Scanning acoustic short-term memory: Evidences for two subsystems with different time-course and memory strength.

We studied the cognitive processes involved in pure acoustic memory retrieval using ERPs and independent component analysis (ICA) of the ERPs. We used a modified Sternberg task with pure tones as memory items, with 3 memory set size conditions (2, 4, or 6 tones), and a control condition in which no task-relevant tone was presented (tones replaced by white-noise segments). Participants reported the presence or absence of a probe tone in the memory set. EEG signals were recorded using a BioSemi ActiveTwo system, with 64 active electrodes following the 10-10 system and average-mastoids reference. A P3 component, at electrode POz, was elicited by the probe when items were to be held in memory, no P3 was observed in control trials. With an increase in memory load, P3 amplitude decreased, accuracy was lower, and response time (RT) longer. There was a clear recency effect, in the electrophysiological and behavioral results reflected in a larger P3, shorter RT, and higher accuracy, when the probe matched the last tone of a sequence than when it matched any of the other presented tones, regardless of memory load. P3 amplitude and behavioral results were similar across all the other serial positions and loads. The results suggest that acoustic memory for pure tones is mediated by two subsystems: one encoding the last item with high probability, and another for all other serial positions. ICA of the ERPs isolated parieto-occipital ICA components with larger activations for probes that matched the last serial position than for earlier ones, and an anterior component that had little activation for the last serial positions and large activations for all earlier ones, corroborating the two-systems theory.