Tuned responses to visual short-term memory load in a fronto-parietal topographic map hierarchy
Poster No:
935
Submission Type:
Abstract Submission
Authors:
Ben Harvey1, Martijn van Ackooij2, Joeri van Helden3, Evi Hendrikx2, Nathan van der Stoep2, Surya Gayet2, Jacob Paul4
Institutions:
1Utrecht University, Utrecht, Nederland, 2Utrecht University, Utrecht, Utrecht, 3University of Birmingham, Birmingham, West Midlands, 4University of Melbourne, Melbourne, Victoria
First Author:
Co-Author(s):
Introduction:
Visual short-term memory (VSTM) is essential for tasks in which visual information must be briefly remembered. Behavioral studies often focus on quantifying the amount or complexity of information that participants can remember (i.e., how much can be remembered). This reflects performance in executive tasks. Conversely, neuroimaging studies have investigated how the memory content is represented in the brain (i.e., what is remembered). This reflects properties of the stimulus. Here, we asked how the brain's responses change with executive task demands, specifically the quantity of visual information that is remembered (i.e., VSTM load). As more information must be remembered to perform tasks accurately, we must allocate more cognitive and neural resources. It may follow that neural response amplitudes in brain areas supporting these executive tasks would monotonically increase with VSTM load. On the other hand, sensory processing is distributed across neural populations using tuned neural responses, organized in hierarchical networks of topographic maps. Here we asked whether these same principles apply to executive processes like task demands.
Methods:
We designed a task that keeps visual stimulation as similar as possible while varying VSTM load, specifically the number of unique orientations that had to be remembered (Figure 1A-B). During ultra-high field (7T) fMRI, we gradually changed the VSTM load (Figure 1C, top). This elicited remarkably different profiles at nearby recording sites (voxels) (Figures 1C, 1D & 2A). We compared these responses to the predictions of (linear and logarithmic) monotonic and tuned neural response functions of VSTM load. We then analyzed how the parameters of the best fitting response functions differed within and between responsive areas.
·Task description, stimuli, and responses
Results:
The model which best described the observed responses consisted of a logarithmic Gaussian function with two free parameters (Figure 1D): VSTM load preference and tuning width (the range of VSTM loads producing a response). These models explained the different responses of different voxels (Figure 1C) using different parameters of VSTM load-tuned response functions (Figure 1D). Projecting the VSTM load preference of each voxel onto each participant's cortical surface revealed topographic maps organized by the VSTM load preferences of their constituent neural populations (Figure 2A). We defined maps of VSTM load preferences by identifying regions where preferences consistently and progressively increased from low to high VSTM loads in all participants. We identified 11 maps in each hemisphere, located in regions that are implicated in visual processing (lateral and dorsal occipital areas, superior parietal lobule, postcentral and precentral sulci), control of spatial attention (superior parietal lobule, intraparietal sulcus, postcentral and precentral sulci) and executive control (precentral sulcus, middle frontal gyrus). VSTM load preferences were significantly correlated with cortical distance across the map in 109 out of 132 maps (Figure 2B). We also found systematic differences in responses to VSTM load between maps. The mean VSTM load preferences of the voxels in each map increased from posterior to anterior maps. Furthermore, responses in superior parietal cortices were clearer and subtended a larger cortical area then elsewhere. VSTM load preferences were not correlated with visual field position or numerosity preferences. The responses were absent if participants viewed the same stimuli without performing the task.
·VSTM load preferences progressed across widespread topographic maps
Conclusions:
Our results show that tuned, topographically organized, and hierarchically processed neural responses distribute complex tasks across neural populations throughout the brain. Therefore, these neural encoding schemes, common in sensory processing, generalize to complex higher-order executive functions. As a result of this generalization, established methods to investigate sensory neural tuning can reveal new mechanisms of underlying executive processes.
Higher Cognitive Functions:
Executive Function, Cognitive Control and Decision Making 1
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 2
Keywords:
Cognition
Design and Analysis
FUNCTIONAL MRI
Memory
Vision
1|2Indicates the priority used for review
Provide references using author date format
Dijkstra, N., Bosch, S.E., and van Gerven, M.A.J. (2019). Shared Neural Mechanisms of Visual Perception and Imagery. Trends Cogn Sci 23, 423–434.
Tsouli, A., Harvey, B.M., Hofstetter, S., Cai, Y., van der Smagt, M.J., te Pas, S.F., and Dumoulin, S.O. (2022). The role of neural tuning in quantity perception. Trends in Cognitive Sciences. 10, 1016
Harvey, B.M., Klein, B.P., Petridou, N., and Dumoulin, S.O. (2013). Topographic Representation of Numerosity in the Human Parietal Cortex. Science. 341, 1123–1126.