Extinction of conditioned meaning: support for a classical conditioning model of word meaning

Abstract

A major theoretical proposition has been that the emotional meaning response elicited by a word will be conditioned to other verbal stimuli with which the word is paired. Considerable evidence of higher-order conditioning of meaning has been found. At the same time, however, several lines of contradictory data have developed: the possible role of awareness in determining meaning change; the absence of an interstimulus interval effect; and the absence of extinction. A review of the three issues suggested that the apparent lack of extinction effects constituted the most crucial empirical challenge. A more detailed analysis suggested that prior extinction studies had used specific values of several parameters of the conditioned meaning procedure which may have: (1) promoted the occurrence of labeling responses to the CS-UCS contingencies; and (2) reduced appropriate attentional behaviors. Both these variables might obscure or negate the effects of extinction. The present study was an attempt to better test the possibility of the extinction of conditioned attitudinal (meaning) responses by modifying the basic Staats' higher-order conditioning procedure so as to control for some possible confounding effects present in earlier studies. The procedural changes included the use of: a larger number of nonsense syllables; fewer conditioning trials; a partial reinforcement schedule; a reduced correlation of particular syllables with specific classes of evaluative words; a single assessment phase; and a moderate number of extinction trials. Using the modified procedures, subjects in Experiment I received 0, 5, or 15 extinction trials with each conditioned stimulus nonsense syllable after receiving higher-order conditioning trials. The same procedures were used in Experiment II, with the exception that a two-week delay was interpolated between conditioning and extinction trials. Conditioning was indexed by evaluative ratings of the syllables on a semantic differential scale by only those subjects presumed to be unaware on the basis of a questionnaire. The results provided some support for a classical conditioning model of word meaning. In Experiment I, the strength of conditioned meaning responses was reduced after 5 and 15 nonreinforced presentations of each verbal conditioned stimulus, although the number of extinction trials did not produce a differential effect. Experiment II did not provide the conditions sufficient to test the hypothesis. When conditioning was measured after a two-week delay, there was no evidence of a significant degree of conditioning in any group. Thus, this investigation presents evidence to suggest that extinction of classically conditioned attitudinal responses established in a higher-order conditioning procedure can occur under certain conditions.