Reaction times in the study of strategies in sentence-picture verification: A reconsideration

Investigations of cognitive strategies based on chronometric data may appear to contradict findings from verbal-report-based approaches. These inconsistencies may be due to misuse of RTs to infer subjects' behaviour. To test this hypothesis, 15 subjects were successively required to adopt 4 different strategies in a sentence-picture verification task. Trend analyses provide evidence that (1) highly dissimilar strategies can nevertheless yield similar mean RT curves as well as similar individual subjects' RT curves; (2) subjects using similar strategies can produce individual RT curves presenting vastly different features due partly to random variation and partly to idiosyncrasies. These findings suggest that RTs are an unreliable behavioural indicator for inferring individual or group strategies on certain experimental tasks and should be analysed in conjunction with other sources of information. An approach consisting of using RTs to test hypotheses based on subjects' verbal reports is described along with the appropriate controls. Results show that this approach provides an improved account for observed RTs.