Abstract
Cognitive neuroscience involves the simultaneous analysis of behavioral and neurological data. Common practice in cognitive neuroscience, however, is to limit analyses to the inspection of descriptive measures of association (e.g., correlation coefficients). This practice, often combined with little more than an implicit theoretical stance, fails to address the relationship between neurological and behavioral measures explicitly. This article argues that the reduction problem, in essence, is a measurement problem. As such, it should be solved by using psychometric techniques and models. We show that two influential philosophical theories on this relationship, identity theory and supervenience theory, can be easily translated into psychometric models. Upon such translation, they make explicit hypotheses based on sound theoretical and statistical foundations, which renders them empirically testable. We examine these models, show how they can elucidate our conceptual framework, and examine how they may be used to study foundational questions in cognitive neuroscience. We illustrate these principles by applying them to the relation between personality test scores, intelligence tests, and neurological measures.
Acknowledgments
This article is original, and the material has not been published elsewhere. The participants were tested in accordance with the ethical guidelines of the American Psychological Association, and this research was approved by the University of Amsterdam Ethical Committee.
Notes
1We refer to the discipline here as cognitive neuroscience, as it is the broadest and most common name for the concurrent study of psychological behavior and physiological properties. However, we do not aim to restrict our perspective to merely cognitive phenomena such as attention, memory, or intelligence: The issues we raise are equally of interest for fields such as social neuroscience or affective neuroscience. Wherever we state cognitive neuroscience, we mean to encompass such more specific branches.
2A similar position can be found in D. Davidson (1980, p. 111).
3Given the exact formulation as a SEM, one should construe this to mean that variability in the underlying attribute causes variability in both the P- and the N-indicators.
4A tetrad is the difference of the products of the covariances of four measured indicators.
5The idea that epistemological speculation can gradually be replaced by empirical science, sometimes termed naturalism or naturalized epistemology, has a long history. See, for instance, CitationQuine (1969) for a philosophical motivation.
6Eighty of the participants in the personality data set were also analyzed in the intelligence data set, albeit on different behavioral and neurological measurements.