Andrew Radford - Linguistics An Introduction [Second Edition]

(488) a.

The elephant squeezed into a telephone booth collapsed

b.

The elephant which was squeezed into a telephone booth collapsed

Despite their well-formedness, many listeners are confused by garden-path

sentences such as (488a). The illusion can be explained in terms of processing considerations, specifically by the parser’s preference for making local processing decisions. To be maximally efficient the parser attempts to close phrases and clauses as soon as possible. In the case of (488a), however, this strategy leads the parser up a garden path: taking the sequence the elephant squeezed into a telephone booth from (488a), the parser assumes that the clause is closed, and this requires that the elephant is the (logical) subject of squeezed, a clear mistake – in fact, the elephant has to be interpreted as the complement of the passive participle form squeezed, and to have undergone movement from complement to subject

position, as described in section 21. Thus, when the parser reaches the second verb (collapsed in this case), time-consuming and laborious reprocessing is necessary to escape from the illusion. Indeed, such is the strength of this illusion that some native speakers experience considerable difficulty in escaping from it at all (exercise 5).

In this section, we have looked at some aspects of how people assign struc-

tures to strings of words with two main themes in mind. Firstly, we wanted to establish that the grammatical constructs developed by linguists as part of their theory of grammar do play a role in sentence processing. Of course, it would be a

Sentence processing

375

puzzling situation if a mentally represented grammar (theory of competence)

were not put to work in sentence perception and production (linguistic performance). Nevertheless, it is reassuring to find experimental evidence which

indicates that constituent structure and antecedent–trace relations are actively involved in processing.

Secondly, we have acknowledged that the theory of grammar does not provide a complete account of sentence processing, and we have looked at different kinds of sentences that cause processing difficulties, even though they are perfectly grammatical. Processing considerations which go beyond the rules and principles of grammar are necessary to understand these phenomena. The idea that the human parser has a strong preference for operating with local operations is a key idea in this area of research.

Exercises

1.

Most psycholinguistic studies on the processing of empty categories

have been done on English. Cross-modal priming experiments, for

instance, have shown that a moved constituent is reactivated at the

hypothesised trace position in sentences such as (a).

(a) Which book did you buy [which book] last week?

Some researchers have argued, however, that such reactivation effects

do not necessarily indicate that a trace copy must be present in the

listener’s mental representation of the sentence but can equally well be

explained in terms of direct lexical association: on encountering the

verb buy, listeners automatically reconstruct all the verb’s arguments,

including the displaced direct object which book. How (if at all) might

it be possible to dissociate the Trace Reactivation Hypothesis and the

Direct Association Hypothesis empirically? Consider, for example,

how a sentence such as (b) might be processed on-line.

(b) To which butcher did the woman who had just inherited a large

sum of money give the very expensive gift the other day?

2.

Object relative clauses (i.e. those where the relative clause modifies a

direct object, as in b below) are more difficult to process than subject

relative clauses (i.e. those where the relative clause modifies a subject,

as in (a) below):

(a) The reporter who attacked the senator admitted the error

(b) The reporter who the senator attacked admitted the error

How can we explain this difference?

3.

Compare the sentence pairs in (a), (b) and (c) and explain why the sentences in (i) are more difficult to process than those in (ii). Discuss what these contrasts might mean for the idea that syntactic parsing is

376

senten ces

autonomous, i.e. independent of other sources of information (e.g.

lexical information); see also section 14.

(a) i. John warned his mother was dangerous

ii. John knew his mother was dangerous

(b) i. Even before the police stopped the driver was getting nervous

ii. Even before the truck stopped the driver was getting nervous

(c) i. The secretary didn’t quit because of her large raise

ii. The secretary didn’t quit because of her low salary

4.

What problems do sentences like the following pose for sentence

processing, and how can they be explained?

(a) Mary figured that Susan wanted to take the train to Liverpool out

(b) The woman the man the girl loved met died of cholera

(c) The brother of the girl who was famous came to visit us

5.

Explain why and how garden-path sentences pose processing pro-

blems for sentence comprehension. Consider the following sentences

and any others that might be useful in your discussion:

(a) While Mary was mending the sock fell off her lap

(b) John told the girl that Bill liked about the problem

(c) Sue gave the man the dog bit the package

(d) The dealer sold the forgery complained

26

Syntactic disorders

The study of syntactic errors in language-disordered patients is an area in which linguists, psychologists and speech therapists have collaborated extensively.

Recent syntactic theories have been applied to neurolinguistic data and have led to a better understanding of patients’ linguistic problems; in turn, theoretical linguists have gained a new source of data from syntactic errors to test their theories.

Generative linguists in particular have shown interest in syntactic disorders.

Recall that many generative linguists (particularly Noam Chomsky and his followers) claim that humans possess a language-specific cognitive system (embody-ing principles of Universal Grammar) that underlies the production and

comprehension of sentences. Syntactic principles are said to be unique to language, and autonomous of non-linguistic cognitive systems such as vision, hearing, reasoning, or memory (see the introduction, p. 11). This view of syntax makes two interesting predictions about language disorders. Firstly, we would expect to find cases of language disorders in which knowledge of syntax is

impaired while other cognitive systems remain unaffected: if the syntactic system is indeed autonomous, then it should be possible for it to be selectively impaired, for example as a result of brain lesions or genetic deficits. The second prediction is that syntactic disorders should involve impairments of both language production and language comprehension. If the linguistic view is correct, and there is indeed only one underlying system of syntactic principles which is crucially involved in both sentence production and sentence comprehension, then an impairment of the underlying system should manifest itself not only in sentence production but also in sentence comprehension and in grammaticality judgement tasks.

These predictions have mainly been tested in the context of the phenomenon of agrammatism, which typically occurs in Broca’s aphasics, and (to a lesser extent) on the so-called paragrammatic errors from Wernicke’s aphasics. In addition to these two areas of enquiry, some years ago psycholinguists started to investigate developmental language disorders, particularly Specific Language Impairment

(SLI), from a syntactic perspective. These three cases of language disorders are unique, in that patients show syntactic impairment while, at the same time, other cognitive functions seem to be unimpaired. In this section, we will describe the syntactic errors that typically occur in agrammatism, paragrammatism and SLI, and we will show what we can learn from applying syntactic theory, as it has been introduced in this part of the book, to the study of these disorders.