Conducted 6 experiments to examine the importance of flanker priming in the selective-attention paradigm of Eriksen and Eriksen (1974). Ss classified target letters presented in the relevant center position of a 3-letter visual display, ignoring simultaneously presented irrelevant flanker letters on each side of it. Although Ss had been instructed to ignore the flankers and later seemed to have no awareness of their correlation with correct responses, response times indicated that Ss learned and were influenced by the predictive relationship of the flanker to the response. These findings contradict the filter and attention-and-priming early selection models and provide examples of semantic processing of unattended stimuli that have not been primed by either task relevance or long-term relevance and of associative learning that is both incidental and apparently unconscious.