Abstract
While visual working memory (VWM) has limited storage capacity, maintaining stored information for a few seconds, its predecessor, iconic memory (IM), has no such limit but holds stored information for only a few hundred milliseconds. Despite this difference, the exact nature of this trade-off has been relatively unclear. To examine this trade-off, we compared the precision of IM and VWM representations in memory (i.e., s.d.) and the probability of holding test-relevant items in memory (i.e., Pm), by utilizing the mixture model proposed by Zhang and Luck (2008). In the experiment, each trial displayed 3 or 6 colored squares as a sample for 500 ms, followed by a 30- or 900-ms blank display (i.e., IM versus VWM). After the blank display, a retro-cue for a recall item in the sample appeared at one of the sample item locations for 970 or 100ms. Therefore, leaving the same length of a 1s-delay interval on both IM and VWM conditions, the retro-cue yet differing depending on if it appeared within the interval of iconic persistence (<350ms) or beyond it. The recall performance corresponding to the model demonstrated a higher Pm for the IM condition than for the VWM condition, which was more evident in the set size 6 condition than in the set size 3 one. The s.d. was also lower for the IM condition overall, but this advantage was constant across the set size conditions. These results indicate that IM outperforms VWM, by holding more items with better precision overall, but their efficiencies for the representational precision do not much differ. These findings further suggest that the VWM capacity limit may arise primarily from the overall degradation of the iconic representations, forcing a selective storage of the representations, but their precision is set constant, without being traded off for a larger capacity.
Meeting abstract presented at VSS 2014