P. Gupta et Ds. Touretzky, CONNECTIONIST MODELS AND LINGUISTIC THEORY - INVESTIGATIONS OF STRESSSYSTEMS IN LANGUAGE, Cognitive science, 18(1), 1994, pp. 1-50
We question the widespread assumption that linguistic theory should gu
ide the formulation of mechanistic accounts of human language processi
ng. We develop a pseudo-linguistic theory for the domain of linguistic
stress, based on observation of the learning behavior of a perceptron
exposed to a variety of stress patterns. There are significant simila
rities between our analysis of perceptron stress learning and metrical
phonology, the linguistic theory of human stress. Both approaches att
empt to identify salient characteristics of the stress systems under e
xamination without reference to the workings of the underlying process
or. Our theory and computer simulations exhibit some strikingly sugges
tive correspondences with metrical theory. We show, however, that our
high-level pseudo-linguistic account bears no causal relation to proce
ssing in the perceptron, and provides little insight into the nature o
f this processing. Because of the persuasive similarities between the
nature of our theory and linguistic theorizing, we suggest that lingui
stic theory may be in much the same position. Contrary to the usual as
sumption, it may not provide useful guidance in attempts to identify p
rocessing mechanisms underlying human language.