Derivational Theory of Complexity

The present essay deals with the rise and fall of the Derivational Theory of Complexity (DTC). Before 1957, when Noam Chomsky came up with the concepts of Generative Grammar and later of Transformational Generative Grammar, psychologists had little to start their work on psychological processes involved in the perception, understanding and/or production of sentences. However, Chomsky’s work provided a simple (though with its own complexity) linguistic model, which psychologists took as a basic framework to start their work. The TG (Transformational Generative) model was ‘thought to provide the basis for a description of how speakers and listeners actually produce and understand sentences—it had potential psychological reality.’ (Gleason et al. 1993)

Chomsky’s idea of Phrase Structure Rules—rules that govern the generation of basic English sentences such as simple active affirmative declarative; and Transformational Rules—used to produced various surface realities (structures), though with different meanings, of same kernel sentence, had profound influence on the way people thought about the understanding and perception of sentences. TG provided a framework for dividing even simple sentences into their component parts, which in turn constitute of more smaller units. A simple set of phrase structure rules to generate a sentence such as ‘the boy ate an apple’ would be

S -> NP VP

NP -> Det N

VP -> V (NP)

N -> boy, apple

V -> ate

Det -> the, an

TG was not just introduced as a linguistic or syntactical theory only. It also shed light on theories of sentence processing. It was asserted that listeners’ extraction of meaning in a sentence involved the perception of such rules as mentioned above. The recognition of sentence was considered to be involving a device which computed deep structures from the strings of words. In other words TG proposed that the more grammatical operations involved in generating a sentence, the more difficult it ought to be to understand the sentence. They started measuring the complexity of a sentence by the number of grammatical rules employed in its derivation (Fodor 1974). This was called The Derivational Theory of Complexity. For example if Gi is the Transformational Generative Grammar of a language L and Si is a sentence of L produced/generated by n number of rules, DTS ‘claims that n is an index of relative sentential complexity. In particular, other things being equal, two sentences assigned the same value of n should be equally complex, and of two sentences with different value, the one assigned the larger number must be the more complex.

Since according to TG, a simple active affirmative declarative is the basic or kernel sentence and other complex structures are produced by applying the transformational rules, their apparent complexity owes its origin to the sentence’s difference from its kernel structure. In other words, it is the number of transformations (transformational rules).

Many people tested this hypothesis. Miller and McKean (1964) (quoted by Harley 1996) and Fodor 1974) argued that the more transformations there are in a sentence, the more difficult it would be to process. They looked at ‘detransformation’ reaction times to the sentences such as:

1. The robot shoots the ghost. (0 transformations, active affirmative)

2. The ghost is shot by the robot. (1 transformation, passive affirmative)

3. Is the ghost not shot by the robot? (3 transformations, passive, negative, question)

Miller and McKean discovered that the time it took to comprehend sentences was directly related to the number of transformations in them. ‘This was interpreted as supporting the psychological reality of transformational grammar.’ (Harley 1996)

Another research by Savin and Perchonock (1965) found that complex structures involving more transformations took more memory space. Mehler (1963) found out hat when subjects made errors while remembering the sentences, the errors were more that of missing out some transformations, rather than adding extra transformations. Further, it was observed that subjects remembered sentences in their kernel forms and applied transformations separately. For example for a given sentence ‘Was the boy hit by the girl’, they reproduced ‘The girl hit the boy’.

Another research by McMohen used active, negative, and passive sentences and required the subject’s response in terms of true or false. He used sentences like ‘5 precedes 13’, ‘3 is preceded by 7’, or ’13 is not preceded by 6’. He found that both negatives and passives took longer to evaluate than active affirmatives. (Fodor 1974). In a research conducted by Gough, which involved whether a given sentence was true or false (given the cue in the picture format), it was observed that short and long passive sentences without the agent took longer to be verified than those with an agent. It shows that for agentless passives the subject had to take account of the absent agent as well, thus resulting in longer processing time.

Although a great amount of evidence in the immediate aftermath hailed the validity of DTC and that of TG, the field became controversial very soon. Some researchers found anomalous results in their studies. For example Slobin’s (1966) study of reversible and irreversible passives, claimed that results achieved by Miller and McKeane could only be obtained for reversible passives. This brought into focus the role played by semantics in the understanding of sentences. For the first time it was noticed that syntactic complexity may not be the only reason for longer processing, it could also be ‘semantic’ complexity. ‘…detransformational parsing only appears to be necessary when there are not sufficient cues to the meaning of the sentence from elsewhere. This result means that DTC is not always true: complexity only involves parsing in the absence of semantic cues’ (Harley 1996).

The idea of semantic intervention was supported by an experiment by Wason on the relationship between the structure of a sentence and it function. He found out that the time it takes to process a syntactic structure such as a negative-formation depends upon the semantic context. A syntactically complex but semantically simple construction will probably take less time than a syntactically simple but semantically complex construction. According to Fodor (1974) ‘there is a residual interaction of response time with the truth or falsity of the sentence, which suggests that some purely semantic variable is operative (Fodor 1974).

Another area of objection on DTC came from computer processing models. The notion that complex structures will take longer to process because the have more underlying transformation to be sorted out assumes that these underlying transformations are processed in a linear way. People argued that this may not necessarily be the case. What about parallel processing in which multiple computations are done at the same time?

The apparent failure of the DTC had some profound after effects. According to Harley (1996) this ‘has effectively led to a separation of linguistics and psycholinguistics’. Miller who earlier on provided the first ever empirical support for the psychological reality of TG grammar, later came to believe that all the time taken in sentence processing was used in semantic operations. Another group of scholars held Transformational Generative Grammar responsible for the failure of DTC by asking whether the failure was due to the poor syntactic theory? This led to a substantial modification of TG theory. It is rather complex to give a verdict on the validity or otherwise of DTC with the help of argument given above. It certainly has its strong points and weaknesses. A further research combing the Government and Binding Theory (the modified form of TG) and the psychological realities of language processing may prove helpful in deciding the extent of the validity or truth of DTC.

Bibliography:

Clark, H.H., and Clark, E.V. (1977). Psychology and Langauge: An Introduction to Psycholinguistics. Harcourt Brace Jovanovich. New York.

Fodor, J. (1974) The Psychology of Language. McGraw-Hill Book Company. New York.

Gleason, J and Ratner, N B. (1993) Psycholinguistics. Wadsworth.

Harlye, Trevor A. (1996) The Psychology of Language: From Data to Theory. Psychology Press.