Therefore, it is unlikely that the spatiotopic learning directly engages peri-saccadic updating of stimulus representations. As discussed above, an explicit spatiotopic map and

peri-saccadic find protocol updating of visual representation are unlikely to be directly engaged in encoding of the spatiotopic learning effect that we observed. As these non-retinotopic mechanisms are mainly seen in the frontoparietal areas, which are also responsible for saccade control and attention allocation (Colby & Goldberg, 1999; Corbetta & Shulman, 2002; Moore & Armstrong, 2003; Shipp, 2004), we cannot exclude the possibility that these non-retinotopic mechanisms could be indirectly involved in spatiotopic perceptual learning by interacting with attentional and saccadic control mechanisms. It has been shown that attention (Connor et al., 1997; Gottlieb et al., 1998; Womelsdorf et al., 2006; Crespi et al., 2011)

and eye movements (Tolias et al., 2001) are critical in generating non-retinotopic properties of visual representations. This is consistent with our finding of the dependence of learning-induced spatiotopic effects GSI-IX price on attention allocated to the first stimulus (Fig. 6). In fact, although attention can be maintained at the same retinotopic location immediately after saccadic eye movements (Talsma et al., 2013), attention deployment also shows some non-retinotopic properties that parallel those of visual representations: attention to a cued location can be predictively remapped, immediately before a saccade, to the retinotopic location that will match the cued spatiotopic location after the saccade (Mathôt & Theeuwes, 2010; Hunt & Cavanagh, 2011; Rolfs et al., 2011); attention can also be allocated to a cued spatiotopic location across saccades

(Golomb et al., 2008, 2010a,b, 2011; Mathôt & Theeuwes, 2010), or to a cued location relative to a reference stimulus (Boi et al., 2011). Despite its importance in non-retinotopic representation, spatial attention or its remapping alone cannot account for the dependence of spatiotopic specificity on simple stimulus attributes that are AZD9291 encoded by the specialized visual cortex. Although a single process is unable to account for our data, the spatiotopic learning effect can be well explained by taking into account interactions between bottom-up and top-down processes (Fig. 7). In our experiments, initial attention allocated to the first stimulus can serve as a reference for subsequent remapping of attention to the retinotopic location corresponding to the second stimulus. This attentional remapping process, which could be based on corollary discharge associated with gaze shift and/or on a gaze-invariant spatiotopic map in higher-order cortical areas, is dependent on the saccade direction and/or the spatiotopic stimulus relation (congruent or incongruent) in our experiments.