Pigeons and the Magical Number Seven

In 1956 George Miller called our attention to the still unexplained fact that humans can identify only about seven (plus or minus two) tones, or sound intensities, without error. Beyond the “magical number seven” increasing the number of such stimuli to be identified produces little gain in information transmitted. Stimuli that are identified perfectly as part of a set of four or five are confused when they are part of a larger set. Since confusions do not occur if the number of stimuli is small, they cannot be completely due to sensory limitations. Further evidence that the limitations on accuracy are not solely due to lack of sensitivity is provided by the range effect. As the separation between adjacent stimuli in a set is increased there is, at first, a gain in accuracy. This, however, quickly levels off. In fact, Pollack (1952) found that the range of tones to be identified can be increased by a factor of about 20 without much improvement in accuracy. Extensive work on the range effect, and other problems of intensity resolution, has lead Durlach and Braida (1969) and their collaborators to conclude that the main source of error in absolute identification of sound intensities is memory (see Berliner, Durlach and Braida, 1978; Braida and Durlach, 1972; Purks, Callahan, Braida, and Durlach, 1980). Confronted with the task of identifying stimuli presented in random order without benefit of a standard to which they can be compared, the subject operates in the context coding mode. That is, “the subject attempts to compare the sensation with the general context of sounds in the experiment.” (Durlach and Braida, 1969, p. 374). In most of the experiments to be described below pigeons were trained to make absolute identifications of lights that varied only in luminance. In many respects their performance is not unlike that of humans confronted with a similar task. Like humans, pigeons seem to be severely limited in their ability to identify such stimuli when they are members of a large set. The limitation appears to be the result of an imperfect memory coupled with a decision process that uses only a small sample of the information in this memory. Although Heinemann (1983a) developed this model to account for choice behavior of pigeons in a variety of twochoice situations, it applies, with very little modification, equally well to situations that involve more than two choices, for example: the categorization and the absolute identification experiments to be described here.