On the Combination of Associative Probabilities in Linguistic Contexts

Abstract

It is a commonplace that meanings of words depend upon the contexts in which they occur. This dependence sets a fundamental problem in the psychology of language: calculation of the psychological effects of a word in its context from the individual properties of the word and of the contextual elements. Recognizing a conventional distinction between the linguistic and non-linguistic contexts of a person's speech, we may conveniently subdivide the former into (a) the homogeneous linguistic contextthe context provided by his own previous language behavior; and (b) the heterogeneous linguistic context-that provided by the utterances of other persons in his environment. Recent work by Shannon has provided a statistical model for describing certain problems of homogeneous sequences and has stimulated several experimental studies in that area.' In the present paper we shall be concerned solely with heterogeneous linguistic contexts; i.e. with the prediction of the language behavior of an experimental subject from the language behavior of another person in his environment. For experimental purposes we take a sequence of four words spoken by an experimenter (E) and measure as a dependent variable the probability that a given word will be emitted as an association to the last word of the sequence. This is a modified form of the conventional word-association experiment. Since the sequence is spoken by E, the properties of the words constituting it can be controlled as independent variables. The strengths of the associative effects of each of the first three words of the sequence upon the subject's (S's) response to the fourth word is the property investigated in the following experiments. Three studies will be reported. If we designate the first three words of the four-word sequence as the context and the fourth word as the test-word, the independent variables defining the three