Marantz (1984) notes an asymmetry between subjects and objects with respect to determining the meaning of the whole verb: while the theta role assigned to the object is determined by the verb, the theta role assigned to the subject is determined by the verb
and its internal argument(s). ' Support for this claim comes from the fact that many English verbs differ radically in their interpretation, depending on their choice of object. Some examples o f this are:
(2) a. kill a cockroach, b. kill a conversation.
- In some theories, the terms internal and external refer to variable at the level o f PAS. Under the assumptions made here, I will assume that an argument is external or internal if it appears at a specific syntactic position.
^ Marantz thus argues that the object is theta-marked by the verb, while the subject receives a compositional theta role and is marked by the whole V .
c. kill a bottle (i.e., empty it). d. kill an audience (i.e., wow it)
(3) a. throw a baseball.
b. throw support behind the candidate. c. throw a party. d. throw a fit. (4) a. raise a hand. b. raise money. c. raise children. d. raise com. e. raise a toast.
Raise changes its interpretation according to the object it takes: raise a hand
means lift one's hand up. Raise money means achieve donations. Raise children means
bring up, raise corn means cultivate and raise a toast means drink in honour o f someone.
No such variation occurs if we substitute the subject. The verb will always denote the same type o f event. The role o f determining the type of semantic event together with the verb seems thus to be reserved to the direct object:
With ditransitive verbs, sometimes the two internal arguments determine the type o f reading together, compare, for example, the following:
(i) a. Carla gave Nina a book.
b. Carla gave Nina a headache / the flu.
c. Cardamon pods gave the rice pudding a sharp, distinctive taste.
However, if we keep the direct object as a constant and only change the indirect object, no such change arises:
(ii) a. Carla gave Nina / the postman / every student a book.
(5) a. Nina raised a hand. b. The king raised a hand. c. Every student raised a hand.
So far it was shown that the choice of an object may affect the semantics o f the whole VP: raise a hand differs from raise money, and both differ from raise children. In all these cases the external argument is interpreted as an agent, or at least an active participant. However, in some cases the choice of an object may affect not only the specific role borne by the subject (fund-raiser, hand-raiser etc.) but also the type of role it bears (agent /patient / goal). Consider the following:
(6) a. She caught a thief. (she = agent)
b. She caught the flu. (she = patient /goal)
(7) a. She took a cookie from the tray. (she = agent) b. She took flight. (she = experiencer) c. She took the matter into consideration.(she = experiencer)
(8) a. She got a letter from New York. (she = goal) b. She got Janet to do the job. (she =agent)
This pattern is not an idiosyncrasy o f English. Across languages, verbs tend to exhibit similar properties when being combined with certain object types:
(9) a. Elle a attrapé un voleur. 'She caught a thief,
b. Elle a attrapé un rhume / un coup de soleil.
(10) a. Hi xatfa et ha mixtav / et ha yeled.
'She snatched the letter'. / She kidnapped the child', b. Hi xatfa makat shemesh.
' She got a sunstroke'. (Hebrew)
To conclude this discussion we can say that objects have an effect on the meaning of the whole VP, which other arguments do not have. ^ The verb may be assigned a special meaning according to its choice o f an object, but not according to its choice of a subject. Given this, it is not surprising (as noted, again, by Marantz), that internal and external arguments differ with respect to participation in idiom formation. There are many idioms with a fixed object, (or other internal arguments):
(11) a. eat humble pie. b. kick the bucket. c. give the cold shoulder. d. get cold feet.
e. bury the hatchet.
On the other hand, idioms with a fixed subject are hard to find. ^
This seems to hold at least in all cases o f standard, non stative verbs. With psych verbs, which will be discussed in chapter five, the choice o f a subject does seem to affect the meaning o f the predicate. In particular, whenever the subject cannot be interpreted as agentive, then the verb will have to be interpreted as a stative psych verb if it is interpretable at all:
(i) a. Oedipus killed his father.
b. This show really killed the audience (i.e., wowed them). However, we are dealing here, for the time being, with standard active verbs.
^ Both Bresnan (1982) and Ruwet (1991) argue against Marantz and bring up a number o f idioms with a fixed subject. Marantz (1984, 1997), in response, argues that these are either (i) sentential idioms
{Noblesse oblige), or (ii) idioms which depend on the interrogative mood {What's eating X?), or, in any
case, the free NP in such cases is not an object, and is optional anyway {Time's up (for NP), The ro o f
The head projecting the agent (or the external argument) seems to act as a boundary for idiomatic meanings. Anything generated above this head cannot form part of an idiom:
(12)
X boundary for idiomatic meaning
V P
head projecting D P V ... the agent
Idiom formation data thus indicate that the domain for special meaning is syntactic rather than lexical. ^ I will use this observation as a tool in identifying the VP structure o f some verbs (in particular, causatives - see chapter four): if a certain argument can form part of an idiom then it is probably not generated as an agent.
2.2.2 "Stricter truth conditions" on objects
Consider the following examples:
(13) a. Brits consume 100,000 tons o f beef yearly.
b. The twenty guests at my party drank twenty bottles o f wine. c. The asylum inmates painted the wall.
Suppose we find out that twenty five per cent o f British people do not eat any beef. (13a) could still be true. However, if it turns out that over the last year only 99,000 tons o f beef were consumed in Britain, then (13a) is necessarily false. Similarly, (13b) could be true in a situation in which some o f the guests had no alcohol at all. But if not all the twenty bottles are completely consumed, then it is false. (13c) could be
^ This contrasts with a prevailing view in many theories, lexicalism, which takes the word level to be the domain where special, idiomatic meaning operates (cf. Marantz 1997).
considered true if one inmate did not take part in the painting, but not if part of the wall remained unpainted.
What the sentences in (13) seem to indicate is that the change effected in the object forms an inherent part of the meaning of the verb, while the subject is related to the verb somewhat more "loosely". Pragmatic (or semantic) considerations allow us to take a sentence as true if only a part of the denotation of the subject took part in the action. If, however, only part of the denotation o f the object is affected, we cannot invoke pragmatics to rescue the sentence: it is strictly false.
The object, I assume, forms part of what is assertedhy the verb. Any negation of this would end up in a contradiction, hence the falseness. Subjects, on the other hand, do not force such a requirement. If I say that the guests drank up a certain amount of wine, it is implied, but not asserted that all (or most) of them drank the wine. (13b) would be perhaps odd in a situation in which only two out of the twenty guests had any wine, but it does not assert that all the guests had wine in the same way it asserts that all the amount o f wine was consumed.
Note that the same holds also for less canonical objects, i.e., objects which are not affected, consumed or created:
(14) a. 150 schoolchildren raised over a thousand pounds for charity.
b. Three thousand girl scouts visited five hundred home-bound patients. c. The audience in the festival liked especially five Georgian films.
In (14a) it may be the case that some children did not do any fund raising, but it is necessary that over a thousand pounds were collected.
Putting together these data with those from the former sub-section, I conclude that the verb and its object form a complex which serves as a closed domain for two purposes:
1. The combination o f the verb and its object can affect the meaning o f the VP in a way other arguments cannot.
2. The object forms an inherent part of the denotation o f the event the VP refers to. The change affected in it is what is asserted by the verb. This seems to hold for most non- stative verbs. ^
In the next section I will note a further characteristic which is unique to objects: their aspectual role in the event structure o f the verb.