Human cognitive neuroscience seminar

Speaker: Professor R.H. Logie (University of Edinburgh)

Title: In this book there will be no statistics whatsoever (Bartlett 1932/1961): Cognitive strategies in working memory tasks

Abstract: In his book ‘Remembering’, Bartlett (1932) stated ‘In this book there will be no statistics whatsoever’. His argument was that because different individuals may perform a memory task in different ways, or using different strategies, it is misleading to draw general conclusions from data aggregated across a group of individuals. This issue was raised by others over the last 90 years, who demonstrated the range of strategies that participants adopt when performing many cognitive tasks. In research on human cognition, it is common to explore whether the aggregate data from multiple participants have sufficient statistical power to reach a standard, but arbitrary level of significance. Statistics provide a set of useful research tools, but should not be the only basis on which to draw conclusions. Effectively, statistics reflect how many participants show an hypothesised difference between experimental conditions compared with participants who do not show that difference. Data from the latter are treated as statistical noise and often are ignored. However, those data could indicate that some participants are performing the task in a different way from the majority. This suggests that it may be misleading to draw general conclusions from aggregate data and the associated statistics about the characteristics of human cognition. A number of different experiments will be described, demonstrating variation across participants in strategies used by participants in the experimental psychology lab, and in brain imaging studies. Regarding the latter, brain activation patterns may vary depending on the cognitive strategy adopted by each participant in the scanner. I will argue that it is essential to develop paradigms that can identify HOW participants are performing a task if we are to have confidence in our interpretation of experimental data.