A common approach in studies of human cognition is to derive conclusions from the aggregate performance of groups of normal subjects. These group-based results are then used in the development of models of a cognitive architecture. What is often missed is that in many experiments there are some subjects who do not consistently show the pattern of results observed for the group as a whole. In this case, the aberrant pattern is often treated as random variation. The issue becomes more acute when the aberrant pattern of performance is obtained from single cases of patients with neurological damage. In this case however, the data are interpreted as indicative of some cognitive deficit, and their modification following a given treatment as indicative of the effect of that treatment. The aim of this study is to tackle the issue of random variation in normal control groups. We argue that one way to avoid biased results is to specify the generality and the reliability of normal group effects before attempting to apply them to neuropsychological patients. The point we want to stress is a broad one. However, to make this issue concrete we will discuss findings in the context of a specific area of cognition, namely, short-term verbal memory. To make our pointwe summarise some ofourownfindings from groupstudies in normal controls, and examine why some individuals, with no overt neuropathology, fail to show widely replicated, well-established phenomena linked with verbal short-term memory. In particular we focus on (1) the possible influence of memory span, (2) the effect of strategy choice, (3) the presence of control samples of subjects with possible hidden cerebral lesions, and (4) the problem of poor reliability. We conclude by arguing for the general importance of test-retest reliability, and of strategy choice in both clinical studies and theoretically driven studies of cognitive deficits.
Read full abstract