Quantal Effects in Language Learning and Change
Many studies of language learning and change have observed relatively
abrupt changes in linguistic behavior. Some prominent examples are:
the so-called word "spurt" at the very earliest stage of vocabulary
acquisition, the U-shaped learning curve in English past tense
acquisition, the logistic trajectory in many instances of morpho-
syntactic change, etc. These effects often have been taken as
evidence for domain-specific linguistic processes and constraints--
components of the Language Acquisition Device--which guides the
learner to look for things like the default rule, or to perform
diachronic reanalysis.
In this talk, I will show that the three aforementioned quantal
effects fall out of the dynamics of learning mechanisms--extrapolated
over time, as in the case of language change--that may turn out to be
used in other cognitive and perceptual tasks and motivated by general
computational principles. These mechanisms, however, can only operate
under specific representations and constraints that appear to be
unique to language.
Two papers that may be used for background:
Yang, C. (2004). Universal grammar, statistics, or both. Trends in
Cognitive Sciences, 8, 451-456.
Yang, C. (2005). On productivity. Language Variation Yearbook, 5,
333-370.
both are available at