Late guidance resolves the search slope paradox in contextual cueing

Visual search is facilitated by statistical learning of repeated search contexts, termed 'contextual cueing'. Repeated displays are thought to enhance attentional guidance, but this has been challenged by the absence of search-slope differences between repeated and novel displays. Here we use eye-tracking to resolve this paradox by calculating a measure of when during search the contextual cueing benefit emerges. In 24 human participants we observe typical reaction time and fixation count benefits for repeated contexts, but no slope differences between repeated and novel search contexts. Eye-tracking showed that the attentional guidance benefit emerged over time, occurring later for larger set sizes, and producing similar response time benefits for small and large set sizes. We argue that repeated and novel contexts have similar slopes because learning benefits are confined to target-adjacent regions of roughly equivalent area across set sizes. This finding rules out one of the strongest pieces of evidence against an attentional account of contextual cueing.