Multidimensional Encoding of the Prefrontal Cortex During Decision Making

Post by Elisa Guma

What's the science?

The prefrontal cortex plays a critical role in context-dependent decision making. It receives sensory, cognitive, and motor signals, which must all be integrated to make decisions. Understanding the prefrontal cortex’s role in this process is confounded by the fact that neurons within this structure respond to multiple types of input signals in a heterogeneous manner, making it difficult to understand the neural representation of different sensory and cognitive variables. This week in Nature Neuroscience, Aoi and colleagues applied novel modeling techniques to dissect the specific context and temporal evolution of neural activity in the prefrontal cortex during decision making.

How did they do it?

The authors acquired electrophysiological recordings of neural activity from the frontal eye-fields (a region of the prefrontal cortex) of two adult male monkeys while they performed a context-dependent, forced-choice decision making task. In each trial of the task, monkeys were cued to either make a choice based on the colour or the motion of moving dots by focusing their gaze on one of the two targets; thus, if the colour was cued, they were required to ignore motion, and vice versa.

The authors analyzed the neural activity recorded from the prefrontal cortex during stimulus presentation using a model-based targeted dimensionality reduction (mTDR) technique. This allowed them to identify the dimensions of neural activity that encode information about different task variables over time (i.e. response to sensory or motion cues). They used colour strength, motion strength, context and choice, and absolute values of colour and motion strength as their dimensions. They then applied a new data dimensionality reduction technique to the outputs of the mTDR — sequential principal component analysis (seqPCA) — to decompose the data into a set of axes ordered by the time at which each variable becomes available. This allowed them to determine if and when a neuronal population became active during the task period. Finally, given that neurons exhibit heterogeneous activity, and response to both motion and color, the authors then tested the “rotational dynamics” of the neural activity, i.e. does the neural response shift from one stimulus to another over time.

What did they find?

The authors found a heterogeneous tuning to the different variables (colour, motion) when examining the pre-stimulus time histograms, suggesting that the neural population recorded encodes several task variables, not just a single one. The authors found that the mTDR technique was able to accurately model the heterogeneous neural response, as most neurons had their variance differently distributed across components (e.g. color strength, motion strength, context, etc).

Using seqPCA to decompose neural activity into different temporal axes, the authors found that most of the variance in neural trajectories occurred in the time immediately after stimulus onset (early axis). They also identified a middle and late axis within the stimulus viewing window. Interestingly, neurons did “rotate” through this multidimensional space, (from early to middle or late axes) all during the stimulus viewing session; this process may reflect an internal, cognitive cue and may also reflect the moment in which the monkey has to make a judgment (before the end of the stimulus). Further, they found that the differences in neural activity did not arise from a distinct population of neurons, but from a single population with broad tuning characteristics. A rotation was indicative of a commitment to a decision, and not an accumulation of further sensory information, suggesting that the PFC maintains information about a choice as well as relevant and irrelevant sensory information over the course of a single trial.  

What's the impact?

This study applied novel modeling techniques to more accurately decode prefrontal cortex neural activity in response to a decision making task. This work highlights that the prefrontal cortex has a multidimensional code for types of stimuli involved in decision making, that these different codes may change as a result of the decision-making process, and that these codes are not neuron-specific.

Aoi et al. Prefrontal cortex exhibits multidimensional dynamic encoding during decision-making. Nature Neuroscience (2020). Access to the original scientific publication here.