38
infants preferred the passages from the beginnings rather than the middles of the familiar-
ized sonatas. Again, we tested a separate control group to ensure that biased test passages
could not account for the experimental group’s performance; these infants did not show a
significant preference for either type of test passage. Along with Experiment 1, the results
of Experiment 2 strongly suggest that infants do remember the music that they hear; the
existence of a preference for passages drawn from the beginnings, rather than the middles,
of the familiar music indicates that the infants remembered the pieces last heard at home
two weeks before. Moreover, the infants preferred music that was not taken out of context.
The control infants showed no such bias; even out of context, these passages sounded
coherent to listeners who were not familiar with the pieces from which they were drawn.
These results suggest that infants are sophisticated musical listeners. Their representa-
tions of music in long-term memory are not undifferentiated sequences of notes. Instead,
passages are linked together to form coherent musical events. Other recent studies investig-
ating infant long-term memory for music similarly suggest that infants’representations are
quite nuanced. For example, infants can represent far more complex pieces of music (such
as Ravel) in long-term memory.^34 Moreover, the content of infants’memories includes
some extremely specific aspects of musical performances. For example, 10-month-olds
represent acoustic patterns specific to the particular performances with which they were
previously familiarized.^35 Six-month-old infants remember the specific tempo and timbre
of music with which they are familiarized, failing to recognize pieces when they are played
at new tempos or with new timbres.^36 It thus appears that infant representations are
extremely specific, without the flexibility to generalize to include changes in tempo or
timbre. This level of specificity must change with age, as either a function of experience or
development, else listeners would not be able to recognize familiar music played on different
instruments or at different rates. Indeed, preschool children can recognize familiar tunes
across many different types of transformations, suggesting that children are in the process
of developing more flexible musical representations.^37
Interestingly, infants can recognize previously familiarized melodies when they are
played in a new key,^36 suggesting that for highly familiar pieces, infants represent the relat-
ive pitches of melodies. Such interval-level information is the appropriate level of spe-
cificity for generalizing memories from one experience to another, and is the preferential
level of processing for most adult listeners. These results reinforce the suggestion that task
demands determine the level of processing of pitch cues in musical materials: infants rep-
resent pitch contour and/or relative pitches more readily given familiar pieces, or given
familiarization with repeating brief tunes (for review, see Ref. 38), but appear to rely on
absolute pitches given unsegmented stretches of previously unfamiliar tone sequences.24,39
Infants have access to multiple levels of pitch representations, much like adults (e.g. Refs 40
and 12); the structure of the task and the familiarity of the stimulus domain may influence
which levels of representation are prioritized. For example, when absolute pitch cues are
rendered uninformative in a task which previously elicited the use of absolute pitch informa-
tion, infants shift to rely on relative pitch cues (Saffran, Reeck, Niebuhr, and Wilson, in
prep.). These results mirror data from an avian species—starlings—who can switch from
relying on absolute pitch cues to using relative pitch cues when necessitated by the struc-
ture of the task.^41