Working memory in spelling: evidence from backward typing

Theories of spelling (Margolin, 1984; Nolan & Caramazza, 1983) propose a working memory system for storing order and identity information of letters during the spelling process. Capacity limitations related to the use of such a graphemic buffer were explored. Participants had to type words backwards. Longer pauses between key presses were assumed to signal points at which graphemic buffer contents were refreshed. Five- and six-letter words were divided by a major pause into chunks of two and three letters, partly coinciding with syllables. Articulatory suppression had no effect on performance. Increasing the length of the stimuli to seven to eight letters resulted in major pauses occurring at syllable boundaries, and performance becoming vulnerable to articulatory suppression but not foot tapping. Forward typing resulted in a similar pause pattern. The results suggest that chunks of approximately three letters can be handled at any one time. For short words the task seems to rely on non-phonological modes of coding, whereas longer words appear to require the use of a phonological code, possibly for keeping track of progress through the word.