Dislocation of Tones in a Musical Sequence: a Memory Illusion

COMPUTER-GENERATED tonal sequences have several advantages for investigations of immediate memory: the stimulus parameters are simple and can be exactly controlled , and the items cannot be readily rehearsed. Previous studies of memory for pitch have investigated the effect on the differential threshold of blank intervals between the standard and comparison tones 1-3 and of one interpolated tone 4-5. In the recognition of simple musical sequences the order of the component tones is usually supposed to be well retained; and this assumption is especially important to musical analyses based on information theory 6. I have, however, found that there can be a rapid loss of such order information in a very simple tonal sequence; namely , the traditional musical scale. The procedure was as follows. A subject listened to a test tone, which was followed by a sequence of four intervening tones and, after a pause, by a probe tone. The subject was instructed to try to remember the test tone, ignore the four intervening tones if he wished, and then to judge whether the probe tone was or was not of the same pitch as the test tone. All tones lasted for 200 ms. The first five tones were separated by intervals of 300 ms, and there was a 2 s pause before the probe tone (see Fig. 1). The tones were all equally loud and taken from an equal-tempered scale (A = 435) ranging just over an octave, from middle C to the C# an octave above. When the test and probe tones differed in pitch, the difference was always a semitone (higher in half the instances, and lower in the other half): this is well above the recognition threshold in the absence of intervening tones, even with much longer delays between test and probe 1-3. The intervening tones were chosen randomly except that no sequence contained repeated tones unless they were specified by the experimental conditions. All the test and probe tone pitches were equally represented in all conditions. Tones were generated by a Wavetek oscillator controlled by a PDP9 computer, and were recorded on tape. Subjects were selected on the basis of obtaining a score of at least 80 per cent correct on a short tape containing similar sequences. The data from the two days of the experiment were averaged. The experiment investigated the effect on recognition memory of inserting, in the intervening sequence, tones of …