UNSPECIFIED. (1995) CONNECTIONIST MODELS OF ANEMIA - A COMMENT ON NICKELS. LANGUAGE AND COGNITIVE PROCESSES, 10 (1). pp. 47-58. ISSN 0169-0965

Abstract

This paper examines the claim by Nickels (this issue) that interactive models of lexical access in speech production are unable to account for naming data from a group of anemic patients. In particular, she argues that connectionist models predict that effects of variables associated with a particular level of processing should be manifested at all other levels. So, for example, imageability effects that are hypothesised to originate at the semantic level should also be found in errors that arise elsewhere. Hence phonological errors, which are hypothesised to arise late in lexicalisation, should show an imageability effect. She presents data that appear to be inconsistent with these predictions. I show that these predictions are not necessarily those made by an interactive model of lexicalisation. This is because the effects of interaction are quickly dissipated by intervening processing levels. This paper reiterates the general point that the behaviour of connectionist models is not easily predictable without running the appropriate simulations. In particular, this paper focuses upon the specific point that interaction between adjacent levels of processing need not lead to variables located at one processing level necessarily having global effects. In addition, this paper discusses the role of frequency in lexical access in speech production.

Data sourced from Thomson Reuters' Web of Knowledge

Request changes to a record

Actions (login required)

View Item