The Dynamics of Representational Change Underlying Recall

Overview

The goal of this clinical trial is to examine how repeated memory retrieval changes visual representations in the brain in healthy adult participants. Specifically, the study aims to determine whether repeated recall strengthens or transforms neural representations in visual cortex compared to a time-matched control condition. The main questions it aims to answer are: 1) Does repeated retrieval produce cumulative changes in visual cortical representations compared to a control condition matched for elapsed time? 2) Do these representational changes differ for simple spatial stimuli and complex natural images? Researchers will compare neural activity patterns during repeated retrieval to those during a control retrieval condition to determine whether repeated recall leads to systematic changes in representational structure beyond those attributable to the passage of time. Participants will: 1) complete two fMRI sessions. 2) Undergo a localizer session including anatomical imaging, population receptive field (pRF) mapping, and a visual category localizer task. 3) Study cue-stimulus pairs consisting of either simple spatial patterns or natural images. 4) Recall previously studied stimuli multiple times during repeated and control retrieval conditions while undergoing fMRI scanning. Brain activity patterns during study and recall will be compared to assess how repeated retrieval influences the structure and tuning of visual representations.

Two fMRI sessions will be conducted per subject. In the localizer session, the investigators will collect anatomical images and fMRI data necessary to define visual areas in each individual's native brain space. In addition to two high-resolution anatomical images, two short functional tasks will be acquired from every participant: 1) population receptive field (pRF) mapping; 2) visual category localizer. The pRF mapping task will be several runs of a bar aperture drifting across the central 8 degrees of the visual field. This will allow us to estimate the retinotopic sensitivity of individual voxels in visual cortex and draw boundaries between different retinotopic field maps. The visual category localizer will be a block design with different image categories. This will allow us to define higher-level category-selective brain areas. In the experimental session, the investigators will collect fMRI data while subjects are undergoing perception and recall of the same stimuli. In order to assess whether our findings generalize across simple and complex visual stimuli, two classes of stimuli will be used: simple patterns in discrete spatial locations, following previous work, and natural images. In study blocks, subjects will study cue-stimulus pairs. Colored fixation dots will be used as cues. Stimuli will be from one of the aforementioned classes. In recall blocks, participants will be asked to recall the stimulus given the cue. Recall will happen under one of two conditions: repeated and control. The repeated condition allows us to examine how memory representations evolve with repeated retrieval, while the control condition will provide a matched baseline to assess changes over the same true time interval. For pairs in the repeated condition, participants will recall each stimulus 5 times, interleaved with recalls from other pairs. For pairs in the control condition, participants will recall each stimulus 2 times. Critically, recalls from this condition will be spaced such that they are matched to recalls 1 and 5 from the repeated condition, providing a control for elapsed time. Study and recall blocks will be performed in interleaved order. Analyses will compare BOLD activity patterns from repeated retrieval, control retrieval, and study conditions. Data from study trials will be averaged to form a single pattern for each stimulus. Retrieval data will be analyzed separately for each stimulus and retrieval attempt in both conditions. For both spatial stimuli and natural images, the investigators will focus on comparing representational changes between repeated and control conditions. Although the specific models used to characterize brain representations differ by stimulus type, the analyses will otherwise follow parallel procedures. For spatial stimuli, the investigators will use a previously published approach that combines pRF estimates with task-based BOLD data to quantify spatial tuning in visual areas. This method assesses the amplitude of the task-evoked BOLD response in each voxel as a function of the distance between the voxel's pRF and the spatial location of the stimulus. The investigators will compare spatial tuning across repeated and control retrieval conditions, as well as with the study condition, to determine whether changes in spatial tuning accumulate over repeated retrieval. For natural images, the investigators will compare the representational structure of BOLD activity patterns to that of a convolutional neural network (CNN) model. Using representational similarity analysis (RSA), the investigators will assess the similarity between CNN-derived embeddings and BOLD activity patterns during both study and recall. Higher correlations with early CNN layers suggest that low-level visual features are represented, while higher correlations with deeper layers indicate the representation of higher-level visual features. The investigators will use this approach to compare correlations from the repeated and control conditions, testing whether representational changes strengthen with repeated retrieval.