Visuomotor Association Orthogonalizes Visual Cortical Population Codes

Stimuli trigger a pattern of activity across neurons in cortex, whose firing rates define a stimulus's representation in a high-dimensional vector space. Learning a visuomotor task can affect the responses of visual cortical neurons, but how and why training modifies population-level representations is unclear.

One hypothesis is that representational plasticity in visual cortex facilitates visuomotor associations by downstream motor systems. Learning systems exhibit "inductive biases", meaning they form some stimulus-motor associations more easily than others. An animal's inductive biases presumably reflect its neuronal representations; its ability to form distinct motor associations for different stimuli depends on the representational similarity of the stimuli. Thus, the plasticity of sensory cortical representations may change inductive bias: for an animal to make different associations to two stimuli, the cortical representations of the stimuli must differentiate, such as if the evoked firing vectors were orthogonalized. A second hypothesis is that task training increases the fidelity of stimulus coding in sensory cortex, which improves decoding accuracy by downstream regions. However, this hypothesis presupposes that the population code in naive cortex suffers from low fidelity, which recent recordings of large cortical populations have questioned.

We used two-photon calcium imaging to study how the tuning of V1 populations changes after mice learn to associate opposing actions with differently oriented gratings. Training did not improve the fidelity of stimulus coding, as it was already perfect in naive animals thanks to a subpopulation of highly reliable neurons. Instead, training caused the population's responses to motor-associated stimuli to become more orthogonal. The basis of this training-evoked orthogonalization was the sparsening of stimulus representations, an effect which could be summarized by a simple nonlinear transformation of naive neuronal firing rates and whose convexity was largest for motor-associated stimuli.