anthem, which have no objective key standard, and are likely to be sung in a
variety of different keys.
The subjects were mostly introductory psychology students (and a few grad-
uate students), they were not specially selected for musical ability or inability,
and they didn’t know ahead of time that they’d be participating in a music
experiment. After they selected a song, they were asked to imagine that it was
playing in their heads, and to sing or hum along with it when they were ready.
The subjects could sing as much or as little of the song as they wanted, and
they could start wherever they wanted. The first five tones they sang were
analyzed, then compared with the five corresponding tones on the CD. There
was no difference in accuracy among the five tones or the average of the five
tones. Octave errors were ignored (as is customary in absolute pitch research),
and how many semitones they were away from the correct tone on the CD was
recorded. Thus, the subjects could deviate from the correct pitch by six semi-
tones in either direction.
Figure 13.3 is a plot of the distribution of the subjects’ errors. If subjects were
no good at this task, their errors would be uniformly distributed among the
error categories. In fact, as the top portion of the figure shows, the modal re-
sponse was to sing the correct pitch. Notice also that the errors cluster around
the correct pitch in a mound-shaped distribution. In fact, 67 percent of the
Figure 13.3
Results of pitch memory for the first tone of rock songs. (Upper) Trial 1; (Lower) trial 2.
Memory for Musical Attributes 305