Abstract
<p>A classic demonstration of long-term learning by repetition is the Hebb effect: immediate recall improves for memory sets presented repeatedly amidst non-repeated sets. Although long regarded as a form of implicit learning, recent work combining individual-participant analysis with computational modeling has shown that learning emerges only after participants become aware of the repetition. The reminding account of the Hebb effect explains these recent findings by proposing that repetition learning requires retrieving a prior encounter of the repeated material from episodic memory during re-encoding; repetitions that go unrecognized are encoded as novel episodes and produce no learning. Here, we tested this explanation using EEG. Participants performed a visual working memory task on 150 visual arrays per block, two of which recurred every ~5th trial. During the task, we measured recall performance, trial-by-trial repetition awareness, and EEG during encoding of arrays. Behaviorally, we replicate that learning of repeated arrays emerged only after participants recognized their repetition. At the neural level, the late positive component (LPC), an established marker of episodic recollection, was enhanced for repeated relative to novel arrays — but only after recognition, and already before recall improved. In contrast, the FN400, an index of familiarity, and a whole-scalp topographic analysis revealed no difference between unrecognized repeated and novel arrays, indicating that unrecognized repetitions are processed like new information. We conclude that repetition learning in the Hebb paradigm is gated not by the gradual, implicit accumulation of experience, but by the explicit retrieval of prior encounters from episodic memory at reencoding.</p>