Eight experiments are reported that examined the contrast between spontaneous and directed recognition using the flanker paradigm. The rationale was that spontaneous recognition of a flanking word would be reflected by the influence that word had on recognition of a target word. Spontaneous recognition, as indexed by flanker effects, was found but only under a restricted set of conditions. When attention was divided at test, recognition decisions for target words were faster when the flanker and target word were congruent (old flanker, old target word; new flanker, new target word) rather than incongruent (new flanker, old target word; old flanker, new target word) with regard to the decision they dictated (Exps. 1, 2, 4, 7, 8a, 8b). However, if old targets were easily identified as old, flanker effects did not emerge, even when attention was divided (Exps. 5 and 6). Surprisingly, increasing the number of prior presentations before a word served as a flanker decreased its likelihood to produce effects when old targets were words that had been presented once at study (Exps. 2 and 4). Words that had been presented to be solved as anagrams at study were effective as flankers, despite their change in physical characteristics from study to test (Exp. 7). Further, rather than the form of modality-specific transfer, flankers were more effective if they matched the modality in which old targets had been presented at study (Exps. 8a and 8b). In combination, the results show that it is the relation between the processing history of the target and flanking words, rather than the absolute history of the flanking word, that determines whether flanker effects will be observed. These findings are discussed in terms of the relativity of automaticity (cf., Neumann, 1984).