Structural complexity is not the (big) issue: A reply to Roelofs (2007)

Abstract

In the last two decades, La Heij and colleagues have presented accounts of a number of context effects in Stroop-like word-production tasks. Roelofs (2007 this issue) criticises various aspects of our proposals, ranging from the number of processing stages assumed to details of simulation results. In this reply we first argue that we do not challenge spoken-word production models developed in the psycholinguistic tradition for being ‘too complex’, as Roelofs asserts. Next we discuss Roelofs’ detailed criticisms on our proposed solutions. Finally, in response to Roelofs’ argument that increasing the structural complexity of our model would render it similar to its main competitor, WEAVER++, we discuss the crucial differences that would still remain.

Notes

¹We may add, however, that the results of our implemented model are in line with the pattern of distractor-naming errors in a speeded-naming variant of the PWI task, reported by Starreveld and La Heij (1999).

²Glaser and Glaser (1982) were aware of this problem and performed a control experiment in which the participants were asked to name the position of the target word (above or below the central fixation point). However, this task is not an appropriate control, for the simple reason that having two identical words in the display is probably very helpful in a word reading task (incorrect selections even go unnoticed), whereas having two identical words in the display does not provide any information concerning the position of the target.

³One possible way in which a strong form of the discreteness assumption could be rescued is by assuming that the phonological facilitation effect observed in these paradigms is due to feedback from the word-form level. That is, the activation of the initial phonemes of the response word may induce the activation of the name of the context stimulus (see Costa, La Heij, and Navarrete, 2006, for a similar argument). Although such an account is hard to defend, it may be useful to obtain corroborating evidence in favour of the cascading view using paradigms in which this form of feedback is less likely to occur.