The effect of selective attention on implicit learning was tested in four e
xperiments using the "contextual cueing" paradigm (Chun & Jiang, 1998, 1999
). Observers performed visual search through items presented in an attended
colour (e.g., red) and an ignored colour (e.g., green). When the spatial c
onfiguration of items in the attended colour was invariant and was consiste
ntly paired with a target location, visual search was facilitated, showing
contextual cueing (Experiments 1, 3, and 4). In contrast, repeating and pai
ring the configuration of the ignored items with the target location result
ed in no contextual cueing (Experiments 2 and 4). We conclude that implicit
learning is robust only when relevant, predictive information is selective
ly attended.