Choice behavior in a verbal recognition task as a function of induced associative strength.

Abstract

Variations in associative strength between letter-digit pairs were produced experimentally using a probabilistic paired-associate procedure. Each pair consisted of 1 of 4 letters and 1 of 4 digits. Ss were then given a series of triplets. Each triplet was a unique combination consisting of 2 of the 4 letters and 1 of the 4 digits. Ss were instructed to choose the letter that "best goes" with the digit on the basis of what they had learned. Choice behavior was related to the relative frequency of item pairings during learning using the Bradley-Terry model for pair comparisons. The model was found to fit best when associative strength of a letter-digit pair was defined as a power transformation of the relative frequency with which the digit was paired with the letter. These results parallel those found for wordassociation data and are consonant with a relative frequency interpretation of associative strength for intraverbal and referent-name associations.

Tests of a two-stage "speaker" communication model using induced respone hierarchies.

“…Since C is essentially a "noise" parameter to be added to the associative proportion (see Equation 4), such large values have little meaning in terms of the model. Second, variations in the value of C produced very little change in the minimum sum of squared deviations, particularly within the range of C found in previous research (Rosenberg & Cohen, 1966;Rosenberg & Donner, 1968). Since within this low range of C there was almost no change in the sum of squared deviations, while variation in B produced considerable change, this strongly suggests that C is not a meaningful parameter for the data of these experiments.…”

“…An approach which allows experimental control of the content of individual response repertoires is that of establishing the response repertoires in the laboratory (Camp, 1960;Goss & Cobb, 1966;Goss & Sugarman, 1961;Osgood & Anderson, 1957;Peterson, 1956;Rosenberg & Donner, 1968 ;Wiggins, 1957). In such experiments, a paired-associate procedure has been used to establish from two to as many as eight responses to a stimulus.…”

“…In such experiments, a paired-associate procedure has been used to establish from two to as many as eight responses to a stimulus. The induced response repertoires are tested with variants of the basic free association task (e.g., Goss & Cobb, 1966) and by a verbal recognition task ("listener" task) used by Rosenberg and Donner. The present study is an extension of the Rosenberg and Donner (1968) procedure. In their study, they established associations between four single-letter stimuli and four single-digit responses.…”

Identification of facial expressions from affective descriptions: A probabilistic choice analysis of referential ambiguity.

“…First, in the case of b the question is whether the fit of a linear transformation of associative probability is statistically indistinguishable from the fit of a general power transformation. To answer this question, an iterative least-squares procedure was used to estimate c-the only parameter which has 5 A detailed description of the fitting procedure was given in Rosenberg and Donner (1968).…”

“…That is, the scale value of an object that the listener is considering as a possible referent of the speaker's utterance is the associative strength of that object to the particular utterance. The Bradley-Terry model (Bradley & Terry, 1952), which relates scale values and choice probability for two-alternative sets, was successfully fitted in previous studies where stimuli and responses were both verbal (Rosenberg & Cohen, 1966;Rosenberg & Donner, 1968). Also the Bradley-Terry model can be derived from Luce's choice axiom, although tests of the choice axiom can be made without estimating the scale values of the stimulus objects involved (Ely, 1967).…”