Another important point that has bearing on the use of tone in speech
is the observation of categorical perception of tone.House (1990)
presented his experiments with Swedish speakers and reviewed the
literature with regard to Chinese lexical tone,German categories of
intonational meaning,and English pitch accent,and concluded that
“results from perception experiments in four different languages support
the concept of linguistic categories (both lexical and semantic) being per-
ceived in terms of tonal levels during maximum spectral change after the
CV [consonant-vowel] boundary and as tonal movement during relative
spectral stability.The synchronization of tonal movement with vowel
onset seems to be important for the perception of linguistically relevant
tonal categories”(p.81).Thus for both intonation languages and tone
languages,cognitive experiments show that people tend to perceive level
tones in a more or less categorical fashion,in support of autosegmental
models of intonation and lexical tone.
What are the implications of these important findings for the musi-
language model? Three basic implications bear mentioning.First,the
production and perception of pitchedvocalizations is a necessary char-
acteristic of such a system,in contrast to vocalizations based purely on
portamentos (glides,slides,etc.).As most primate vocalizations systems
rely heavily on unpitched grunts and pants (e.g.,chimpanzee pant-hoots,
vervet monkey alarm calls) or on high-contoured pitch glides (gibbon
song),the musilanguage theory posits that a pitched vocalization system
involving at least two pitch states would have had to evolve at some point
in the hominid line.This theory does not demand evolution of new artic-
ulatory capacities to form novel types of segmental phonemes but simply
the cognitive capacity to use level tones in a meaningful fashion.Nor
does this argument have any bearing on the types of transitions that
occur between level tones;they are just as likely to be pitch glides as
pitch jumps.All that is important is that some notion of level tones be
involved.
Second,the idea of lexical tone,as seen from the autosegmental per-
spective,suggests that level tones are just as important for intonation
languages as they are for tone languages.Therefore,discrete pitch levels
and pitch-scaling mechanisms are not merely features of tone languages
and music but are important features of intonation languages as well.
Speech,like music,is based on discrete pitch levels that themselves
are scaled,although variably so.This is supported by experiments
showing that normalizing approaches explain pitch-range effects better
than do initializing approaches as well as by studies demonstrating the
categorical perception of tone in both intonation languages and tone
languages.
Third,any evolutionary expansion of this system to generate phrases
will follow,at least to an important extent,localist rules whereby strings283 The “Musilanguage”Model of Music