Infants Get Five Stars on Iconic Memory Tests

Abstract

By the late 19th century, tests of adults' shortterm visual memory—how much one retains from a briefly presented display—had acquired a familiar character: An alphanumeric array was presented, removed, and participants were asked to report the items (Wundt, 1912). These ‘whole report’ tests revealed capacities of 3-4 items (Cattell, 1886; Sperling, 1960). However, participants often felt they had seen more items, but forgotten them before report. A crucial innovation was to ask for just a ‘partial report’, cuing participants to sample from their memory (e.g. a high-pitched tone might cue report of the middle of three rows of letters) (Sperling, 1960). If the cue occurs after display offset (a ‘post-cue’) and the sampled subsets are random, then accurate reports mean all the presented items had been stored. This clever methodological change exposed a new—early, high-capacity (∼9 items), and fast-decaying (∼200 ms half-life)—memory system dubbed ‘iconic memory’ (Neisser, 1967). In this sense, then, the study of infants' visual memory is still rooted in the conventions of the 19th century: No one has yet asked infants for a partial report. This leaves a conspicuous gap in our understanding of visual memory development. Research on infants' visual memory has a similarly long history (e.g. Hunter, 1917; Fantz, 1964; Fagan, 1970; for reviews see Nelson, 1995; Rose, Feldman & Jankowski, 2004). And while there has been ample work on infants' short-term memory, for example recency/primacy effects (Olson, 1979; Cornell & Bergstrom, 1983), short-term capacity has been studied only recently (and iconic memory not at all). These studies though have revealed a striking limitation: young infants' short-term memory capacity seems limited to a single object (see Ross-Sheehy, Oakes, & Luck, 2003 for evidence from visual short-term, <300 ms retention interval, memory; Pelphrey et al., 2004 and Kaldy & Leslie, 2005 for evidence from visual working memory). 6-month-olds, then, have sufficiently well-developed visual acuity to distinguish arguably dozens of objects in a single fixation (Teller, 1997), but short-term memory for only one—a curiously inimical constraint unless this object were very well chosen (e.g. by virtue of high salience, task relevance, or cueing). We argue here that iconic memory is the buffer holding the choices.