The Extension of Marbe's Law to the Recall of Stimulus-Words

Abstract

In one of the early studies of free association Thumb and Marbe reported an inverse curvilinear relationship between individual reaction-times for freely associated responses and the frequency of occurrence of these responses derived by pooling group data.' This negative relationship between frequency of responses and the speed of their emission has come to be known as Marbe's Law. Woodworth has reviewed evidence of its confirmation by several investigators.2 The present study reports findings indicating that Marbe's law may be extended to the recall of the words of stimulus-lists. In general terms our procedure was that of presenting the words of stimulus-lists in serial order for learning to groups of undergraduate student Ss who could be assumed to have fairly similar cultural and educational backgrounds. These Ss were instructed to write as many of the stimulus-words as they could recall in the order of their occurrence in memory. Within the framework of these operations we undertook to test the following hypothesis: The rank order of recall of the twords of a stimulus-list by individual Ss should be a negative function of the frequency of recall of the words by the group. In accounting for this deduction, we assume that both Marbe's law and our extension of it should apply when the Ss comprising the group have more or less similar habit-strengths for the individual words comprising the stimulus-list. Such similarities should be a function of the degree of homogeneity in the social and educational backgrounds of the Ss. In the so-called free association experiment, the S responds to a stimulus-word by drawing from a pool of competing associates having the common characteristic of being related to the stimulus-word. Under these conditions the individual S produces the associate having the greatest strength. This strength, in turn, may be measured by reaction-time. Within the conditions of the present study, the Ss were required during recall also to draw upon a pool of associates. It would follow that associates having the greatest strength should be produced first. The order of recall of the subsequent words, moreover, should be a function of their strengths. It should be noted here that in treating our data we split the recall sequences of the Ss into decile intervals. This method is consistent with our theorizing especially if we can accept the reasonable assumption that decile rank and response latency are directly related by some monotonic function.