Dissociation of striosome and matrix activation in the human striatum during the cue and execution phases of working memory: A probabilistic tractography and task fMRI assessment

Alishba Sadiq; Jeff L Waugh

doi:10.1016/j.neuroimage.2026.122212

. Author manuscript; available in PMC: 2026 Oct 18.

Published in final edited form as: Neuroimage. 2026 Sep 9;340:122212. doi: 10.1016/j.neuroimage.2026.122212

Dissociation of striosome and matrix activation in the human striatum during the cue and execution phases of working memory: A probabilistic tractography and task fMRI assessment

Alishba Sadiq ^a, Jeff L Waugh ^a,^b,^*

PMCID: PMC13675866 NIHMSID: NIHMS2216064 PMID: 42716361

Abstract

The striatum comprises two neurochemically and anatomically distinct tissue compartments, the striosome and matrix, that are hypothesized to support different aspects of cognition and action. Animal studies link striosome to reward evaluation, emotional learning, and decision-making under conflict, whereas matrix is associated with sensorimotor integration and task execution. However, evidence for compartment-specific function in humans remains limited, in part because the gold standard for identifying striosome and matrix, immunohistochemistry, requires ex vivo tissue sections. We instead utilized an inferential method, probabilistic tractography guided by prior histologic assessments, to identify voxels with striosome-like and matrix-like patterns of structural connectivity in healthy adults. Striosome-like and matrix-like voxels identified through this probabilistic method are highly reliable (test-retest error: 0.14%) and match the spatial distribution, extra-striate connectivity, tendency for separation vs. clustering, and relative abundance of striosome and matrix in tissue. Using these inferential compartment-like voxels, we then examined compartment-like responses during an fMRI n-back working-memory task using four visual stimulus categories (body part, face, place, and tool), comparing low (0-back) and high (2-back) working-memory load. Functional activation was temporally segregated, consistent with prior motor-task findings: striosome-like voxels were preferentially engaged during cue and preparatory phases, whereas matrix-like voxels dominated during task execution. Trial accuracy strongly modulated striatal responses, with both compartments activating more during correct than error trials (+71% in striosome-like voxels, +29% in matrix-like voxels). However, accuracy-related modulation did not significantly differ between the compartments. Both compartments showed increased activation from 0-back to 2-back, indicating sensitivity to working-memory load, with larger load-related increases in matrix-like voxels. Category-selective responses differed by compartment and load. Under low load, stimulus-category effects were modest and similar between compartments. Under high load, striosome-like voxels retained category selectivity, whereas matrix-like voxels showed reduced specificity. Together, these findings suggest that striosome–matrix specialization extends beyond motor functions to cognitive domains, reflecting a conserved division between preparatory and execution-related processes that varies systematically with task demands, stimulus category, and performance accuracy.

Keywords: Striatum, Compartment, Striosome, Matrix, Connectivity-based parcellation, Task-based fMRI, Working memory

1. Introduction

Historically, the basal ganglia have been primarily associated with motor control, a view shaped by early 20th-century work by Wilson (1912, 1914) and Vogt (1911), who observed motor impairments following damage to this region. The cardinal motor symptoms of Parkinson’s disease (PD) arise from the progressive degeneration of dopaminergic neurons in the substantia nigra pars compacta (SNc), leading to dopamine (DA) depletion in the striatum (Surmeier et al., 2014). However, anatomical and functional evidence accumulated over the past several decades has established that the striatum supports not only motor control but also cognitive and limbic functions. As the major input nucleus of the basal ganglia, the striatum is thought to mediate the selection and sequencing of motor actions (Mink, 1996; Graybiel, 2008), it evaluates cortical “action plans,” integrates sensory and motivational context with prior experience, and transmits this evaluation to downstream nuclei that shape motor output through temporally patterned inhibitory signaling to the thalamus and brainstem. Anatomical studies show that the striatum forms part of several cortico–striato–thalamo–cortical circuits, each responsible for processing different types of information such as motor, cognitive, or limbic functions (Middleton and Strick, 2000). Haber et al. (2000) further demonstrated in macaques that the ventral midbrain acts as an interface, enabling information exchange between these striatal regions and integrating signals related to movement, cognition, and emotion (Haber et al., 2000). These findings highlight the role of the striatal regions to cognitive as well as motor functions.

Basal ganglia dysfunction contributes not only to classic movement disorders but also to a range of cognitive and psychiatric impairments. For example, in Parkinson’s disease and Huntington’s disease (Wichmann and Dostrovsky, 2011) motor symptoms are typically accompanied by, and are often preceded by, significant cognitive and mood symptoms. In addition, conditions such as obsessive–compulsive disorder (OCD; (Burguiere et al., 2015)), attention-deficit/hyperactivity disorder (ADHD; (Singh et al., 2024)), Tourette’s syndrome (TS;(Albin, 2006)), anxiety (Funk and Waugh, 2026) and depression (Dunlop and Nemeroff, 2007; Tieu and Waugh, 2025) highlight the basal ganglia’s central involvement in cognitive control and behavioral regulation. Within this system, the striatum (comprised of caudate and putamen) integrates afferent signals from the cortex, thalamus, and brainstem to guide the selection and execution of discrete behaviors. Striatal spiny projection neurons (SPNs) are organized into two distinct tissue compartments, the striosome and matrix, which differ in their developmental origins, neurochemical markers, pharmacological properties, and connectivity with other brain regions (Crittenden and Graybiel, 2011). The striosome is a labyrinthine, 3-dimensional structure that is surrounded by the matrix (Brimblecombe and Cragg, 2017). The compartments are interdigitated and the precise location of the striosome branches varies between individuals, precluding the use of region-of-interest approaches to distinguish the functions of striosome and matrix.

The striosome receives dense inputs from limbic and prefrontal regions and exerts direct inhibitory control over dopaminergic neurons in the substantia nigra pars compacta, positioning the striosome to regulate motivational salience and the gating of task-relevant information (Watabe-Uchida et al., 2012; Friedman et al., 2015; Crittenden et al., 2016). In contrast, the matrix receives inputs from dorsolateral prefrontal, parietal association, and sensorimotor areas, and projects to basal ganglia output nuclei, supporting the sustained maintenance and manipulation of information necessary for cognitive performance (Lévesque and Parent, 2005; Watabe-Uchida et al., 2012). Dopaminergic inputs further differentiate the compartments, decreasing striosome firing while increasing matrix firing (Prager et al., 2020). Together, this architecture suggests that striosome and matrix, and the functional networks in which they are embedded (Sadiq et al., 2025), form complementary but distinct substrates for higher-order cognitive functions, including working memory (Weglage et al., 2021).

Despite detailed descriptions of this two-compartment architecture in animal models and human histology, its functional significance for memory has yet to be elucidated. Evidence from animal and human studies supports functional specialization between the striosome and matrix compartments. The striosome directly regulates nigral dopamine signaling and mediates decision-making under conflict (White and Hiroi, 1998; Friedman et al., 2015), aversive learning, and negative-valence memory (Jenrette et al., 2019; Nadel et al., 2020). In contrast, the matrix receives extensive sensorimotor and associative inputs and is critical for movement execution, habit formation, and behavioral routines (Aosaki et al., 1994; Lévesque and Parent, 2005). Through its strong projections to both the direct and indirect pathways, the matrix compartment contributes to an oppositional mechanism in which the direct pathway facilitates desired motor programs while the indirect pathway suppresses competing or inappropriate programs, thereby enabling precise action selection and control (Mink, 1996). Importantly, this dynamic has also been proposed to extend to cognitive domains, including attention, working memory, and rule selection (Swerdlow and Koob, 1987; Graybiel, 1997; Bejjani et al., 1999). This balance between stabilization and suppression provides the neural substrate for selective flexibility – the ability to maintain focus on current goals while efficiently shifting to alternative strategies when required.

We recently demonstrated, using probabilistic diffusion tractography, that connectivity-based parcellation can delineate striatal voxels with striosome-like or matrix-like properties in vivo, based on their distinct cortical and subcortical connectivity profiles (Waugh et al., 2022). This approach enables individualized mapping of compartment-like organization in the living human brain, while recognizing that “striosome-like” and “matrix-like” voxels are defined probabilistically and are not the equivalent of compartment mapping by immunohistochemical markers. Nevertheless, MRI-based parcellations recapitulate all of the anatomic features of striosome and matrix architecture described in tissue. First, the relative abundance of matrix-like (~85%) and striosome-like (~15%) voxels closely matches histological estimates from human and primate studies (Desban et al., 1993; Holt et al., 1997; Waugh et al., 2022; Funk et al., 2023, 2024; Sadiq et al., 2025; Tieu and Waugh, 2025). Second, their spatial distributions mirror histological observations, with striosome-like voxels concentrated in the rostroventral striatum and matrix-like voxels more evenly distributed in dorsolateral and caudal regions (Graybiel and Ragsdale, 1978). Third, connectivity of these voxels is organized somatotopically (Funk et al., 2023), consistent with tract tracing studies in animals (Flaherty and Graybiel, 1993; Eblen and Graybiel, 1995). Fourth, matrix-like voxels tend to form large contiguous clusters, whereas striosome-like voxels are more often isolated. Finally, this parcellation method is highly reliable, with a test–retest error rate of only 0.14% across repeated scans (Waugh et al., 2022). Importantly, both structural and functional analyses have shown that these compartment-like biases depend on the precise spatial location of voxels rather than their striatal “neighborhood”: even small (2–3 mm) random displacements abolish compartment-specific structural connectivity patterns (Funk et al., 2023; Sadiq et al., 2025) and resting state functional connectivity networks (Sadiq and Waugh, 2025). Together, these findings establish a reproducible framework for probing the functional significance of striosome- and matrix-like voxels in humans, including their roles in working memory.

Although striosome and matrix have been well described in animal studies and human histology, their contributions to higher-order cognition remain poorly understood. It is unclear how the compartments participate in the dynamic processes of working memory, such as preparing to update task-relevant information, sustaining representations across delays, and executing responses based on those representations. We previously demonstrated that striosome-like and matrix-like voxels are embedded in segregated structural and resting-state functional networks (Funk et al., 2023, 2024; Sadiq et al., 2025). However, whether striosome- and matrix-like voxels are differentially recruited during working memory tasks in humans has not been established. To address this gap, we used compartment-like voxels as seeds in task-based fMRI, testing whether the two compartments exhibit distinct activation profiles during working memory tasks. Specifically, we asked: (1) Are striosome- and matrix-like voxels differentially engaged during working memory demands? (2) Do these differences depend on distinct phases of the task, such as cue-related processing versus information maintenance and response execution? (3) Do the observed dynamics align with proposed roles of the striosome in motivational and evaluative processing, and of the matrix in sustaining task-relevant cognitive representations? This study provides the first direct evidence linking microcircuit-specific striatal activation patterns to higher-order cognition and neuropsychiatric symptoms.

2. Materials and methods

Overview: we combined structural connectivity and task-based functional MRI approaches to investigate compartment-specific activation in the human striatum during working memory tasks (Fig. 1). We performed structural connectivity-based parcellation using probabilistic diffusion tractography to identify striatal voxels with striosome-like or matrix-like connectivity biases. To assess functional activation, we extracted task-evoked BOLD signals from each subject’s striosome- and matrix-like masks. Each participant’s response to working memory events was modeled using a general linear model (GLM), a standard framework for estimating task-related brain activity. We calculated contrast parameter estimates (COPEs) at each voxel to quantify activation differences between task and baseline conditions, yielding compartment-specific activation measures. This framework allowed us to compare how the magnitude and timing of activation differed between striosome- and matrix-like voxels across varying working-memory demands.

Fig. 1. — Schematic of the experimental workflow for evaluating compartment- and load-specific activation in the human striatum. Individual striosome-like and matrix-like compartments were delineated using structural connectivity–based parcellation. Because precise striosome location varies between individuals, uniform region-of-interest approaches cannot reliably distinguish the compartments. Therefore, subject-specific compartment-like masks were used to extract activation during a working-memory task.

2.1. Study population

This study was a secondary analysis of MRI data from the Human Connectome Project (HCP) S1200 release (Van Essen et al., 2013), which includes high-quality multimodal neuroimaging data from a large cohort of healthy young adults. From the original sample of 1206 participants, we included only those with complete diffusion MRI and task-based (working memory) fMRI data. Participants were excluded if they reported any lifetime history of illicit or addictive substance use (including cocaine, hallucinogens, cannabis, nicotine, opiates, sedatives, or stimulants). We also excluded individuals who met DSM-5 criteria for Alcohol Use Disorder (Alcohol Abuse or Dependence) or who reported consuming, on average, more than four alcoholic drinks per week in the year prior to scanning. These exclusions resulting in a final cohort of 964 participants. All participants provided written informed consent as part of their enrollment in the HCP study (Van Essen et al., 2012). In a related study using an overlapping experimental cohort, we previously identified resting-state networks that covaried with striosome-like versus matrix-like voxels (Sadiq et al., 2025).

2.2. MRI acquisition protocols

Task-based functional MRI (tfMRI) and diffusion tensor imaging (DTI) scans were obtained from the Human Connectome Project (HCP) S1200 dataset. All were collected on 3T MRI systems using standardized protocols across three participating sites. The tfMRI scans were acquired with echo-planar imaging (EPI) sequences matched to the resting-state fMRI (rs-fMRI) acquisitions, ensuring identical spatial and temporal resolution between modalities. A multiband gradient-echo EPI sequence was employed with the following acquisition parameters: repetition time (TR) of 720 ms, echo time (TE) of 33.1 ms, flip angle of 52°, field of view (FOV) of 208 × 180 mm, and matrix dimensions of 104 × 90, yielding 72 contiguous axial slices. The multiband acceleration factor was 8, with an echo spacing of 0.58 ms and a bandwidth of 2290 Hz per pixel. The tfMRI resolution was 2.0 mm isotropic, which we have previously demonstrated is sufficient to resolve striosome-like and matrix-like structural connectivity (Waugh et al., 2022). Each working memory task condition included two runs of 405 frames, each lasting approximately 5 min and 1 s. Separation of cue and execution phases was enabled by the task design, in which brief cue events and longer execution periods occurred at different times and showed low correlation across the time series. DTI data for S1200 subjects was acquired at 1.25 mm isotropic resolution using 200 directions (14 B0 vol, 186 vol at noncolinear directions) with the following parameters: repetition time = 3.23 s, echo time = 0.0892. DTI scans included both anterior-posterior and posterior-anterior acquisitions, allowing for correction of susceptibility artifacts.

2.2.1. Working memory-task design

The experiment used a working-memory task that incorporated stimuli from multiple visually-presented categories (e.g., faces, places, body parts, tools) under both 0-back and 2-back conditions. In the 0BK task, participants identified a pre-specified target, whereas in the 2BK task they determined whether the current stimulus matched the one presented two trials earlier. Stimuli consisted of pictures of places (e.g., outdoor scenes or buildings), tools, faces, and non-mutilated body parts (with no nudity), presented in separate blocks within each run. Each run included eight task blocks: half using the 0-back and half the 2-back task, and four fixation blocks. At the start of each block, a 2.5-s cue indicated the task type (and the target for 0-back). Because each condition (defined by load and stimulus category) occurred once per run (left-to-right (LR) and right-to-left (RL)), this yielded two cue events per condition at the individual subject level. Although cue responses were modeled over a short duration (2.5 s), these timepoints represent a single event per block and do not increase the number of independent observations. As a result, cue-related estimates are based on relatively sparse sampling within subjects. However, the large sample size of the HCP dataset enables reliable estimation of condition-specific effects at the group level, consistent with prior work demonstrating that event-related fMRI responses can be estimated at the level of individual trials, even in studies with brief stimuli and limited repetitions per condition (Mumford et al., 2012; Turner et al., 2012; Prince et al., 2022). Task blocks contained 10 trials (each 2.5 s: 2 s stimulus plus 500 ms inter-task interval), totaling 25 s per block, while fixation blocks lasted 15 s. The fixation blocks served as baseline condition, during which participants fixed on a centrally presented cross without performing any task. All memory task events were modeled relative to the average of the four 15-s fixation blocks. This approach allowed us to contrast memory-related activation with a stable, aggregated baseline estimate.

For subsequent analyses, we examined activation patterns as a function of both task performance and compartment type. For the accuracy-based analysis, we extracted activation values separately for correct and error trials within each compartment and aggregated across stimulus categories. We aggregated error trials across stimulus categories because the HCP task design provides error events as a single regressor per condition (e.g., 2-back errors), without category-specific separation. This design, combined with the relatively low frequency of error trials, necessitated aggregation to enable reliable estimation of error-related activation. Not all participants contributed data to every condition: all 964 participants had valid 0-back correct trials, whereas only 844 participants had at least one 2-back error event and were therefore included in the error-related analysis. Because 0-back error trials were rare, we restricted our assessment of error-related activation to the 2-back condition. Category-selectivity analyses contrasted activation within each compartment for each stimulus category (e.g., place) against the mean of the other three categories (e.g., tools, faces, and body parts).

We implemented two separate first-level general linear models (GLMs) in FSL FEAT to address the cue–execution and accuracy analyses. For the cue–execution analyses, the first-level GLM modeled the cue (2.5 s) and execution (25 s) phases separately for each combination of memory load (0-back and 2-back) and stimulus category (body, face, place, and tool). This model included 16 task regressors (eight execution and eight cue regressors), each convolved with FSL’s canonical double-gamma hemodynamic response function. We included temporal derivatives for each regressor to account for small variations in hemodynamic response timing, resulting in 32 regressors in the final design matrix. We generated separate contrast estimates for each cue and execution condition. For the accuracy analyses, a second first-level GLM modeled 0-back correct, 2-back correct, and 2-back error trials as separate regressors. We again included temporal derivatives for each regressor, resulting in six regressors in the final design matrix. Because 0-back error trials were rare (too few for a reliable analysis), we did not model them separately.

To assess the separability of cue- and execution-related responses, we computed Pearson correlations between the corresponding cue and execution regressors for each stimulus category and task run (LR and RL). Correlations close to zero were interpreted as evidence of minimal collinearity and reliable estimation of cue- and execution-related activation.

2.3. fMRI preprocessing

We analyzed the minimally preprocessed HCP task-fMRI data (Glasser et al., 2013). The HCP minimal preprocessing pipeline applies gradient-distortion correction, motion correction (rigid-body realignment), EPI distortion correction using spin–echo field maps, and registration of each subject’s functional data to their T1-weighted anatomical image and subsequently to MNI space via the fMRIVolume pipeline. The data is further cleaned using ICA-FIX, which removes structured noise components, including motion-related artifacts. No additional preprocessing steps, such as temporal filtering or spatial smoothing were applied before first-level modeling.

To evaluate residual motion, we computed framewise displacement (FD) following Power et al. (2012) using the HCP-provided movement regressors from both the LR and RL runs. We then calculated the average FD across all participants, which was low (group mean FD = 0.16 mm). Given the overall low motion in the sample and the extensive motion handling already implemented in the HCP pipeline, no subjects were excluded based on individual FD values.

2.4. Striatal parcellation

We recently developed a method for identifying striosome-like and matrix-like voxels in the human striatum based on their distinct in vivo connectivity profiles (Waugh et al., 2022). We use the terms “striosome-like” and “matrix-like” to emphasize that these parcellations are inferential and probabilistic, in contrast with the gold standard for identifying striosome and matrix in tissue, immunohistochemical staining. To parcellate the striatum, we generated composite target masks representing regions with connectivity biases toward one compartment, as established by prior tract-tracing studies in animals and by diffusion tractography in humans (summarized in Waugh et al., 2022). Striosome-favoring regions included the posterior orbitofrontal cortex, anterior insula, basolateral amygdala, basal operculum, and posterior temporal fusiform cortex, whereas matrix-favoring regions included the inferior frontal gyrus pars opercularis, primary motor cortex, supplementary motor area, primary somatosensory cortex, and superior parietal cortex.

Striatal parcellation was performed using the FSL tool probtrackx2, which allowed us to evaluate the relative connectivity of each striatal voxel to striosome-favoring versus matrix-favoring target masks. Tractography was conducted in each subject’s native diffusion space using standard parameters: curvature threshold = 0.2, step length = 0.5 mm, number of steps per sample = 2000, number of samples per seed voxel = 5000, and distance correction enabled to prevent target proximity from biasing connection strength. For each voxel, we compared the number of streamlines projecting to striosome-favoring versus matrix-favoring targets, and the resulting ratio of streamline counts was used as an index of compartmental bias. This produced a probability value for each voxel (P = 0–1), indicating the degree of striosome- or matrix-like connectivity. Voxels were classified as compartment-biased when P > 0.55 toward either set of targets, yielding continuous subject- and hemisphere-specific maps of striosome-like and matrix-like connectivity. Because the precise spatial distribution of striosome is unique to each individual, these parcellations must be defined for each subject; standardized striatal region-of-interest masks are not suitable for studying compartmental organization.

Because the resolution of our diffusion voxels matched the upper limit of the diameter of striosome branches (1.25 mm), all voxels had the potential to contain both striosome and matrix tissue. In fact, many striatal voxels exhibited no or only modest compartment-like connectivity bias. To maximize the contrast between compartments, we excluded voxels with low or indeterminate bias using an iterative thresholding procedure applied to the compartmental probability maps. For each subject and hemisphere, we selected voxels from the most strongly biased portion of the distribution (favoring either striosome-like or matrix-like connectivity) and iteratively lowered the threshold until the mask volume reached a predetermined target. This target was set at 13% of the original striatal mask, corresponding to 1.5 standard deviations above the mean of a Gaussian distribution. To ensure equality of mask volumes, which reduced size-based biases, we matched striosome-like and matrix-like mask volume within each subject and hemisphere. We have previously demonstrated that these high-bias, equal-volume masks reproduce the characteristic spatial distribution, differential connectivity, and somatotopy of striosome and matrix observed in histological studies (Waugh et al., 2022; Funk et al., 2023, 2024; Sadiq et al., 2025; Sadiq and Waugh, 2025; Tieu and Waugh, 2025). Because the diffusion data were acquired at 1.25-mm isotropic resolution (voxel volume = 1.953 mm³) and the fMRI data at 2-mm isotropic resolution (voxel volume = 8 mm³), a given anatomical volume is represented by fewer voxels in fMRI space than in diffusion space. Based on the volumetric ratio of approximately 4.1 (8 ÷ 1.953), 311 diffusion-space voxels translated to roughly 76 voxels at fMRI resolution. This volume-based conversion provides a consistent and biologically grounded reference for comparing compartment sizes across modalities and ensures appropriate interpretation when projecting diffusion-derived masks into fMRI space. Each subject’s high-bias compartment masks were then registered from diffusion space to structural (T1-weighted) space using FSL’s FLIRT, and subsequently nonlinearly transformed into fMRI space with the precomputed functional-to-structural registration matrices from the HCP pipeline. All transformations were visually verified for accuracy. The resulting striosome-like and matrix-like masks in fMRI space were used as seeds in subsequent functional connectivity analyses.

2.5. Validation of striatal parcellation

To test whether our parcellated voxels reproduced the anatomic patterns of striatal compartmentalization established through histology, we quantified their spatial organization within the equal-volume 1.5 SD masks. For each subject and hemisphere, Cartesian coordinates (x, y, z) were extracted for every voxel and referenced to the centroid of the corresponding nucleus (caudate or putamen). We then assessed spatial distribution by calculating within-plane dispersion and the root-mean-square (RMS) distance from the nucleus centroid, providing voxelwise measures of compartment organization in three-dimensional striatal space. Consistent with prior histological descriptions, we have previously shown that striosome-like voxels are preferentially localized to rostral, medial, and ventral striatum (Graybiel and Ragsdale, 1978; Goldman-Rakic, 1982; Donoghue and Herkenham, 1986; Ragsdale and Graybiel, 1990; Desban et al., 1993; Eblen and Graybiel, 1995; Waugh et al., 2022).

Striosomal branches are embedded within the surrounding matrix (Graybiel and Ragsdale, 1978; Holt et al., 1997). In coronal histologic sections, striosome appear as discrete “islands” within a continuous “sea” of matrix tissue, though the striosome is actually a contiguous, 3-dimensional structure. To compare the spatial organization of our MRI-derived parcellations with this observation from tissue, we applied the fsl-cluster command at a high bias threshold (P > 0.87) to isolate voxels with strong compartment-like bias. We analyzed the largest cluster within each compartment-like probability distribution, comparing the degree of separation (contiguity vs. isolation) in striosome- and matrix-like voxels.

We previously showed, in a similar HCP-derived cohort, that shifting voxel location by only 2–3 mm was sufficient to eliminate striosome-like bias in structural connectivity, indicating that compartment-specific biases depend on precise voxel position rather than the striatal “neighborhood” in which a voxel is located (Funk et al., 2023). More recently, we demonstrated that shifting voxel location also eliminated compartment-specific biases in resting-state functional connectivity (Sadiq et al., 2025). Based on these findings, we hypothesized that task-evoked functional biases would similarly depend on precise voxel location. To test this, we assessed the spatial specificity of task-based activation by jittering the locations of striosome- and matrix-like voxels by ±0–3 voxels in each anatomical plane, randomly and independently for each voxel. For every subject, we confirmed that although individual voxels were displaced, the mean location of the jittered masks remained nearly identical to the mean of the original masks, and importantly, no jittered voxel overlapped with voxels from the original masks. The average root-mean-square displacement was small: 2.9 voxels for the randomized striosome-like mask and 3.1 voxels for the randomized matrix-like mask. When shifts were averaged within each anatomical plane (combining positive and negative displacements), the mean displacement was minimal (0.37 voxels; range: 0.07–1.2). Thus, the jittering procedure produced small but measurable displacements at the voxel level without altering the overall spatial neighborhood of the striatal masks. In our prior motor task fMRI study (Sadiq and Waugh, 2025), we showed that such jittering was sufficient to abolish compartment-specific activation differences specifically in the striosome-like compartment, confirming that observed biases were spatially precise and not attributable to local neighborhood effects. Consistent with this framework, we used the jittered masks here as a negative control, directly comparing task-evoked activation in the original striosome- and matrix-like voxels with activation in the randomized voxel location masks.

2.6. Analysis of task-based functional MRI (tfMRI)

We performed a task-based fMRI analysis to characterize condition-specific activation within striosome-like and matrix-like striatal compartments during working memory tasks. At the first level, statistical modeling was carried out using a GLM, with each working memory condition (0-back and 2-back) represented as a 25-s block defined by onset timings from the HCP-provided explanatory variable (EV) files. To account for acquisition variability, data LR and RL phase-encoding runs were combined in a second-level analysis, yielding averaged estimates. For each subject, contrasts of parameter estimate (COPEs) were extracted for the working memory conditions (0-back and 2-back), and mean activation values were calculated within the striosome-like and matrix-like masks. This activation measure corresponded to COPE estimates from the subject-level GLMs and are expressed in arbitrary units (AU), reflecting the relative amplitude of BOLD signal change.

Unlike resting-state functional connectivity, which measures spontaneous correlations among networks, this task-based approach isolates activity that is aligned with specific task events and characterizes how that activity unfolds across time. In doing so, it provides a direct test of how striosome- and matrix-like voxels are differentially recruited across distinct phases of memory performance.

2.7. Slice-wise voxel sampling for sensitivity analysis

Our primary voxel-selection strategy emphasized voxels with the strongest compartmental bias but did not explicitly control for their spatial distribution within the striatum. As a consequence, highly biased voxels, particularly within the matrix-like distribution, could be spatially clustered, raising the possibility that observed compartment-specific effects might reflect regionally concentrated signals rather than striatum-wide organizational principles. Our previously described striosome-like masks utilized all, or nearly all voxels in the striosome-like distribution (target: 13% of striatal volume; expected from tissue: 15% of striatal volume). Re-selecting striosome-like voxels was unlikely to substantially change these masks. In contrast, the matrix-like distribution included many voxels that were unsampled in our primary matrix-like masks, and these tended to be substantially more clustered than striosome-like voxels. To evaluate whether matrix-like effects were robust to this voxel-selection assumption, we performed a sensitivity analysis using a slice-wise voxel sampling approach to define our matrix-like masks.

Specifically, for each axial slice of the striatum, we selected an equal number of the most-biased matrix-like voxels. This procedure ensured that matrix-like voxels were sampled across the full superior–inferior extent of the striatum while preserving our goal of using high-bias voxels to represent each compartment. This approach reduced the influence of localized clustering and tested the sensitivity of compartment-specific activation effects to voxel selection strategy. Persistence of matrix-like effects under this alternative sampling scheme would indicate that these effects reflect striatum-wide compartmental organization rather than localized regional bias.

2.8. Statistical analysis

We assessed the accuracy of our striatal parcellations – the intra-striate position, clustering, and mean bias of our compartment-like voxels – using a series of two-tailed, paired-samples t-tests. We compared the volume within streamline bundles seeded by compartment-like voxels using two-tailed, paired-samples t-tests.

For the task-based analyses, separate comparisons tested (1) compartment-specific activation during cue presentation, (2) load-dependent activation (2-back vs. 0-back), (3) accuracy-related modulation (correct vs. error trials), and (4) category selectivity (each stimulus type vs. the mean of all other stimulus types). To control for multiple comparisons, we applied the Benjamini-Hochberg (BH) procedure (Benjamini and Hochberg, 1995) across each family of related tests (e.g., stimulus categories × task loads × compartments), using a false discovery rate (FDR) threshold of Q = 0.05. For transparency, corrected significance thresholds are reported in the Results. We defined a result as trending toward significance if p < 0.05 but the result did not survive multiple comparisons correction.

Because there were very few errors in 0-back trials, we restricted our accuracy-based analyses to the 2-back condition (n = 844 participants with sufficient error trials). All activation values were averaged across hemispheres to yield mean compartment-level activation per subject.

To formally test whether compartmental activation depended on working-memory load, stimulus category, behavioral phase, or response accuracy, we performed repeated-measures ANOVAs using within-subject factors appropriate for each analysis (Compartment × Load × Category; Compartment × Phase; and Compartment × Accuracy). Because the Human Connectome Project (HCP) cohort includes related participants, we performed an additional sensitivity analysis using linear mixed-effects models to account for the potential non-independence of observations arising from family structure. These models included fixed effects of Compartment, Load, Category, and their interactions, together with a random intercept for Family_ID and subject-specific random intercepts and slopes for the within-subject factors (Compartment, Load, and Category). We assessed statistical significance using Satterthwaite approximations for denominator degrees of freedom.

3. Results

After applying exclusion criteria, the final sample included 964 healthy adults (mean age = 29.3 ± 3.7 years; 526 females, 438 males). Participants were excluded primarily due to incomplete diffusion or task-fMRI data or due to self-reported substance or alcohol use. This cohort partially overlapped with our previously reported resting-state sample (Sadiq et al., 2025), with 674 participants (70%) common to both analyses.

3.1. Validation of cue–execution regressor separability and sensitivity analysis accounting for family relatedness

Each cue condition contained only two events. To assess whether cue- and execution-related responses could be reliably estimated despite this limited sample, we quantified the correlation between the corresponding cue and execution regressors for each stimulus category and task run. Cue and execution regressors exhibited consistently low correlations across all conditions (overall mean absolute correlation ≈ 0.02), indicating minimal collinearity. The largest average correlation observed was −0.132 (0-back body condition, LR run), whereas all remaining conditions were close to zero (Supplementary Table S1). These findings demonstrate good separability of the cue and execution regressors, supporting direct comparison of cue- and execution-related activation estimates.

Because the Human Connectome Project cohort includes related individuals, we performed an additional sensitivity analysis using linear mixed-effects models with a random intercept for Family_ID and subject-specific random intercepts and random slopes for the within-subject factors (Compartment, Load, and Category). The principal findings were largely preserved. For cue-related activation, the significant main effects of Load (p = 1.2 × 10⁻⁶), Category (p = 9.5 × 10⁻⁹), and Compartment (p = 1.1 × 10⁻⁷), as well as the significant Load × Category (p = 1.3 × 10⁻¹⁷) and Load × Category × Compartment (p = 2.7 × 10⁻⁴) interactions, were preserved. However, the Load × Compartment interaction was no longer significant. For task-related activation, the principal compartment-related effects remained significant, including Compartment (p = 0.047), Load × Compartment (p = 1.7 × 10⁻⁶), Category × Compartment (p = 0.0021), and Load × Category × Compartment (p = 2.9 × 10⁻⁴). These findings indicate that the principal conclusions were robust and were not driven by the inclusion of related participants.

3.2. Comparing MRI-parcellated voxels to striosome and matrix in tissue

To validate our MRI-based parcellations (Supplemental Fig. 1), we first tested whether striosome-like and matrix-like voxels reproduced the well-established spatial organization of these compartments within the striatum. Across species ranging from rodents to primates, striosome branches are consistently enriched in the rostral, ventral, and medial striatum, whereas the matrix compartment predominates in caudal, dorsal, and lateral regions (Graybiel and Ragsdale, 1978; Goldman-Rakic, 1982; Donoghue and Herkenham, 1986; Ragsdale and Graybiel, 1990; Desban et al., 1993; Eblen and Graybiel, 1995; Waugh et al., 2022). Our voxelwise location analyses replicated this bias bilaterally. In the caudate, matrix-like voxels were significantly more lateral (1.3 mm; p = 8.6 × 10⁻⁵²), caudal (− 9.4 mm; p < 1 × 10⁻²⁶⁰), and dorsal (8.7 mm; p < 1 × 10⁻²⁶⁰) than striosome-like voxels. Similarly, in the putamen, matrix-like voxels were more lateral (0.7 mm; p = 8.8 × 10⁻⁵), caudal (− 5.8 mm; p < 1 × 10⁻²⁶⁰), and dorsal (5.9 mm; p < 1 × 10⁻²⁶⁰) than striosome-like voxels. Distance-to-centroid RMS analyses confirmed this spatial dissociation: in the caudate, striosome-like voxels clustered 12.2 mm medio-rostro-ventral to the centroid, whereas matrix-like voxels were positioned 13.7 mm latero-caudo-dorsal; in the putamen, striosome-like voxels were 10.2 mm medio-rostro-ventral, compared to 8.0 mm latero-caudo-dorsal for matrix-like voxels.

Histological studies in both animal and human tissue found that striosome and matrix comprise roughly 15% and 85% of the striatal volume, respectively (Johnston et al., 1990; Desban et al., 1993; Holt et al., 1997). To evaluate whether our MRI-based parcellations approximated this ratio, we quantified compartment-like volume across a range of bias thresholds. Because diffusion voxels (1.25 mm isotropic) sample tissue without regard to compartment boundaries, many voxels inevitably contain mixtures of striosome and matrix. Increasing the bias threshold excludes more of these blended voxels, yielding volume estimates based on voxels with stronger compartment-specific connectivity. At the highest cutoff (P > 0.87), striosome-like voxels represented 5.9% of biased voxels and matrix-like voxels 94.1%, while the majority of striatal voxels (82%) remained unclassified (i.e., indeterminate bias). At the lowest cutoff (P > 0.55), 25.5% of biased voxels were striosome-like and 74.5% were matrix-like, with 45% of the striatum showing sufficient compartment-like bias to be classified. The threshold that most closely matched histological estimates (15:85) was P > 0.7, which yielded 17.3% striosome-like voxels and 82.7% matrix-like voxels, encompassing 32% of total striatal volume. Although proportions shifted with the chosen threshold, all analyses demonstrated consistent volumetric relationships (matrix-like much larger than striosome-like) that paralleled histological observations. This correspondence supports the validity of our MRI-based parcellation approach for identifying compartment-like organization in vivo.

Finally, we examined whether striosome-like and matrix-like voxels differed in their tendency to cluster with other voxels of the same type. In histologic sections, each striosome branch is surrounded by contiguous matrix tissue (Graybiel and Ragsdale, 1978; Holt et al., 1997). In contrast, diffusion MRI samples the striatum using a rigid voxel grid that does not conform to individual striosome architecture, introducing partial volume effects that blur striosome-like signal across adjacent voxels. Although this resolution is insufficient to resolve the connectivity of individual striosome branches, we evaluated whether striosome- and matrix-like voxels differed in their clustering tendencies. Striosome-like voxels tended to form smaller, more scattered clusters, whereas matrix-like voxels aggregated into larger, more contiguous clusters. In the right hemisphere, matrix-like clusters were on average 4.2 times larger than striosome-like clusters, and in the left hemisphere they were 3.4 times larger. In the right hemisphere, striosome-like clusters had a mean volume of 234.2 mm³ (SEM ± 6.1; 95% CI [222–246]), substantially smaller than the mean matrix-like cluster volume of 985 mm³ (SEM ± 10.9; 95% CI [963–1006]; p = 1.5 × 10⁻²⁷⁵). Similarly, in the left hemisphere, striosome-like clusters averaged 301 mm³ (SEM ± 6.7; 95% CI: [287–314]), compared to 1009 mm³ for matrix-like clusters (SEM ± 12.4; 95% CI [984–1033]; p = 6.0 × 10⁻²²⁶). These results parallel histological observations, in which striosome branches appear as spatially dispersed islands embedded within the continuous matrix.

3.3. Cue-evoked compartment-specific activation in 0-back and 2-back tasks

We assessed cue-specific activation in striosome- and matrix-like voxels during the 0-back and 2-back tasks (Fig. 2). During cue periods in the 0-back condition, striosome-like voxels exhibited significantly greater activation than matrix-like voxels across all stimulus categories (body part, face, place, and tool), indicating a consistent compartment-like bias during task preparation. In the 2-back condition, cue-related striosome-like activation remained higher than matrix-like activation for body part and face cues, though compartment differences were no longer significant for place and tool stimuli. Overall, cue-evoked activation across all categories was markedly higher in striosome-like voxels than in matrix-like voxels (Fig. 2A). When averaged across categories, mean activation was 83.1 in striosome-like voxels and 52.7 in matrix-like voxels, corresponding to a mean paired difference of 30.4 activation units (95% CI [24.8, 35.9], Cohen’s d_z = 0.17, p = 1.5 × 10^–26). Thus, striosome-like activation was 1.6-fold greater than matrix-like activation. During the 2-back cue period, a significant compartment effect was observed only for body -part and face stimuli, where striosome-like activation exceeded matrix-like activation (Fig. 2B). For body-part cues, mean activation was 62.5 in striosome-like voxels and 27.8 in matrix-like voxels, corresponding to a mean paired difference of 34.7 activation units (95% CI [20.78, 48.58], Cohen’s d = 0.16, p = 6.5 × 10^–8). For face cues, mean activation was 64.3 in striosome-like voxels and 51.5 in matrix-like voxels, corresponding to a mean paired difference of 12.8 activation units (95% CI [1.18, 24.50], Cohen’s d = 0.07, p = 2.3 × 10^–2). Multiple comparisons correction yielded a corrected significance threshold of p = 3.8 × 10⁻².

To formally evaluate whether cue-related compartment differences depended on working-memory load and stimulus category, we performed a repeated-measures ANOVA with within-subject factors of Compartment (striosome-like, matrix-like), Load (0-back, 2-back), and Category (body, face, place, tool). Significant main effects were observed for Compartment (F₍₁₉₆₃₎ = 216.5, p = 2.3 × 10⁻⁴⁴), Load (F₍₁₉₆₃₎ = 59.7, p = 2.8 × 10⁻¹⁴), and Category (F_(3,2889) = 5.4, p = 0.001). Significant Compartment × Load (F₍₁₉₆₃₎ = 8.3, p = 0.004), Load × Category (F_(3,2889) = 42.8, p = 4.9 × 10⁻²⁷), and Compartment × Load × Category (F_(3,2889) = 6.8, p = 0.0001) in teractions indicated that compartmental differences varied with working-memory load and, furthermore, that the load-dependent compartment effect differed across stimulus categories. The Compartment × Category interaction approached but did not reach significance (F_(3,2889) = 2.5, p = 0.055).

3.4. Compartment-specific activation patterns during working-memory tasks

Next, we assessed the post-cue activation in striosome- and matrix-like voxels (Fig. 3). Matrix-like voxels exhibited significantly greater activation during working memory tasks, with differences becoming more pronounced under higher cognitive load (2-back). In the 0-back condition, a compartment effect was observed only for face stimuli, where matrix-like activation exceeded striosome-like activation (matrix = 11.6, striosome-like = 6.6), corresponding to a mean paired difference of 4.95 activation units (95% CI [3.14, 6.77], Cohen’s d = 0.17, p = 4.1 × 10^–9). In the 2-back condition, however, activation across all stimulus categories was consistently higher in matrix-like voxels (Fig. 3B). When averaged across categories for 2BK, mean activation was 16.6 in matrix-like voxels and 12.4 in striosome-like voxels, corresponding to a mean paired difference of 4.27 activation units (95% CI [3.25, 5.29], Cohen’s d = 0.13, p = 1.9 × 10^–19). Thus, matrix-like activation was approximately 1.3-fold greater than striosome-like activation during the 2-back task. Multiple comparisons correction yielded a corrected significance threshold of p = 3.1 × 10^–2.

While the direction of compartment-specific effects was consistent across load conditions, activation magnitude showed opposite scaling across task epochs. Execution-related activation increased from the 0-back to the 2-back condition, whereas cue-locked activation was markedly larger in the 0-back condition and reduced under higher load.

To formally evaluate whether task-related compartment differences depended on working-memory load and stimulus category, we performed a repeated-measures ANOVA with within-subject factors of Compartment (striosome-like, matrix-like), Load (0-back, 2-back), and Category (body, face, place, tool). Significant main effects were observed for Compartment (F₍₁₉₆₃₎ =27.9, p = 1.6 × 10⁻⁷), Load (F₍₁₉₆₃₎ = 197.5, p = 6.0 × 10⁻⁴¹), and Category (F_(3,2889) = 6.7, p = 0.0001), indicating overall differences in task-evoked activation between compartments, working-memory loads, and stimulus categories. Significant Compartment × Load (F₍₁₉₆₃₎ = 11.1, p = 0.0008), Compartment × Category (F_(3,2889) = 8.9, p = 7.7 × 10⁻⁶), and Compartment × Load × Category (F_(3,2889) = 6.7, p = 0.0002) interactions indicated that compartmental differences varied according to both working-memory load and stimulus category, and that the load-dependent compartment effect differed across stimulus categories. The Load × Category interaction was not significant (F_(3,2889) = 2.0, p = 0.11).

Taken together, these findings suggest distinct compartmental activation profiles during cue preparation and task execution. To formally evaluate whether these compartment differences varied between behavioral phases, we performed a repeated-measures ANOVA with within-subject factors of Compartment (striosome-like, matrix-like) and Phase (cue, task). Significant main effects were observed for Compartment (F₍₁₉₆₃₎ = 99.6, p = 2.1 × 10⁻²²) and Phase (F₍₁₉₆₃₎ = 886.3, p = 1.3 × 10⁻¹³⁸). Importantly, a significant Compartment × Phase interaction was observed (F₍₁₉₆₃₎ = 377.0, p = 4.1 × 10⁻⁷¹), demon strating that striosome- and matrix-like voxels were differentially activated during cue preparation and task execution.

3.5. Sensitivity analysis using less-biased voxels

To assess the robustness of our findings, we next characterized how the slice-wise voxel sampling procedure altered the properties of the matrix-like masks relative to the primary equal volume 1.5 SD masks. As intended, the slice-wise approach produced matrix-like masks with a modest but consistent reduction in compartmental bias, with mean bias decreasing from 0.955 in the 1.5 SD masks to 0.950 in the slice-wise masks (0.48% decrease). This confirms that the slice-wise masks were less dominated by the most extreme matrix-biased voxels while still retaining strong compartmental specificity.

More importantly, the slice-wise procedure substantially altered the spatial organization of matrix-like voxels. Average cluster volume was reduced from 143.9 to 46.3 voxels, a 67.8% decrease, and the average number of clusters decreased from 14.5 to 6.5 (54.7% decrease). These changes indicate a marked reduction in localized voxel clustering and a redistribution of matrix-like voxels across the striatum, demonstrating that the slice-wise masks differed meaningfully from the primary masks in both spatial distribution and organization.

Despite these substantial changes in voxel selection and spatial organization, the compartmental dissociations observed in the primary analysis were largely preserved. Matrix-like voxels continued to show greater activation during task execution, whereas striosome-like voxels exhibited stronger cue-period activation. These effects were most robust for body- and face-related conditions across both 0-back and 2-back tasks.

Several contrasts, most notably those involving place stimuli did not retain statistical significance under the slice-wise voxel definition, and in a small subset of conditions (e.g., 2-back place execution and cue), effect directions were not preserved. This pattern suggests reduced robustness rather than systematic reversal and indicates that place-related compartmental effects are more sensitive to the intra-striate location of matrix-like voxels.

3.6. Impact of specific voxel location on striatal compartmentalization

Striosome and matrix compartments are distributed in distinct regions of the striatum, and our striosome-like and matrix-like voxel masks showed a similar spatial distribution. This raises the possibility that the observed compartment-specific differences in functional activation might not reflect intrinsic properties of the compartments themselves, but instead the activation patterns of the striatal territories where those voxels were predominantly located – a potential “neighborhood effect.” To address this concern, we compared activation values obtained from the original striosome- and matrix-like masks (Section 3.2) with those derived from spatially shifted versions of the same masks, thereby testing whether compartment-specific effects depended on the precise voxel locations selected. We applied voxelwise jittering to each compartment-like mask, such that the position of individual voxels was randomized while the overall displacement of the mask remained minimal. On average, the center of gravity of the jittered masks shifted only 0.37 voxels from the original masks. In matrix-like voxels, mean bias decreased from 0.95 in the original mask to 0.83 in the jittered version (p < 1 × 10⁻²⁶⁰). In striosome-like voxels, bias declined more sharply, from 0.78 to neutral (0.47; p < 1 × 10⁻²⁶⁰).

These findings indicate that compartment-specific bias cannot be attributed to regional location (neighborhood effects) but instead reflected the structural connectivity of the precisely defined voxels within our compartment-like masks. We therefore set these location-shifted voxels as negative controls to assess the specificity of compartment-specific activation during working memory. In spatially shifted striosome-like masks, functional activation during both 0-back and 2-back working memory tasks was reduced across all eight task conditions, compared with the original, precisely-selected striosome-like masks (Table 1). In the 0-back condition, shifting voxel locations reduced activation for all stimulus categories, with decreases ranging from −8.9% (place, p = 2.5 × 10⁻²) to −32.1% (body part, p = 2.5 × 10⁻⁶). In the 2-back condition, activation was again reduced across all categories, with reductions of −13.6% (place, p = 2.0 × 10^–9) to −30.8% (body part, p = 5.0 × 10^–14). These findings demonstrate that even when location-shifted voxels occupied the same striatal “neighborhood,” their striosome-like activation profile was critically dependent on precise voxel placement.

Table 1.

Randomly shifting the location of striosome-like voxels by 2–3 voxels significantly reduced functional activation during working-memory tasks, whereas shifting matrix-like voxels had minimal effects. This matches the expectation from histology: since striosome is surrounded by matrix, shifting from striosome-like voxels is likely to select a matrix-like voxel and thus change the pattern of activation. In contrast, shifting the location of a matrix-like voxel is likely to select another matrix-like voxel. Δ Activation (Shifted – Original), percentage change, and p-values indicate how voxel displacement altered mean activation within each compartment. Bolded p-values denote significant differences (corrected for multiple comparisons).

Compartment	Task	Δ Activation (Shifted – Original)	% Change	p-value
0-back Striosome	Body part	−1.8	−32.1	2.5 × 10 ⁻⁶
	Face	− 2.1	−31.6	4.9 × 10 ⁻⁹
	Place	−0.7	−8.9	2.5 × 10 ⁻²
	Tool	−1.8	−25.9	9.9 × 10 ⁻⁸
0-back Matrix	Body part	−0.4	−5.3	2.7 × 10⁻¹
	Face	−0.3	−2.3	4.5 × 10⁻¹
	Place	−0.4	−4.6	1.7 × 10⁻¹
	Tool	−0.2	−2.0	6.2 × 10⁻¹
2-back Striosome	Body part	−2.8	−30.8	5.0 × 10 ⁻¹⁴
	Face	−2.7	−22.3	1.1 × 10 ⁻¹³
	Place	−2.1	−13.6	2.0 × 10 ⁻⁹
	Tool	−3.5	−27.7	1.1 × 10 ⁻¹⁸
2-back Matrix	Body part	−0.3	−1.9	3.7 × 10–1
	Face	−0.73	−4.8	3.8 × 10–2
	Place	−0.2	−1.4	5.0 × 10–1
	Tool	−1.2	−8	9.7 × 10–4

Open in a new tab

In the matrix-like compartment, shifting voxel locations had minimal impact on activation. For the 0-back condition, activation changes were uniformly small (−2.0% to −5.3%) and non-significant across all stimulus categories (all p ≥ 0.17). For the 2-back condition, reductions were similarly minimal (−1.4% to −4.8%), except for the tool category, which showed a significant decrease (−8.0%, p = 9.7 × 10⁻⁴). This stability likely reflects the abundance and widespread distribution of matrix-like voxels, such that shifting the location of a matrix-like voxel most often results in selecting another matrix- like voxel with slightly weaker bias. This control analysis highlights the biological specificity of compartment-related differences: shifting striosome-like voxels removes their striosome-like properties, whereas shifting matrix-like voxels only reduces the strength of their matrix bias. Thus, compartment-specific effects reflect precise voxel selection rather than nonspecific activation patterns of the surrounding striatal “neighborhood.” Notably, following voxel displacement all activation differences were negative, indicating that voxels with stronger structural connectivity bias also exhibited stronger functional activation, further supporting a tight coupling between structural compartmentalization and functional engagement.

3.7. Activation changes from 0-back to 2-back within striatal compartments

Next, we investigated within-compartment differences between the 0-back and 2-back conditions to assess how increasing working memory load modulated striatal activation. During task execution, both striosome- and matrix-like voxels exhibited significantly higher activation during the 2-back condition compared to the 0-back condition (Fig. 4). Both striosome- and matrix-like voxels showed robust increases in activation with the increased effort of the 2-back condition, with fold-changes ranging from ~1.7- to 1.9-fold in the striosome-like and ~1.3- to 2.3-fold in the matrix-like voxels, reflecting proportional increases in mean BOLD signal within each compartment (Fig. 2). Specifically, striosome-like activation rose by 1.7 × for body part, 1.8 × for face, 1.9 × for place, and 1.8 × for tool stimuli, while matrix-like activation increased by 2.3 ×, 1.3 ×, 1.8 ×, and 2.0 × for the same categories, respectively. Since all comparisons survived correction, the effective corrected significance threshold was p < 0.05.

Fig. 4. — Activation in 2-back memory tasks was higher for every condition, in both compartments, than activation in 0-back tasks. Mean BOLD activation is shown separately for striosome-like (red/orange) and matrix-like (blue shades) voxels across four stimulus categories (body part, face, place, tool). For both compartments, activation increased significantly in the 2-back relative to the 0-back condition, indicating sensitivity to working memory load. Striosome-like voxels (left) showed modest but reliable increases for body part, face, place, and tool stimuli, suggesting a limited but consistent contribution under higher task demands. Both striosome-like and matrix-like voxels showed significant increases in activation from the 0-back to the 2-back condition across all stimulus categories. The magnitude of these load-dependent increases varied across stimulus categories and compartments. These findings demonstrate that both compartments are sensitive to increasing working-memory load, with the magnitude of load-related modulation varying by compartment and stimulus category. Error bars illustrate standard error of the mean. *, p < 0.05; **, p < 1 × 10^–8.

3.8. Compartment-specific activation in correct and error trials

We next investigated whether trial accuracy correlated with compartment-specific task activation by comparing correct and error trials (Fig. 5). Notably, we assessed activation differently in this experiment – unlike other analyses that showed overall or stimulus-locked task activation, this analysis estimated activity that was uniquely attributable to specific trial types relative to the implicit baseline. Analyses were limited to participants with sufficient data for each condition: all subjects contributed 0-back correct trials (n = 964), but very few subjects made errors in this simple task. In contrast, most participants had both correct and incorrect responses in the 2-back trials (n = 844). Activation values for correct and error trials were analyzed separately rather than as difference scores (as in our other analyses), allowing direct comparison of mean BOLD responses between accuracy conditions within each compartment. Correct trials were available for both the 0-back and 2-back conditions and were aggregated across all stimulus categories (body, face, place, and tool). Since 0-back error trials were absent for most participants, we examined error-related activation only for the 2-back condition. Matrix-like voxels exhibited significantly greater activation than striosome-like voxels during correct trials in both the 0-back and 2-back tasks (Fig. 5A). In the 0-back correct condition, matrix-like activation was 9.3-fold higher than striosome-like activation (matrix-like = 7.4, striosome-like = 0.8), corresponding to a mean paired difference of 6.6 activation units (95% CI [5.38, 7.78], Cohen’s d = 0.37, p = 2.3 × 10^–25). Because sustained task-related engagement during the 0-back condition was present throughout the block and was not uniquely time-locked to individual correct events, this ongoing activity was absorbed into the baseline. As a result, striosome-like responses during 0-back correct trials were near zero. That is, the robust task-related activation observed in striosome-like voxels in our other analyses was sustained in 0-back trials, but did not increase further in correct trials. These compartment-specific differences were further amplified during correct 2-back trials, where matrix-like activation exceeded striosome-like activation (matrix-like = 14.8, striosome-like = 8.2), corresponding to a mean paired difference of 6.59 activation units (95% CI [5.35, 7.82], Cohen’s d = 0.25, p = 8.1 × 10^–25). Importantly, even during error trials in the 2-back task, matrix-like activation remained significantly greater than striosome-like activation (matrix-like = 10.5, striosome-like = 2.4), corresponding to a mean paired difference of 8.13 activation units (95% CI [5.95, 10.32], Cohen’s d = 0.18, p = 4.5 × 10^–13).

To determine how trial accuracy influenced activation, we compared correct and error trials within each compartment (Fig. 5B). Analyses were restricted to the 844 participants who had valid activation estimates for both correct and error trials in the 2-back condition. Both compartments showed reduced activation during error trials, but the magnitude of this reduction differed between the compartments. In striosome-like voxels, mean activation declined sharply in error trials, from 8.2 to 2.4 (Correct to Error, 71% reduction; p = 3.7 × 10⁻⁹), whereas in matrix-like voxels, activation decreased from 14.8 to 10.5 (29% reduction; p = 1.1 × 10⁻⁵). Thus, trial accuracy was associated with activation in both compartment-like populations.

To formally evaluate whether compartment-specific activation differed between correct and error trials, we performed a repeated-measures ANOVA with within-subject factors of Compartment (strio-some-like, matrix-like) and Accuracy (correct, error) using 2-back trials only. Significant main effects were observed for Compartment (F₍₁₈₄₃₎ = 73.1, p = 5.7 × 10⁻¹⁷) and Accuracy (F₍₁₈₄₃₎ = 22.6, p = 2.4 × 10⁻⁶), indicating overall differences in activation between striosome- and matrix-like compartments and between correct and error trials. However, the Compartment × Accuracy interaction did not reach significance (F₍₁₈₄₃₎ = 1.9, p = 0.17), indicating that the effect of response accuracy on activation did not differ significantly between striosome-like and matrix-like voxels. Thus, although compartment-specific and accuracy-related differences were present independently, accuracy did not differentially modulate activation between the compartments.

3.9. Category-specific activation during 0-back and 2-back tasks in striosome-like and matrix-like voxels

Next, we evaluated whether category-specific information was preserved within striosome-like and matrix-like voxels as working-memory load increased. This analysis aimed to determine whether compartmental specialization extends beyond temporal differences (cue vs. execution) to include content selectivity, that is, whether each compartment retains distinct patterns of activation for specific visual categories under varying task demands.

For each visual category (body part, face, place, tool), we compared activation during that category with the mean activation across the other three categories (hereafter, MeanOther; Fig. 6). At low load (0-back), striosome-like voxels showed reduced body and increased place activation relative to MeanOther. Body activation was lower than MeanOther (body = 5.5, MeanOther = 7.3), corresponding to a mean paired difference of 1.78 activation units (95% CI [0.31, 3.25], Cohen’s d = 0.054, p = 0.018). In contrast, place activation exceeded MeanOther (place = 8.3, MeanOther = 6.4), corresponding to a mean paired difference of 1.87 activation units (95% CI [0.39, 3.35], Cohen’s d = 0.057, p = 0.01). Face and tool responses did not differ significantly from MeanOther. Matrix-like voxels similarly showed reduced body activation relative to MeanOther (body = 7.5, MeanOther = 9.7), corresponding to a mean paired difference of 2.15 activation units (95% CI [0.48, 3.81], Cohen’s d = 0.058, p = 0.01). In contrast, face activation exceeded MeanOther (face = 11.5, MeanOther = 8.4), corresponding to a mean paired difference of 3.21 activation units (95% CI [1.67, 4.75], Cohen’s d = 0.093, p = 4.3 × 10⁻⁵). Place activation did not differ significantly from MeanOther, whereas tool activation showed a trend toward significance. Thus, at 0-back, body responses were suppressed in both compartments, whereas the stimulus category showing enhanced activation differed between compartments, with place preferentially enhanced in striosome-like voxels and face preferentially enhanced in matrix-like voxels.

Fig. 6. — Category-specific activation relative to the mean of all other categories in striosome-like and matrix-like voxels during 0-back and 2-back tasks. Mean BOLD activation is shown for each stimulus category (*body* part, *face, place, tool*) relative to the average activation across the remaining three categories (MeanOther), plotted separately for striosome-like (red) and matrix-like (blue) voxels. **(A) 0-back task**. In striosome-like voxels, *body* activation was significantly lower than MeanOther, whereas *place* activation was significantly higher; *face* and *tool* did not differ significantly from MeanOther. In matrix-like voxels, *body* activation was significantly lower than MeanOther, whereas *face* activation was significantly higher. *Place* did not differ significantly from MeanOther, and *tool* showed a trend toward significance. **(B) 2-back task**. In striosome-like voxels, *body* activation was significantly lower than MeanOther, whereas *place* activation was significantly higher. Matrix-like voxels did not show significant category-specific differences from MeanOther. Together, these findings indicate that category-related modulation was present in both compartment-like populations under low working-memory load, whereas under higher load, significant category-specific modulation was observed in striosome-like but not matrix-like voxels. Error bars illustrate the standard error of the mean. *, p < 0.022; τ, trend toward significance.

At higher cognitive load (2-back), category modulation became less distinct in matrix-like voxels but was preserved in striosome-like voxels. Specifically, body activation remained lower than MeanOther (body = 9.4, MeanOther = 13.4), corresponding to a mean paired difference of 4.0 activation units (95% CI [2.59, 5.41], Cohen’s d = 0.127, p = 3.0 × 10⁻⁸). In contrast, place activation exceeded MeanOther (place = 15.6, MeanOther = 11.3), corresponding to a mean paired difference of 4.35 activation units (95% CI [2.82, 5.88], Cohen’s d = 0.127, p = 2.9 × 10⁻⁸). These findings indicate a compartmental dissociation: whereas both compartments exhibited some degree of category sensitivity under low cognitive load, increasing task demands preserved category-specific tuning in striosome-like voxels, while matrix-like responses shifted toward more generalized task engagement. Multiple comparison correction yielded a significance threshold of p = 2.2 × 10⁻².

4. Discussion

Differences in development, pharmacology, and connectivity predict distinct functions for striosome and matrix (Graybiel and Ragsdale, 1978; Crittenden and Graybiel, 2011; Brimblecombe and Cragg, 2017). Animal studies support compartment-specific roles across valuation/threat decisions, habit learning, and motor control (Friedman et al., 2015; Xiao et al., 2020; Nadel et al., 2021; Okunomiya et al., 2025). In humans, evidence for compartment-specific functionality has been largely inferential, based on neuroimaging studies demonstrating distinct striatal connectivity and functional organization consistent with striosome–matrix architecture. Building on our prior work, which described compartment-specific intrinsic functional networks at rest and a temporal dissociation of compartment activation during motor behaviors (Sadiq et al., 2025; Sadiq and Waugh, 2025), we asked whether this striatal compartment specialization extends to cognition.

In the n-back task, striosome-like voxels preferentially activated during preparatory/cue-related demands, whereas matrix-like voxels were more engaged during on-task maintenance and response execution (0-back/2-back blocks). This echoes our motor-task results, where striosome-like voxels were preferentially active during motor preparation and matrix-like voxels during movement execution (Sadiq and Waugh, 2025), and complements our resting-state finding that the two compartments occupy distinct intrinsic functional networks (Sadiq et al., 2025). These converging observations indicate that compartment-based specialization is expressed both at rest and dynamically during behavior. This framework suggests that imbalances in compartment-specific engagement may contribute to motor and cognitive symptoms in neuropsychiatric disease (Tippett et al., 2007; Crittenden and Graybiel, 2011; Tieu and Waugh, 2025; Waugh et al., 2025, 2026). In this context, future work could examine the correspondence between whole-brain cue- and task-related activation patterns and the cortical regions used to define striosome- and matrix-like voxels. Such analyses may help further clarify how compartment-specific organization relates to large-scale functional networks.

It is important to acknowledge several limitations of this study. Our method identified striosome- and matrix-like voxels indirectly, based on biases in structural connectivity derived from probabilistic tractography. Consequently, the validity of our functional analyses is contingent upon the anatomical precision of these parcellations. Previous work has shown that our approach to striatal parcellation is highly reliable, with a test-retest error rate of 0.14% (Waugh et al., 2022), and that compartment-like biases are highly spatially specific: shifting voxel locations by only a few millimeters eliminates all compartment-like bias in both structural (Funk et al., 2024; Sadiq et al., 2025; Sadiq and Waugh, 2025) connectivity. Furthermore, these connectivity-based parcellations reproduce key anatomical features observed in human and animal tissue, including the relative abundance, spatial distribution, contiguity, and extra-striatal connectivity of striosome and matrix (Waugh et al., 2022; Funk et al., 2023, 2024). Nonetheless, this technique cannot substitute for direct histological identification of striosome and matrix. While voxels parcellated through differential connectivity share all of the anatomic properties of striosome and matrix identified in tissue, the extent to which compartment-like voxels match the location of the underlying tissue compartments remains untested.

Another limitation concerns the spatial resolution of diffusion MRI. Although the voxel size used in this study (1.25 mm isotropic) approximates the diameter of the largest human striosome branches (Graybiel and Ragsdale, 1978; Holt et al., 1997), each striosome-like voxel inevitably contains some proportion of matrix tissue. Nevertheless, our previously reported validation studies demonstrated that compartment-like voxels reproduce the anatomical features of striosome and matrix identified through immunohistochemistry, even when derived from diffusion datasets with lower resolution than employed here (Waugh et al., 2022; Funk et al., 2023, 2024). The fMRI dataset used here has a comparable resolution to those prior diffusion MRI datasets (2 mm isotropic), and we have shown that this resolution is sufficient to detect robust, widespread compartment-specific patterns of functional connectivity (Sadiq et al., 2025; Sadiq and Waugh, 2025). In addition, transformation of the diffusion-derived compartment masks from 1.25-mm diffusion space to the 2-mm functional MRI space necessarily required interpolation during image registration. This resampling may introduce spatial blurring and partial-volume effects, thereby reducing the spatial specificity of the compartment-like masks. Because such effects would tend to increase signal mixing between neighboring striosome- and matrix-like voxels, resampling would be expected to reduce, rather than exaggerate, compartment-specific differences. Still, it is important to emphasize that these non-invasive, inferential methods cannot achieve the fine-grained histological detail available in post-mortem tissue. This limitation applies broadly across all current in vivo anatomical mapping approaches in humans.

Since the architecture and precise location of the striosome differs between individuals, standard region-of-interest approaches are inadequate to assess the striatal compartments in vivo. Therefore, we quantified mean activation within striosome-like and matrix-like masks that were uniquely generated for each subject, rather than performing voxelwise activation analyses. Although averaging within compartments facilitated meaningful group-level comparisons, this method may have obscured more spatially restricted activation patterns. Given that corticostriatal projections are somatotopically organized (Flaherty and Graybiel, 1993; Waugh et al., 2022; Sadiq et al., 2025; Sadiq and Waugh, 2025), it is possible that the functional specializations we observed are confined to specific subregions within each compartment, and do not generalize to the whole striosome, or the whole matrix.

It is also important to acknowledge potential distortions introduced by our experimental design in assessing functional connectivity. Histological studies consistently show that the matrix occupies roughly six times more volume than the striosome (Desban et al., 1993; Holt et al., 1997). To avoid size-related biases in probabilistic connectivity analyses, we generated equal-volume masks, which required assessing only a subset of matrix-like voxels. Consequently, matrix-like voxels outside the most strongly biased subset may exhibit functional activation patterns not captured in our analysis. While this natural volume asymmetry between striosome and matrix provides essential anatomical context, it does not alter the robustness of the functional activation patterns observed in the present study.

These experiments revealed a clear functional dissociation between striosome-like and matrix-like voxels during the cue period. Across both 0-back and 2-back tasks, striosome-like voxels consistently showed stronger activation than matrix-like voxels, suggesting that the striosome may play a specialized role in anticipatory processes related to working memory. Such processes may include attentional orienting, evaluating potential outcomes, and assessing the motivational or emotional significance of upcoming events, all of which align with functions previously attributed to the striosome (Friedman et al., 2017; Karunakaran et al., 2021). Importantly, this compartment bias was evident across multiple stimulus categories (body part, face, place, and tool), highlighting the domain-general nature of striosome-like engagement during task preparation. In contrast, matrix-like voxels showed lower cue-related activity, consistent with their stronger association with execution phases in motor tasks. Together, these findings suggest that during working memory, striosome-like voxels preferentially support cue-dependent processes that establish task context and readiness, while matrix-like voxels may contribute more prominently during subsequent response implementation.

Interestingly, we observed a striking reversal of the compartmental dissociation seen during cue processing. During the active task blocks, matrix-like voxels exhibited significantly greater activation than striosome-like voxels across nearly all stimulus categories, with especially robust differences for body part, face, and tool conditions. This reversal in relative amplitude indicates that matrix-like voxels are preferentially engaged once participants transition from preparatory to execution phases, consistent with the proposed role for the matrix in sustaining online processing demands (Flaherty and Graybiel, 1993). By contrast, striosome-like voxels showed relatively modest activity during task blocks, supporting the interpretation that their primary contribution lies in anticipatory or evaluative processes rather than in ongoing task execution.

Beyond these phase-dependent dissociations, cognitive load systematically modulated the magnitude of striatal responses across task phases. The opposite scaling of activation across task phases appears to reflect how cognitive effort is distributed over time rather than a change in the roles of striosome- and matrix-like compartments. In the 0-back condition, the task is simple and the cue clearly signals what response will be required. As a result, the cue itself carries substantial information and evokes strong preparatory responses. In contrast, during the 2-back condition maintenance of the presented stimulus is a persistent task demand. Since activation is continuously present and less dependent on the brief cue block, the cue may encode less information, leading to reduced cue-related activation. Concurrently, activation during task execution increased with higher load, consistent with the greater demands of maintaining information, selecting responses, and monitoring performance during the 2-back task. When directly contrasting 2-back with 0-back conditions, both compartments demonstrated load-dependent increases in activation, with the magnitude of this effect differing by compartment and stimulus category (Fig. 4). Overall, these findings are consistent with greater matrix-like engagement during sustained task execution. Although prior anatomical studies have proposed roles for the matrix in encoding, manipulation, and execution of processes based on its extensive associative and sensorimotor connectivity (Flaherty and Graybiel, 1993), these functions have not been directly demonstrated in humans or animals. Together, these results highlight a complementary division of labor: striosome-like voxels appear specialized for anticipatory and evaluative roles, while matrix-like voxels dominate the sustained processing required for ongoing and/or demanding memory and motor tasks (Okunomiya et al., 2025; Sadiq and Waugh, 2025).

We next examined how compartmental activation differed between correct and error trials. Both striosome- and matrix-like voxels showed reduced activation during error relative to correct 2-back trials. The numerical reduction was larger in striosome-like voxels (71%) than in matrix-like voxels (29%), suggesting a possible compartment-specific relationship with performance accuracy. However, the interaction of Compartment × Accuracy was not significant after covariate adjustment. Thus, the present findings do not establish a striosome-specific effect of accuracy, but they raise the possibility that compartment-specific relationships with performance may exist. A future study designed specifically to assess accuracy, with sufficient error trials and appropriate control of behavioral covariates, is needed to directly test this possibility. We also examined whether compartment-specific responses were driven by stimulus categories. In the 0-back condition, category-selectivity was most apparent in matrix-like voxels, which showed stronger activation for body part and face stimuli, while striosome-like activity remained relatively undifferentiated. This suggests that under low working memory demand, the matrix may be more sensitive to content-specific inputs that guide immediate responses. In contrast, under the 2-back condition, matrix activity became uniformly elevated across categories, consistent with its role in supporting sustained, domain-general task execution. Striosome-like voxels, however, displayed category-selective modulation at higher load, with significant enhancements for place stimuli and reduced activation for body part stimuli. These results indicate that the contributions of striosome-like voxels may become increasingly specialized under cognitive challenge, reflecting sensitivity to motivationally or contextually salient categories (Beck and Friedman, 2026), while matrix responses shift toward generalized support of task performance. However, the slice-wise sensitivity analysis demonstrated that several place-related contrasts were less consistently preserved than the corresponding body- and face-related effects. Accordingly, the place-specific findings should be interpreted with greater caution, as they appear to be more sensitive to the precise spatial definition of matrix-like voxels.

In conclusion, this study provides the first evidence that in humans, the striatal compartments have specialized roles in working memory that vary with effort and the type of stimulus. Whereas our prior resting-state analyses demonstrated that these compartment-like voxels participate in segregated intrinsic functional networks (Sadiq et al., 2025), the current findings extend this framework by showing that compartmental organization also shapes activation dynamics under varying cognitive demands. Specifically, striosome-like voxels were preferentially engaged during cue-related processes, while matrix-like voxels dominated during sustained task execution and scaled robustly with increasing working memory load. Notably, this cue–execution dissociation parallels our prior motor task findings (Sadiq and Waugh, 2025) suggesting that compartmental specialization may reflect a general organizing principle of striatal function across both motor and cognitive domains. These results support the emerging view that striatal microarchitecture contributes to phase-specific and demand-dependent specialization, with striosome- and matrix-like voxels differentially engaged across anticipation, execution, and performance accuracy. This compartmental specialization aligns with animal studies demonstrating that striosome and matrix circuits mediate distinct phases of motivated behavior. For instance, Friedman et al. (2015) showed that corticostriatal projections to striosome are selectively recruited during decision-making under conflict, whereas matrix circuits support ongoing action execution (Friedman et al., 2015). Future studies will be necessary to link these patterns of functional activation to underlying cellular and network mechanisms, and to examine their generalizability across additional cognitive domains and in populations with neuropsychiatric diseases.

Supplementary Material

NIHMS2216064-supplement-Supplementary_Material.docx^{(280.4KB, docx)}

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.neuroimage.2026.122212.

Funding

Dr. Waugh was supported by: the CTSA Pilot Award; the Elterman Family Foundation; NINDS grant 1K23NS124978-01A; the Brain and Behavior Research Foundation Young Investigator Award; and the Children’s Health CCRAC Early Career Award. The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of these funding agencies.

Footnotes

Declaration of competing interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.

Data availability statement

Publicly available datasets were analyzed in this study. HCP data can be found here: https://www.humanconnectome.org/study/hcp-young-adult/document/1200-subjects-data-release. This dataset is BIDS compliant. The code, bait, seed, and exclusion masks necessary to complete striatal parcellation can be accessed here: github.com/jeff-waugh/Striatal-Connectivity-based-Parcellation.

References

Albin RL, 2006. Neurobiology of basal ganglia and Tourette syndrome: striatal and dopamine function. Adv. Neurol. 99, 99. -NEW YORK-RAVEN PRESS. [PubMed] [Google Scholar]
Aosaki T, Graybiel AM, Kimura M, 1994. Effect of the nigrostriatal dopamine system on acquired neural responses in the striatum of behaving monkeys. Science 265 (5170), 412–415 (1979). [DOI] [PubMed] [Google Scholar]
Beck DW, Friedman A, 2026. Prediction error correlates in the striosome-dopamine circuit emerge from information gain. Nat. Commun. 17 (1), 1–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
Bejjani B-P, Damier P, Arnulf I, Thivard L, Bonnet A-M, Dormont D, Cornu P, Pidoux B, Samson Y, Agid Y, 1999. Transient acute depression induced by high-frequency deep-brain stimulation. N. Engl. J. Med. 340 (19), 1476–1480. [DOI] [PubMed] [Google Scholar]
Benjamini Y, Hochberg Y, 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 57 (1), 289–300 (Methodological). [Google Scholar]
Brimblecombe KR, Cragg SJ, 2017. The striosome and matrix compartments of the striatum: a path through the labyrinth from neurochemistry toward function. ACS Chem. Neurosci. 8 (2), 235–242. [DOI] [PubMed] [Google Scholar]
Burguiere E, Monteiro P, Mallet L, Feng G, Graybiel AM, 2015. Striatal circuits, habits, and implications for obsessive–compulsive disorder. Curr. Opin. Neurobiol. 30, 59–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
Crittenden JR, Graybiel AM, 2011. Basal Ganglia disorders associated with imbalances in the striatal striosome and matrix compartments. Front. Neuroanat. 5, 59. [DOI] [PMC free article] [PubMed] [Google Scholar]
Crittenden JR, Tillberg PW, Riad MH, Shima Y, Gerfen CR, Curry J, Housman DE, Nelson SB, Boyden ES, Graybiel AM, 2016. Striosome–dendron bouquets highlight a unique striatonigral circuit targeting dopamine-containing neurons. Proc. Natl. Acad. Sci. USA 113 (40), 11318–11323. [DOI] [PMC free article] [PubMed] [Google Scholar]
Desban M, Kemel M, Glowinski J, Gauchy C, 1993. Spatial organization of patch and matrix compartments in the rat striatum. Neuroscience 57 (3), 661–671. [DOI] [PubMed] [Google Scholar]
Donoghue JP, Herkenham M, 1986. Neostriatal projections from individual cortical fields conform to histochemically distinct striatal compartments in the rat. Brain Res. 365 (2), 397–403. [DOI] [PubMed] [Google Scholar]
Dunlop BW, Nemeroff CB, 2007. The role of dopamine in the pathophysiology of depression. Arch. Gen. Psychiatry 64 (3), 327. [DOI] [PubMed] [Google Scholar]
Eblen F, Graybiel AM, 1995. Highly restricted origin of prefrontal cortical inputs to striosomes in the macaque monkey. J. Neurosci. 15 (9), 5999–6013. [DOI] [PMC free article] [PubMed] [Google Scholar]
Flaherty AW, Graybiel AM, 1993. Two input systems for body representations in the primate striatal matrix: experimental evidence in the squirrel monkey. J. Neurosci. 13 (3), 1120–1137. [DOI] [PMC free article] [PubMed] [Google Scholar]
Friedman A, Homma D, Gibb LG, Amemori K. i., Rubin SJ, Hood AS, Riad MH, Graybiel AM, 2015. A corticostriatal path targeting striosomes controls decision-making under conflict. Cell 161 (6), 1320–1333. [DOI] [PMC free article] [PubMed] [Google Scholar]
Friedman A, Homma D, Bloem B, Gibb LG, Amemori K. i., Hu D, Delcasso S, Truong TF, Yang J, Hood AS, 2017. Chronic stress alters striosome-circuit dynamics, leading to aberrant decision-making. Cell 171 (5), e1128. [DOI] [PMC free article] [PubMed] [Google Scholar]
Funk AT, Waugh JL, 2026. Striosome-like structural connectivity is reduced in the rostral putamen in anxiety disorder. Depress. Anxiety 2026, 7682055. [DOI] [PMC free article] [PubMed] [Google Scholar]
Funk AT, Hassan AA, Brüggemann N, Sharma N, Breiter HC, Blood AJ, Waugh JL, 2023. In humans, striato-pallido-thalamic projections are largely segregated by their origin in either the striosome-like or matrix-like compartments. Front. Neurosci. 17, 1178473. [DOI] [PMC free article] [PubMed] [Google Scholar]
Funk AT, Hassan AA, Waugh JL, 2024. In humans, insulo-striate structural connectivity is largely biased toward either striosome-like or matrix-like striatal compartments. Neurosci. Insights 19, 26331055241268079. [DOI] [PMC free article] [PubMed] [Google Scholar]
Glasser MF, Sotiropoulos SN, Wilson JA, Coalson TS, Fischl B, Andersson JL, Xu J, Jbabdi S, Webster M, Polimeni JR, 2013. The minimal preprocessing pipelines for the human connectome project. Neuroimage 80, 105–124. [DOI] [PMC free article] [PubMed] [Google Scholar]
Goldman-Rakic PS, 1982. Cytoarchitectonic heterogeneity of the primate neostriatum: subdivision into island and matrix cellular compartments. J. Comp. Neurol. 205 (4), 398–413. [DOI] [PubMed] [Google Scholar]
Graybiel AM, Ragsdale CW Jr, 1978. Histochemically distinct compartments in the striatum of human, monkeys, and cat demonstrated by acetylthiocholinesterase staining. Proc. Natl. Acad. Sci. USA 75 (11), 5723–5726. [DOI] [PMC free article] [PubMed] [Google Scholar]
Graybiel AM, 1997. The basal ganglia and cognitive pattern generators. Schizophr. Bull. 23 (3), 459–469. [DOI] [PubMed] [Google Scholar]
Graybiel AM, 2008. Habits, rituals, and the evaluative brain. Annu. Rev. Neurosci. 31 (1), 359–387. [DOI] [PubMed] [Google Scholar]
Haber SN, Fudge JL, McFarland NR, 2000. Striatonigrostriatal pathways in primates form an ascending spiral from the shell to the dorsolateral striatum. J. Neurosci. 20 (6), 2369–2382. [DOI] [PMC free article] [PubMed] [Google Scholar]
Holt DJ, Graybiel AM, Saper CB, 1997. Neurochemical architecture of the human striatum. J. Comp. Neurol. 384 (1), 1–25. [DOI] [PubMed] [Google Scholar]
Jenrette TA, Logue JB, Horner KA, 2019. Lesions of the patch compartment of dorsolateral striatum disrupt stimulus–response learning. Neuroscience 415, 161–172. [DOI] [PMC free article] [PubMed] [Google Scholar]
Johnston JG, Gerfen CR, Haber SN, van der Kooy D, 1990. Mechanisms of striatal pattern formation: conservation of mammalian compartmentalization. Dev. Brain Res. 57 (1), 93–102. [DOI] [PubMed] [Google Scholar]
Karunakaran KB, Amemori S, Balakrishnan N, Ganapathiraju MK, Amemori KI, 2021. Generalized and social anxiety disorder interactomes show distinctive overlaps with striosome and matrix interactomes. Sci. Rep. 11 (1), 18392. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lévesque M, Parent A, 2005. The striatofugal fiber system in primates: a reevaluation of its organization based on single-axon tracing studies. Proc. Natl. Acad. Sci. USA 102 (33), 11888–11893. [DOI] [PMC free article] [PubMed] [Google Scholar]
Middleton FA, Strick PL, 2000. Basal ganglia output and cognition: evidence from anatomical, behavioral, and clinical studies. Brain Cogn. 42 (2), 183–200. [DOI] [PubMed] [Google Scholar]
Mink JW, 1996. The basal ganglia: focused selection and inhibition of competing motor programs. Prog. Neurobiol. 50 (4), 381–425. [DOI] [PubMed] [Google Scholar]
Mumford JA, Turner BO, Ashby FG, Poldrack RA, 2012. Deconvolving BOLD activation in event-related designs for multivoxel pattern classification analyses. Neuroimage 59 (3), 2636–2643. [DOI] [PMC free article] [PubMed] [Google Scholar]
Nadel JA, Pawelko SS, Copes-Finke D, Neidhart M, Howard CD, 2020. Lesion of striatal patches disrupts habitual behaviors and increases behavioral variability. PLoS One 15 (1), e0224715. [DOI] [PMC free article] [PubMed] [Google Scholar]
Nadel J, Pawelko S, Scott J, McLaughlin R, Fox M, Ghanem M, van der Merwe R, Hollon N, Ramsson E, Howard C, 2021. Optogenetic stimulation of striatal patches modifies habit formation and inhibits dopamine release. Sci. Rep. 11 (1), 19847. [DOI] [PMC free article] [PubMed] [Google Scholar]
Okunomiya T, Watanabe D, Banno H, Kondo T, Imamura K, Takahashi R, Inoue H, 2025. Striosome circuitry stimulation inhibits striatal dopamine release and locomotion. J. Neurosci. 45 (4), e0457242024. [DOI] [PMC free article] [PubMed] [Google Scholar]
Power JD, Barnes KA, Snyder AZ, Schlaggar BL, Petersen SE, 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. Neuroimage 59 (3), 2142–2154. 10.1016/j.neuroimage.2011.10.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
Prager EM, Dorman DB, Hobel ZB, Malgady JM, Blackwell KT, Plotkin JL, 2020. Dopamine oppositely modulates state transitions in striosome and matrix direct pathway striatal spiny neurons. Neuron 108 (6), 1091–1102.e5. 10.1016/j.neuron.2020.09.028. [DOI] [PMC free article] [PubMed] [Google Scholar]
Prince JS, Charest I, Kurzawski JW, Pyles JA, Tarr MJ, Kay KN, 2022. Improving the accuracy of single-trial fMRI response estimates using GLMsingle. Elife 11, e77599. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ragsdale CW Jr, Graybiel AM, 1990. A simple ordering of neocortical areas established by the compartmental organization of their striatal projections. Proc. Natl. Acad. Sci. USA 87 (16), 6196–6199. [DOI] [PMC free article] [PubMed] [Google Scholar]
Sadiq A and Waugh JL. (2025). “In humans, fMRI reveals that striosome-like and matrix-like striatal voxels are engaged in different phases of movement.” bioRxiv: 2025.2009.2001.673509. [DOI] [PMC free article] [PubMed] [Google Scholar]
Sadiq A, Funk AT, Waugh JL, 2025. The striatal compartments, striosome and matrix, are embedded in largely distinct resting-state functional networks. Front. Neural Circuits 19, 1514937. [DOI] [PMC free article] [PubMed] [Google Scholar]
Singh M, Skippen P, He J, Thomson P, Fuelscher I, Caeyenberghs K, Anderson V, Hyde C, Silk TJ, 2024. Developmental patterns of inhibition and fronto-basal-ganglia white matter organisation in healthy children and children with attention-deficit/hyperactivity disorder. Hum. Brain Mapp. 45 (15), e70010. [DOI] [PMC free article] [PubMed] [Google Scholar]
Surmeier DJ, Graves SM, Shen W, 2014. Dopaminergic modulation of striatal networks in health and Parkinson’s disease. Curr. Opin. Neurobiol. 29, 109–117. [DOI] [PMC free article] [PubMed] [Google Scholar]
Swerdlow NR, Koob GF, 1987. Dopamine, schizophrenia, mania, and depression: toward a unified hypothesis of cortico-striatopallido-thalamic function. Behav. Brain Sci. 10 (2), 197–208. [Google Scholar]
Tieu AN, Waugh JL, 2025. Depression-associated reductions in putaminal volume are accompanied by a shift from matrix-like to striosome-like structural connectivity. Front. Psychiatry 16, 1647240. [DOI] [PMC free article] [PubMed] [Google Scholar]
Tippett LJ, Waldvogel HJ, Thomas SJ, Hogg VM, Roon-Mom W.v., Synek BJ, Graybiel AM, Faull RL, 2007. Striosomes and mood dysfunction in Huntington’s disease. Brain 130 (1), 206–221. [DOI] [PubMed] [Google Scholar]
Turner BO, Mumford JA, Poldrack RA, Ashby FG, 2012. Spatiotemporal activity estimation for multivoxel pattern analysis with rapid event-related designs. Neuroimage 62 (3), 1429–1438. [DOI] [PMC free article] [PubMed] [Google Scholar]
Van Essen DC, Ugurbil K, Auerbach E, Barch D, Behrens TE, Bucholz R, Chang A, Chen L, Corbetta M, Curtiss SW, 2012. The human connectome project: a data acquisition perspective. Neuroimage 62 (4), 2222–2231. [DOI] [PMC free article] [PubMed] [Google Scholar]
Van Essen DC, Smith SM, Barch DM, Behrens TE, Yacoub E, Ugurbil K, Consortium W-MH, 2013. The WU-Minn human connectome project: an overview. Neuroimage 80, 62–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
Vogt C. (1911). Quelques considérations générales à propos du syndrome du corps strié. [Google Scholar]
Watabe-Uchida M, Zhu L, Ogawa SK, Vamanrao A, Uchida N, 2012. Whole-brain mapping of direct inputs to midbrain dopamine neurons. Neuron 74 (5), 858–873. [DOI] [PubMed] [Google Scholar]
Waugh JL, Hassan AA, Kuster JK, Levenstein JM, Warfield SK, Makris N, Brüggemann N, Sharma N, Breiter HC, Blood AJ, 2022. An MRI method for parcellating the human striatum into matrix and striosome compartments in vivo. Neuroimage 246, 118714. [DOI] [PMC free article] [PubMed] [Google Scholar]
Waugh JL, Hassan AO, Funk AT, Maldjian JA, 2025. The striatal matrix compartment is expanded in autism spectrum disorder. J. Neurodev. Disord. 17 (1), 8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Waugh JL, Tieu A, Price R, Stern E, 2026. A diffusion MRI-based structural assessment of the striatal compartments, striosome and matrix, in obsessive compulsive disorder. Res. Sq. [Google Scholar]
Weglage M, Wärnberg E, Lazaridis I, Calvigioni D, Tzortzi O, Meletis K, 2021. Complete representation of action space and value in all dorsal striatal pathways. Cell Rep. 36 (4), 109437. [DOI] [PubMed] [Google Scholar]
White NM, Hiroi N, 1998. Preferential localization of self-stimulation sites in striosomes/patches in the rat striatum. Proc. Natl. Acad. Sci. USA 95 (11), 6486–6491. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wichmann T, Dostrovsky JO, 2011. Pathological basal ganglia activity in movement disorders. Neuroscience 198, 232–244. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wilson SK, 1912. Progressive lenticular degeneration. Br. Med. J. 2 (2710), 1645. [Google Scholar]
Wilson SK, 1914. An experimental research into the anatomy and physiology of the corpus striatum. Brain 36 (3–4), 427–492. [Google Scholar]
Xiao X, Deng H, Furlan A, Yang T, Zhang X, Hwang G-R, Tucciarone J, Wu P, He M, Palaniswamy R, 2020. A genetically defined compartmentalized striatal direct pathway for negative reinforcement. Cell 183 (1), 211–227.e20 e220. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material

NIHMS2216064-supplement-Supplementary_Material.docx^{(280.4KB, docx)}

Data Availability Statement

[R1] Albin RL, 2006. Neurobiology of basal ganglia and Tourette syndrome: striatal and dopamine function. Adv. Neurol. 99, 99. -NEW YORK-RAVEN PRESS. [PubMed] [Google Scholar]

[R2] Aosaki T, Graybiel AM, Kimura M, 1994. Effect of the nigrostriatal dopamine system on acquired neural responses in the striatum of behaving monkeys. Science 265 (5170), 412–415 (1979). [DOI] [PubMed] [Google Scholar]

[R3] Beck DW, Friedman A, 2026. Prediction error correlates in the striosome-dopamine circuit emerge from information gain. Nat. Commun. 17 (1), 1–30. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] Bejjani B-P, Damier P, Arnulf I, Thivard L, Bonnet A-M, Dormont D, Cornu P, Pidoux B, Samson Y, Agid Y, 1999. Transient acute depression induced by high-frequency deep-brain stimulation. N. Engl. J. Med. 340 (19), 1476–1480. [DOI] [PubMed] [Google Scholar]

[R5] Benjamini Y, Hochberg Y, 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 57 (1), 289–300 (Methodological). [Google Scholar]

[R6] Brimblecombe KR, Cragg SJ, 2017. The striosome and matrix compartments of the striatum: a path through the labyrinth from neurochemistry toward function. ACS Chem. Neurosci. 8 (2), 235–242. [DOI] [PubMed] [Google Scholar]

[R7] Burguiere E, Monteiro P, Mallet L, Feng G, Graybiel AM, 2015. Striatal circuits, habits, and implications for obsessive–compulsive disorder. Curr. Opin. Neurobiol. 30, 59–65. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R8] Crittenden JR, Graybiel AM, 2011. Basal Ganglia disorders associated with imbalances in the striatal striosome and matrix compartments. Front. Neuroanat. 5, 59. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R9] Crittenden JR, Tillberg PW, Riad MH, Shima Y, Gerfen CR, Curry J, Housman DE, Nelson SB, Boyden ES, Graybiel AM, 2016. Striosome–dendron bouquets highlight a unique striatonigral circuit targeting dopamine-containing neurons. Proc. Natl. Acad. Sci. USA 113 (40), 11318–11323. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R10] Desban M, Kemel M, Glowinski J, Gauchy C, 1993. Spatial organization of patch and matrix compartments in the rat striatum. Neuroscience 57 (3), 661–671. [DOI] [PubMed] [Google Scholar]

[R11] Donoghue JP, Herkenham M, 1986. Neostriatal projections from individual cortical fields conform to histochemically distinct striatal compartments in the rat. Brain Res. 365 (2), 397–403. [DOI] [PubMed] [Google Scholar]

[R12] Dunlop BW, Nemeroff CB, 2007. The role of dopamine in the pathophysiology of depression. Arch. Gen. Psychiatry 64 (3), 327. [DOI] [PubMed] [Google Scholar]

[R13] Eblen F, Graybiel AM, 1995. Highly restricted origin of prefrontal cortical inputs to striosomes in the macaque monkey. J. Neurosci. 15 (9), 5999–6013. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R14] Flaherty AW, Graybiel AM, 1993. Two input systems for body representations in the primate striatal matrix: experimental evidence in the squirrel monkey. J. Neurosci. 13 (3), 1120–1137. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R15] Friedman A, Homma D, Gibb LG, Amemori K. i., Rubin SJ, Hood AS, Riad MH, Graybiel AM, 2015. A corticostriatal path targeting striosomes controls decision-making under conflict. Cell 161 (6), 1320–1333. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R16] Friedman A, Homma D, Bloem B, Gibb LG, Amemori K. i., Hu D, Delcasso S, Truong TF, Yang J, Hood AS, 2017. Chronic stress alters striosome-circuit dynamics, leading to aberrant decision-making. Cell 171 (5), e1128. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] Funk AT, Waugh JL, 2026. Striosome-like structural connectivity is reduced in the rostral putamen in anxiety disorder. Depress. Anxiety 2026, 7682055. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] Funk AT, Hassan AA, Brüggemann N, Sharma N, Breiter HC, Blood AJ, Waugh JL, 2023. In humans, striato-pallido-thalamic projections are largely segregated by their origin in either the striosome-like or matrix-like compartments. Front. Neurosci. 17, 1178473. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R19] Funk AT, Hassan AA, Waugh JL, 2024. In humans, insulo-striate structural connectivity is largely biased toward either striosome-like or matrix-like striatal compartments. Neurosci. Insights 19, 26331055241268079. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R20] Glasser MF, Sotiropoulos SN, Wilson JA, Coalson TS, Fischl B, Andersson JL, Xu J, Jbabdi S, Webster M, Polimeni JR, 2013. The minimal preprocessing pipelines for the human connectome project. Neuroimage 80, 105–124. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R21] Goldman-Rakic PS, 1982. Cytoarchitectonic heterogeneity of the primate neostriatum: subdivision into island and matrix cellular compartments. J. Comp. Neurol. 205 (4), 398–413. [DOI] [PubMed] [Google Scholar]

[R22] Graybiel AM, Ragsdale CW Jr, 1978. Histochemically distinct compartments in the striatum of human, monkeys, and cat demonstrated by acetylthiocholinesterase staining. Proc. Natl. Acad. Sci. USA 75 (11), 5723–5726. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R23] Graybiel AM, 1997. The basal ganglia and cognitive pattern generators. Schizophr. Bull. 23 (3), 459–469. [DOI] [PubMed] [Google Scholar]

[R24] Graybiel AM, 2008. Habits, rituals, and the evaluative brain. Annu. Rev. Neurosci. 31 (1), 359–387. [DOI] [PubMed] [Google Scholar]

[R25] Haber SN, Fudge JL, McFarland NR, 2000. Striatonigrostriatal pathways in primates form an ascending spiral from the shell to the dorsolateral striatum. J. Neurosci. 20 (6), 2369–2382. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R26] Holt DJ, Graybiel AM, Saper CB, 1997. Neurochemical architecture of the human striatum. J. Comp. Neurol. 384 (1), 1–25. [DOI] [PubMed] [Google Scholar]

[R27] Jenrette TA, Logue JB, Horner KA, 2019. Lesions of the patch compartment of dorsolateral striatum disrupt stimulus–response learning. Neuroscience 415, 161–172. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] Johnston JG, Gerfen CR, Haber SN, van der Kooy D, 1990. Mechanisms of striatal pattern formation: conservation of mammalian compartmentalization. Dev. Brain Res. 57 (1), 93–102. [DOI] [PubMed] [Google Scholar]

[R29] Karunakaran KB, Amemori S, Balakrishnan N, Ganapathiraju MK, Amemori KI, 2021. Generalized and social anxiety disorder interactomes show distinctive overlaps with striosome and matrix interactomes. Sci. Rep. 11 (1), 18392. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] Lévesque M, Parent A, 2005. The striatofugal fiber system in primates: a reevaluation of its organization based on single-axon tracing studies. Proc. Natl. Acad. Sci. USA 102 (33), 11888–11893. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R31] Middleton FA, Strick PL, 2000. Basal ganglia output and cognition: evidence from anatomical, behavioral, and clinical studies. Brain Cogn. 42 (2), 183–200. [DOI] [PubMed] [Google Scholar]

[R32] Mink JW, 1996. The basal ganglia: focused selection and inhibition of competing motor programs. Prog. Neurobiol. 50 (4), 381–425. [DOI] [PubMed] [Google Scholar]

[R33] Mumford JA, Turner BO, Ashby FG, Poldrack RA, 2012. Deconvolving BOLD activation in event-related designs for multivoxel pattern classification analyses. Neuroimage 59 (3), 2636–2643. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R34] Nadel JA, Pawelko SS, Copes-Finke D, Neidhart M, Howard CD, 2020. Lesion of striatal patches disrupts habitual behaviors and increases behavioral variability. PLoS One 15 (1), e0224715. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R35] Nadel J, Pawelko S, Scott J, McLaughlin R, Fox M, Ghanem M, van der Merwe R, Hollon N, Ramsson E, Howard C, 2021. Optogenetic stimulation of striatal patches modifies habit formation and inhibits dopamine release. Sci. Rep. 11 (1), 19847. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] Okunomiya T, Watanabe D, Banno H, Kondo T, Imamura K, Takahashi R, Inoue H, 2025. Striosome circuitry stimulation inhibits striatal dopamine release and locomotion. J. Neurosci. 45 (4), e0457242024. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R37] Power JD, Barnes KA, Snyder AZ, Schlaggar BL, Petersen SE, 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. Neuroimage 59 (3), 2142–2154. 10.1016/j.neuroimage.2011.10.018. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R38] Prager EM, Dorman DB, Hobel ZB, Malgady JM, Blackwell KT, Plotkin JL, 2020. Dopamine oppositely modulates state transitions in striosome and matrix direct pathway striatal spiny neurons. Neuron 108 (6), 1091–1102.e5. 10.1016/j.neuron.2020.09.028. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R39] Prince JS, Charest I, Kurzawski JW, Pyles JA, Tarr MJ, Kay KN, 2022. Improving the accuracy of single-trial fMRI response estimates using GLMsingle. Elife 11, e77599. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R40] Ragsdale CW Jr, Graybiel AM, 1990. A simple ordering of neocortical areas established by the compartmental organization of their striatal projections. Proc. Natl. Acad. Sci. USA 87 (16), 6196–6199. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R41] Sadiq A and Waugh JL. (2025). “In humans, fMRI reveals that striosome-like and matrix-like striatal voxels are engaged in different phases of movement.” bioRxiv: 2025.2009.2001.673509. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R42] Sadiq A, Funk AT, Waugh JL, 2025. The striatal compartments, striosome and matrix, are embedded in largely distinct resting-state functional networks. Front. Neural Circuits 19, 1514937. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R43] Singh M, Skippen P, He J, Thomson P, Fuelscher I, Caeyenberghs K, Anderson V, Hyde C, Silk TJ, 2024. Developmental patterns of inhibition and fronto-basal-ganglia white matter organisation in healthy children and children with attention-deficit/hyperactivity disorder. Hum. Brain Mapp. 45 (15), e70010. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R44] Surmeier DJ, Graves SM, Shen W, 2014. Dopaminergic modulation of striatal networks in health and Parkinson’s disease. Curr. Opin. Neurobiol. 29, 109–117. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R45] Swerdlow NR, Koob GF, 1987. Dopamine, schizophrenia, mania, and depression: toward a unified hypothesis of cortico-striatopallido-thalamic function. Behav. Brain Sci. 10 (2), 197–208. [Google Scholar]

[R46] Tieu AN, Waugh JL, 2025. Depression-associated reductions in putaminal volume are accompanied by a shift from matrix-like to striosome-like structural connectivity. Front. Psychiatry 16, 1647240. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R47] Tippett LJ, Waldvogel HJ, Thomas SJ, Hogg VM, Roon-Mom W.v., Synek BJ, Graybiel AM, Faull RL, 2007. Striosomes and mood dysfunction in Huntington’s disease. Brain 130 (1), 206–221. [DOI] [PubMed] [Google Scholar]

[R48] Turner BO, Mumford JA, Poldrack RA, Ashby FG, 2012. Spatiotemporal activity estimation for multivoxel pattern analysis with rapid event-related designs. Neuroimage 62 (3), 1429–1438. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R49] Van Essen DC, Ugurbil K, Auerbach E, Barch D, Behrens TE, Bucholz R, Chang A, Chen L, Corbetta M, Curtiss SW, 2012. The human connectome project: a data acquisition perspective. Neuroimage 62 (4), 2222–2231. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R50] Van Essen DC, Smith SM, Barch DM, Behrens TE, Yacoub E, Ugurbil K, Consortium W-MH, 2013. The WU-Minn human connectome project: an overview. Neuroimage 80, 62–79. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R51] Vogt C. (1911). Quelques considérations générales à propos du syndrome du corps strié. [Google Scholar]

[R52] Watabe-Uchida M, Zhu L, Ogawa SK, Vamanrao A, Uchida N, 2012. Whole-brain mapping of direct inputs to midbrain dopamine neurons. Neuron 74 (5), 858–873. [DOI] [PubMed] [Google Scholar]

[R53] Waugh JL, Hassan AA, Kuster JK, Levenstein JM, Warfield SK, Makris N, Brüggemann N, Sharma N, Breiter HC, Blood AJ, 2022. An MRI method for parcellating the human striatum into matrix and striosome compartments in vivo. Neuroimage 246, 118714. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R54] Waugh JL, Hassan AO, Funk AT, Maldjian JA, 2025. The striatal matrix compartment is expanded in autism spectrum disorder. J. Neurodev. Disord. 17 (1), 8. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R55] Waugh JL, Tieu A, Price R, Stern E, 2026. A diffusion MRI-based structural assessment of the striatal compartments, striosome and matrix, in obsessive compulsive disorder. Res. Sq. [Google Scholar]

[R56] Weglage M, Wärnberg E, Lazaridis I, Calvigioni D, Tzortzi O, Meletis K, 2021. Complete representation of action space and value in all dorsal striatal pathways. Cell Rep. 36 (4), 109437. [DOI] [PubMed] [Google Scholar]

[R57] White NM, Hiroi N, 1998. Preferential localization of self-stimulation sites in striosomes/patches in the rat striatum. Proc. Natl. Acad. Sci. USA 95 (11), 6486–6491. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R58] Wichmann T, Dostrovsky JO, 2011. Pathological basal ganglia activity in movement disorders. Neuroscience 198, 232–244. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R59] Wilson SK, 1912. Progressive lenticular degeneration. Br. Med. J. 2 (2710), 1645. [Google Scholar]

[R60] Wilson SK, 1914. An experimental research into the anatomy and physiology of the corpus striatum. Brain 36 (3–4), 427–492. [Google Scholar]

[R61] Xiao X, Deng H, Furlan A, Yang T, Zhang X, Hwang G-R, Tucciarone J, Wu P, He M, Palaniswamy R, 2020. A genetically defined compartmentalized striatal direct pathway for negative reinforcement. Cell 183 (1), 211–227.e20 e220. [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Dissociation of striosome and matrix activation in the human striatum during the cue and execution phases of working memory: A probabilistic tractography and task fMRI assessment

Alishba Sadiq

Jeff L Waugh

Roles

Abstract

1. Introduction

2. Materials and methods

Fig. 1.

2.1. Study population

2.2. MRI acquisition protocols

2.2.1. Working memory-task design

2.3. fMRI preprocessing

2.4. Striatal parcellation

2.5. Validation of striatal parcellation

2.6. Analysis of task-based functional MRI (tfMRI)

2.7. Slice-wise voxel sampling for sensitivity analysis

2.8. Statistical analysis

3. Results

3.1. Validation of cue–execution regressor separability and sensitivity analysis accounting for family relatedness

3.2. Comparing MRI-parcellated voxels to striosome and matrix in tissue

3.3. Cue-evoked compartment-specific activation in 0-back and 2-back tasks

Fig. 2.

3.4. Compartment-specific activation patterns during working-memory tasks

Fig. 3.

3.5. Sensitivity analysis using less-biased voxels

3.6. Impact of specific voxel location on striatal compartmentalization

Table 1.

3.7. Activation changes from 0-back to 2-back within striatal compartments

Fig. 4.

3.8. Compartment-specific activation in correct and error trials

Fig. 5.

3.9. Category-specific activation during 0-back and 2-back tasks in striosome-like and matrix-like voxels

Fig. 6.

4. Discussion

Supplementary Material

Funding

Footnotes

Data availability statement

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases