Lexical preferences
Under a syntax-first approach, when each word in a sentence is encountered it is built into the sentence tree as a member of a particular syntactic category verb, noun, and so on, just as long as the syntactic category membership of a word is known see Chapter 10. Any preferences for particular syntactic structures are a result of parsing operations and not of any differences in the structures associated with individual words. However, it has been claimed that not all words are equal in this respect.
For example, some verbs occur more typically in certain constructions than in others. As another example, the type of noun that is the subject of a verb can sometimes affect the interpretation of that verb. The following material gives examples of each of these.
Verb preferences
Recall that in Chapter 10 we looked at sentences like those repeated in (11.27) and (11.28). A partial syntactic tree is also given for each sentence. These sentences contain verbs whose position is shown by the circles that can be used in either of the two structures shown. In (11.27) the verb is followed by a noun phrase object, which is the preferred structure according to the Minimal Attachment strategy. In the more complex structure in (11.28), the verb is followed by a subordinate clause in object position. Note the extra S node between the NP the schedule and the VP node. This is a non-minimal attachment of the NP.
In one study, Holmes (1987) pre-tested a list of verbs to see which kind of object each verb was more likely to occur with. Two sets of verbs were then selected. In one set the verbs were more likely to be used with a noun phrase object (including verbs such as hear, read, see, answer). In the other, the preference was for a clause as the object (daim, know, doubt, believe etc.). Pairs of sentences were made for each verb. All sentences were used in the experiment with a complement clause structure see (11.28), i.e. the structure that is dispreferred by Minimal Attachment. The members of each pair differed according to whether they included the complementiser that. Examples are given in (11.29)–(11.32).
The task used was self-paced word-by-word reading. The crucial measure is the reading time for words in the position of was in (11.30) or would in (11.32). Before this point these sentences are perfectly compatible with the structure preferred by Minimal Attachment, but was and would make it clear that the sentence in fact involves a clause as the object of the verb e.g. the reporter in (11.30) did not literally see her friend, but saw – i.e. realised – that her friend was not succeeding. Therefore these words are the point at which a garden path should be experienced. Note that in (11.29) and (11.31) the structure with a clause as its object becomes obvious earlier, because of the inclusion of the complementiser that, and so no garden path should be experienced. These sentences in (11.29) and (11.31) provide a baseline, against which (11.30) and (11.32) can be compared, to see if there is any evidence for a garden path.
The results showed that there was indeed a strongly experienced garden path. Importantly, though, this was found only for the verbs for which the pre-test had shown a bias towards a noun phrase object construction, i.e. for sentences like (11.30). When participants read sentences containing verbs with a clause object preference, such as (11.32), there was no garden path, i.e. there was no difference in reading times for would in this sentence compared with (11.31). This indicates that the early syntactic analysis of sentences does not simply build structures using the syntactic categories of the input words, but actually takes into account the structural preferences associated with individual lexical items.
Kennison 2002 used both self-paced phrase-by-phrase reading and an eye movement tracking study to measure participants’ processing of sentences e o with two sets of transitivity-ambiguous verbs that showed a preference towards either transitive usage (e.g. read, as in (11.33)) or intransitive usage (e.g. perform, as in (11.34)). The NP immediately following the verb was either an argument of the verb, as in these examples, or an adverbial adjunct, which was every week for the counterparts to these examples. Kennison found that after transitive-bias verbs (read) argument NPs (every play) were read faster than adjunct NPs (every week), but that after intransitive-bias verbs (perform) argument NPs were read more slowly than adjunct NPs.
Noun preferences
Nouns also enter into different structures, depending on the semantic properties of the noun. For example, some nouns denote animate entities (people, animals, etc). that are capable of performing certain conscious and deliberate actions, while others denote inanimate entities physical objects, etc. that would not be expected to perform such actions. In addition, English has many verbs that exhibit transitivity ambiguity. When such verbs are used with animate entities as their subjects, then they can be used comfortably in both intransitive (11.35) and transitive (11.36) structures.
When the subject of such verbs is inanimate, the intransitive reading is preferred, as in (11.37), with the transitive reading in (11.38) being less plausible.
The parsing strategy of Late Closure predicts that verbs which can be either transitive or intransitive will always be interpreted as transitive if they are followed by a NP, because that following NP will be attached into the current syntactic tree as the verb’s object. This means that garden pathing should be found at the position of as in both (11.35) and (11.37), because this word indicates that the driver is not the direct object of stopped, but is instead the subject of a new clause.
In another self-paced reading task, however, Stowe (1989) found garden pathing for sentences in which the subject of the first verb was animate, such as (11.35), but not for sentences with inanimate subjects, such as (11.37). This indicates that semantic properties of the subject noun in (11.35) affected the syntactic interpretation of the verb in the first clause, i.e. trucks are not usually expected to stop something else, but the police are. Again, this is evidence that the initial interpretation of sentences takes into account the structural preferences associated with lexical items, in this case the combination of the animacy of the subject noun and the transitivity or otherwise of the verb.
Thematic roles
Earlier in the chapter we encountered sentences that were used to demonstrate that the plausibility of a sentence did not have an effect on parsing preferences. The examples used the verb form examined in combination with different preceding nouns, as in the fragments repeated here in (11.39) and (11.40).
The eye movement data reported for these materials was taken to indicate that the past tense verb interpretation of examined was preferred following syntactic principles in both cases, regardless of plausibility (Ferreira & Clifton, 1986).
In a further examination of such materials, Trueswell et al. (1994) argued that the reason no plausibility effect was found using these mate rials was that the plausibility difference was not strong enough. In effect, they claimed that both defendants and evidence are more likely to be examined than to do the examining. That is, they are more likely to be understood in the thematic role of THEME or object of the verb than in the role of AGENT (subject, in this case). To test this claim, Trueswell et al. carried out further experiments where the contrast between the nouns was stronger. For example, the expected role of archeologist in (11.41) is much more clearly that of the AGENT of examined than could be argued for defendant in (11.39).
When reading patterns for sentences beginning with the fragments in (11.41) and (11.42) were compared, a clear difference was found in the likelihood of garden-pathing, reflecting this difference in the plausibility of the noun being the AGENT of the verb. The lack of such an effect in the original experiment with the materials in (11.39) and (11.40) could therefore be a result of an insufficiently distinct difference in plausibility.
The examples just discussed have shown that the same constituent, syn tactically defined, can play a different role in a sentence. Another case that has been studied involves prepositional phrases PPs. The PP (for Chris) in (11.43) can modify either the verb (bought) or the noun (book). Following the parsing strategy of Minimal Attachment, the preferred analysis is that the PP modifies the verb. This claim has been supported in many reading studies.
However, even when PPs are in the same syntactic relationship to the rest of the sentence, they can differ in their thematic roles. This is illustrated in the examples in (11.44) and (11.45), taken from Taraban and McClelland (1988). Note that different words appear in the PP with the … but in both of these sentences the PP tells us something about the cleaning operation. Nevertheless, even though they are in the same structural relationship to the rest of the sentence, the roles played by the PPs in the action denoted by deaned are different. Broom is an INSTRUMENT (a thing used to carry out the action of the verb), while manager is some sort of ACCOMPANIMENT.
The sentence in (11.46) contrasts with those in (11.44) and (11.45), because the relationship of the PP to the rest of the sentence is different. The PP now tells us about the cupboard, rather than about the cleaning. (11.44) and (11.45) are preferred by Minimal Attachment, while (11.46) is dispreferred.
A self-paced reading study showed that (11.44) was easier than both (11.45) and (11.46), and that (11.45) and (11.46) did not differ. This shows that it is not just a syntactic preference that determines the interpretation of the PP, but also the thematic role. In other words, a PP modifying the verb in this sentence is expected to be in the role of INSTR MENT rather than ACCOMPANIMENT. Other sentences in this experiment also demonstrated that the non-minimal i.e. noun-modifying attachment of PPs was in some cases the preferred attachment, depending on the words used in the sentence especially the verb and the object of the verb, and that the strength of this preference also depended on the role of the PP in the meaning structure of the sentence.
As another example of how thematic role preferences can affect interpretation, consider the example in (11.47). This comes from an announcement sent out by email to staff at the author’s university, warning them about impending roadworks outside the university buildings:
The idea of a road surface made out of heavy machinery rather than the more conventional asphalt caused some amusement. The effect is a result of the interpretation of the PP . . . as showing what the replacement will be rather than showing as intended the INSTR MENT of the verb action.
The examples discussed in this section show that the processing of sentences does not depend on syntactic structure alone, but that the roles that parts of a sentence play in the meaning of that sentence are important, as too are the types of structure that individual words typically occur in.
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
