Brain Scientists Find a Rosetta Stone for How We Think

Cognition is implemented by distributed neural populations whose collective activity evolves through structured, low-dimensional state spaces. This review surveys tools for studying latent population dynamics in biological neural networks—from single-unit coding to population vectors, cosine similarity, PCA and related methods, multivariable linear regression axes, demixed PCA, and single-trial latent-variable models—and distinguishes operationally defined population subspaces from neural manifolds with reproducible geometry or topology.

It discusses sensory coding, working memory, decision-making, temporal processing, and sensorimotor transformations, emphasizing two organizing principles: cognitive variables can be expressed in separable or near-orthogonal population dimensions, reducing interference among sensory, mnemonic, decision-related, and motor signals; and such geometric separation does not require anatomically or functionally discrete cell classes, because separable population readouts can coexist with a continuum of mixed-selective single-neuron responses within a shared biological substrate. The review also considers recurrent neural networks as testbeds for mechanistic hypotheses about how circuits produce low-dimensional dynamics, and intracortical brain-computer interfaces as a translational domain in which population geometry becomes an operational control signal. Overall, it argues that population dynamics now provide a common language for linking neurophysiology, computational modelling, and neurotechnology.