Initial conditions combine with sensory evidence to induce decision-related dynamics in premotor cortex

We used a dynamical systems perspective to understand decision-related neural activity, a fundamentally unresolved problem. This perspective posits that time-varying neural activity is described by a state equation with an initial condition and evolves in time by combining at each time step, recurrent activity and inputs. We hypothesized various dynamical mechanisms of decisions, simulated them in models to derive predictions, and evaluated these predictions by examining firing rates of neurons in the dorsal premotor cortex (PMd) of monkeys performing a perceptual decision-making task. Prestimulus neural activity (i.e., the initial condition) predicted poststimulus neural trajectories, covaried with RT and the outcome of the previous trial, but not with choice. Poststimulus dynamics depended on both the sensory evidence and initial condition, with easier stimuli and “fast” initial conditions leading to the fastest choice-related dynamics. Together, these results suggest that initial conditions combine with sensory evidence to induce decision-related dynamics in PMd.

Methods
Subjects

Experiments were performed using two adult male macaque monkeys (Macaca Mulatta; monkey T, 7 years, 14 kg & monkey O, 11 years, 15.5 kg) trained to touch visual targets for a juice reward. Monkeys were housed in a social vivarium with a normal day/night cycle. Protocols for the experiment were approved by the Stanford University Institutional Animal Care and Use Committee.  Animals were initially trained to come out of their housing and to sit comfortably in a chair. After initial training (as described in Chandrasekaran et al, 2017), monkeys underwent sterile surgery where cylindrical head restraint holders (Crist Instrument Co., Inc., Hagerstown, MD, United States) and standard circular recording cylinders (19 mm diameter, Crist Instrument Co., Inc.) were implanted. Cylinders were placed surface normal to the cortex and were centered over caudal dorsal premotor cortex (PMdc; +16, 15 stereotaxic coordinates). The skull within the cylinder was covered with a thin layer of dental acrylic.

Unit Selection and Classification

The electrophysiological recordings consist of  996 units (546 units in T and 450 units in O, including both single neurons and multi-units) recorded from PMd of the two monkeys as they performed the task over 141 sessions. Chosen units were included as they were well isolated from other units/separated from noise and modulated activity in at least one task epoch.

U-probes were useful for recording from isolated single neurons as U-probes are low impedance (~100 kohm) with a small contact area. A conservative threshold was used to maximize the number of well defined waveforms and to minimize contamination from spurious non-neural events. Single neurons were delineated online by the ‘hoops’ tool of the Cerebus system software client (Blackrock Microsystems, Salt Lake City, UT, United States) after the electrodes had been in place for 30 - 45 minutes. When a spike was detected via thresholding, a 1.6 ms snippet was stored and used for subsequent evaluation of the clusters as well as modifications needed for spike sorting. 

Some electrodes in U-probe recordings captured mixtures of 2 or more neurons, well separated from each other and noise. In the majority of cases the waveforms were separable and labeled as single units. These separations were verified by viewing the waveforms in principal component (PC) space using custom code in MATLAB (The MathWorks, Inc., Natick, MA, United States). MatClust the MATLAB based clustering toolbox or Plexon Offline Sorter (Plexon, Inc.) were used to adjust the clusters that were isolated online.

Recording activity labeled as ‘multi-units’ were mixtures of 2 or more neurons not separable using a PCs method or consisted of recordings with waveforms only weakly separable from noise.

The number of interspike interval (ISI) violations after clustering and sorting was used to mitigate subjectivity in the classification of units. A unit was labeled as a single neuron if the percentage of ISI violations (refractory period of <= 1.5 ms) was <=  1.5%, otherwise it was labeled as a multi-unit. 801/996 PMd units were labeled as single neurons (T: 417, O: 384, median ISI violation = 0.28%, mean ISI violation = 0.43%, ~0.13 additional spikes/trial). Therefore 195/996 units were labeled as multi-unit (T: 129, O: 66, mean ISI violation = 3.36%, ~1.4 additional spikes/trial). 

Units from both monkeys were pooled together as the electrophysiological characteristics were similar. Change-of-mind trials (~2-3%) were excluded from averaging as the change in reach direction mid-movement execution made the assignment of choice ambiguous. Incorrect and correct trials arranged by choice were averaged together.

For futher details on methods please refer to preprint on bioRxiv: 

https://doi.org/10.1101/2022.06.30.497070