Identification of integral stimuli.

Abstract

The attempt to define the stimulus from the observer's viewpoint is an old and difficult problem. One aspect of this problem is deciding if stimulus aspects that the experimenter manipulates independently are also independent aspects for the observer. If aspects are independent in some experiments but not in others, then models of information processing that are general over tasks will not become available unless we can account for the differing effects of stimulus attribute combinations. The portion of this problem studied in this article is, Do observers analyze the stimulus into its logical components, as the experimenter does when constructing them, and judge these components separately, or do observers initially judge the entire stimulus in a holistic manner ?Experiment 1 evaluated identification performance (accuracy and latency) on a set of integral stimuli. Analytic models, which call for independent judgments of stimulus attributes, and holistic models, which call for a single unitary judgment of the entire stimulus, make several different predictions of relative reaction times to identify different members in the stimulus subsets studied. None of the analytic predictions were supported. All of the results are consistent with a holistic model that had been suggested earlier (Lockhead, 1972).Experiment 2 compared identification performance as a function of the relations between stimuli in physical space and between the same stimuli in similarity space. The results provide comparison between several possible analytic models and the suggested holistic, or blob, model. The data support the holistic model and call for rejection of any model that requires detection of or judgments about aspects of integral stimuli before a unitary identification has been made.We conclude that it is the separation between stimuli in similarity space and not the experimenter-defined differences between stimuli that predicts object-identification performance. Increases in redundancy do not produce faster and more accurate object identification. It is the concomitant increase in psychological distance between stimuli, which often accompanies redundancy, that allows for improved performance.The stimuli used in Experiment 2 do not fit all previously suggested definitions of what makes an integral stimulus, and a broadened definition of integrality is suggested. The essence of the suggestion is that relations between the attributes as well perhaps as the attributes themselves are important considerations.

A psychophysical approach to dimensional separability

“…It might therefore be argued that all the dimensions manipulated in Experiment 1 were integral and that the apparent absence of interference for the psychological dimension was due to the relatively small degree of dissimilarity that the irrelevant dimension generated between stimuli in the same category. Such an explanation would in fact follow from the model of integrality proposed by Lockhead and King (1977;Monahan & Lockhead, 1977). This model distinguishes between two types of representations (integral and separable), but postulates that performance on integral dimensions is predicted not by differences on physical dimensions, but by psychological overall similarity, as measured by Euclidean distances in a multidimensional configuration of similarity judgments.…”

“…In contrast, our correspondence hypothesis postulates only one type of representation and attempts to explain performance by psychological similarity as measured by distances along separable psychological dimensions. Although overall similarity has been shown to predict performance for "integral" dimensions, similarity along psychological dimensions was not controlled in such studies (Lockhead & King, 1977;Monahan & Lockhead, 1977); overall similarity may therefore have been confounded with similarity along psychological dimensions.…”

Knowledge partitioning in categorization: Boundary conditions

“…There have been repeated suggestions in the literature that people may not consider those stimuli according to their height and width but according to some other variables, such as area and shape (Krantz & Tversky, 1975) or some emergent unidimensional property based on the ratio of their height and width (see, e.g., Monahan & Lockhead, 1977). Our response to this problem is twofold: First, the problem does not appear to be particularly widespread.…”

“…For example, rectangles that differ in height and widthtwo integral dimensions-may be classified on the basis of their "tallness" or "thinness," expressed as the ratio of the two dimensions (see, e.g., Monahan & Lockhead, 1977). Given that knowledge partitioning relies on participants' recognition of how dimensions interact-specifically, that context gates the nature of the relationship between the remaining two dimensions-it follows that knowledge partitioning, known to occur with separable stimuli, may not be detectable with integral stimuli.…”

Orthogonal stimulus variation and attention in dimensional contrast

“…For example, in our work and work by others (e.g., D. S. Blough, 1988), rectangles appear to have integral dimensions. Rectangles also appear to have integral dimensions for human observers (Monahan & Lockhead, 1977). On the other hand, a modified rectangular form with an inverted U as its lower aspect appeared to have dimensions that could be selectively ignored.…”