The authors hypothesize that in the perception of temporal
patterns listeners internally generate a hierarchical clock,
consisting of units and simple subdivisions whose durations are
flexible and are suggested by the distribution of accents in the
pattern itself. For the purposes of this article, accents are
defined only in terms of the temporal grouping or "clustering" of
note-onsets, since only artifically generated sequences of
identical tones are considered here. Using a few simple rules of
accent placement, a computer program considers all possible two-level
clocks for a given temporal pattern and selects the one
that is least contradicted by the locations of accents in the
pattern.With further computer assistance, the authors develop a
large repertoire of temporal patterns, categorized by the degree
to which even the "best choice" of internal clock is still
contradicted by the accentual structure of the pattern. The more
the accentual structure contradicts even the best clock (i.e.,
the less strongly the pattern suggests a clock), the authors
predict, the more difficult the clock will be to generate and
hence the more difficult the pattern will be to reproduce. Two
experiments are carried out in order test this prediction. In the
first, subjects were able to reproduce patterns more readily and
accurately the better the accentual structure of the pattern
suggested a clock. In the second, subjects who were presented
with patterns identical to those in the first experiment, except
that the clock chosen by the computer was presented
literally as part of the stimulus, were even better able
to reproduce the pattern than they were in the first
experiment.
The authors further hypothesize that the presence of an internal
clock allows for more efficient encoding of the temporal pattern,
and they suggest that the most efficient encoding of a single
clock unit occurs when the unit is either completely empty or
completely full of equally-spaced events. Thus, the complexity of
a pattern can be estimated according to how many of the pattern's
clock units meet these criteria. In a third experiment, subjects
were presented with pairs of temporal patterns, in which
the two paired patterns were actually identical, but were
accompanied by two different clock patterns, suggesting two
different metrical interpretations of the pattern. Subjects were
asked to choose the simpler of the two patterns. In the cases
where one of the metrical interpretations indicated a more
efficient encoding according to the model, subjects generally
chose the pattern accompanied by that metrical interpretation as
the simpler pattern. (Subjects rarely noticed that the two
patterns were actually identical.)
