What does a patient with semantic dementia remember in verbal short-term memory? Order and sound but not words

Abstract

In this study, we explored capacities for three different aspects of short-term verbal memory in patients with semantic dementia. As expected, the two patients had poor recall for lexico-semantic item information, as assessed by immediate serial recall of word lists. In contrast, their short-term memory for phonological information was preserved, as evidenced by normal performance for immediate serial recall of nonword lists, with normal or increased nonword phonotactic-frequency effects, and increased sensitivity to phonological lures in a delayed probe recognition task. Furthermore, the patients appeared to have excellent memory for the serial order of the words in a list. These data provide further support for the proposal that language knowledge is a major determining factor of verbal STM capacity, but they also highlight the necessary distinction of processes involved in item and order recall, as proposed by recent models of STM.

Acknowledgement

We are extremely grateful to the patients for their patience in participating in the experiments described in this article. We also thank Professor John R. Hodges, Department of Neurology, University of Cambridge, for access to the patients and for permission to publish details of their case histories and neuroradiological information, as well as Martial Van der Linden for helpful discussion. Finally, we gratefully acknowledge the assistance of Jane Hall and Jo Drake during data collection. This study was conducted while the first author was a visiting scientist at the Cognition and Brain Sciences Unit, Medical Research Council, Cambridge, UK. He was supported by a Travel Grant from the Ministry of Research of the French Speaking Community of Belgium and the Belgian National Fund of Scientific Research (FNRS). Steve Majerus is a postdoctoral researcher at the FNRS.

Notes

¹ The cut-off score of the ACE is 88, with 93% sensitivity and 71% specificity for dementia.

² The grammatical ambiguity values provided here represent the average number of different grammatical classes that a same word can be a part of; the majority of words in both lists could be either nouns or verbs.

³ A sequence length of 6 words is the standard sequence length used in experiments exploring the influence of imageability on immediate serial recall in healthy adults (e.g., Walker & Hulme, 1999). This relatively long sequence length is chosen in order to reach the limits of STM capacity and hence increase the reliance on long-term semantic knowledge during performance on this task. However, for patients with impaired semantic knowledge, this would have proven too difficult given that a number of the words to be recalled would not be known anymore, and hence the limits of STM capacity would be reached earlier. Preliminary testing had shown that a sequence length of 5 words for the patients was a more appropriate length avoiding floor effects.

⁴ For example, for a target sequence such as “map, cat, mat, ham, ram” and a response such as “mat, cat, map, ram, ham” we scored 4 order errors; however, “mat” in Serial Position 1 of the response sequence could also have been a blend error resulting from the combination of the onset of “map” and the coda of “cat”; the same applies to “order” errors “map”, “ram”, and “ham” in the response sequence.