Abstract
The target template is often described as the mental representation that drives attentional selection, for instance, in visual search. However, this template is not necessarily a veridical representation of a sought-after target. According to Optimal Tuning, the attentional template shifts to an exaggerated target value to maximise the signal-to-noise ratio when the target is similar to the non-targets. By contrast, the Relational Account states that attention is tuned to the relative target feature that specifies how the target differs from the other items in the context (e.g. all redder items or the reddest item). Both theories are empirically supported, but used different paradigms (perceptual decision tasks vs. visual search), and different attentional measures (probe response accuracy vs. gaze capture). Here, we incorporated both paradigms to provide a critical test of these accounts. The results revealed Optimal Tuning shifts in probe trial accuracy (when participants had to indicate the location of the target), but this did not drive early attention or gaze behaviour in visual search. Instead, attentional guidance followed the Relational Account, selecting all items with the relative target colour (e.g., redder). This suggests that the masked probe trials used in Optimal Tuning do not probe the attentional template that guides visual attention. Moreover, we found that optimal tuning shifts could be explained by simultaneous contrast effects: Surrounding an odd-coloured target with similar-coloured nontargets can shift the appearance of the target (e.g., make the target appear slightly redder). This suggests that the optimal tuning shift in probe responses may in fact be a perceptual artefact rather than a strategic adaptation to optimise the signal-to-noise ratio. In sum, the results demonstrate that the attention-guiding target template contains relative features while the template guiding decision-making contains a veridical representation of the target that can be shifted due to simultaneous contrast effects.