Implicit learning in visual search: Implications for complex task performance

Implicit learning, defined here as learning without intention, was demonstrated during the performance of a search task, thereby extending the finding of implicit learning in relatively simple tasks. Implicit learning was primarily limited to unique associations in a target feature sequence (i.e., associations in which a particular element is always followed by the same element, as 2 follows 1 in the sequence 125124). Implicit learning was similar under both distributed and focused visual attention and was unaffected by added workload, indicating that implicit learning utilized minimal attentional resources. In contrast, when subjects were instructed to learn the sequence (i.e., were to use an explicit learning mode), sequence learning was adversely affected by added workload, and the impairment was similar for both unique and ambiguous associations. Implicit learning in a distributed-attention search task was similar to implicit learning in a simpler serial reaction time task. However, in a focused-attention search task that was more perceptually-demanding, sequence learning benefits did not appear until late in task performance. This pattern suggests that implicitly-learned sequence information may have been retrieved and used only after becoming explicit. This research also demonstrated that associations between sequence elements are not learned implicitly if a sequenced stimulus feature is not assigned to responses. This indicates that associations between stimulus-response pairs, but not between stimuli alone, were learned. In addition to ambiguous sequence associations, high perceptual demands, and the absence of response assignments for patterned stimuli, low accuracy rates, low event rates, and cumbersome response codes were implicated as potential sources of interference in implicit learning. An instance-based model of implicit learning is indirectly supported, according to which implicit learning benefits reflect the storage and subsequent activation (i.e., retrieval) of instances that contain response-relevant stimulus information, the response, and information about the next event or immediate consequence of that response