ner,’’ you will probably scan through the lyrics at a rate faster than you nor-
mally sing them. This does not necessarily raise your mental representation of
the pitch.
In addition, different sections of songs seem to carry ‘‘flags’’ or ‘‘markers’’
that serve as starting points. If you were asked to sing the third verse of ‘‘The
Twelve Days of Christmas,’’ you might start right on the line: ‘‘On the third day
of Christmas, my true love gave to me ...’’ without having to start from the
very beginning. Markers in songs are to some extent idiosyncratic, and depend
on what parts of a song are salient, and how well you know the song. Few
people are able to jump immediately to the word ‘‘at’’ in ‘‘The Star Spangled
Banner,’’ but some might be able to start singing it from the phrase ‘‘whose
broad stripes and bright stars’’ without having to start from the beginning.
With respect to the other attributes of songs, most people can imagine a song
being played loud or soft, being heard in their left ear or right ear or both, be-
ing performed inside or outside a large church, and the main melody being
carried by various instruments. Most of these things can be imagined even if
they have never been experienced before, just as we can imagine a polka-dot
elephant, although it’s unlikely we’ve ever seen one.
It is striking to listen to the tapes of non-musical subjects singing, super-
imposed on the corresponding passage from the CD. They are only singing
along with their memory, but it appears that they hear the recording in their
head. Enormous amounts of detail appear to be remembered—the subjects re-
produce vocal affectations and stylistic nuances, so that it’s hard to imagine
they could perform any better if they were singing along with the CD.
It wasn’t immediately obvious that people would encode tempo with great
accuracy, but the data shown in figure 13.4 suggest that they do. As shown in
that plot of subject-produced versus actual tempo, 72 percent of the subject’s
productions were within 8 percent of the correct tempo. How close is 8 percent?
Carolyn Drake and Marie-Claire Botte (1993) found that the perceptual thresh-
old for changes in tempo (the just-noticeable difference, or JND) was 6.2–8.8
percent. Thus it appears that people encode tempo information in memory with
a high degree of precision.
We have seen that music has a number of different attributes, and that some
of these attributes appear to be stored in memory in two forms: a relative
encoding of relations and an absolute encoding of sensory features. The preci-
sion with which other attributes of musical performances, such as timbre and
loudness, are encoded in memory, is the topic of experiments currently under
way.
13.10 Summary
The modern view is that memory is distributed throughout various parts of the
brain, and that different types of memory engage separate neural structures.
Memory for music, just like memory for prose or pictures, probably comprises
different cognitive subsystems to encode the various aspects of music. There is
a growing consensus that memory serves a dual function: it abstracts general
rules from specific experiences, and it preserves to a great degree some of the
details of those specific experiences.
Memory for Musical Attributes 307