A Neural Population Mechanism for Rapid Learning

Abstract

SUMMARY Long-term learning of language, mathematics, and motor skills likely requires cortical plasticity, but behavior often requires much faster changes, sometimes even after single errors. Here, we propose one neural mechanism to rapidly develop new motor output without altering the functional connectivity within or between cortical areas. We tested cortico-cortical models relating the activity of hundreds of neurons in the premotor (PMd) and primary motor (M1) cortices throughout adaptation to reaching movement perturbations. We found a signature of learning in the “output-null” subspace of PMd with respect to M1 reflecting the ability of premotor cortex to alter preparatory activity without directly influencing M1. The output-null subspace planning activity evolved with adaptation, yet the “output-potent” mapping that captures information sent to M1 was preserved. Our results illustrate a population-level cortical mechanism to 18 progressively adjust the output from one brain area to its downstream structures that could be exploited for rapid behavioral adaptation.

Independent representations of ipsilateral and contralateral limbs in primary motor cortex

“…This seems unlikely given that we observed a substantial proportion of neurons that were active during both contexts. Nonetheless, we addressed this issue following a similar procedure to Perich et al (2018). We summed the absolute value of the weights from the top-ten ipsilateral (wipsi) and contralateral (wcontra) principal components.…”

“…It is interesting to note that the subspaces related to the contralateral limb may not be fixed. M1 activity remains in the same subspace during motor learning (Golub et al, 2018; Perich et al, 2018; Vyas et al, 2018), and during reaching tasks with different load conditions (Gribble and Scott, 2002; Gallego et al, 2018) and initiation cues (Lara et al, 2018). However, Miri et al (2017) demonstrated that activity during locomotion and reaching occupied orthogonal subspaces that may allow motor cortex to engage separate spinal circuits for each behavior.…”

Dissociable mechanistic contributions of limb and task related errors during human sensorimotor learning

“…Indeed, changes in these regions following learning (Mandelblat-Cerf et al, 2011;Paz et al, 2003;Perich et al, 2018) may partially reflect changes in motor plans driven by such processes. Future work could probe the interactions between these two systems, which, neuroanatomically could be sustained by reciprocal connections between the basal ganglia and the cerebellum (Bostan and Strick, 2018).…”

Engrams of Fast Learning

“…Importantly, cortical representations acquired slowly over time are also essential in guiding future fast learning, which usually relies on prior knowledge and existing schemata such as when learning associations between known elements or novel words (Tse et al, 2007(Tse et al, , 2011Hebscher et al, 2019). Thus, one should keep in mind that fast and slow cortical dynamics are closely intertwined as exemplified in tasks necessitating high levels of cognitive flexibility (Pasupathy and Miller, 2005;Tse et al, 2007;Durstewitz et al, 2010;Siniscalchi et al, 2016;Perich et al, 2018;Remington et al, 2018;Rikhye et al, 2018).…”