Humans show an extraordinary ability to tune their behavior to the environment, much akin to how chameleons change color to match their surroundings. Consider, for example, how quickly a tennis player can adjust from clay to grass courts in just a matter of minutes. The basis for such adaptability is largely unknown. At some level, the answer must lie in how the brain dynamically changes its patterns of activity to accomplish specific behavioral adjustments. Such changes are often equated with plasticity, i.e., the capacity to alter synaptic connections between neurons. Whether other mechanisms, perhaps less expensive and more reversible, can support adaptive behavior remains an intriguing possibility.
In this talk, I will take a dynamical systems perspective of the brain, viewing the collective activity of neurons as a state evolving in a high-dimensional space, to suggest a new way of thinking about sensorimotor adaptation. In this framework, flexible control of neural dynamics can be accomplished in two complementary ways associated with distinct timescales and neural signatures. On the one hand, changing synaptic couplings within the system modifies its latent dynamics, effectively creating new activity patterns in the state space. On the other hand, adjusting external inputs enables the system to quickly survey unexplored regions of the state space. Because synaptic modifications occur on relatively long timescales, I hypothesize that input-control strategies may be crucial in achieving rapid adaptation.
I will leverage recent findings from our lab to directly test this hypothesis in the specific setting of a time reproduction task. I will introduce two variants of the task which can elicit sensorimotor adaptation at different timescales. Recordings of population activity in the macaque frontal cortex during fast adaptation will be examined through the lens of input versus latent dynamics, and I will provide evidence for an input-control strategy. I will close by proposing that certain forms of sensorimotor adaptation may not require changes of synaptic couplings within cortical areas, but may instead rely on inputs from elsewhere in the brain to uncover suitable regions of cortical latent dynamics.