On the origin of the cumulative semantic inhibition effect

CUMULATIVE SEMANTIC INHIBITION

Table 2: Regrouping of a subset of the original items in supra-categories.

^a cat. = category

^b N = participants (/24) that named that co-category before the other one

^c Capitani, Laiacona, Mahon, and Caramazza (2003)
terials listed in Table 2 were included. We proceed in three steps. First, we show that the
original effect of interest is similarly present in the restricted dataset. Second, we show that
having named the co-category earlier in the experiment affects naming latencies in a sys-
tematic fashion. Finally, we contrast two assumptions about the underlying representations
that may cause the influence of co-categories on one another.

First step: Is the cumulative inhibition effect present in the restricted dataset?

The data from the ten categories of Table 2 were entered in a linear regression model
very similar² to H-model 3. This model is summarized under N-model 1 on Table 3. The
linear inhibition effect driven by Ordinal position is also present in this restricted dataset,
with an estimated size of the same order of magnitude than in the complete dataset. This
conclusion is strengthened by an inspection of Figure 1, which shows that individual esti-
mates of the inhibition effect for each of the categories in the restricted data set (right panel)
are very similar to those obtained previously from the complete dataset (left panel). Here
again there is a significant increase of the model’s fit with the inclusion of a variable in-

² The variables Lag between trials and Lag between co-categories were included in previous versions of
the analysis. They never contributed significantly, and hence are omitted here for simplicity.

<<   <   3   4   5   6   7   8   9   >   >>