How do we use prior experiences to shape future actions?

Organisms use prior experiences to shape their current motor repertoire. Disruptions in these processes have broad impacts on daily function and can be found across neurodevelopmental and neuropsychiatric disorders. In our lab, we investigate this in the context of value-based choice and early instrumental learning.

Tracking and Acting on Value Information

History of choices and outcomes can be used to produce estimates for the values of future choices. This requires distributed neural processing that represents and tracks fluctuating values as well as monitoring selected actions to properly attribute credit. This integrated information is used to select actions and modulate how they are performed. Our lab studies prefrontal cortical circuits, basal ganglia nuclei, as well as relevant neuropeptides and neuromodulators mediating value-guided action.

(A) Joystick-based task with changing outcome probabilities or reward volumes. (B) Joystick trajectories for a single session. (C) Session-averaged choice latency and peak displacement vary with the reward volume. (D) Example of within-trial joystick position variability that fluctuates with underlying value - grey trace is a hard value comparison (push and pull are similar values) and black is an easy comparison (push is clearly better). 

Relevant Papers: Choi et al., Nature Communications 2023;  Linares-Garcia and Iliakis et al., eNeuro 2025;  Alabi et al., eLife 2020;  Alabi et al., Frontiers in Neuroscience 2019

Early Instrumental Learning

Learning context-appropriate actions requires trial-and-error interactions with the environment. In stringing together sequences of actions, animals must navigate the balance between continued motor exploration versus exploitation of known action-outcome patterns. Our lab discovered that the somatostatin-positive low-threshold spiking (LTS) class of striatal interneuron acts as a break on early instrumental learning, delaying acquisition until rewards are being consistently obtained.

(A) Instrumental learning task. (B, left) Plot of early learning separated first (red) and last (blue) ten rewards. (B, right) Population photometry signal in striatal LTS interneurons during reward collection showing that the signal decreases across learning. (C) Optogenetic inhibition of LTSIs during rewarded outcomes increases the rate of instrumental learning.

Relevant Papers: Holly et al., Neuron 2019; Holly et al. Current Biology 2021