Temporal stability and representational distinctiveness: Key functions of orthographic working memory

Abstract

A primary goal of working memory research has been to understand the mechanisms that permit working memory systems to effectively maintain the identity and order of the elements held in memory for sufficient time as to allow for their selection and transfer to subsequent processing stages. Based on the performance of two individuals with acquired dysgraphia affecting orthographic working memory (WM; the graphemic buffer), we present evidence of two distinct and dissociable functions of orthographic WM. One function is responsible for maintaining the temporal stability of letters held in orthographic WM, while the other is responsible for maintaining their representational distinctiveness. The failure to maintain temporal stability and representational distinctiveness gives rise, respectively, to decay and interference effects that manifest themselves in distinctive error patterns, including distinct serial position effects. The findings we report have implications beyond our understanding of orthographic WM, as the need to maintain temporal stability and representational distinctiveness in WM is common across cognitive domains.

Keywords:

• Working memory
• Spelling
• Dysgraphia
• Orthographic representations

Acknowledgments

This research was made possible with the support of NIH (National Institutes of Health) Grant DC006740 to B.R., a predoctoral research fellowship from the William Orr Dingwall Foundation to S.F.B., the support of PAT (Provincia Autonoma di Trento) to R.C. and G.M., and a doctoral research fellowship from University of Trento to V.C.

Notes

¹ Evidence from a number of dysgraphic individuals reported in the literature indicates that double (geminate) letters are represented by a single letter with separate representation of doubling information. For example, patient L.B. (Caramazza & Miceli, 1990) never made errors involving only one of the geminate consonants (pa ll a → p l a l a; see also, Miceli et al., 1995; Schiller et al., 2001; Tainturier & Caramazza, 1996; for a review, see Miceli & Capasso, 2006). G.S.I.'s and C.R.I.'s spelling errors were consistent with these reports. In fact, substitutions and transposition errors occurred on both letters of a geminate and never involved only one of the two letters.

² While Cox and Snell pseudo-r ² values cannot be interpreted as the proportion of variance explained, as in a linear regression, they can be used to evaluate multiple models predicting the same outcome on the same dataset, with the higher pseudo-r ² value indicating the model that better predicts the outcome (e.g., Freese & Long, 2006).