1. The Development of New Tests of Source Memory and a New Approach to the Testing of Equivalence of Parallel Versions.
- Author
-
van Niekerk, Jan K., Nielsen, Tom A., Pasu, Saruban, Shore, Sarah, and Huppert, Felicia A.
- Subjects
MEMORY testing ,COGNITIVE ability ,LONGITUDINAL method ,PSYCHOMETRICS ,DEMENTIA ,NEUROBEHAVIORAL disorders ,MEDICAL research - Abstract
The principal aim of this study was to develop tests of source memory in both the verbal and visuo-spatial domains, which would be brief and easy to use in clinical and population settings. Source memory, that is, memory for context or attributes, is very vulnerable to the effects of ageing and to the early stages of dementia. Because longitudinal studies of cognitive function, including clinical trials, involve repeat testing, practice effects may hamper interpretation when identical tests are administered on each occasion. Accordingly, we developed parallel versions of each of the source memory tests. Rather than assuming equivalence by design, we tested the parallel versions to determine whether they were in fact equivalent in difficulty. We also explored the use of different analytic approaches to establishing equivalence. We report the results of a study in which we compared within- and between-subject performance on three parallel versions of visuo-spatial and verbal source memory tests, as well as three parallel versions of standard tests of episodic memory (word list recall) and executive function (letter fluency, category fluency, ideational fluency) in a sample of healthy elderly volunteers. Participants were tested on three occasions separated by 1 week, and test versions were counterbalanced across test sessions. We also examined performance in a group of participants tested on one occasion only, with test versions being allocated at random. We found that for a number of tests, versions which were designed to be parallel, failed to produce equivalent performance in our sample. We show how the results of this study may be of general use to researchers wishing to select particular versions of the tests in our battery. We also urge researchers to be cautious when using parallel versions of other tests. The careful matching or random allocation of items to parallel versions of a test is no guarantee that performance on the different versions will be equivalent. [ABSTRACT FROM AUTHOR]
- Published
- 2004
- Full Text
- View/download PDF