Semantic contributions to immediate serial recall: Evidence from two contrasting aphasic individuals

Abstract

This paper examines the effect of semantic variables on serial recall in two contrasting aphasic cases and a group of controls. Experiment 1 manipulates word imageability and Experiment 2 manipulates semantic similarity. Controls not only showed better recall of imageable/semantically grouped lists, but under some conditions they also produced proportionately fewer phonological errors. These findings suggest that increasing the effectiveness of lexical/semantic support reduces reliance on phonological support. Consistent with this proposal, case TV, whose phonological impairment should increase his reliance on lexical/semantic support, produced abnormally low rates of phonological errors under some conditions. Conversely, case NP, who had a lexical/semantic impairment, produced abnormally high rates of phonological errors under some conditions. Analysis of serial recall curves in both aphasics and controls support the hypothesis that phonological processes are particularly critical for the recall of list-final items. However, there was no evidence that semantic support is especially crucial for list-initial recall. Controls did not exhibit stronger effects of semantic variables at list-initial position. Case NP (lexical/semantic impairment) performed disproportionately poorly on these items, but only under certain conditions.

Keywords:

• Aphasia
• Language
• Short-term memory
• Serial recall
• Spoken word production
• Anomia

Acknowledgments

We thank NP and TV and all our controls for participating in this study. Thanks also to Liz Stuart for her help with the testing of TV and NP for Experiment 2, and to Kate Smith for her feedback on an earlier version of the manuscript.

Notes

¹We performed a General Linear Mixed Model analysis on the entire combined data set for TV and controls, incorporating participant group (TV vs. Control) as an independent variable and including participant identity and item identity as random effects.

²Values for the “middle” position were calculated by averaging the values for all non-initial, non-final positions for each particular sequence.

³This pre-exposure phase was intended to provide a baseline reading/repetition score for each participant on the stimulus word set. It was also intended to provide some exposure to the stimulus words prior to the commencement of the experimental sequences, to reduce the possible contribution of word recognition errors to performance on the first few word sequences.

⁴Six observations were missing for NP, due to experimenter error.

⁵List-initial items, being the first to be produced, may also be subject to less decay. Indeed, the concepts of interference and decay are difficult to distinguish empirically (see. e.g., Cowan et al., 2002).