Contingency representations in prefrontal cortex unify goal-directed planning and working memory
Poster No:
1118
Submission Type:
Abstract Submission
Authors:
Jacob Miller1,2, Daniel Ehrlich3, Yvette Afriyie-Agyemang4, Nicole Santamauro1, Zailyn Tamayo1, Alan Anticevic1, John D. Murray5,1
Institutions:
1Yale University School of Medicine, New Haven, CT, 2Wu Tsai Institute, Yale University, New Haven, CT, 3University of California, Berkeley, Berkeley, CA, 4University of Pittsburgh, Pittsburgh, PA, 5Dartmouth College, Hanover, NH
First Author:
Jacob Miller
Yale University School of Medicine|Wu Tsai Institute, Yale University
New Haven, CT|New Haven, CT
Yale University School of Medicine|Wu Tsai Institute, Yale University
New Haven, CT|New Haven, CT
Co-Author(s):
Introduction:
Working memory (WM) is critical in guiding our adaptive behavior based on immediate and future demands (van Ede & Nobre, 2023). The prefrontal cortex (PFC) and a connected network of areas are consistently active during WM (Sreenivasan & D'Esposito, 2019). But, it remains difficult to parse out the differential contributions of brain areas to WM function, especially when the maintenance and usage of WM content is often intertwined. How do representations for stimuli, task rules, and future behavior all contribute to WM, and what is their brain circuit organization? A recently developed computational framework and task paradigm helps unify the representational geometry for goal-directed planning and WM (Ehrlich & Murray, 2022). In this WM task, patterns of human behavior and activity in neural networks suggest the emergence of combined representations of stimuli and task rules into response contingencies. Here, we tested for differential neural substrates of stimulus, rule, and response information, and newly predicted contingency representations to clarify the role of PFC during WM.
Methods:
During functional magnetic resonance imaging (fMRI), human participants completed 4 runs (32 trials each) of a conditional delayed logic WM task (Fig. 1a). On each trial, participants were shown a rule cue (colored box), then a gabor filter stimulus (vertical/horizontal), followed by a jittered delay period, and then a second gabor (vertical/horizontal), after which participants responded with a left or right hand button. The correct response on each trial was determined based on the rule and the first and second stimulus (Fig. 1b-c).
fMRI data was acquired via 8x multi-band sequence with 2mm isotropic voxels, along with field maps and T1w/T2w anatomical scans. Functional data were minimally pre-processed in native anatomical space for each participant using fMRIprep (Esteban et al., 2019). We then used a general linear modeling framework in Nilearn (Abraham et al., 2014) with separate regressors for each combination of cue 1 stimulus and rule, modeling the initial cue period (1.6 s) and delay (3.2-8.0 s). Separate F-test contrasts were constructed to test for voxels that responded highly to different levels of stimulus, rule, response, and contingency information (Fig. 1b). Contrast maps were passed through probabilistic Threshold-Free Cluster Enhancement (pTFCE) to incorporate cluster information into voxel inference (Spisák et al., 2019) and corrected at FWE level or an arbitrary threshold for visualization.
fMRI data was acquired via 8x multi-band sequence with 2mm isotropic voxels, along with field maps and T1w/T2w anatomical scans. Functional data were minimally pre-processed in native anatomical space for each participant using fMRIprep (Esteban et al., 2019). We then used a general linear modeling framework in Nilearn (Abraham et al., 2014) with separate regressors for each combination of cue 1 stimulus and rule, modeling the initial cue period (1.6 s) and delay (3.2-8.0 s). Separate F-test contrasts were constructed to test for voxels that responded highly to different levels of stimulus, rule, response, and contingency information (Fig. 1b). Contrast maps were passed through probabilistic Threshold-Free Cluster Enhancement (pTFCE) to incorporate cluster information into voxel inference (Spisák et al., 2019) and corrected at FWE level or an arbitrary threshold for visualization.
Results:
During the cue period, stimulus (orientation: horizontal vs. vertical) and rule (GRTR vs. XNOR vs. MEMO vs. NAND) information was present across voxels in early and high-level visual areas, extending into the lateral occipital complex and intraparietal sulcus. During the WM delay, a distributed set of voxels in frontal, parietal, and temporal cortex responded highly to different levels of contingency across trials (LL vs. LR vs. RL vs. RR), but most strongly in lateral frontal cortex (pFWE < 0.05, Fig. 2a). Maps of contingency representations often overlapped with rule representations, however, contingency sensitive voxels showed specific, non-overlapping territory in more anterior PFC areas (Fig. 2b), extending ventrally beyond the inferior frontal sulcus and dorsally along superior frontal sulcus/gyrus (Miller et al., 2021), or even wrapping around the frontal pole onto the medial PFC surface.
Conclusions:
Building on classic and modern studies of rule coding in PFC (Milner, 1963; Vallentin et al., 2012), we show that the most anterior PFC areas in the human brain display unique representations of task contingencies in WM by transforming stimulus and rule information to future responses. Tying in perspectives of hierarchical cognitive control in PFC (Badre & Nee, 2018), these results suggest anterior PFC subserves WM by integrating different task information to plan future behavior.
Higher Cognitive Functions:
Executive Function, Cognitive Control and Decision Making 2
Higher Cognitive Functions Other
Learning and Memory:
Working Memory 1
Keywords:
Cognition
Computational Neuroscience
Cortex
FUNCTIONAL MRI
Learning
Memory
Other - Prefrontal Cortex
1|2Indicates the priority used for review
Provide references using author date format
Badre, D., & Nee, D. E. (2018). Frontal Cortex and the Hierarchical Control of Behavior. Trends in Cognitive Sciences, 22(2), 170–188.
Ehrlich, D. B., & Murray, J. D. (2022). Geometry of neural computation unifies working memory and planning. Proceedings of the National Academy of Sciences of the United States of America, 119(37), e2115610119.
Esteban, O., Markiewicz, C. J., Blair, R. W., Moodie, C. A., Isik, A. I., Erramuzpe, A., Kent, J. D., Goncalves, M., DuPre, E., Snyder, M., Oya, H., Ghosh, S. S., Wright, J., Durnez, J., Poldrack, R. A., & Gorgolewski, K. J. (2019). fMRIPrep: a robust preprocessing pipeline for functional MRI. Nature Methods, 16(1), 111–116.
Miller, J. A., Voorhies, W. I., Lurie, D. J., D’Esposito, M., & Weiner, K. S. (2021). Overlooked Tertiary Sulci Serve as a Meso-Scale Link between Microstructural and Functional Properties of Human Lateral Prefrontal Cortex. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience, 41(10), 2229–2244.
Milner, B. (1963). Effects of Different Brain Lesions on Card Sorting: The Role of the Frontal Lobes. Archives of Neurology, 9(1), 90–100.
Spisák, T., Spisák, Z., Zunhammer, M., Bingel, U., Smith, S., Nichols, T., & Kincses, T. (2019). Probabilistic TFCE: A generalized combination of cluster size and voxel intensity to increase statistical power. NeuroImage, 185, 12–26.
Sreenivasan, K. K., & D’Esposito, M. (2019). The what, where and how of delay activity. Nature Reviews. Neuroscience, 2019/05/16. https://doi.org/10.1038/s41583-019-0176-7
Vallentin, D., Bongard, S., & Nieder, A. (2012). Numerical rule coding in the prefrontal, premotor, and posterior parietal cortices of macaques. The Journal of Neuroscience: The Official Journal of the Society for Neuroscience, 32(19), 6621–6630.
van Ede, F., & Nobre, A. C. (2023). Turning Attention Inside Out: How Working Memory Serves Behavior. Annual Review of Psychology, 74, 137–165.