Provided by Cognitive Sciences ePrint Archive
On the origin of the cumulative semantic inhibition effect
______ . _ . '
F.-Xavier ALARIO & Fermln MOSCOSO DEL PRADO MARTIN
Aix-Marseille Université & C.N.R.S.
Marseille, France
Abstract
We report an extension of the cumulative semantic inhibition effect found by
Howard, Nickels, Coltheart, and Cole-Virtue (2006). Using more sensitive
statistical analyses, we found a significant variation in the magnitude of the
effect across categories. This variation cannot be explained by the naming
speed of each category. In addition, using a sub-sample of the data, a second
cumulative effect arouse for newly-defined supra-categories, over and above
the effect of the original ones. We discuss these findings in terms of the
representations that drive lexical access, and interpret them as supporting
featural or distributed hypotheses.
In a recent article, Howard et al. (2006) reported a remarkable observation concerning
lexical access during speech production. They asked speakers to name pictures of common
objects presented in a continuous series of trials. The sequence of trials had an underlying
structure; the members of a given category were separated by a variable, yet carefully
controlled, number of trials. Such design allowed contrasting semantic effects driven by
the ordinal position of every response within the sequence of items from its category, from
those driven by the distance (in trials) between related items.1 The remarkable observation
was that the ordinal position within the category had a very systematic effect, whereas the
distance of the previous related item (measured in trials) did not affect performance. The
relationship between ordinal position and average naming latency was reported to be linear.
A unit increase in ordinal position lead to an average increase of 30 ms in naming latency
(95 confidence interval ± 8.2 ms). As shown in their Figure 1, the average latency data
plotted against ordinal position are remarkably aligned.
In the theoretical discussion of this phenomenon, Howard et al. (2006) focus on
the cumulative property of the effect. These authors rightly highlight that none of the
current models of lexical access in language production predicts this observation. They then
1More specifically, each list comprised 120 items drawn from 24 different categories (i.e. 5 items per
category), plus a set of 45 filler pictures. The items of a given category were separated by 2, 4, 6 or 8 trials;
this arrangement was counterbalanced across participants and categories.
The authors thank David Howard for kindly providing the complete dataset of the original publication.
Correspondence can be addressed to [email protected]