonstrated by the middle morphology which marks many of these verbs cross-linguistically (e.g. nidbaq ‘be infected-mid’/ hidbiq ‘infect-caus’, nidham ‘be amazed-mid/ hidhim ‘amaze-caus’, nivhal ‘be frightened-mid’/ hivhil ‘frighten-caus’, nignav ‘be excited-mid’/ higniv ‘excite-caus’, ne’elav ‘be oVended-mid’/ he’eliv ‘oVend-mid’, nig’al ‘be dis- gusted-mid’ / hig’il ‘disgust-caus’). Example structures with psych/locative verbs are shown in (37) and (38).
(37) a.
y paxad
y fear-SIMPL from x ‘y feared x’
R
λe [fear (e,y) & Cause (e,x)] λe [Cause (e,x)] mi-x R λe [fear (e,y)]
y [R pxd] λy λe [fear (e,y)] b.
y nadaf
y emanated-SIMPL from x ‘y emanated from x’
λe [Cause (e,x)] mi-x R λe [emanate (e,y)] R
mi- x
mi- x
λe [emanate (e,y) & Cause (e,x)]
(38) a.
x hifxid et y
x fear-CAUS ACC y ‘x frightened y’
λe [fear (e,y) & Cause (e,x)] x
R λe [fear (e,y)] y
g
g
g
λx λe [fear (e,y) & Cause (e,x)] λx λe [Cause (e,x)]
[R pxd] λx λe [fear (e,y)] b.
x hidif et y
x emanated-CAUS ACC y ‘x emanated y’
x
R
g
g
g
λe [emanate (e,y) & Cause (e,x)] λx λe [emanate (e,y) & Cause (e,x)] λx λe [Cause (e,x)] λe [emanate (e,y)]
y [R ndf] λx λe [emanate (e,y)] We now turn to the second pattern of transitive verb causativization, where the causee is accusative. It turns out that these are what I will call verbs of consumption (e.g. safag ‘absorb’, axal ‘eat’, gama’ ‘drink’, yanaq ‘suck’, nasˇam ‘breathe’, ta’an/ amas ‘carry’, xatam ‘sign/ undertake obligation’, lavasˇ ‘wear’). These verbs diVer from the locative/experiencer subject verbs, where the locative subject is the oblique argument of the root. In consumption verbs, the subject is the direct argument of the root, and it is the object which is the root’s oblique argument. The object is assigned structural accusative Case by a simple verb, as in (39a), but in the causative and adjectival passive construc- tions it is inherently case-marked by the preposition be-‘with’, as in (39b) and (39c) respectively:
(39) a. ha-ripud safag et ha-mayim
the upholstery absorb-simpl accthe-water ‘The upholstery absorbed the water.’
b. dani hispig et ha-ripud be-mayim Dani absorb-caus accthe upholstery with water ‘Dani soaked the upholstery with water.’
c. ha-ripud safug be-mayim
the upholstery absorb-simpl-pass-part with water ‘The upholstery is soaked with water.’
Consumption verbs are like psych/locative verbs in that the direct argument of the root surfaces as the subject. Here too adjectival passives may be predicated of the subject: axul ‘someone who has eaten’, sˇatuy ‘drunk’, lavusˇ ‘dressed’, safug ‘soaked’, ta’un/ amus ‘loaded’, xatum ‘signatory’. There are other languages as well, such as Marathi, where the subject of consumption verbs such as ‘eat’ is an internal argument (see Alsina and Joshi 1991). (40) y safag et x
y absorb-SIMPL ACC x ‘y absorbed x’
R λe [absorb (e,y) & with (e,x)] y R λy λe [absorb (e,y) & with (e,x)]
x [R spg] λx λy λe [absorb (e,y) & with (e,x)] (41) z hispig et y be-x
z absorb-CAUS ACC y with x ‘z drenched y with x’
λe [absorb(e,y) & with (e,x) & Cause (e,z)] z λz λe [absorb (e,y) & with (e,x) & Cause (e,z)] λz λe [Cause (e,z)] R λe [absorb (e,y) & with (e,x)]
λz λe [absorb (e,y) & with (e,x)]
x [R spg] λx λy λe [absorb (e,y) & with (e,x)]
g
g
g
R y
In sum, the present system accounts for the two patterns of causativization of Hebrew transitive verbs described by Cole (1976), according to whether it is the subject or the object of the simple verb which is obliquely case-marked.
7.5 Conclusion
This chapter has provided evidence for the realization of the aspectual dimen- sion of agency by functional heads which syntactically merge with roots. One such functional head which has already been argued for in the literature is the light verb v, which introduces the Agent. The present work has provided evidence, based on the morphology of Semitic verbs, for two agency-heads which determine whether the thematic role of the external argument of the verb is Actor or Cause. Neither Actor nor Cause is a role assigned by the root or the light verb v. Morphologically, these two agency-heads mark the verb with intensive or causative morphology. The intensive agency-head is a mod- iWer of the root. The argument of the root that it modiWes is not a participant in the event, but the event itself, which it classiWes as an Action. The causative agency-head merges with a fully constructed verb. Semantically, it is not a modiWer, but introduces its own argument.
The agency dimension adds the marked thematic relations of Actor and Cause to the unmarked thematic relation of Agent, and creates a thematic classiWcation of verbs. As is well known, the temporal aspectual classiWcation is based on the concepts of change and culmination. This chapter has shown that the agency aspectual classiWcation, on the other hand, is based on the concepts of action and causality.