An investigation of the role of context in retrieval of information from semantic memory

Abstract

The research reported here examines how a person' s knowledge of the world is used in language recognition and production. Essentially it is concerned with the importance of a word's meaning as a factor in its recognition by a listener or reader and is its production by a speaker or writer. This area of research overlays with a great many areas in psychology, drawing upon research in attention, pattern recognition, memory, psycholinguistics and thought . It is necessary to give some working definitions of the terms used. The definition of semantic memory used here is that supplied by Tulving (1972, p 386): "Semantic memory is the memory necessary for the use of a language. It is a mental thesaurus, the organized knowledge a person possesses about words and other verbal symbols, their meanings and referents, about relations among them and about rules, formulas and algorithms for the manipulation of these symbols, concepts and relations." The contents of semantic memory are typically what a person would say that he "knows" rather than what he "remembers". e . g. a person might say "I know canaries are yellow" whereas "I remember canaries are yellow" would not "sound right" to most native English speakers. This also illustrates an important property of semantic memory. The knowledge it contains is to a large extent common to members of a given culture. There will of course be individual differences but a sufficient body of knowledge will be shared in order to allow communication between persons. Retrieval from semantic memory is used here to refer to any process that involves making use of such stored knowledge. This may range from simply deciding that a particular sound pattern has occurred in speech before to verifying complex propositions. Context is restricted here to linguistic context. The question asked is how information provided by previous linguistic input affects processing of later input or output of language. The view of language comprehension taken here is similar to Goodman's (1967) approach to reading. This approach is described as follows: " ... Reading is a psycholinguistic guessing game. It involves an interaction between thought and language. Efficient reading does not result from precise perceptions and identification of all elements but from skill in selecting the fewest, most productive cues necessary to produce guesses which are right first time. The ability to anticipate that which has not been seen, of course, is vital in reading, just as the ability to anticipate what has not yet been heard is vital in listening." (p 260) It is assumed here that a person's ability to anticipate is dependent upon the knowledge stored in semantic memory. The way this knowledge is used will in turn depend upon how it is organized. Since the Ancient Greeks the importance of organization in memory has been recognized but it is only relatively recently that psychologists have attempted to determine the principles underlying this organization. Since Quillian (1966) a number of models of how semantic memory is organized have been proposed. These will be discussed in the following sections. Many of the experiments reported here are concerned with what might be called "micro-context", that is how individual words, phrases and sentences affect recognition of incoming stimuli. Of course, the use of context goes far beyond the immediately preceding input but as yet there are no satisfactory theories, linguistic or psychological, that can deal with these wider aspects of language use. In fact there is still considerable disagreement over the processes involved in the recognition of single words, (see, for example, Rubenstein, Lewis and Rubenstein, 1971; Baron, 1973) . The approach taken to word recognition here is similar to Norman (1968) and Morton (1969). The notion which is central to both these authors and Goodman (see above) is the realization that no process can be analysed in isolation. The language system cannot decode the incoming sensory information without reference to stored knowledge. As Norman (1969, p 3) describes the role of memory, "it provides the information about the past necessary for proper understanding of the present". Thus context indicates to the memory system what knowledge is relevant to the analysis of the current input. To summarize this approach the information provided by context (immediate past) is referred to semantic memory (past) which in turn helps to produce the best guess as to the nature of the current sensory input (present) or even the nature of input which has not yet arrived (future) . The problem examined in this research is how the organizational structure of knowledge in semantic memory influences this guessing process . Whether such guessing is an active process as suggested by some investigators (e . g. Liberman, Stevens and Halle) or a passive process suggested by others (e . g. Morton, Treisman) will be discussed in a later section.