Abstract
Characterizing inter-individual differences is critical to realize neuroimaging full potential, but can hardly be achieved without accurately assessing the statistical dependencies between inter-individual differences in behavior and inter-individual differences in neural activity. In this manuscript, we consider two hypotheses: 1) BOLD signal scales linearly with behavioral variables across individuals and 2) BOLD signal encodes behavioral variables on a similar scale across individuals. We formally show that these two hypotheses induce opposite brain-behavior correlational results in group-level analyses, illustrating the importance of explicitly testing interindividual brain-behavior scaling before engaging in the study of functional inter-individual differences. To further evidence the relevance of this framework, we illustrate its practical consequences for model-based fMRI using computational simulations, and demonstrate its empirical robustness in four fMRI studies investigating values coding in the prefrontal cortex. This may constitute an important step forward in our conceptualization, analysis and interpretation of inter-individual differences in cognitive neurosciences.
