(2002) used computer simulations to claim that many current models of metacontrast masking can account for the findings of V. Di Lollo, J. T. Enns, and R. A. Rensink (2000). They also claimed that notions of reentrant processing are not necessary because all of V. Di Lollo et al. 's data can be explained by feed-forward models. The authors show that G. Francis and F. Hermens's claims are vitiated by inappropriate modeling of attention and by ignoring important aspects of V. Di Lollo et al. 's results.We note with interest Francis and Hermens's (2002) article, which purports to show that the findings reported by Di Lollo, Enns, and Rensink (2000) can be explained by other models of metacontrast masking. To buttress their claim, Francis and Hermens reported computer simulations showing that some of our results can be modeled by the theories of Bridgeman (1978), Francis (2000) and Weisstein (1968). This claim has a good deal of surface appeal because it is parsimonious. It argues that our results can be explained without recourse to the new concept of object substitution. Parsimony, however, is achieved at the cost of inappropriate modeling of attention and modeling an incomplete portion of our masking data. Here, we reiterate our original claim that reentrant modeling is necessary for explaining our findings. We do so by showing that a plausible case for the sufficiency of feed-forward processes has not been made by Francis and Hermens.
Modeling of AttentionModeling of the effects of attention in Di Lollo et al. 's (2000) study was based on the large literature on set-size effects in visual perception (Duncan & Humphreys, 1989; Eriksen,1995;Sperling, 1960;Treisman & Gelade, 1980). Accordingly, set size was modeled as though it influenced the time required for attention to be deployed to the location of the target. The larger the set size, the longer it takes to find the target.Current models of masking do not encompass attentional effects. This is one reason why they cannot account for Di Lollo et al. 's (2000) data. Francis and Hermens (2002) modeling attention as affecting the intensity of the mask. As set size is increased, the mask representation is assumed to become stronger whereas the target representation remains the same. We believe this modeling of attention to be inappropriate.It has long been known that the strength of metacontrast masking is directly related to mask intensity (Alpern, 1953). It is, therefore, unsurprising that weakening the mask reduces the metacontrast effect in all extant models. There is no need to expand current models along these lines because they already include the effect of mask intensity. The question that Francis and Hermens (2002) did not address is why anyone should model attention in terms of mask intensity unless they were aware of Di Lollo et al. 's (2000) results and did not have any other convenient mechanism for reducing the magnitude of the metacontrast effect. In more general terms, there is a substantial difference between claiming that current models can a...