Hulleman & Olivers' (H&O's) focus on simulating how slope gradients are influenced by the difficulty of search provokes a thoughtful discussion. However, limiting simulations to these data alone can mistakenly suggest that the FVT framework's usefulness is itself limited. To help address this perceived limitation, we have reanalysed data from a study of the Prevalence Effect (PE; Godwin et al. Reference Godwin, Menneer, Cave, Thaibsyah and Donnelly2015a; Wolfe et al. Reference Wolfe, Horowitz and Kenner2005).

The PE refers to the influence that target probability has on both target selection and verification (e.g., Godwin et al. Reference Godwin, Menneer, Riggs, Cave and Donnelly2015b; Hout et al. Reference Hout, Walenchok, Goldinger and Wolfe2015). Frequently occurring targets tend to be found and verified quickly. In contrast, their absence is reported slowly. The presence of infrequent targets is reported slowly and their absence reported quickly.

The target article accounts for the modulating effect of target discriminability on search reaction times solely by changes in the size of the FVF. Might changes in the size of the FVF also contribute to the PE? Specifically, high target prevalence might lead participants to initially adopt a broader FVF than when target prevalence is low. A relatively broad FVF would allow the presence of targets to be detected quickly whereas a relatively narrow FVF would lead to slowed target detection. In deriving these hypotheses, we have made two assumptions. First, and to account for slow target-absent responses when target prevalence is high, we assume that failure to find evidence of target presence when the FVF is broad leads to a dynamic resizing of the FVF to allow, at the limit, item-by-item analysis (note that a global-to-local fixation pattern is consistent with recent consideration of search, Godwin et al. Reference Godwin, Reichle and Menneer2014; Over et al. Reference Over, Hooge, Vlaskamp and Erkelens2007). Second, we assume that the fixation point of a broadened FVF is more likely to be centrally than peripherally positioned. For a broad FVF, a central fixation will encompass more items than a non-central fixation will. These reduce to a hypothesis that, early in search, fixations are more centrally biased in high-prevalence than low-prevalence search.

To test this hypothesis, we reanalysed data on target-present trials from Godwin et al. (Reference Godwin, Menneer, Cave, Thaibsyah and Donnelly2015a). Space restrictions prohibit a full account of these data and analyses. Briefly, to assess the patterns quantitatively, the distribution of fixation locations across displays were normalised within high- and low-prevalence conditions and split into fixations made early and late in search (as defined by median split). Z-scores for the differences between high- and low-prevalence conditions were calculated for these normalised data. We found that increasing prevalence is associated with more fixations to the centre of search displays early in search. A centre bias (Tseng et al. Reference Tseng, Carmi, Cameron, Munoz and Itti2009) is present early in low-prevalence search, but the bias is significantly stronger under high prevalence.

These data, then, are consistent with the FVF framework. However, we do not claim that our reanalysis provides unequivocal support. Rather, the framework prompts us to reconsider data in a way that provide an additional account of how search patterns might change with target prevalence.

The current utility of the FVF is tempered, in our view, by two limitations. First, the current focus on numbers of fixations ignores the influence of fixation duration. Increasing cognitive demands affects both the number and duration of fixations (Liversedge & Findlay Reference Liversedge and Findlay2000). Consequently, any comprehensive framework of search behaviour must explain both fixation number and duration. Recent evidence suggests that fixation durations during visual search are controlled on the basis of a trade-off between making rapid fixations and allowing time to examine objects in the display (Godwin et al. Reference Godwin, Reichle and Menneerin press). As a consequence, there have been calls for a greater understanding of fixation duration variability during visual search tasks (Reingold & Glaholt Reference Reingold and Glaholt2014).

Second, the authors rightly wish to extend consideration to searching in scenes. As the search environment becomes richer in contextual information, equation of selection time, processing time, and dwell time to fixation time (sect. 6.3) becomes more challenging. In reading, spillover effects are frequently observed (whereby a linguistic influence of one word is seen to affect fixations on it and later words in the sentence; Rayner & Duffy Reference Rayner and Duffy1986). By extension, visual search in scenes may also be subject to partial dissociation between fixation location and the set of locations from which information is currently being processed. To this extent, evaluation of effects across temporally contiguous fixations as well as spatially contiguous fixations is a critical issue for theoretical development.

In sum, we consider the FVF framework as a useful prompt to rethink visual search. Here we have provided some provisional data that might further support this framework. In addition, two areas of concern to be addressed in future developments have been noted.