A connectionist view of language production

Tracking the time course of multidimensional stimulus discrimination: Analyses of viewing patterns and processing times during “same”-“different“ decisions

“…Relevance of multidimensional stimulus discrimination on higher-order processes In the referential communication task, which shall serve here as an example of higher-order processes that rely on multidimensional stimulus discrimination, speakers have to refer to multidimensional objects in the context of other multidimensional objects by specifying a set of features that clearly distinguish the intended object from the surrounding objects (Eikmeyer & Ahlse n, 1998;Eikmeyer, Schade, Kupietz, & Laubenstein, 1999;Ford & Olson, 1975;Herrmann & Deutsch, 1976;Mangold-Allwinn, Baratelli, Kiefer, & Koelbing, 1995;Pechmann, 1989Pechmann, , 1994Schriefers & Pechmann, 1988). In situated referential communication, multidimensionality means variation between objects in terms of dimensions such as colour, size, and object class in most of the cases (Danks & Schwenk, 1972;Eikmeyer & Ahlse n, 1998;Ford & Olson, 1975;Olson, 1970;Pechmann, 1989;Whitehurst, 1976).…”

“…The ''same''-''different'' experiment presented earlier revealed the main characteristics of the first stage. We also know much about how speech production processes work and how the verbalisation stage might be modelled (Bock & Levelt, 1994;Dell, 1986;Eikmeyer et al, 1999;Levelt, 1989;Levelt, Roelofs, & Meyer, 1999;Schade, 1999). But what about the evaluation of stimulus dimensions with regard to their relevance?…”

Recency Effects for Meaning and Form in Word Selection

“…Although interactive activation models (e.g., Dell, 1986Dell, , 1990Dell et al, 1997;Eikmeyer et al, 1999;Harley & MacAndrew, 1995;MacKay, 1982;Stemberger, 1985) have predicted a number of results for which discrete two-stage models (e.g., Butterworth, 1982;Garrett, 1982;Levelt et al, 1999) lack convincing accounts (see Dell & Reich, 1981;Peterson & Savoy, 1998), the present results are more readily explained by the latter. However, both classes of models have yet to explicitly address how the production of a word impacts future processing of its representations.…”

“…Simpson and Kang (1994) reviewed evidence in word recognition suggesting that committing to or selecting one homophone meaning makes the other meaning less accessible. In a production model with connections from phonological information to lemmas and within-layer inhibitory connections (e.g., Eikmeyer et al, 1999;Stemberger, 1985), homophone lemmas might come to strongly inhibit one another to prevent the unintended meaning's lemma from competing with the intended one. Arguing against inhibitory connections between homophone lemmas are the results of Cutting and Ferreira (1999).…”

“…For example, a word such as rat is more likely to substitute for cat than one would predict based on the probabilities of semantically related word substitutions (dog) and phonologically related word substitutions ( fat). Within interactive activation models (e.g., Dell, 1986Dell, , 1990Dell et al, 1997;Eikmeyer, Schade, Kupietz, & Laubenstein, 1999;Harley & MacAndrew, 1995;MacKay, 1982MacKay, , 1987Stemberger, 1985), this mixed error effect is driven by the convergence of top-down semantic and bottom-up phonological activation on the substituting lemma (Dell & Reich, 1981;Harley, 1993). A system in which the ease of phonological encoding is taken into account in early word production may better support fluent speech than a system that committed itself to using words without regard for their phonological availability.…”