freely in pitch space, starting anywhere one wants them to. ‘‘Happy Birthday’’
is an example of this. Now, you might object to all this and say that Beetho-
ven’s String Quartet in F Major ought to be played in F major, and that it loses
something when it is transposed. The timbre of the stringed instruments
changes with range, and if the piece is played in C major, the overall spectrum
of the piece sounds different to the careful listener. But listeners will still rec-
ognize the melody because the identity of the melody is independent of pitch.
A number of controlled laboratory experiments have confirmed that people
have little trouble recognizing melodies in transposition (Attneave and Olson
1971; Dowling 1978, 1982; Idson and Massaro 1978). Also, at different times and
different places, the tuning standard has changed; our present A440 system is
arbitrary and was adopted only during the twentieth century. The absolute pitch
of the melody’s tones is not the most important feature. It is the pattern, or
relation of pitches, that is important.
Note the parallel here with our earlier discussion of generalization and ab-
straction in memory. One of the reasons we are able to recognize melodies is
that the memory system has formed an abstract representation of the melody
that is pitch-invariant, loudness-invariant, and so on. We take for granted that
our memory system is able to perform this important function. Recent evidence
suggests that memory retains both the ‘‘gist’’ and the actual details of experi-
ence. But what about melodies? Do we retain pitch details, like the absolute
pitch information, alongside the abstract representation? This is an interesting
question that we will take up in section 13.9, after first reviewing research
on memory for contour, lyrics, and failures of musical perception known as
amusias.
13.6 Contour
Recall that the termcontourrefers to the shape of a melody when musical in-
terval size is ignored, and only the pattern of ‘‘up’’ and ‘‘down’’ motion is con-
sidered. At first, the idea of contour being an important attribute of melody
seems counterintuitive. Contour is a relatively gross characterization of a
song’s identity. However, its utility has been shown in laboratory experiments.
There is evidence that for melodies we do not know well (such as a melody we
have only heard a few times), the contour is remembered better than the actual
intervals (Massaro, Kallman, and Kelly 1980). In contrast, the exact interval pat-
terns of familiar melodies are well remembered, and adults can readily notice
contour-preserving alterations of the intervallic pattern (Dowling 1994). Infants
respond to contour before they respond to melody; that is, infants cannot dis-
tinguish between a song and a melodic alteration of that song, so long as con-
tour is preserved. Only as the child matures is he/she able to attend to the
melodic information. Some animals show a similar inability to distinguish dif-
ferent alterations of a melody when contour is preserved (Hulse and Page 1988).
One explanation of why the contour of a melody might be more readily pro-
cessed is that it is a more general description of the melody, and it subsumes
the interval information. It is only with increasing familiarity, or increasing cog-
nitive abilities, that the intervallic details become perceptually important.
300 Daniel J. Levitin