Does perceptual simulation explain spatial effects in word categorization?

In three experiments we investigated the origin of the effects of the compatibility between the typical location of entities denoted by written words (e.g., "up" for eagle and "down" for carpet) and either the actual position of the words on the screen (e.g., upper vs. lower part of the screen), or the response position (e.g., upper- vs. lower- key presses) in binary categorization tasks. Contrary to predictions of the perceptual simulation account (Barsalou, 1999), conceptual spatial compatibility effects observed in the present study (faster RTs when the typical position of the stimulus referent in the real word was compatible with either the stimulus or response physical position) seem to be independent of whether there was an overlap between simulated processes possibly triggered by the presented stimulus and sensory-motor processes actually required by the task. Rather, they appear to depend critically on whether the involved stimulus and/or response dimensions had binary, variable (vs. fixed) values. Notably, no stimulus-stimulus compatibility effect was observed in Experiment 3, when the stimulus physical position was presented in a blocked design (i.e., it was kept constant within each block of trials). In contrast, in all three experiments, a compatibility effect between response position and another (non-spatial) conceptual dimension of the stimulus (i.e., its semantic category) was observed (i.e., an effect analogous to the MARC [linguistic markedness of response codes] effect, which is usually observed in the number domain; Nuerk et al., 2004). This pattern of results is fully accounted for by the polarity principle, according to which these effects originate from the alignment of the polarities of either different stimulus dimensions or stimulus and response dimensions.