Return now to the main claim of 4.4.2: Hebrew differs from some other languages not in its well-formedness rules (which are given by UG) but in the range of lexical items which allow causativization. This is reflected by children's mastery of causativization. I concentrate on Hebrew and English speaking children. My assumptions then are as follows:
(56) a. UG constraints (including FI) are available to children.
b. Particular lexical items, or lexicalization patterns which are particular to their language, may not be available to children.
Children should expect, therefore, that there should be no verbs with two originators or two measurers. For example, there should be no Hebrew verb o f the form *heherig
(cause to kill) or *himriax (cause to smear), or no English verbs such as * smear, as in
(57):
(57) I am going to smear him butter on the bread
(meaning "I am going to make him smear butter on the bread").
Forms such as (57) are not reported for English or Hebrew. When Hebrew speaking children come across a verbal form which has both event participants, originator and measurer, they can expect one of the following:
(58) a. A causative form in hiCCiC does not exist (e.g. *heherig, cause to kill). b. The event can be causativized, if the former external argument can be made into a measurer (as with feed, pass, etc.).
c. A hiCCiC form exists, but it denotes a different event (zarak, throw vs.
hizrik, inject).
If children know the meaning of zarak (throw), they can expect that hizrik, if it exists, does not denote a throwing event.
On the other hand, when there is no originator there should be no problem to causativize the verb. This is the case of unaccusatives, which have only one participant, a measurer. Unaccusatives are indeed easily causativized by children in both English and Hebrew (Berman and Sagi 1981, Bowerman 1982, Borer and Wexler 1987, Pinker 1989).
However, it is predicted that children might make mistakes whenever non-UG factors are involved: particular lexical items in their language, particular lexicalization patterns or language-particular morphology. I will go through each of these cases.
Children do not know particular lexical items:
The pattern o f adding an originator to an unaccusative predicate is allowed by UG, but this does not mean that any unaccusative verb can be causativized. There is a group o f unaccusative verbs which do not causativize in English or Hebrew, including, among others, come, arrive and faint (cf. HK 1997b). Other verbs, such as disappear or
die, allow causativization in Hebrew, but not in English. Children may not know whether in their language disappear may be causativized. We thus find utterances such as (59) (from Bowerman 1982, in Pinker 1989):
(59) a. C, 7;8: Did they vanish "knock-knock" cups? (Noticing Dixie cups in new pack no longer have knock-knock jokes on them)
b. Scot, 5:0: I disappeared a bear in the back of the car; that's why you can't see him.
c. C, 5;0: ...Eva's gonna die it. She's gonna make it die.
Similarly, children might not know that the causativized form in their language is, in fact, a different lexical item (i.e. pairs such as rise-raise, eat-feed). This may result in utterances as in (60) (ibid.):
(60) a. C, 3;3: But I can't eat her! (=feed) b. Rachel, 2;0: Don't eat it me (=feed)
Children do not know language-particular patterns:
Languages might differ in allowing action verbs to have their arguments turned into measurers. I noted in section 4.4.3 that verbs o f manner o f motion {run, jump, etc.) are used as transitives in both Hebrew and English. In Hebrew, this pattern is extended to a larger set o f verbs: laugh, eat, see, breathe, etc.
Children know that in such cases adding an originator is, in principle, allowed, but they do not know if their language allows such "internalization" o f an external argument. We thus get an over-generalization of causativization in English (Bowerman 1982):
(61) a. E, 3;0: Don't giggle me. (As D tickles E)
b. E, 5:3: You cried her! (After M drops E's doll and it squeals) c. E, 5;3: This is aching my legs (As she climbs a long flight o f stairs)
* Borer and Wexler (1987) argue that internalization o f an external argument is marked in English while it is unmarked in Hebrew. In any case, internalization is a possible (i.e. allowed by UG) process, which is not used equally by all languages.
d. E, 2;1: 1 wanta swim that (Holding an object in the air and wiggling it as if it were swimming)
e. C, 3; 1 : I'm singing him (Pulling string on cow-shaped music box). f. E, 2;2: I'm talking my birdie (Pulling string on bird-shaped music box).
Note that all the causative forms in (61) involve putting someone into action, rather than making someone or initiate an action (as in analytic causatives). In other words, children know that the lexical causative is associated with a manipulative interpretation (which is, presumably, a universal pattern, derived from the syntactic properties o f this form). In (6Id) swim behaves like run in (51b): an activity that can be initiated by an outside cause. When wanting to "swim" the object, the child presumably means an activity in which the object undergoes something, rather than initiates the swimming. As expected, children do not know language-specific patterns, but they do know the universal constraints on causativization and other syntactic operations. Overgeneralization of the causative is exhibited in Hebrew as well: saxa (swim) and
calal (dive) do not have a causative hiCCiC form, but children use them in such a way nonetheless (Borer 1995):
(62) a. ba yam aba masxe oti
in the sea daddy make-swim me
b. at kimat hiclalt oti you almost make-dive me
Children know verb argument-structure but not verb morphology:
In Hebrew the transition from a non-causative into a causative form involves a change in verb morphology (CaCaC/niCCaC vs. hiCCiC). Children exhibit knowledge of the argument structure associated with causative and non-causative forms, but their mastery o f verb morphology is not complete. Borer (1995) presents two types of morphological deficiency. In the first, children use the CaCaC form instead o f hiCCiC.
Note, however, that they know the argument structure associated with causatives, and thus include two event participants:
(63) a. ra'iti et ha ciyurim le aba
I saw OM the paintings to daddy
{ra 'iti, I saw, used for her'eti, I showed)
b. zuzi li et ha kise
move-intrans. form e OM the chair
{zuzi, move-intrans., for tazizi, move-trans.)
The second type of morphological mistakes is overgeneralization of hiCCiC forms for all transitive verbs, including those in CaCaC:
(64) a. lama at madxifa oti kaxa
why you push me this way
{madxifa for doxefet, push)
b. ze mamas masrif oti, ha semes
it really bums me, the sun
{masrif for soref bum)
These forms are hyper-corrections: hidxif is used for daxaf as is evident from the argument stmcture the child uses (one participant who pushes, another who is pushed). It does not mean "cause someone to push".
To sum up: children's use of verbs makes it evident that they are aware of the change in the event stmcture associated with causativization. Although they are not always familiar with the morphological marking of such changes in their language, with language-specific pattems (intemalization) or with particular lexical items, they never violate universal principles such as FI.
Chapter 5: Psych notes - the syntax of psychological predicates
Abstract:
In this chapter I suggest an analysis of psych verbs. Taking Object Experiencer verbs (GbJBxp verbs) as my starting point, I will show that they participate in two types of alternations. The first alternation is event-structure based: these verbs can have a stative, eventive or agentive interpretation. Examining data from six languages, I will show that Object Experiencer verbs exhibit specific syntactic properties only on their stative reading. On their agentive reading they behave like standard transitive predicates. This establishes aspectual properties such as stativity and agentivity as crucial to the interface with the syntax.
The second alternation that Object Experiencer verbs participate in is between a psych interpretation and a standard, "physical" interpretation. I will use these data to argue that there is no such thing as an "experiencer" or "a psych construction": (just about) any argument can be interpreted as an experiencer, and just about any verb can have a psych interpretation.
Based on the structural properties of psych verbs in English, French, Hebrew and Irish I will motivate an analysis of psych verbs which is essentially similar to locatives or datives. I will argue that this analysis is more adequate than the unaccusative analysis suggested by Belletti and Rizzi (1988).
My conclusion will be that the syntax is sensitive to aspectual properties, and that it may narrow down the set of event types associated with a certain structure (causation, location change of state etc.). However, it is not sensitive to thematic labels such "experiencer": the interpretation of an argument as an experiencer, a locative or a goal will be determined by the specific properties of the verb and its arguments.
The chapter is organized as follows:
Section 5.1 introduces the three readings of ObjExp verbs.
In section 5.2 I show that all the psych properties disappear once an agentive reading is forced.
Section 5.3 discusses the properties of the eventive reading. Section 5.4 suggests a syntactic analysis of ObjExp verbs.
Section 5.5 argues against the unaccusative analysis of ObjExp verbs. Section 5.6 discusses the structure of Subject Experiencer verbs.
In section 5.7 I discuss Italian ObjExp verbs specifically, suggesting a way to account for their syntactic irregularities. This section also discusses the properties of a particular group of Italian ObjExp verbs, the piacere group.
Finally, in section 5.8 I point out the consequences of this chapter for the theory of psych verbs and for the syntax-lexicon interface in general.