Presuppositions, plausibility and parsing
المؤلف:
Paul Warren
المصدر:
Introducing Psycholinguistics
الجزء والصفحة:
P182
2025-11-10
56
Presuppositions, plausibility and parsing
Recall sentence (11.20), which was introduced in Chapter 10 as a famous example of a garden path sentence (Bever, 1970).
In the original research using sentences such as this, it was claimed that a structural preference results in the interpretation of the verb ae as a past tense form, rather than as a past participle in a reduced relative construction. This analysis turns out to be wrong when the final word ell is encountered; the reader is garden pathed and has to revise their analysis. No such effect would be predicted in (11.21), since the final word confirms that the past tense interpretation is the correct one.
In the preceding section we saw several examples of sentences from studies claiming syntax-driven processing which also involve this ambiguity between past tense verbs and reduced relatives. Critics of the syntax-first approach to sentence processing maintain that there are additional issues to be considered with respect to many such sentences. For instance, Crain and Steedman (1985) claim that reduced relative clause constructions such as that in (11.20) carry some assumptions or presuppositions that are not found in the past tense verb interpretation. Moreover, these presuppositions make the relative clause interpretation less plausible, in the absence of further contextualising information. These presuppositions arise primarily because the relative clause in (11.20) is a restrictive relative, as shown by the absence of commas see sidebar. By way of contrast, (11.22) is not a restrictive relative.
Under their Referential Hypothesis Crain and Steedman argue that for the structure in (11.20) to be motivated we have to assume that there is a set of possible horses, and that one of these horses is being identified by the use of the restrictive relative. In contrast, in (11.21) and (11.22) no other horses are presupposed. These sentences simply introduce one definite horse. It is claimed that it is this difference between (11.20) on the one hand and (11.21) and (11.22) on the other that explains the relative processing difficulties for 11.20, and not a garden path resulting from syntactic preferences.
If this alternative explanation is correct, then one way in which it might have its effect in terms of a processing account is as follows. When the horse raced has been read in 11.20 or 11.21, a syntactic ambiguity is detected which involves the interpretation of ae as either a past tense verb or a past participle. Because more than one syntactic analysis is available, a call is made for assistance from other information sources, including mechanisms that assess the plausibility of a sentence. The past tense verb reading is more straightforward, because the alternative, i.e. the restrictive relative reading, carries the more complex presupposition as pointed out above. The past tense verb reading is therefore the preferred interpretation. So the processing difficulty with (11.20) is argued to result from a plausibility constraint, rather than from a syntactic one.
A clear difference between this account and the parsing strategy approach outlined in Chapter 10 is the relationship between syntactic and non-syntactic information sources. In the parsing strategy account, the syntactic analysis has to have made a misanalysis i.e. it has to be garden pathed before other information is used. The account sketched above makes an earlier appeal to non-syntactic information, at the point where a structural ambiguity is detected.
There are other garden path effects that might be explained in a similar way to the example above. Take for instance the sentence in (11.23). This contains a subordinate clause (that he was having trouble with). This particular subordinate clause is a relative clause, modifying the wife (i.e. it tells us which wife; this would be made more obvious if who were used instead of that).
Research results show that this sentence is much harder to process than the sentence in (11.24) where the string of words after wife also forms a subordinate clause, but this time an object clause, i.e. a clause which is the object of tell. It is the X in The psychologist told the wife X, and it could answer the question What did the psychiatrist tell the wife’ Note that in this case that cannot be replaced by who.
In syntactic terms, it is argued that the Minimal Attachment strategy leads readers to prefer to interpret the string that he was having trouble with … as part of an object clause, and that this is why (11.23) is more difficult to process than (11.24). However, the relative clause in (11.23), like that in (11.20), is a restrictive relative clause, and so the sentence presupposes that there is more than one wife involved. The competing explanation is therefore that it is the presence of this presupposition that makes interpretation of the sentence more complex, and not the mistaken application of a parsing strategy.
This alternative explanation is backed up by further tests by Crain and Steedman 1985 in which they altered the presuppositions associated with a sentence by preceding it with different kinds of context. These could either bias the reader towards a relative clause interpretation of that he was having trouble with, as would be the case with (11.25), or towards an object clause interpretation (11.26).
In different conditions of the experiment, each of these contexts was combined with each of the sentences in (11.23) and 911.24). Participants had to decide whether (11.23) and (11.24) were grammatical or not. Overall, there were no more ungrammatical’ responses to relative clause structures (11.23) than there were to object clause structures (11.24). Importantly, though, the greatest number of ungrammatical’ responses for each type resulted from a mismatch between the context and the test sentence i.e. when (11.23) followed (11.26) or (11.24) followed (11.25).
These findings contradict the reading studies outlined earlier, which concluded that syntactic analysis is not directly affected by semantic fac tors such as plausibility. Is the difference between these two groups of findings due to differences in the types of sentence used, or perhaps to differences between offline and online tasks After all, grammaticality judgements are made some time after the sentence has been read. This means that they could be influenced by a range of other factors that may have a late effect on interpretation. These results would therefore be compatible with the notion that an initial syntactic analysis is subsequently checked by the output of a thematic, or meaning-based, processor.
The next section presents evidence of more immediate effects on syn tactic processing, as the outcome of biases that result from the specific words used in structurally ambiguous sentences.
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
