Annotation for Rosenthal, David
A Model of the Process of Listening to Simple Rhythms
Annotation (by Bill Tilghman):
- This article discusses the structure and workings of a computer
program that models a possible process of listening to and
learning simple rhythms. The primary structural entity in the
program is a "recongnizer," a simple object constructed by the
program as it "listens" and designed for a very specialized
purpose: to examine all subsequent input for a specific, unique
rhythmic pattern. If the pattern is encountered, the recognizer
"succeeds" and makes a record of the success; otherwise the
rhythmic input is used to contruct a new unique recognizer. The
lowest-level recognizers are constructed using an implementation
of some of Lerdahl
& Jackendoff's "Grouping Well-Formedness
Rules" and "Grouping Preference Rules" and can
recognize chunks consisting of just a few events. Patterns of
success among these recognizers, resulting from parallelisms or
recurrences in the music, are used to construct higher-level
recognizers recursively, forming a hierarchical structure that
represents an abstract description of rhythmic grouping in the
music. While the model has certain limitations (e.g., it ignores
the effects of pitch structure and of possible upper limits in
human cognitive processing, and it fails to accomodate humans'
abilities to retain conflicting interpretations of grouping and
to recognize rhythmic variation), it does seem to display some of
the same aesthetic idiosyncracies as human listeners do. In
particular, the model "prefers" patterns of departure and return
(Meyer's "law of return"), in that such patterns are represented
efficiently by the program in a procedure that is a plausible
parallel to human processing.
- Keywords:
