Abstract
Perception, working memory, and long-term memory each evoke neural responses in the visual cortex. While previous neuroimaging research on the role of the visual cortex in memory has largely emphasized similarities between perception and memory, we hypothesized that responses in the visual cortex would differ depending on the origins of the inputs. Using fMRI, we quantified spatial tuning in the visual cortex while participants (both sexes) viewed, maintained in working memory, or retrieved from long-term memory a peripheral target. In each condition, BOLD responses were spatially tuned and aligned with the target's polar angle in all measured visual field maps including V1. As expected given the increasing sizes of receptive fields, polar angle tuning during perception increased in width up the visual hierarchy from V1 to V2, V3, hV4, and beyond. In stark contrast, the tuned responses were broad across the visual hierarchy during long-term memory (replicating a prior result) and during working memory. This pattern is consistent with the idea that mnemonic responses in V1 stem from top-down sources, even when the stimulus was recently viewed and is held in working memory. Moreover, in long-term memory, trial-to-trial biases in these tuned responses (clockwise or counterclockwise of target) predicted matched biases in memory, suggesting that the reinstated cortical responses influence memory-guided behavior. We conclude that feedback widens spatial tuning in the visual cortex during memory, where earlier visual maps inherit broader tuning from later maps thereby impacting the precision of memory.
Significance Statement
We demonstrate that remembering a visual stimulus evokes responses in the visual cortex that differ in spatial extent compared with seeing the same stimulus. Perception evokes tuned responses in early visual areas that increase in size up the visual hierarchy. Prior work showed that feedback inputs associated with long-term memory originate from later visual areas with larger receptive fields resulting in uniformly wide spatial tuning even in the primary visual cortex. We replicate these results and show that the same pattern holds when maintaining in working memory a recently viewed stimulus. That trial-to-trial difficulty is reflected in the accuracy and precision of these representations suggests that the visual cortex is flexibly used for processing visuospatial information, regardless of where that information originates.
Introduction
While visual maps in the occipital cortex have been known to be essential for vision for over a century (Inouye, 1909), recent evidence supports the provocative idea that these visual maps also play important roles in both working and long-term memory. Surprisingly, the contents of visual working memory (Harrison and Tong, 2009; Serences et al., 2009; Curtis and Sprague, 2021) and long-term memory (Bosch et al., 2014; Naselaris et al., 2015; Vo et al., 2022) can be decoded from the patterns of voxel activity in the human primary visual cortex (V1). Such results support influential theories of how the encoding mechanisms used for perception might also be used to store working memory representations (D’Esposito and Postle, 2015; Serences, 2016) and recall visual properties from long-term memory (Tulving and Thomson, 1973; Schacter et al., 1998; Rugg et al., 2008). It is no longer a question as to “whether” the visual cortex participates in cognitive functions beyond perception but a question of “how.”
Patterns of evoked activity during perception can be used to predict the contents of memory (Albers et al., 2013; Bosch et al., 2014; Rademaker et al., 2019), supporting the idea that stimulus features are encoded in visual cortex activity in a similar manner in perception and memory. Moreover, recalling large objects evokes activity that encroaches into more peripheral portions of visual field maps (Kosslyn et al., 1995), as if recalled objects that are larger encompass more of the visual field just like it does for seen objects. Despite these parallels, we hypothesized that these responses in the striate and extrastriate cortex also differ because the origins of their inputs differ. During perception, visual information is transmitted through the eyes and the retinogeniculate pathway to a cluster of retinotopically organized maps, including V1, V2, and V3 (Van Essen and Maunsell, 1983). Neural activity in these early visual maps strongly influences one's perception (Tong et al., 1998). During working memory, information enters the visual cortex in the same feedforward way, but after stimulus offset, the maintenance of visual information depends on interactions between higher-order cortical regions like the prefrontal cortex and sensory areas like V1. Such interactions support the storage of working memory representations (Curtis and D’Esposito, 2003; D’Esposito and Postle, 2015; Curtis and Sprague, 2021). During long-term memory, information is thought to be reconstructed in the visual cortex through the retrieval of a memory stored in other brain structures such as the hippocampus (Schacter et al., 1998).
Most of the research described above focused on similarities between memory and perceptual representations. Some recent work, however, has observed systematic differences in visual cortex activity between perception and long-term memory (Breedlove et al., 2020; Favila et al., 2022) and between perception and working memory (Kwak and Curtis, 2022; Chunharas et al., 2023; Duan and Curtis, 2024). Several important questions remain. There have been no comparisons of visual cortex responses in perception, working memory, and long-term memory with the same study parameters, no direct estimates of tuning width in the visual cortex during working memory, and limited links between the precision of long-term memory reactivation and single-trial behavior.
Here, we use the receptive field properties of neural populations within voxels (Dumoulin and Wandell, 2008) in visual field maps to quantify and compare the spatial tuning of responses during perception and memory. As a preview, fMRI responses during perception matched the spatial position of seen targets and increased in tuning width up the visual hierarchy consistent with increases in receptive field sizes (Smith et al., 2001; Dumoulin and Wandell, 2008). During working and long-term memory, tuning widths were large in the early visual cortex as if they were inherited from feedback from higher-order areas with large receptive fields. This confirms studies of long-term memory (Breedlove et al., 2020; Favila et al., 2022) and shows now that the same pattern holds in working memory. Critically, errors in spatially tuned responses during visual memory aligned with errors in memory behavior, suggesting that memory behavior depends on these responses in the early visual cortex.
Materials and Methods
Subjects
Eight human subjects (five males, 25–32 years old) were recruited to participate in the experiment and were compensated for their time. Subjects were recruited from the New York University community and included author R.W. Other subjects were naive to the purpose of the experiments. All subjects gave written informed consent to procedures approved by the New York University Institutional Review Board prior to participation. All subjects had normal or corrected-to-normal visual acuity, normal color vision, and no MRI contraindications. No subjects were excluded from the main data analyses.
Stimuli
Target stimuli
For each of the three conditions (perception, working memory, long-term memory), 16 polar angles were selected from 16 evenly spaced bins along an isoeccentric ring at 7° eccentricity from a central fixation point (0–22.5°, 22.5–45°, … 337.5–360°). Within each bin, the precise target location was randomly assigned. The 48 target locations were uniquely generated for each participant. The target stimulus at each spatial location consisted of a drifting Gabor patch (SD = 0.33°; spatial frequency, 2 cyc/deg; truncated at 5 SD). Each Gabor patch drifted radially toward fixation at a rate of 3 Hz (1.5°/s), with an orientation tangential to the isoeccentric ring.
Object stimuli
A small image of an object (1° diameter) was shown at fixation at the beginning of each trial. For the long-term memory condition, each object stimulus is paired with one of the 16 target locations. These object–target pairings remained consistent throughout the prescan behavioral training (see below) and during the main experiment, and therefore a particular object served as a cue for the location to retrieve from long-term memory. For the perception and working memory conditions, the object–target pairings were unique for each trial, with each object only ever presented once for each subject, ensuring that subjects could not form associations between objects and target locations. Because the long-term memory condition required pairing target Gabors to fixation stimuli that were easy to learn and because the perception and working memory conditions required the presentation of a unique fixation stimulus on every trial, this required a large number of easily recognizable cues with a low probability of confusion. Therefore, the fixation stimuli were selected randomly without replacement from a bank of colored images consisting of everyday objects (BOSS dataset; Brodeur et al., 2010). For each subject, a total of 144 unique everyday objects were sampled without replacement from this dataset, of which 16 were assigned to the long-term memory condition and the other 128 were assigned to each trial in the perception and working memory conditions.
Experimental procedure
Main experiment
The main experiment cycled through three scans: one each for perception, working memory, and long-term memory, repeated twice per scan session (Fig. 1). Each subject participated in two scan sessions spaced no more than a week apart, for a total of 12 scans (4 per condition). Each scan had 16 trials corresponding to the 16 target locations, in random order. Across both sessions, this culminated in a total of 64 trials per condition, 192 trials total per subject. The subjects were unaware of the trial design.
Main fMRI task design. Subjects participated in two fMRI sessions with six scans each: perception, working memory, and long-term memory scans, repeated twice (top). Each 292 s scan included 16 trials corresponding to 16 polar angles, in random order, with the Gabor targets always at 7° eccentricity. Objects are shown centrally for 0.5 s but are only meaningful for the long-term memory trials, for which they serve as cues for the target. In perception trials, the Gabor target remains on screen for the 11.5 s delay. For working memory trials, the Gabor target disappears after 0.5 s. For long-term memory trials, the Gabor target is not shown at all. At the end of the delay, the fixation cross turns green, and participants make a saccade to the location of the Gabor target. The conditions are matched except for how the Gabor target location is accessed and how it is maintained throughout the delay period.
For perception trials, we instructed participants to fixate on a central cross. The cross was briefly replaced by an object for 0.5 s. After the object disappeared, the cross reappeared, and the target stimulus remained visible for 11.5 s, during which subjects maintained central fixation. After this delay period, the fixation cross changed from black to green, indicating to the subject to make a saccade to the target location. Participants were instructed to maintain their fixation at the expected location until the fixation cross changed color back to black, at which point they returned their gaze to the central fixation cross. This saccade response period lasted 1.5 s for each trial.
The working memory block was the same except the target stimulus disappeared when the object disappeared. The subject was instructed to “hold the target in mind” throughout the delay while centrally fixating. At the end of the delay, the subject was similarly cued to make a saccade response to the location of the target stimulus.
The long-term memory block was the same except that the target stimulus was not shown at all. The subject was instructed to retrieve from memory the target stimulus associated with the object, with these associations learned during a prescan behavioral training session (see below, Long-term memory training). Similar to the working memory and perception blocks, participants maintained central fixation during the delay period of 11.5 s and made a saccade to the target location when the cross turned green.
Long-term memory training
Subjects learned associations between 16 object stimuli and 16 target locations by completing study and retrieval blocks before each scan session (outside the scanner). Following each study block, subjects completed three retrieval blocks. In each trial of the study block, an object stimulus was briefly presented at fixation simultaneously with its corresponding Gabor target stimulus (Fig. 2A). Subjects were instructed to fixate a central cross and learn the association between each object stimulus and its corresponding target location. The study block was self-paced with a minimum 1 s intertrial interval (ITI). Each of the 16 object–target pairs was presented five times per block (80 trials), with at least two study blocks per behavioral training session.
Prescan long-term memory training. A, Study phase, Self-paced brief (0.5 s) viewings of the paired precue at fixation and the associated target Gabor in the periphery, followed by an intertrial interval (minimum of 1 s). Each participant had their own set of 16 unique pairs to memorize. B, Retrieval phase, The bulk of learning happened during this training phase. The retrieval phase was similarly self-paced, where following an intertrial interval (minimum of 1 s), a paired cue was briefly presented at fixation without the target in the periphery. After a brief delay period, the participant was cued to make a saccade response to the target's location. Feedback was then given in the form of (1) the fixation cross either remaining green if the saccade was closest to the target's location, or changing to red if the response landed closer to a different target's location, and (2) a brief presentation of the peripheral target Gabor, irrespective of accuracy. C, Example layout of all target stimuli shown in a single image: For each condition, target stimuli are presented at 7° eccentricity, sampled around the visual field within each of 16 nonoverlapping 22.5° bins. Every participant had their own unique set of target locations. This led to some targets spaced near each other (i.e., near the bin borders marked by the dashed lines), while others were spaced far apart. Note that only a single target was shown in each trial, as in B.
In the retrieval block, subjects were instructed to fixate on the central cross while the object stimulus was flashed briefly, followed by a short delay where the subject was instructed to retrieve from memory the associated target stimulus (Fig. 2B). This was followed by the fixation cross changing color to green, which indicated to the participant to make an eye movement to the target's expected location. Feedback followed each retrieval trial in the form of the fixation cross remaining green if correct or changing color to red if incorrect. A saccade was considered correct if it was closer to the correct target than to any of the other 15 possible targets; otherwise, it was considered incorrect. Simultaneously, the target stimulus was revealed at its true location, with the subject instructed to make a corrective saccade to the target stimulus. The retrieval block was also self-paced with a 1 s minimum ITI. Each of the 16 object/target pairs was presented four times per block (64 trials), with at least six retrieval blocks for the first training session and four for the second session. Subjects were asked after six blocks if they felt they learned all the associations or if they needed to do more practice. If they needed more practice, they did at least one more retrieval block.
Retinotopic mapping procedure
In addition to the main experiment, each participant completed 10–12 retinotopic mapping scans in a single separate scan session. The stimuli and procedures are the same as the ones used by Himmelberg et al. (2021), described here in brief. Each scan consisted of contrast patterns windowed by bar apertures (1.5° width) that swept across the visual field within a 12° radius circle. There were eight sweeps along different directions, four cardinal (left, right, up, down) and four diagonal. The cardinal sweeps traveled the entire extent of the circular aperture in 24 s. The diagonal sweeps stopped halfway (12 s) and were then replaced by blank periods of 12 s. Hence, each bar sweep took 24 s to complete. At eight sweeps for each functional run, each scan took 192 s in total to complete. The contrast patterns were pink noise (grayscale) backgrounds with randomly placed and sized items. The stimuli and background were updated at 3 Hz. Participants were instructed to report any observed change in fixation dot color with a button box press. Color changes occurred around once every 3 s. The contrast patterns for the mapping stimuli were first used by Benson et al. (2018).
MRI acquisition
Imaging was conducted at the New York University Center for Brain Imaging using a 3 T Siemens Prisma MRI system and a Siemens 64-channel head/neck coil. We acquired functional images with a T2*-weighted multiband echo planar imaging (EPI) sequence with whole-brain coverage (repetition time, 1 s; echo time, 37 ms; flip angle, 68°; 66 slices; 2 × 2 × 2 mm voxels; multiband acceleration factor, 6; phase-encoding, posterior–anterior). We collected spin echo images with anterior–posterior and posterior–anterior phase-encoding to estimate, and correct for, the susceptibility-induced distortion in the functional EPIs. We also acquired one to three whole-brain T1-weighted MPRAGE 3D anatomical volumes (0.8 × 0.8 × 0.8 mm voxels) for each of the eight subjects.
MRI processing
All original MRI data [Digital Imaging and Communications in Medicine (DICOM) files] were defaced to anonymize them using pydeface (https://github.com/poldracklab/pydeface). The DICOM data were then converted to neuroimaging informatics technology initiative format and organized into the Brain Imaging Data Structure format (K. J. Gorgolewski et al., 2016) using Heuristic DICOM Converter (Halchenko et al., 2018). The data were then preprocessed using fMRIPrep 20.2.7 (Esteban et al., 2018, 2019), which is based on Nipype 1.7.0 (K. Gorgolewski et al., 2011; K. J. Gorgolewski et al., 2018).
Anatomical data preprocessing
The following sections on anatomical and functional data preprocessing are provided by the fMRIPrep boilerplate text generated by the preprocessed scan output.
Each of the one to three T1w images was corrected for intensity nonuniformity with N4BiasFieldCorrection (Tustison et al., 2010), distributed with ANTs 2.3.3 (Avants et al., 2008). The T1w reference was then skull-stripped with a Nipype implementation of the antsBrainExtraction.sh workflow (from ANTs), using OASIS30ANTs as target template. Brain tissue segmentation of cerebrospinal fluid, white matter, and gray matter was performed on the brain-extracted T1w using FAST (FSL 5.0.9; Zhang et al., 2001). A T1w-reference map was computed after registration of the T1w images (after intensity nonuniformity correction) using mri_robust_template (FreeSurfer 6.0.1; Reuter et al., 2010). Brain surfaces were reconstructed using recon-all (FreeSurfer 6.0.1; Dale et al., 1999), and the brain mask estimated previously was refined with a custom variation of the method to reconcile ANTs-derived and FreeSurfer-derived segmentations of the cortical gray matter of Mindboggle (Klein, 2017).
Functional data preprocessing
For each of the 12 BOLD runs found per subject (across all tasks and sessions), the following preprocessing was performed. First, a reference volume and its skull-stripped version were generated by aligning and averaging a single-band reference. A B0 nonuniformity map (or fieldmap) was estimated based on two EPI references with opposing phase-encoding directions, with 3dQwarp (Cox and Hyde, 1997). Based on the estimated susceptibility distortion, a corrected EPI reference was calculated for a more accurate coregistration with the anatomical reference. The BOLD reference was then coregistered to the T1w reference using bbregister (FreeSurfer) which implements boundary-based registration (Greve and Fischl, 2009). Coregistration was configured with 6° of freedom. Head-motion parameters with respect to the BOLD reference (transformation matrices and six corresponding rotation and translation parameters) are estimated before any spatiotemporal filtering using MCFLIRT (FSL 5.0.9; Jenkinson et al., 2002). BOLD runs were slice-time corrected to 0.445 s (0.5 of slice acquisition range, 0–0.89 s) using 3dTshift from AFNI 20160207 (Cox and Hyde, 1997). First, a reference volume and its skull-stripped version were generated using a custom methodology of fMRIPrep. The BOLD time series were resampled onto the fsnative surface. The BOLD time series (including slice-timing correction) were resampled onto their original, native space by applying a single, composite transform to correct for head-motion and susceptibility distortions. These resampled BOLD time series will be referred to as preprocessed BOLD. All resamplings can be performed with a single interpolation step by composing all the pertinent transformations (i.e., head-motion transform matrices, susceptibility distortion correction, and coregistrations to anatomical and output spaces). Gridded (volumetric) resamplings were performed using antsApplyTransforms (ANTs), configured with Lanczos interpolation to minimize the smoothing effects of other kernels (Lanczos, 1964). Nongridded (surface) resamplings were performed using mri_vol2surf (FreeSurfer).
Many internal operations of fMRIPrep use Nilearn 0.6.2 (Abraham et al., 2014), mostly within the functional processing workflow. For more details of the pipeline, see the section corresponding to workflows in fMRIPrep's documentation.
GLM analyses
From each subject's surface-based time series, we used GLMsingle (Prince et al., 2022) to estimate the neural pattern of activity during the 11.5 s delay periods of the main experiment for each trial. GLMsingle is a three-step process where (1) an optimal hemodynamic response function (HRF) is fit to each vertex's time series from a bank of 20 HRF functions obtained from the Natural Scenes Dataset (Allen et al., 2021) via an iterative linear fitting procedure. (2) Noise regressors are computed from the data by identifying noisy vertices defined by negative R2, deriving noise regressors from this noise pool using principal component analysis, and then iteratively removing each noise regressor from all vertices’ time series. The optimal number of regressors is determined via a cross-validated R2 improvement for the task model. (3) GLMsingle implements fractional ridge regression as a way to improve the robustness of single-trial beta estimates, particularly useful here as our design yields a limited number of trials per target position within each condition.
We constructed our design matrices to have 48 regressors of interest (16 polar angle bins multiplied by three conditions), with the events modeled as boxcars corresponding to the 11.5 s delay periods. We estimated one model for each participant with these designs in GLMsingle, resulting in single-trial estimates for each trial in each surface vertex.
Fitting population receptive field models
Using the data from the retinotopy session, we fit population receptive field (pRF) models for each vertex on the cortical surface, as described by Himmelberg et al. (2021). In brief, for each surface vertex, we fit a circular 2D-Gaussian linear population receptive field (pRF) to the BOLD time series, averaged across identical runs of the bar stimulus. The software was implemented in Vistasoft as described in Dumoulin and Wandell (2008), with a wrapper function to handle surface data (https://github.com/WinawerLab/prfVista). The models are parameterized by the Gaussian center (x, y) and standard deviation (σ).
Visual field map definitions
Visual field maps were defined by drawing boundaries at polar angle reversals on each subject's cortical surface using an early version of the visualization tool, cortex-annotate (https://github.com/noahbenson/cortex-annotate), which is built on neuropythy software (https://github.com/noahbenson/neuropythy; Benson and Winawer, 2018). We followed common heuristics to define seven maps spanning early to midlevel visual cortex: V1, V2, V3 (Himmelberg et al., 2021; Benson et al., 2022); hV4 (Winawer and Witthoft, 2015); V3A and V3B (grouped into one ROI, V3ab) and IPS0 (Mackey et al., 2017); and LO1 (Larsson and Heeger, 2006).
We defined experiment-specific regions of interest for each visual field map composed of vertices whose pRF centers were near the target eccentricity and whose variance explained by the pRF model was above 10%. Specifically, we included only those vertices whose pRF centers were within one σ of 7° (the target eccentricity in the experiments). For example, a vertex with pRF center at 6° and pRF size (σ) of 1.5° would be included, but a vertex with pRF center at 6° and pRF size of 0.5° would not be included. We imposed the eccentricity restriction for the purpose of examining polar angle activation profiles, described in the next section. These measures are based on the retinotopy scans only and are therefore independent of the main experiment.
Analyses quantifying perception, working memory, and long-term memory activity
To examine the BOLD response within the 11.5 s delay, we constructed polar angle activation profiles for each visual field map and each condition (perception, working memory, long-term memory). Some analyses averaged over the delay period. For these analyses, we obtained the response amplitudes from GLMsingle (one beta weight per trial for each surface vertex). The visual field coordinates for each vertex came from pRF mapping, which was conducted in a separate fMRI session. We binned the response amplitudes by the polar angle distance between each vertex's pRF and the target location on that trial. Binning by polar angle distance from the target enabled us to average across trials with different target locations, resulting in an activation profile as a function of distance from the target. This results in a distinct polar angle activation profile for each subject, condition, and visual field map. Prior to averaging across subjects, we normalized each activation profile by dividing by the vector length of the perceptually evoked activation profile for that map. To preserve meaningful units, we then rescaled the activation profile by the average of vector lengths across subjects. Visual inspection of the average profiles showed a peak near 0 and negative responses far from 0. We fit the functions with a difference of two von Mises distributions, constrained so that the centers were the same.
A separate procedure was used to derive 2D activation profiles, which include time as well as polar angle. To derive these, we extracted the denoised BOLD time series for each vertex on each trial, rather than a single beta weight per trial, expressed as percent signal change from the mean of each scan. The denoised data were derived from the output of a GLM analysis using the GLMdenoise algorithm (Kay et al., 2013a). This algorithm is closely related to the GLMsingle algorithm used for the static analyses but has greater flexibility in specifying the trial structure (e.g., events with different durations) than the GLMsingle algorithm, but doesn't return estimates for single trials. The event beta weights from this GLM were not used for any analyses; instead, we only used it to derive denoised BOLD time series for subsequent temporal analyses.
For these 2D analyses, time was sampled at seconds from −3 to 17 s in each trial, relative to the start of the delay period. We then computed polar angle activation profiles independently for each time point, using the procedure described above: binning by polar angle distance from the target, averaging across trials within a condition, and visual field map. This results in 2D activation profiles, which span polar angle and time within a trial.
For temporal analyses, these 1D polar angle activation profiles were computed at each timepoint (TR = 1 s) from 3 s before the onset of the paired cue to 17 s after, resulting in a 2D heatmap of the polar angle activation profiles over time. As in the static activation profiles, the vertices were binned by polar angle distance from the target and averaged across trials within each visual map, separately for each condition and subject. This results in a matrix that is polar angle by time. We fit the polar angle profile independently for each time point, again using a difference of von Mises distributions, constrained so that the centers were the same.
For both analyses, we estimated the amplitude (trough to peak), the peak location, and full-width at half-maximum (FWHM) from each of the differences of von Mises fits. We bootstrapped across subjects (with replacement) 10,000 times to obtain 68 and 95% confidence intervals for these location, amplitude, and FWHM parameter estimates.
All of the analyses above were applied to all three experimental conditions: perception, long-term memory, and working memory. The pRF estimates, which came from a separate retinotopy experiment, were of course derived from a mapping experiment with a visual stimulus. In principle, one could also derive pRFs from separate experiments in which the mapping stimuli were retrieved from long-term memory or maintained in working memory, in order to have separate pRF estimates from each condition. However, memory pRFs are much more difficult to estimate and have a much lower signal-to-noise ratio. Moreover, if we used separate pRF estimates for each condition, then the polar angle response functions for the three conditions would not map to the same voxels, precluding us from addressing our primary question of how memory-driven BOLD responses differ from perceptually driven responses.
Temporal analyses
We characterized the time course for each 2D polar angle activation profile by fitting logistic functions to the amplitude estimate at each time point from −3 to 14 s relative to the onset of the paired cue. This is because, after 14 s, the saccade response dominated the signal introducing noise to the target-aligned response. The logistic fits resulted in estimates of four parameters for each 2D polar angle activation profile: t0, the rise-time to reach the function's midpoint; L, the upper asymptote of the amplitude estimates; k, the logistic growth rate of the function (i.e., the steepness of the curve); and c, a baseline. All four parameters were constrained to have a lower bound of 0; the upper bound was unconstrained for L and c, 15 s for t0, and 5% signal change/s for k. This fit was sufficient for the perception and long-term memory conditions, which generally showed a rise and then a steady response. For working memory, the response rose transiently and then declined to a lower value. This pattern was accurately captured by fitting a multiplication of two logistic functions rather than a single logistic function. The two functions were constrained to have the same L parameter but could differ in t0, c, and k. The parameter bounds for the second logistic function were the same as the first, except where k was constrained to be negative instead of positive. We repeated these logistic fits for each of the 10,000 bootstraps, computing 68% confidence intervals of the estimated logistic time series.
Saccade analyses
We used two EyeLink eye trackers (SR Research) with a 1,000 Hz sampling rate, one in the scanner and one outside the scanner. In the scanner, the EyeLink 1000 plus was mounted onto a rig in the magnet bore that holds the projection screen. In the psychophysics room for long-term memory training, we used an EyeLink Tower mount with a 25 mm lens mounted on the EyeLink camera to allow close viewing.
During both the training and the fMRI experiment, saccades were labeled by EyeLink's default saccade detector, which classifies saccades as eye movements with velocity and acceleration exceeding 30°/s and 8,000°/s2, respectively. Saccade responses were collected during the response window at the end of each trial. For all behavioral analyses, the saccade responses used are those which landed nearest the target eccentricity during the saccade response window.
Subjects often make multiple saccades to get to the target. We defined the endpoint as the saccade whose eccentricity was closest to the target eccentricity (7°), irrespective of the polar angle. We then measured the angular distance between this point and the target (ignoring eccentricity). We excluded saccade responses whose eccentricity was <3.5 or >12° visual angle from fixation. One subject was removed from the saccade analysis due to a technical error with the eye tracker during the scan sessions. Data from three scans from another subject were also excluded due to a calibration error.
For comparison between BOLD data and saccades, we divided the saccade data for each subject and each condition into tertiles for counterclockwise, center, and clockwise. We repeated the ROI-level analyses on this split saccade data to obtain 1D polar angle activation profiles for each ROI, condition, subject, and tertile.
Resampling statistics
We used the bootstrapped data from each analysis to make inferences on spatial tuning properties as a function of condition and of other trial-level factors. Statistics reported are computed using the bootstrapped data. To assess our main claims, we report the mean and confidence intervals from the bootstrapped data. For comparisons between a measurement and a fixed value, we report the 95% CI, which corresponds to ±2 standard deviations of a normal distribution. For comparisons between two estimates, we report the 68% CI, which corresponds to ±1 standard deviation of a normal distribution.
To assess the relationship between a map's position in the visual hierarchy and spatial tuning width, we assigned each map an ordinal value according to its relative position: 1, V1; 2, V2; 3, V3; 4, hV4, LO1, V3A/B; 5, IPS0. We fit a line of tuning width versus ordinal position for each bootstrap, generating a distribution of slope means for each condition. The rest of our analyses took the form of computing the differences of mean effects between conditions, both for individual visual maps and for the early (V1–V3) versus later (hV4–IPS0) visual cortex.
Software
Data visualization, model fitting, and statistical quantification for all analyses described in this paper were made using Matplotlib 3.5.2 (Hunter, 2007), NiBabel 3.2.2 (Brett et al., 2022), Pandas 1.4.2 (The pandas development team, 2024), scikit-learn 1.0.2 (Pedregosa et al., 2011), SciPy 1.8.0 (Virtanen et al., 2020), and Seaborn 0.11.2 (Waskom, 2021).
Results
We tested how the spatial tuning of cortical visual representations is shaped by viewing a peripheral target, maintaining it in working memory, or retrieving it from long-term memory. We also tested how the cortical representations during memory relate to the metrics of memory-guided saccades.
Responses in the visual cortex are spatially tuned during working memory and long-term memory
We first ask whether activation in the early visual cortex during memory is spatially tuned. To assess this, we parameterized the sensory representations generated during the delay period by remapping GLM estimates of brain activity from the cortical surface to visual space (Fig. 3A,B). We then computed polar angle activation profiles to capture the spatial tuning of sensory representations generated during the delay period (Fig. 3C) and estimated their amplitude, peak location, and tuning width (Fig. 3D). We compared these estimates between perception, working memory, and long-term memory conditions.
Target-aligned averages of sensory representations in visual space. A, Single-trial beta estimates for each vertex on the brain surface are obtained from the main task's delay period. In addition, a separate retinotopic mapping procedure is used to obtain population receptive field (pRF) estimates for each vertex. B, We used the pRF estimates to remap the single-trial estimates on the brain surface to single-trial estimates in visual space, which are then rotated and aligned by trial target position. The pRF estimates are also used to define retinotopic maps for seven regions of interest across the visual cortex and to restrict voxels to those with pRF centers near the target eccentricity. C, Normalizing and averaging the aligned trial estimates for each map yields target-aligned averages in visual space for each condition. A 2D visualization of a subject's target-aligned average is shown above for visual map V3. D, Beta estimates are binned by polar angle distance from target and fit to a difference of von Mises functions to produce polar angle activation profiles. This mapping of BOLD response as a function of polar angle distance from the stimulus captures the spatial tuning profile of cortical responses during perception, working memory, and long-term memory. Panels in this figure and all following figures can be reproduced using code and data located at https://osf.io/4mf8j/, and contained in the /paper/figures folder at https://github.com/rfw256/Woodry_2024_Cortical-tuning-of-visual-memory/.
The clearest difference between perception and the two memory conditions is that the BOLD amplitude is much larger during perception. This difference is particularly evident in earlier visual maps V1–V4. The larger response during perception was in part due to the drifting Gabor target, which further drove the response during the delay period. While the amplitudes were lower during memory, they were all positive, ranging from 0.25 to 0.5% percent signal change across visual maps (Fig. 4B, middle). The 95% confidence interval did not overlap 0% BOLD response in any ROI.
Memory has broader spatial tuning than perception in the earlier visual cortex. A, Polar angle activation profiles fit to brain activity during 11.5 s presentation/delay period. Highlighted regions indicate 68% confidence intervals bootstrapped across subjects. Perception working memory and long-term memory brain activity show spatial tuning to target locations across visual ROIs. B, Spatial tuning metrics obtained from polar angle activation profiles across visual ROIs. The dots are means across bootstraps. The thick shading is the 68% confidence interval. The thin shading is the 95% confidence interval. The dotted line in the leftmost panel represents the target location (0°).
The memory activation profiles, like the perception activation profiles, were spatially tuned to the stimulus. Specifically, for all three conditions and all seven visual field maps, peak location estimates were centered around 0° relative to the stimulus angle (Fig. 4B, left). Polar angle activation profiles which peak at 0° indicate accurate tuning to the true target location. Tuning to the target location is of course expected in the perception condition. But remarkably, we even see tuning to the target location in the earliest visual map, V1, during both memory conditions (the confidence intervals from all three conditions include 0°). Tuning in the two memory conditions confirms prior work showing engagement of visual areas, including primary visual cortex, during memory (Sprague and Serences, 2013; Breedlove et al., 2020; Favila et al., 2022; Vo et al., 2022). The fact that the peak location was close to 0 in both memory conditions means that the preferred polar angle location in memory is the same as it is in perception.
Memory shows broader spatial tuning than perception in the early visual cortex
The lower amplitude but similar peak location during memory compared with perception suggests that memory responses might be similar to perception except for a scale factor. This turns out to not be correct. We find that instead of memory responses resembling perception responses (up to a scale factor), the memory responses show a difference in tuning width, which cannot be achieved by simply increasing or decreasing the response amplitude.
During perception trials, tuning widths increased sharply from early to later visual maps. To quantify this tendency, we assigned each map an ordinal value based on its estimated position in the visual hierarchy: 1, V1; 2, V2; 3, V3; 4, hV4, LO1, V3A/B; 5, IPS0. We fit a line of tuning width versus ordinal position and found that for perception, the tuning width increased ∼21° per position in the hierarchy (slope = 20.9°, CI 18.5–23.8), consistent with what is known about receptive fields increasing in size from early to later visual areas (Smith et al., 2001; Dumoulin and Wandell, 2008). In contrast, there was little increase in tuning width across visual maps during both forms of memory, with a slope nearly two-thirds that measured during perception (working memory: slope = 15.2°, CI 10.6–18.2; long-term memory: slope = 14.8°, CI 7.7–18.6). Because the tuning widths were so similar for long-term memory and working memory, we compared the average of the two memory conditions to perception. This steeper slope for perception than memory was robust (diff = 6.2°, CI 3–13.3).
In early visual maps V1–V3, spatial tuning during memory was broader than during perception (diff = 20°, CI 15.8–29.7), especially in V1 (diff = 28.8°, CI 15.9–61; Fig. 4B, right). In later maps—hV4, LO1, and IPS0—tuning widths were about the same in perception, working memory, and long-term memory (diff = 0.31°, CI −6.7 to 8.5). Therefore, the memory responses are not just a scaled-down version of the perceptual responses but rather show broader tuning in earlier maps.
One possibility is that the broader polar angle response profiles during memory come from the choice of eccentricity band for the analysis. Breedlove et al. (2020) reported that a voxel's preferred eccentricity for remembered stimuli is shifted foveally. Hence, our analyses, which were limited to voxels near the target eccentricity of 7°, might have missed the largest responses during memory, where the polar angle tuning width might have differed. To assess this possibility, we recomputed the polar angle response profiles in different eccentricity bands. Because a shift in receptive fields results in an opposite shift in population response functions (Qian et al., 2023), the foveal shift in memory pRFs reported by Breedlove et al. predicts that the memory responses would be largest in voxels whose receptive fields are more peripheral than the stimulus. Indeed, at eccentricities less than the 7° target, the memory response amplitudes declined sharply. At target eccentricities greater than 7°, the memory response amplitudes were about the same as at 7°, indicating that we did not miss a much larger activation elsewhere in the map (Fig. 5B). Moreover, the polar angle tuning functions in memory were the same at 9 and 12° as at 7°. Hence, these analyses confirm that the broader tuning in memory compared with perception is not an artifact of memory responses shifting in eccentricity.
Broader spatial tuning during memory is robust when aligning to the angular location of saccade responses or to different eccentricities. A, Tuning estimates as a function of assuming greater target eccentricities. While amplitude is higher at increased eccentricity bands, the broader tuning profiles during memory remain consistent. B, Aligning to the polar angle location of the saccade response produces spatial tuning estimates similar to the original analyses. This shows that the broader tuning observed when aligning the memory responses to the target is likely not due to an error in memory recall. If so, this would predict a narrower memory response when aligning responses to the polar angle location of the saccade, assuming it is a reasonable proxy for the participant’s estimate of the target location.
A second possibility was that the broader tuning observed in the early visual cortex is an artifact of averaging across trials: If tuning is as narrow during memory as in perception, but the tuning “center” varies from trial to trial more during memory than perception, then averaging data across trials would inflate the estimated width during memory. To assess this, we aligned each trial's data to the polar angle of the saccade response rather than to the polar angle of the target, assuming that the saccade provides a reasonable estimate of where the participant thought the target was on each trial. Recomputing the tuning functions with this alignment did not alter the main pattern of results, again showing greater tuning width during memory than perception for early visual areas (Fig. 5A). This argues against the possibility that the broader tuning in early visual cortex observed during memory is instead due to larger error during recall.
Because the only differences between conditions are how visual information is accessed, the observed differences in tuning widths confirm our hypothesis that differences in how information enters the visual cortex shape the spatial tuning of the subsequent representations.
Distinct spatial tuning profiles emerge over time
The analysis above showed that the routing of stimulus information to the visual cortex—feedforward in perception versus top-down in memory—affects the spatial representation. Here, we ask how the routing affects the temporal dynamics of the response. Our expectation is that long-term memory responses will be slowest because the responses must be generated entirely internally (no stimulus is viewed). We can also ask whether working memory and long-term memory signals show the same tendency to sustain over the delay. To compare the temporal dynamics across conditions, we averaged the BOLD time series throughout the delay period across trials, binned by polar angle distance from the target (Fig. 6A). We then computed the peak amplitude at each time point and fit a rising logistic function (perception and long-term memory) or the product of a rising and falling logistic (working memory) across the time series (Fig. 6B). The logistic product captures both the transient response evoked by the target at the beginning of the working memory delay period and the decay to a lower, sustained, activation for the remainder of the delay period. Because we fit the working memory response using a different function, we do not compare logistic parameters to the other two conditions. We make two observations that distinguish perception from long-term memory from working memory.
Perception, working memory, and long-term memory have distinct tuning over time. A, BOLD time course during the delay periods. The target location is 0°. Tuning width estimates are plotted as black lines, starting from 3 s after the onset of the delay. The arrows in the V1 working memory panel indicate the time at which the working memory amplitude peaks and the end of the delay period. The working memory responses in V1 and V2 are initially narrow, similar to perception (first black arrow), and then widen, similar to long-term memory (second black arrow). B, Logistic functions fitted to amplitude estimates over time. Shaded regions represent the 68% confidence interval of the logistic fits, bootstrapped across subjects. C, Comparison of rise times to the inflection point of logistic fits between perception and long-term memory. Error bars represent 68% bootstrapped confidence intervals. D, Time course of tuning width (same values as the black lines in A). Error bars are 68% bootstrapped confidence intervals.
First, the working memory responses show a clear transition from a stimulus-driven transient, peaking at ∼4–5 s after the cue, to a lower sustained signal. This is expected because the target stimulus is briefly shown prior to the delay during working memory. Corroborating this interpretation, the tuning width in early visual areas is narrow during working memory at 4–5 s after the cue, similar to the perceptual condition. It then increases in width by stimulus offset, so that the width at offset is similar to the width in long-term memory. The increase in width in V1 is 36.6° (CI 19–125.5°) and in V2 is 35.3° (CI 28.2–45.4°). This is indicated by the arrows in Figure 6A (for V1) and summarized in the time series in Figure 6D. Since the perceptually driven responses are brief, the estimates from the static analyses in Figure 4 are similar to those at the end of trial, visible in the dynamic plots of Figure 6.
In contrast to the working memory dynamics in V1 and V2, there is no clear change in the tuning width of the working memory response in other visual areas, except for in V3ab (diff = 18.1°, CI 4.3–28.3°). This is to be expected given that the tuning widths of the working memory responses are matched to perception in these later visual areas. Moreover, there is little change in tuning width during the long-term memory trials or the perception trials.
Second, we find a general tendency for slower rise-time in long-term memory than perception across visual maps, with longer rise times in all maps except LO1, most prominently in earlier visual areas V1–V4 (diff = −1.4 s, CI −3.6 to −0.3; Fig. 6C). This is consistent with memory responses arising later due to the sluggishness of feedback. Moreover, the rise times are much more variable in memory, as expected from responses that are internally and effortfully generated (memory) rather than from external, stimulus-triggered responses (perception).
Errors in cortical tuning aligned with errors in memory-guided behavior
The results above demonstrate that the visual cortex is engaged in a retinotopically specific way during working and long-term memory. These results do not, however, indicate whether these retinotopic representations are relevant for behavior. This is an important and open question in the field, with many reports claiming the representations are linked to behavior (Bone et al., 2019, 2020; Hallenbeck et al., 2021; Li et al., 2021) though some question their relevance (Xu, 2017). If the visual cortex representations are relevant for behavior, we expect alignment between the memory-driven cortical responses during the delay and subsequent saccade responses. Here, we took advantage of trial variability to test whether the peak location of cortical responses during the delay aligned with the direction of saccade error.
To test the alignment between cortical and saccade responses, for each subject, we split the trials into three groups by their saccade error (Fig. 7B), those trials with saccades near the targets, counterclockwise of the targets, or clockwise of the targets, with angular thresholds set to make the three bins contain equals number of trials. We then repeated our spatial tuning analyses separately for the clockwise and counterclockwise trials to compute the estimates of peak location during the delay period (Fig. 7C). To reduce the number of comparisons, we defined a new region of interest as the union of all seven maps. For this large ROI, there was a strong link between neural tuning and saccade error in long-term memory: The peak estimates for trials with clockwise saccades were 16.5° more clockwise than the trials with counterclockwise saccades (diff = −16.5°, CI −27.6 to −8.8). In each of the separate maps, the same general pattern is found in long-term memory. It is particularly pronounced in V1 (diff = −13.3, CI −49.8 to 9.9), V3 (diff = −14.3°, CI −17.5 to −10.6), V3ab (diff = −12.7°, CI −20.4 to −7.3), LO1 (diff = −38.7°, CI −69.1 to −8), and IPS0 (diff = −27.6°, CI −43 to −12.3). In no map does the tuning go in the opposite direction of the saccades.
Saccade errors align with cortical tuning in long-term memory. A, Saccade response spread in visual space. We aligned saccade responses to the target location by subtracting their polar angle distance from target and added 90° to align to the vertical axis. Saccade responses during memory had greater error along polar angle than during perception, consistent with the broader spatial tuning during memory within early visual maps. Standard deviations of saccade errors in polar angle are shown in the bottom right. B, Histograms of saccade error along polar angle. We grouped trials into tertiles: clockwise (black), center (gray), or counterclockwise (white) to target polar angle location. The tertiles were defined for each subject and condition; hence, there is some overlap in the group-average histograms plotted. C, Peak location estimates for clockwise/counterclockwise trial groups. We fit polar angle activation profiles to the clockwise (black circles) and counterclockwise (white circles) groups. Error bars represent the bootstrapped 68% confidence intervals. D, Peak location estimates for clockwise/counterclockwise groups for each measured visual map. Error bars represent 68% confidence intervals.
In contrast, for working memory, there was no systematic relationship between the peak estimates of spatial tuning and saccade direction for all ROIs as a group (diff = 3.1°, CI −3.8 to 12.2), but there were decent effects in V1 (diff −14.4, CI −27.5 to −2.7). There was a small overall effect in perception (diff = −1.1°, CI −2.5 to 0.4), particularly in V1 (diff = −7.2, CI −13.9 to −1.3), V3 (diff = −4.7, CI −7.2 to −1.7), V3ab (diff = −3.8, CI −7.3 to 0.5), and LO1 (diff = −10.7, CI −15.2 to −4.3). This could be due to a restricted range of saccade errors in perception and to the possibility that when there are errors in working memory, they develop gradually over the trial, as the representation drifts away from the viewed target. Since the analysis of peak location pools across the whole delay, this would diminish the impact of biases near the end of the trials.
The long-term memory results suggest that information reconstructed in visual maps is likely used for memory-guided behavior.
Discussion
We measured spatial tuning of cortical responses during perception, working memory, and long-term memory. We found that both long-term and working memory scales the spatial tuning of responses in early but not late visual field maps. We further demonstrated the importance of memory-driven visual cortex activations by showing that they correlate with behavior: Trial-to-trial variation in the spatial tuning of cortical responses during long-term memory is related to the accuracy of subsequent oculomotor responses.
Spatial tuning in the visual cortex during long-term memory differs from perception
Long-term memory retrieval of information from the hippocampus can reinstate patterns of cortical activity evoked during encoding (Wheeler et al., 2000; Rugg et al., 2008; Johnson et al., 2009; Pearson et al., 2015). This “cortical reinstatement” is supported by neurally inspired models of episodic memory (Damasio, 1989; McClelland et al., 1995; Rolls, 2000; Teyler and Rudy, 2007). These theories do not imply that memory and perceptual representations are identical. For example, memory responses are noisier than perceptual responses. One possibility is that neural responses during memory reinstatement are like those during perception, except for decreased signal-to-noise level (Wheeler et al., 2000; Rugg et al., 2008; Johnson et al., 2009; Pearson et al., 2015).
Replicating two recent studies (Breedlove et al., 2020; Favila et al., 2022), we find only partial support for such reinstatement. On the one hand, spatial tuning during long-term memory retrieval peaks at the expected locations in multiple visual field maps, as early as V1, demonstrating that memory reinstatement is stimulus specific. On the other hand, long-term memory responses in earlier visual maps (V1–V3) were more broadly tuned than perceptual responses, demonstrating that memory-driven activity is not simply a reinstatement of perceptual responses. A likely explanation lies in the architecture of the visual cortex: During feedback, earlier maps inherit the lower spatial precision characteristic of later visual maps where neurons have large receptive fields. Prior work supports this reversal of cortical information flow as an explanation for top-down reinstatement during imagery (Kosslyn, 1994; Dentico et al., 2014; Pearson, 2019; Dijkstra et al., 2020; Breedlove et al., 2020). Lower precision in later stages is expected from a hierarchical encoding process with increasing levels of tolerance to the size and position of stimuli (Ito et al., 1995; Kay et al., 2013b); the greater size and position tolerance in later visual maps reflect a loss of position information likely not recoverable during retrieval. This information loss puts a limit on the precision of cortical reinstatement, independent of the lower amplitude signals during memory.
How much information is lost in long-term memory?
In addition to architectural constraints, the precision of top-down generated signals may also depend on task demands. A comparison between our results and those of Favila et al. (2022) supports this idea. While both studies showed broader tuning for memory than perception in early visual maps, the decreased precision during memory was more pronounced in Favila's study than this one (Fig. 8A). This difference is unlikely due to measurement noise or differences in eccentricity (4 vs 7°) or analysis. We infer this because the tuning width estimates of the two studies match during perception and, in later maps, during memory—the only difference between them being long-term memory in V1–V3. We speculate that the difference arises from encoding demands. The studies differed in the number of remembered targets and their spacing: four targets spaced by 90° in Favila et al. versus 16 targets spaced on average 22.5° here. The narrow spacing in our study likely required more precise encoding, leading to narrow tuning in V1–V3 during memory. A post hoc analysis supports this interpretation. Due to the random placement of stimulus locations within polar angle bins, the spacing between our stimuli varied (Fig. 8C). For some stimuli, the spacing was just 5°. For others, it was over 30°. Although two stimuli were never simultaneously present, the training protocol required saccade responses to be closer to the correct target than to any other target. Hence, nearby targets required greater precision. In V1–V3, tuning was narrower for targets with near neighbors than with far neighbors (Fig. 8B). Tuning width for far-spaced targets is similar to Favila et al.'s results. The implication is that increased competition during memory training invokes a greater demand for spatial precision, resulting in narrower tuning during recall. Further studies are needed to test whether this effect is driven by hippocampal pattern separation (Bakker et al., 2008; Yassa and Stark, 2011) or by selective attention, either at encoding or retrieval.
Systematic differences in spatial tuning of cortical responses. A, Comparison of spatial tuning (FWHM) estimates across measured visual maps between the Favila et al. (2022) study (circles) and our study (squares). Both studies show large agreement in their spatial tuning estimates during the perception condition (filled) and in later visual maps V4-IPS0 during long-term memory retrieval (unfilled). During long-term memory, tuning widths differ in early visual maps V1–V3. The main differences between ours and the Favila et al. study's memory conditions are in the spacing of the stimuli; their study required the retrieval of four targets equally spaced around the visual field, whereas ours involved the retrieval of 16 targets that varied in their separation. B, Spatial tuning width estimates across visual maps, split by whether targets were spaced near to (pink diamonds; spacing, 1.7–12°; mean, 7.7°; SD, 3.3°) or spaced far from (blue diamonds; spacing, 19.2–31.8°; mean, 23°; SD, 3.4°) each other. Also plotted are the mean tuning width estimates from both the Favila et al. study (circles) and this study (squares). The tuning widths from far-spaced targets in our study more closely resemble the broader tuning observed in the Favila et al. study, suggesting the differences between mean width estimates between our studies are in large part due to the increased precision required of more closely spaced targets. C, Example of this study's long-term memory retrieval targets. Sixteen target stimuli, each sampled from within 22.5° bins, vary in their separation from other targets. Some targets were spaced far apart, while others were spaced near each other, meaning those spaced near to each other likely required more precision during later recall. This is reflected in the narrower spatial tuning observed during these trials than for the “far-spaced” trials.
Amplitude and tuning dynamics in the visual cortex during memory
A peripheral visual stimulus occurred at the beginning of working memory but not long-term memory trials. During working memory, the location of this stimulus must be maintained over the delay period such that a memory-guided saccade can be generated to the stored location. During long-term memory, the target of the memory-guided saccade was retrieved from memory and not based on a visual stimulus. Consequently, in all areas, a robust transient response time-locked to the early visual stimulus distinguished working and long-term memory (Fig. 6B). In early areas, this transient response was narrowly tuned and matched to the perception condition. If working memory simply maintained recent activity, one would expect the tuning to stay narrow throughout the delay, but it does not. In fact, the tuning widths of responses in the visual cortex during the later epochs of the trial were remarkably similar during working and long-term memory (Fig. 6D). Despite originating from distinct bottom-up visual and top-down mnemonic sources, perhaps the contents of memory are similar by this time (i.e., the target of the memory-guided saccade). Indeed, visual memories flexibly and efficiently represent visual properties that are most behaviorally relevant (Kwak and Curtis, 2022; Chunharas et al., 2023; Li and Curtis, 2023; Duan and Curtis, 2024), in this case the goal of the memory-guided saccade. What limits the precision appears to be not whether the stimulus was recently viewed (working memory) or retrieved from long-term storage, but rather that the stimulus representation is maintained in the absence of perceptual input.
Surprisingly, responses in V1–V3 were more precise in some long-term memory trials (Fig. 8B, near-spaced stimuli) than working memory, despite the longer delay between encoding and retrieval for long-term memory. This is likely because during our long-term memory experiment, subjects viewed the stimuli many times (∼60 times each during prescan training) and, for near-spaced stimuli, needed to make fine distinctions. Recent work shows that when stimuli are highly learned in long-term memory, the working memory of these stimuli becomes more precise (Miller et al., 2022), similar to the relatively precise representations in V1–V3 for our near-spaced long-term memory stimuli.
Are the memory responses in the visual cortex behaviorally relevant?
Our trial-by-trial analyses demonstrated an alignment between behavioral responses and cortical activation in long-term memory, but not during perception or working memory. Other studies have reported links between trial-to-trial variability in long-term memory responses and self-reported vividness (Bone et al., 2019) and recognition accuracy (Bone et al., 2020). These studies do not predict the specific behavioral errors based on the cortical responses. Quantifying neural responses in stimulus units (polar angle) enables us to tightly link the trial-by-trial responses to behavior: The direction of the error in cortical responses is predictive of the direction of the error in behavior.
Nonetheless, our results are correlative. Demonstrating the “necessity” of visual cortex activity for behavior requires manipulating the neural responses. A recent transcranial magnetic stimulation (TMS) study suggests that disrupting V1 activity during spatial working memory maintenance reduces memory-guided saccade accuracy (Dake and Curtis, 2024), though effects of TMS on V1 during working memory delays did not reduce accuracy for feature judgments (orientation, Rademaker et al., 2017; color, van Lamsweerde and Johnson, 2017). Perhaps it is surprising then that we only find a correspondence between cortical activity and behavior in long-term memory, not working memory. It is unclear why there is a difference between the conditions. One likely possibility is that the saccade errors in working memory are smaller, making it more difficult to separate encoding errors from motor noise. Overall, the combination of a correspondence between behavior and long-term memory cortical responses reported here, between behavior and working memory cortical responses reported elsewhere (Ester et al., 2013; van Bergen et al., 2015; Hallenbeck et al., 2021; Li and Curtis, 2023), and between TMS and behavior (Dake and Curtis, 2024) make a strong case that memory responses in early visual cortex are relevant for behavior.
Is it just attention?
Memory shares features with attention, in that both involve modulation of neural responses that are not purely stimulus-driven. There are shared mechanisms between attention and long-term memory (Aly and Turk-Browne, 2017) and attention and working memory (Jerde et al., 2012; Panichello and Buschman, 2021; Zhou et al., 2022). Indeed, one way to describe sustaining a memory, whether sustained from recent exposure or retrieved from a long-term store, is attending to an internal feature (Cabeza et al., 2008). Thus, our memory tasks surely entail some elements of attention. Nonetheless, there are no results from the attention literature that would have enabled us to predict the pattern of tuning widths we observe in early versus late visual areas among perception, working memory, and long-term memory. Moreover, attention alone could not produce cortical patterns of activation in the visual cortex spatially tuned to the location of a remembered target. In the absence of any visual stimulation at that location during the long-term memory trials, the location information must first be retrieved from memory storage. Further experiments are needed to disentangle the contributions of attention and episodic memory to activation patterns in the visual cortex.
Decoding versus encoding approaches to studying memory representations
Our use of retinotopic models allowed us to quantify and compare sensory activation during perception, working memory, and long-term memory in terms of spatial tuning functions. This differs from the decoding approach, often used for stimulus orientation or other stimulus features which do not vary systematically at the mm to cm scale of fMRI measurement. Decoding fMRI responses in memory or imagery is less accurate than perception (Naselaris et al., 2015). Such results have been coupled with the idea that memory or imagery representations are like weak forms of perception, similar to viewing stimuli at low contrast (Pearson, 2019). But the decrease in decoding accuracy can arise from a decrease in signal-to-noise ratio (expected from a weaker stimulus) or a decrease in precision or a combination. Our stimulus manipulation allowed us to quantify precision independent of amplitude. Precision was lower in memory and also varied with task demands and map location in the visual hierarchy. Separating the precision of the representation from the signal-to-noise ratio is difficult for stimulus features whose representations vary over a submillimeter scale, such as orientation, despite recent advances in studying their encoding (Roth et al., 2018; Sprague et al., 2018; Gardner and Liu, 2019). In contrast, it is straightforward in the spatial domain, making spatial manipulations a useful tool for comparing neural representations across conditions. But spatial manipulations are not just a tool. Space is a fundamental part of visual representations (Wandell and Winawer, 2011), and its role in both perception and memory disorders remains a central topic in cognitive neuroscience (Bisiach and Luzzatti, 1978; Farah, 2003).
Footnotes
We thank the New York University (NYU) Center for Brain Imaging for the technical support. This work was supported by the National Eye Institute (R01 EY033925 and R01 EY016407 to C.E.C.; R01 EY027401 to J.W.; P30 EY013079 Core Grant for Vision) by the National Institute of Mental Health (R01 MH111417 to J.W.), and pilot funds from the NYU Center for Brain Imaging to R.W. This work was supported in part through the NYU IT High Performance Computing resources, services, and staff expertise. We thank Ilona Bloem, Jan Kurzawski, Marc Himmelberg, Rania Ezzo, and Ekin Tunçok for their help with scanning; Serra Favila, Tommy Sprague, and Ekin Tunçok for their helpful comments on the manuscript; and Serra Favila for sharing data and code. Because this is my first submission, I’d like to extend special thanks to Kate Yurgil and Liz Chrastil for taking a chance on me.
The authors declare no competing financial interests.
- Correspondence should be addressed to Robert Woodry at rfw256{at}nyu.edu.
