Parallel encoding of information into visual short-term memory

Humans can temporarily retain information in their highly limited short-term memory. Traditionally, objects are thought to be attentionally selected and committed to short-term memory one-by-one. However, few studies directly test this serial encoding assumption. Here, we demonstrate that information from separate objects can be encoded into short-term memory in parallel. We developed models of serial and parallel encoding that describe probabilities of items being present in short-term memory throughout the encoding process, and tested them in a whole-report design. Empirical data from four experiments in healthy individuals were fitted best by the parallel encoding model, even when items were presented unilaterally (processed within one hemisphere). Our results demonstrate that information from several items can be attentionally selected and consequently encoded into short-term memory simultaneously. This suggests the popular feature integration theory needs to be reformulated to account for parallel encoding, and provides important boundaries for computational models of short-term memory.