Implicit memory for regularities in rapid sound sequences: behavioural evidence

Bianco, R. 1 , Harrison, P. . 2 , Picken, S. . 1 , Pearce, M. 2, 3 & Chait, M. 1

1 UCL Ear Institute, University College London, London, UK
2 School of Electronic Engineering and Computer Science, Queen Mary University of London, UK
3 Department of Clinical Medicine, Aarhus University, Denmark

The ability to detect and store acoustic statistical regularities is fundamental to auditory perception and cognition. Recent work using 'repeated-noise paradigms' demonstrated robust, log-term implicit memory for 'frozen' reoccurring white noise. Results suggested that participants become sensitive to idiosyncratic, local spectral features within the noise samples. Here we introduce a complementary paradigm to study memory for sequences of events. We investigate (1) whether listeners remember sequences of acoustic events (2) the limits on this memory (3) and whether it requires active involvement with the sequence.
Participants detected the transition from random to regular frequency patterns (repeating cycles of twenty 50 ms tone-pips) in novel rapid-sound sequences. Unbeknownst to them, three regular patterns sparsely reoccurred across trials.
Compared with novel sequences, RT to reoccurring regularities became 250ms faster within few reoccurrences. This benefit persisted after 24 hours, and up to 7 weeks; it was robust to doubling the number of regularities to memorise, and to subsequent memorization of other patterns. It could occur without listeners noticing, during passive exposure. A computational model is also provided.
Overall, we show that as regularities reoccur they are implicitly stored in memory, progressively up-weighted and more quickly retrieved to resolve the identity of sensory signals.