Summary Three alternative hypotheses for extending Estes' and Bower's all-or-none learning model to cover the case where S learns two response components concurrently to a nonsense syllable stimulus were tested. Sixty college students each received two six-item lists under each of three schedules of one or two practice trials and two tests. In each list, each of the six combinations of the color adjectives “red,” “green,” and “blue” and the form nouns “circle” and “triangle” was randomly designated as a “correct” response compound for one syllable. The main analysis consisted in computing the probability of a correct response component, conditional on an incorrect response on the other component of that compound. Since these conditional probabilities were consistently above chance, conditioning of the compound as a unit was refuted. However, the finding that the probability of a correct response component was higher when the other component of that item was correct than when the latter was incorrect disputed the hypothesis that the response components are conditioned independently of each other. The data supported the hypothesis that the rate of conditioning of the noun component depends on whether the adjective component is conditioned.
Journal of Verbal Learning and Verbal Behavior