Toward a “Hyperaligned” Brain: Whole-Brain, Correlation-Based Analysis in Functional Neural Alignment

Abstract

Functional magnetic resonance imaging (fMRI) analysis techniques remain limited by neuroanatomical differences among individuals, precluding population inference and accurate pattern analysis. “Hyperalignment” (Haxby et al., 2011) addresses this issue, using function, rather than anatomy, to build a common model of response-pattern vector space. In two, related experiments, we expanded the application of hyperalignment to a whole-brain, correlation-based analysis in the cognitive control domain. In Experiment One, cognitive control and visual tasks demonstrated, respectively, modestly improved and substantially decreased classifiability after hyperalignment. Experiment Two sought to replicate and explain these results, and to use novel neuroimaging techniques to resolve neural representations of cognitive control tasks. Although complications introduced by the imaging protocol confounded results, these questions remain relevant and will continue to be explored. To this end, experimental findings, technical difficulties, and next steps are discussed.