Constraint-based accounts
المؤلف:
Paul Warren
المصدر:
Introducing Psycholinguistics
الجزء والصفحة:
P192
2025-11-11
11
Constraint-based accounts
The preceding sections have introduced a range of factors that influence the interpretation of sentences. Many of the experimental results presented in those sections indicate developments beyond the syntax-first approach introduced in Chapter 10. We can think of the syntax-first approach as an example of a restricted account of sentence processing, in which only certain types of information are used in the initial analysis of sentences, even when other sources of information would potentially be useful. These restricted accounts usually assume that sentence comprehension involves serial processing, in two senses of this term. The first sense is that one source of information is processed before others, e.g. syntactic information before semantic. The second sense is that one sentence structure is considered at a time, meaning that a new revised analysis of the input is required if the first analysis fails.
Unrestricted accounts of processing, by contrast, claim that many or all types of information can be used during processing, including at the earliest stages. Such accounts often also allow multiple sentence interpretations to be considered in parallel, with the different possibilities ranked or weighted according to how well they fit the evidence from the multiple information sources. They are also usually interactive, in that the different information sources are involved in an interplay that moves the processor towards a coherent interpretation. Different versions of this type of account differ in their details, for instance in the nature and frequency of the interaction of information sources.
Weak interactive accounts include those that maintain that the inter action between syntactic and other sources occurs only when the syntactic analysis requires it. This can be seen as an extension of the syntax-first approach, in that the syntactic parser usually remains the central driving force in sentence interpretation. These accounts differ from the syntax first approach, though, because instead of blindly pursuing an analysis based on default parsing strategies, it is assumed that the processor will cast around for information from non-syntactic sources whenever it encounters a structural ambiguity, such as whether an ambiguous verb is being used transitively or intransitively, or whether a PP is modifying a verb or a noun.
In addition there are strong interactive accounts, where the non-syntactic sources play a more determining role in sentence analysis, and are not subservient to a central syntactic processor. These include constraint-based approaches (Boland, Tanenhaus & Garnsey, 1990; Macdonald, 1994; Taraban & McClelland, 1988; Trueswell et al., 1994). These accounts claim that each of the various types of information available to a reader or listener is used to determine the analysis of a sentence. Frequently, but not always, constraint-based accounts assume a parallel processing approach, with multiple alternative analyses being assessed. These alternatives may be in competition with one another, in which case the processing difficulty associated with a particular sentence will depend on the strength of the competition between this and alternative analyses. This is reflected in digging-in’ effects. Commitment to one particular interpretation results in inhibition of competing interpretations. In the absence of correcting information, the competing interpretations become increasingly inhibited, resulting in a more severe garden path experience. In example (11.61), the initial preference to see the town as the object of invaded has to be undone i.e. results in garden path effects when as is encountered. In (11.62), the text that is read after o does not contradict the late closure of the first clause, and the garden path effect on encountering as is even stronger (Ferreira & Henderson, 1991).
Individual investigations of constraint-based sentence processing have tended to focus on specific types of constraint, adding detail to the overall picture of what drives sentence interpretation. Some of these studies have, as in some of the examples referred to earlier, taken as their starting point published materials claimed to exhibit garden-pathing, and have attempted to determine whether there are non-syntactic properties of the sentences that might cause or allow the garden path to occur.
Further factors that have been shown to influence sentence processing include fictional contexts that provide a motivation for a sentence that might otherwise be anomalous. For instance, when participants read the sentence in (11.63), their knowledge of the world will generally tell them that it contains an anomaly, as cats do not pick up chainsaws. In the fictional world of Tom and Jerry cartoons, however, this sentence is not odd.
An experiment presented sentences like (11.63) following neutral or fictional contexts (Filik & Leuthold, 2008). The experiment measured event related brain potentials ERPs as a measure of brain activity while participants performed a reading task. An index of semantic processing, the N400 effect see sidebar, showed a significant reduction in processing load when sentences like (11.63) followed an appropriate fictional context.
الاكثر قراءة في Linguistics fields
اخر الاخبار
اخبار العتبة العباسية المقدسة
