Psycholinguistics
As noted above, the psycholinguist addresses the question of how the mentally represented grammar (linguistic competence) is employed in the production and comprehension of speech (linguistic performance). The most direct way to approach this relationship is to adopt the hypothesis that a generative grammar can simply be regarded as itself providing an account of how we understand and produce sentences in real time. From the point of view of language comprehension, this gives rise to the following (highly simplified) model, where the input is a stretch of spoken or written language such as a particular sentence:
(14) input phonological lexical syntactic semantic output
processor processor processor processor
In terms of this rather crude model, the first step in language comprehension is to use the phonological processor to identify the sounds (or written symbols) occurring in the input. Then, the lexical processor identifies the component words. The next step is for the syntactic processor (also called the parser, and incorporating the syntactic component of the grammar) to provide a syntactic representation of the sentence (i.e. a representation of how the sentence is struc-tured out of phrases and the phrases out of words). The last step is for the semantic processor to compute a meaning representation for the sentence, on the basis of the syntactic and lexical information supplied by earlier stages in the process. The relevant meaning representation serves as the output of the model: once this has been computed, we have understood the sentence.
An important characteristic of (14), as of all models of psycholinguistic proces-sing, is that its various stages are to be viewed as taking place in real time, and a consequence of this is that psycholinguists can utilise their experimental techni-ques to try to measure the duration of specific parts of the process and link these measurements to levels of complexity as defined by the grammar itself. In fact, it is fairly easy to see that the idea that the grammar can, without any additional
10 linguistics
considerations, serve as a model of sentence comprehension is implausible. A sentence such as (15) is known as a garden-path sentence:
(15) The soldiers marched across the parade ground are a disgrace
A common reaction to (15) from native speakers of English is that it is not an acceptable sentence. However, this reaction can often be modified by asking native speakers to consider the sentences in (16) (recall our observation that not all linguistic data have immediately obvious properties):
(16) a. The soldiers who were driven across the parade ground are a disgrace
b. The soldiers driven across the parade ground are a disgrace
c. The soldiers who were marched across the parade ground are a disgrace
Sentence (16a) should be regarded as entirely straightforward, and we can view (16b) as ‘derived’ from it by deleting the sequence of words who were. Now, if we delete who were from sentence (16c), which should also be recognised as an acceptable English sentence, we ‘derive’ (15), and at this point many readers are likely to change their reaction to (15): it is an acceptable English sentence, so long as it is interpreted with the phrase the soldiers as the logical object of marched (see p. 5 above). When we read (15) for the first time, we immediately interpret the soldiers as the logical subject of marched – the soldiers are marching rather than being marched; as a consequence, the sequence the soldiers marched across
the parade ground is interpreted as a complete sentence and the sentence proces-sor doesn’t know what to do with are a disgrace. The sentence processor has been ‘garden-pathed’, i.e. sent down the wrong analysis route (exercise 6).
What is important about garden-path sentences is that they show that sentence comprehension must involve something in addition to the grammar. As far as the grammar is concerned, (15) is an acceptable structure with only one interpretation. However, it appears that this structure and interpretation are not readily available in sentence processing, suggesting that the parser must rely (to its detriment in this case) on something beyond the principles which determine acceptable combina-tions of words and phrases.
There are other aspects of (14) which are controversial and have given rise to
large numbers of experimental psycholinguistic studies. For instance, there is no place in (14) for non-linguistic general knowledge about the world; according
to (14), interpretations are computed entirely on the basis of linguistic properties of expressions without taking any account of their plausibility, and an alternative would allow encyclopaedic general knowledge to ‘penetrate’ sentence perception
and guide it to more likely interpretations. A further assumption in (14) is that the different sub-components are serially ordered (in that the first stage is pho-
nological processing which does its job before handing on to lexical processing, etc.) An alternative would allow syntactic and semantic factors to influence phonological and lexical processing, for semantic factors to influence syntactic processing, etc. These issues, along with several others, will be discussed in sections 14 and 26.
Psycholinguistics
As noted above, the psycholinguist addresses the question of how the mentally represented grammar (linguistic competence) is employed in the production and comprehension of speech (linguistic performance). The most direct way to approach this relationship is to adopt the hypothesis that a generative grammar can simply be regarded as itself providing an account of how we understand and produce sentences in real time. From the point of view of language comprehension, this gives rise to the following (highly simplified) model, where the input is a stretch of spoken or written language such as a particular sentence:
(14) input phonological lexical syntactic semantic output
processor processor processor processor
In terms of this rather crude model, the first step in language comprehension is to use the phonological processor to identify the sounds (or written symbols) occurring in the input. Then, the lexical processor identifies the component words. The next step is for the syntactic processor (also called the parser, and incorporating the syntactic component of the grammar) to provide a syntactic representation of the sentence (i.e. a representation of how the sentence is struc-tured out of phrases and the phrases out of words). The last step is for the semantic processor to compute a meaning representation for the sentence, on the basis of the syntactic and lexical information supplied by earlier stages in the process. The relevant meaning representation serves as the output of the model: once this has been computed, we have understood the sentence.
An important characteristic of (14), as of all models of psycholinguistic proces-sing, is that its various stages are to be viewed as taking place in real time, and a consequence of this is that psycholinguists can utilise their experimental techni-ques to try to measure the duration of specific parts of the process and link these measurements to levels of complexity as defined by the grammar itself. In fact, it is fairly easy to see that the idea that the grammar can, without any additional
10 linguistics
considerations, serve as a model of sentence comprehension is implausible. A sentence such as (15) is known as a garden-path sentence:
(15) The soldiers marched across the parade ground are a disgrace
A common reaction to (15) from native speakers of English is that it is not an acceptable sentence. However, this reaction can often be modified by asking native speakers to consider the sentences in (16) (recall our observation that not all linguistic data have immediately obvious properties):
(16) a. The soldiers who were driven across the parade ground are a disgrace
b. The soldiers driven across the parade ground are a disgrace
c. The soldiers who were marched across the parade ground are a disgrace
Sentence (16a) should be regarded as entirely straightforward, and we can view (16b) as ‘derived’ from it by deleting the sequence of words who were. Now, if we delete who were from sentence (16c), which should also be recognised as an acceptable English sentence, we ‘derive’ (15), and at this point many readers are likely to change their reaction to (15): it is an acceptable English sentence, so long as it is interpreted with the phrase the soldiers as the logical object of marched (see p. 5 above). When we read (15) for the first time, we immediately interpret the soldiers as the logical subject of marched – the soldiers are marching rather than being marched; as a consequence, the sequence the soldiers marched across
the parade ground is interpreted as a complete sentence and the sentence proces-sor doesn’t know what to do with are a disgrace. The sentence processor has been ‘garden-pathed’, i.e. sent down the wrong analysis route (exercise 6).
What is important about garden-path sentences is that they show that sentence comprehension must involve something in addition to the grammar. As far as the grammar is concerned, (15) is an acceptable structure with only one interpretation. However, it appears that this structure and interpretation are not readily available in sentence processing, suggesting that the parser must rely (to its detriment in this case) on something beyond the principles which determine acceptable combina-tions of words and phrases.
There are other aspects of (14) which are controversial and have given rise to
large numbers of experimental psycholinguistic studies. For instance, there is no place in (14) for non-linguistic general knowledge about the world; according
to (14), interpretations are computed entirely on the basis of linguistic properties of expressions without taking any account of their plausibility, and an alternative would allow encyclopaedic general knowledge to ‘penetrate’ sentence perception
and guide it to more likely interpretations. A further assumption in (14) is that the different sub-components are serially ordered (in that the first stage is pho-
nological processing which does its job before handing on to lexical processing, etc.) An alternative would allow syntactic and semantic factors to influence phonological and lexical processing, for semantic factors to influence syntactic processing, etc. These issues, along with several others, will be discussed in sections 14 and 26.
การแปล กรุณารอสักครู่..