The varieties of perceptual independence in identification and similarity judgement tasks

Abstract

Summary: The work reported here is an investigation of various forms of independence in identification and similarity judgement tasks for sudatory stimuli. Experiment 1A and 1B tested various independence properties (single factor and joint factor independence) on similarity data. Both experiments suggested that not only was a dimensional representation appropriate, but also that the ‘differences’ between two stimuli were combined additively. The second sets of experiments (2A and 2B) explored dimensional independence in identification tasks for auditory stimuli presented in noise. It was found that for both pairs of dimensions tested (pitch and duration; and pitch and loudness) there was a lack of independence. Specifically, it was found, that the identifiability of a value on one dimension varied over different levels of the other. In other words there were interaction effects. Another group of experiments (3A, 3B, 3C, and 3D) investigated both independence and metric properties or data derived from similarity judgements of auditory tones. Here the data was required to satisfy simultaneously, not only the dimensional requirements tested in experiments 1A and 1B but the various ordinal conditions which would allow it to be represented by a Minkowski metric. The four experiments established, for the particular values of pitch, duration, and loudness tested, that both a dimensional and metric representation of the data was tenable. One outcome from these experiments was that the similarity data could be represented by either a Euclidean, City Block or Dominance metric. The last experiment, 4A, was in part a replication of Garner and Felfoldy’s (1970) study. These authors had used metric concepts such as the Euclidean and City Block distance functions to account for the results of their identification experiment. It was noted that metric concepts were appropriate only for similarity data (under certain conditions), and their use, to explain results in identification data was entirely objectionable. It was suggested instead that Garner and Felfoldy’s data could have arisen as a result of interaction between the stimulus dimensions. These interaction effects had already been investigated in experiments 2A and 2B. Experiment 4A, did indeed suggest that the results might be due to lack of independence of the dimensions rather that to their alleged metric properties.