Task-specific modulation of PFC activity for matching-rule governed decision-making

Abstract

Storing information from incoming stimuli in working memory (WM) is essential for decision-making. The prefrontal cortex (PFC) plays a key role to support this process. Previous studies have characterized different neuronal populations in the PFC for working memory judgements based on whether an originally presented stimulus matches a subsequently presented one (matching-rule decision-making). However, much remains to be understood about this mechanism at the population level of PFC neurons. Here, we hypothesized differences in processing of feature vs. spatial WM within the PFC during a matching-rule decision-making task. To test this hypothesis, the modulation of neural activity within the PFC during two types of decision-making tasks (spatial WM and feature WM) in comparison to a passive fixation task was determined. We discovered that neural population-level activity within the PFC is different for the match vs. non-match condition exclusively in the case of the feature-specific decision-making task. For this task, the non-match condition exhibited a greater firing rate and lower trial-to-trial variability in spike count compared to the feature-match condition. Furthermore, the feature-match condition exhibited lower variability compared to the spatial-match condition. This was accompanied by a faster behavioral response time for the feature-match compared to the spatial-match WM task. We attribute this lower across-trial spiking variability and behavioral response time to a higher task-relevant attentional level in the feature WM compared to the spatial WM task. The findings support our hypothesis for task-specific differences in the processing of feature vs. spatial WM within the PFC. This also confirms the general conclusion that PFC neurons play an important role during the process of matching-rule governed decision-making.

Keywords: Decision-making; Feature-based attention; Prefrontal cortex (PFC); Task-specific; Trial-to-trial variability; Working memory (WM).

Fig. 1

Behavioral passive and WM tasks. a In the passive task, two sequential stimuli were followed by two delay periods. Stimuli could be match or non-match relative to the spatial location or shape feature aspect. Regardless of stimuli properties, animals needed to maintain their gaze on the fixation point to receive a drop of juice reward. b Spatial set consisted of nine locations (left panel) and feature set included of eight different shapes (right panel). c, d In the WM tasks, monkeys were trained for making a decision regarding the match/non-match status between sample and test stimuli. In the spatial WM task, the shape of sample and test stimuli were identical and monkeys should identify the match or non-match status of their location. In contrast, in the feature WM task, the location of stimuli remained the same and monkeys should identify the matching status based on their shapes. The WM tasks included two additional phases in comparison to the passive task; choice target and saccade response epochs. e The prefrontal cortex (red region) as the target recording area

Fig. 2

Matching-rule decision-making modulates the PFC neural activity in the feature WM task. The top panels show the average spiking activity (based on SDF) during − 0.5 to 4 s relative to the sample stimulus onset (F fixation, S1 sample stimulus, D1 1st delay, S2 test stimulus, D2 2nd delay). They exhibit the neural responses of the match (red line) and non-match (blue line) conditions within spatial passive (a), spatial WM (b), feature passive (c), and feature WM (d) tasks. Black dots on top of the SDFs indicate the time points with a significant difference between the population response of match and non-match conditions. The differences between average responses of the match and non-match conditions were displayed in the lower panels. Gray dashed lines show the period of sample and test stimuli presentation. Red dashed lines indicate the standard error from the mean (SEM) of the chance distribution

Fig. 3

Matching-rule modulation arises during the late test period and its following delay interval. a Time-course of discrimination between firing rate response of match and non-match conditions based on F-statistic values obtained by ANOVA across the population of neurons (left: passive, middle: spatial WM, and right: feature WM). b Histogram of the F-statistic values of each epoch in the passive and WM tasks, separately (F fixation, S1 sample stimulus, D2 1st delay, S2 test stimulus, D2 2nd delay). The integers represent the percentage of points with significant selective responses for matching conditions. Red solid lines show the mean F-statistic in each epoch and indicate the distance of the mean selectivity value from zero (red dashed lines). c Statistical comparisons between the histograms of the three tasks during test stimulus interval (left panel) and its following delay period (right panel). Notation ‘ns’ denotes non-significant effect and *Shows significant effect

Fig. 4

Matching-rule decision-making modulates the PFC spiking variability in the feature WM task. a–c The Fano factor during the time-course of analysis from − 0.5 to 4 s with respect to the sample stimulus onset in the passive (a), spatial WM (b), and feature WM (c) tasks. Each time point shows the middle of the corresponding window that the Fano factor was calculated. Red and blue traces show the averaged Fano factor of the match and non-match conditions, respectively. The black horizontal line indicates the time interval with a significant difference between matching conditions. Dashed lines display stimulus presentation epochs. d The variation of Fano factor across different trials of spatial WM (top panel) and feature WM (bottom panel) tasks. e Scatter plot of Fano factor values of the sample and test stimuli for spatial (green) and feature (purple) WM tasks. f The distribution of Fano factor changes between sample and test stimuli (sample test) for the spatial (green) and feature (purple) WM tasks. *Significant effect

Fig. 5

The different discriminability of WM tasks is due to their differential attentional level. a Mean Fano factor throughout the passive (black line), spatial WM (green line), and feature WM (purple line) tasks. Shaded lines display the standard error of the mean (SEM) across all neuron conditions. b Discrimination strength between the Fano factors of the three tasks. The gray solid line denotes the chance level of the null distribution. Black horizontal lines show the time intervals with a significant discrimination effect. c Distribution of reaction times (RT) in two WM tasks. The inset plot shows the significant difference in behavioral reaction times between the two WM tasks