Abstract
In order to deal with our complex visual environment, human observers have developed the capability to extract a variety of statistical regularities from our ambient array, thereby facilitating attentional orienting. For instance, in visual search, detection of a target is faster when the spatial configuration or the object identities of nontarget items are repeatedly encountered. These results show that both spatial and object-based contextual invariances can implicitly guide attention to the target (contextual cueing; Chun, 2000, Trends Cogn. Sci., for review). Here, we explored how such acquired contextual regularities can be adapted subsequent to an unexpected environmental change (see Zellin et al., 2014, Psychonomic Bull. Rev.). A series of experiments were performed in which, in an initial learning phase, observers learned to associate a given repeated context of nontargets with a given target. A subsequent test phase then introduced identity and/or location changes to the target. Our results show that observers were rather ineffective in adapting to a change of the spatial location, or to the object identity of the target. However, when the change to the target occured both in terms of its identity and in location, then contextual cueing was effective again shortly after the change – showing that successful adaptation occurred. These results suggest that contextual learning is capable to efficiently extract regularities relating to the spatial configuration and to object identity at first. However, contextual learning, once acquired, is rather insensitive to adapt, at least, as long as it does not reveal a "rich", i.e., redundant change signal that combines both spatial- and object-based scene statistics.
Meeting abstract presented at VSS 2016