White matter microstructure predicts effort and reward sensitivity

White matter microstructure predicts effort and reward sensitivity | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results White matter microstructure predicts effort and reward sensitivity View ORCID Profile Nam Trinh , View ORCID Profile Laurence Dricot , View ORCID Profile Pierre Vassiliadis , View ORCID Profile Quentin Dessain , View ORCID Profile Julie Duque , View ORCID Profile Tomas Ward , View ORCID Profile Gerard Derosiere doi: https://doi.org/10.1101/2025.08.19.671080 Nam Trinh 1 Centre for Research Training in Machine Learning (ML-Labs), School of Computing, Dublin City University , D09V209, Dublin, Ireland 2 Insight Research Ireland Centre for Data Analytics, School of Computing, Dublin City University , D09V209, Dublin, Ireland 3 Université Claude Bernard Lyon 1 , CNRS, INSERM, Centre de Recherche en Neurosciences de Lyon (CRNL) , U1028 UMR5292, Impact Team, F-69500, Bron, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Nam Trinh Laurence Dricot 4 Institute of Neuroscience, Université Catholique de Louvain , 1200, Brussels, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Laurence Dricot Pierre Vassiliadis 5 Max Planck University College London Centre for Computational Psychiatry and Ageing Research, University College London , London, UK 6 Department of Brain Sciences, Imperial College London , London, UK Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Pierre Vassiliadis Quentin Dessain 4 Institute of Neuroscience, Université Catholique de Louvain , 1200, Brussels, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Quentin Dessain Julie Duque 4 Institute of Neuroscience, Université Catholique de Louvain , 1200, Brussels, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Julie Duque Tomas Ward 2 Insight Research Ireland Centre for Data Analytics, School of Computing, Dublin City University , D09V209, Dublin, Ireland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Tomas Ward Gerard Derosiere 3 Université Claude Bernard Lyon 1 , CNRS, INSERM, Centre de Recherche en Neurosciences de Lyon (CRNL) , U1028 UMR5292, Impact Team, F-69500, Bron, France 4 Institute of Neuroscience, Université Catholique de Louvain , 1200, Brussels, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Gerard Derosiere For correspondence: gerard.derosiere{at}inserm.fr Abstract Full Text Info/History Metrics Supplementary material Preview PDF ABSTRACT From rodents to humans, animals constantly face a central question: is the reward worth the effort? Effort and reward sensitivity in such situations vary substantially across individuals and ultimately shape goal-directed behavior. Yet, the brain mechanisms underlying this variability across individuals remain unclear. Here, we combined computational modeling of effort and reward sensitivity during decision-making with whole-brain diffusion MRI in 45 healthy participants to identify white matter substrates of individual sensitivity. A data-driven, cluster-based analysis of fractional anisotropy and mean diffusivity revealed 12 clusters: five linked to effort sensitivity, all within tracts connected to major frontal valuation nodes (e.g., supplementary motor area [SMA], dorsal anterior cingulate cortex [dACC], orbitofrontal cortex [OFC]), and seven linked to reward sensitivity, spanning frontal valuation, fronto-parietal, and sensorimotor networks. The strongest associations involved two SMA-connected clusters, one shared across effort and reward sensitivity and another consistent across both microstructural metrics. Critically, microstructural features from the five effort-related and seven reward-related clusters reliably predicted individual effort and reward sensitivity in out-of-sample machine learning analyses, respectively, whereas randomly sampled clusters did not. SMA-connected tracts were the dominant predictors in these decoding analyses, with additional contributions from fronto-parietal and sensorimotor pathways for reward sensitivity. These findings reveal a distributed white matter architecture underlying inter-individual differences in effort and reward sensitivity, with SMA pathways emerging as central hubs. They demonstrate that localized white matter microstructure can robustly predict these individual differences, offering a framework to forecast the impact of lesions or interventions on goal-directed behavior, including apathy and impulsivity. SIGNIFICANCE STATEMENT Why do some people give up easily when faced with high effort demands, while others persist even when rewards are small? Such differences in effort and reward sensitivity shape goal-directed behavior, yet their neural basis is unclear. Using diffusion MRI and computational modeling, we show that white matter microstructure in specific pathways reliably predicts individual differences in these sensitivities. Tracts connected to the supplementary motor area emerged as central hubs, with additional contributions from fronto-parietal and sensorimotor networks. These results demonstrate that variability in effort and reward sensitivity is rooted not only in brain activity but also in structural connectivity, providing a framework to anticipate how white matter lesions or interventions may alter goal-directed behavior, including apathy and impulsivity. INTRODUCTION From rodents to humans, animals constantly face a central question: is the reward worth the effort? Everyday choices – doing chores or watching TV, cooking or ordering fast food – hinge on evaluating effort versus reward ( Hogan et al., 2020 ). The drive to act in these situations depends on one’s sensitivity to each. Abnormal sensitivities disrupt goal-directed behavior: effort hypersensitivity leads to inaction despite potential rewards, contributing to apathy ( Husain and Roiser, 2018 ; Costello et al., 2024 ), while reward hypersensitivity can drive action despite effort costs, fuelling impulsivity ( Long et al., 2022 ; Luijten et al., 2017 ). Even in health, sensitivities vary widely ( Bonnelle et al., 2016 ; Fuentes-Claramonte et al., 2016 ), shaping goal-directed behavior. These inter-individual differences – sometimes termed “computational phenotypes” ( Pessiglione et al., 2018 ; Schurr et al., 2024 ) – can be captured by computational models of decision-making. Yet, the brain mechanisms behind this variability remain unclear. Identifying their anatomical basis could explain this variability and enable prediction of effort and reward sensitivity from brain structure ( Thiebaut de Schotten and Forkel, 2022 ), offering a framework to forecast the impact of structural lesions on goal-directed behavior. So far, insights into effort and reward processing have come from functional neuroimaging studies tracking brain activity across varying effort and reward levels during decision-making ( Husain and Roiser, 2018 ). These highlight a core fronto- striatal network with partially dissociable roles: the supplementary motor area (SMA) tracks preferentially effort ( Bonnelle et al., 2016 ), the orbitofrontal cortex (OFC) tracks reward ( Klein-Flügge et al., 2022 ), while the dorsal anterior cingulate cortex (dACC; Shenhav et al., 2013 ) and nucleus accumbens (NAcc; Suzuki et al., 2021 ) encode both, integrating them into net value signals. Beyond this core, broader fronto-parietal ( Etzel et al., 2016 ) and sensorimotor structures – including the cerebellum ( Kostadinov and Häusser, 2022 ) and primary motor cortex (M1; Derosiere et al., 2025 ) – also show reward-related activity during decision-making. Together, these findings suggest that effort-reward choices rely on computations across distributed regions, with the fronto- striatal valuation network at the core. What remains poorly understood is how individual differences in effort and reward sensitivity emerge from variability across this distributed system. While most studies focus on gray matter activity, growing evidence shows white matter also actively modulate, rather than just relay, neural signals, shaping behavior ( Innocenti et al., 2022 ). Because effort and reward computations require coordination across many regions, structural connectivity may determine how these signals are processed ( Thiebaut de Schotten and Forkel, 2022 ). This hypothesis has been rarely tested, and mainly using tract-of-interest approaches, as in our own work ( Derosiere et al., 2024 ). For instance, subclinical apathy, linked to effort hypersensitivity, relates to anterior cingulum alterations, a tract connecting SMA, OFC and dACC ( Bonnelle et al., 2016 ). More recently, we found that heightened effort sensitivity associates with reduced SMA-NAcc macroscopic connectivity ( Derosiere et al., 2024 ). These findings suggest that white matter in tracts connecting fronto-striatal valuation regions shapes individual differences in effort and reward sensitivity. Still, given the wide distribution of value- related activity, whole-brain, data-driven approaches are now essential to determine the structural architecture underlying these individual differences. Advances in diffusion imaging can help bridge this gap. Voxel-wise microstructure metrics, fractional anisotropy (FA) and mean diffusivity (MD), capture complementary features of axonal organization: FA reflects alignment, density, and myelination, whereas MD indexes extracellular space and reduced barriers ( Beck et al., 2021 ; Song et al., 2025 ). Both are amenable to data-driven analyses, allowing identification of white matter loci with microstructural differences in one or both metrics. Here, we combined these analyses with computational modelling of effort and reward sensitivity during decision-making in 45 healthy participants to identify white matter regions whose microstructure covaries with, and predicts, sensitivity. A cluster-based approach unconstrained by anatomical priors revealed 12 clusters: 5 linked to effort sensitivity, all within tracts connected to frontal valuation nodes (SMA, dACC, OFC), and 7 linked to reward sensitivity, distributed across frontal valuation, fronto-parietal, and sensorimotor pathways. The strongest effects involved two SMA-connected clusters, one shared across effort and reward sensitivity and another consistent across FA and MD. Microstructural metrics from these clusters reliably predicted individual sensitivity in out-of-sample machine learning analyses, whereas randomly sampled clusters did not. Clusters in SMA-connected tracts were the dominant predictors in machine learning analyses, but fronto-parietal and sensorimotor pathways also contributed to the decoding of reward sensitivity. These findings reveal distributed white matter substrates for individual differences in effort and reward sensitivity and show that these sensitivities can be reliably decoded from localized microstructural features, offering a potential avenue to forecast the impact of structural lesions in patients with abnormal goal-directed behaviour, such as apathy and impulsivity. METHODS Participants Fifty healthy adult participants were initially recruited from the Research Participant Pool of the Institute of Neuroscience at the Université catholique de Louvain (Brussels, Belgium). Three participants did not complete the full experimental protocol, and tractography could not be performed in two subjects due to corrupted MRI data. The final sample comprised 45 individuals (mean age = 25.1 ± 0.8 years; 31 females, 14 males), whose data were included in all subsequent analyses. All subjects were right- handed based on the Edinburgh Questionnaire ( Oldfield, 1971 ) and had no history of neurological disorders, psychiatric illness, substance abuse, or use of medications that could affect performance. Participants received financial compensation for their involvement and could earn additional rewards based on task performance (see Task description section below). The study protocol was approved by the institutional review board of UCLouvain, and written informed consent was obtained from all participants. A previous publication using this dataset of 45 participants investigated the relationship between apathy, decision-making and structural connectivity with a tract- of-interest approach, focusing on a restricted set of motor-related circuits, together with effective connectivity measures derived from transcranial magnetic stimulation ( Derosiere, et al., 2024 ). That study relied on streamline-based tractography, which quantifies the reconstructed fiber tracts (streamlines) connecting two regions and reflects the macroscopic “size” or “strength” of a connection, but is largely insensitive to microstructural properties, and is focused on a limited set of tracts. In contrast, the present work addresses a distinct and complementary question by moving beyond predefined tracts to a whole-brain, data-driven framework. We leverage voxel-wise microstructural metrics (i.e., FA and MD) to capture complementary aspects of white matter integrity across the brain, and combine these with machine learning to identify the anatomical predictors of inter-individual variability in effort and reward sensitivity. Data acquisition The data were acquired in two sessions, conducted with a minimum of interval of 24 hours and a maximum of interval of one week. In the first session, MRI data were acquired at the Saint-Luc University Hospitals (Brussels, Belgium). The second session, conducted at the Institute of Neuroscience, UCLouvain (Brussels), included behavioral assessment using an effort-based decision-making task commonly used in the field and allowing computational modelling of effort and reward sensitivities ( Morris et al., 2025 ; Gilmour et al., 2023 ; Pessiglione et al., 2018 ). Detailed procedures for each session are described below. MRI data acquisition Both structural T1-weighted and diffusion-weighted MRI data were acquired for each participant on a 3 Tesla MRI (SIGNA TM Premier, General Electric), equipped with a 48-channel head coil ( Figure 1 .A). 3D T1-weighted anatomical images were obtained with the following parameters: Echo Time (TE) = 2.96 ms, Repetition Time (TR) = 2238.93 ms, Inversion Time TI = 900 ms, 170 slices, slice thickness = 1 mm, in-plane FOV = 256 × 256 mm², matrix = 256 × 256; voxel size = 1 mm 3 isotropic. Diffusion-weighted MRI (DWIs) were acquired in the axial plane with the parameters: TR = 7289 ms, TE = 57.1 ms, 70 slices, slice thickness = 2 mm, in-plane FOV = 220 × 220 mm², matrix size = 110 × 110; 2 mm isotropic voxels, with 64 gradients at b = 1000 s/mm² and one b0 reference image. Download figure Open in new tab Figure 1: Experimental protocol and data analyses. A. Diffusion MRI (dMRI) data acquisition. dMRI data were acquired to assess white matter microstructure, including FA and MD. B. Effort-based decision-making task and computational modelling of behavior. Participants completed an effort-based decision-making task, choosing whether to perform biceps contractions of varying effort levels (20%, 40%, 60%, or 80% of maximal voluntary contraction) to obtain varying monetary rewards (1, 5, 10, or 20 euro cents). Computational modelling of their acceptance rates in this task, yielded individual effort and reward sensitivity parameters (β Effort and β Reward , respectively). C. Whole-brain dMRI data analysis. dMRI data were exploited to compute voxel-wise FA and MD quantification in each subject (top left panel). We then applied a cluster-based analysis to identify significant clusters where FA and MD covaried with β Effort or β Reward (left and bottom panels). Significant clusters were mapped onto white matter tracts and characterized by MNI coordinates and their metrics were used in a predictive machine learning analysis (right panel; see D.). D. Machine learning analysis. Identified microstructural features were used as inputs for machine learning classifiers predicting low and high β Effort and β Reward . Classifier performance was assessed using 5-fold cross-validation, yielding average accuracy and area under the receiving operator curve metrics. Behavioral data acquisition Experiment setup Participants engaged in an effort-based decision-making task on a computer while seated in an ergonomic chair positioned 100 cm from a monitor (refresh rate: 100 Hz). The right arm was flexed to 90 degrees and stabilized on an armrest, with participants gripping a custom-designed handle using the right hand. The left hand remained unrestrained and was used to make finger responses via the left and right arrow keys on a standard keyboard positioned on a table. To minimize movement artifacts and ensure consistent biomechanical positioning during force exertion, the right forearm was secured to the armrest using an adjustable strap. Task description The effort-based decision-making task was programmed using custom-written MATLAB scripts incorporating the Psychtoolbox library ( Brainard, 1997 ). During this task, participants were required to evaluate whether to engage in varying levels of physical effort in exchange for varying monetary rewards ( Figure 1 .B). Effort was operationalized as an isometric contraction of the right biceps, elicited by instructing participants to attempt to flex their arm — bringing the fist toward the shoulder — while maintaining elbow contact with the armrest and gripping a fixed handle. The intensity of muscular contraction was continuously monitored via real-time surface electromyography. Monetary rewards were expressed in euro cents; participants were informed they would obtain the total amount earned during the task at the end of the session. Each trial started with the display of a vertical force gauge accompanied by a horizontal reference line indicating the required force level for that trial. Adjacent to the gauge, the monetary reward was displayed. The full scale of the gauge represented 100% of the participant’s maximal voluntary contraction (MVC), which was individually calibrated prior to the experimental blocks (see Blocks of trials and MVC section, for details). Effort levels were set at 20%, 40%, 60%, or 80% of MVC, paired with potential rewards of 1, 5, 10, or 20 euro cents, resulting in 16 possible effort-reward combinations. Participants had 5 seconds to decide whether to accept or reject each offer, responding with their left index and middle fingers by pressing the right arrow key to accept or the left arrow key to reject. Upon acceptance of an offer, a "Go" signal was presented on the screen following a variable delay of 1.0 to 1.2 seconds, marking the onset of the contraction period. During this period, participants received continuous visual biofeedback of their exertion level, depicted as a dynamic filling of the gauge. The extent of gauge filling was proportional to the rectified amplitude of the EMG signal recorded from the right biceps. Participants were instructed to maintain a contraction at or above the target force level — corresponding to 20%, 40%, 60%, or 80% of their MVC — for a minimum duration of 2.7 seconds within a 4-second window (i.e., >66% of the contraction period). At the end of each trial, participants received visuo-auditory feedback indicating the outcome of the trial (success or failure) and whether the corresponding reward had been earned. Trials in which muscle activity was detected prior to the onset of the "Go" signal were immediately aborted, and an "Anticipated" message was displayed on the screen. Regardless of the outcome, a fixed 2-second inter-trial interval followed each trial to allow for recovery. If participants declined the offer, the trial terminated immediately and was followed by a 2-second blank screen before the onset of the next trial ( i.e. , corresponding to the inter-trial interval). Similarly, if no response was recorded within the initial 5-second decision window, the trial was considered a missed response, and the subsequent trial started after the standard inter-trial interval. Data analyses Computational modelling of acceptance rates To quantify individual differences in effort and reward sensitivity, we applied computational modelling to the acceptance rates in the effort-based decision-making task. As such, for each participant, the model estimated the influence of effort and reward on offer valuation, which was translated into a probability to accept the offer and engage in the action. First, we tested a set of candidate models of subjective value computation informed by prior literature ( Morris et al., 2025 ; Gilmour et al., 2023 ; Le Heron et al., 2018 ; Bonnelle et al., 2016 ), evaluating model fits using Bayesian information criterion (BIC) minimization and visual inspection (Supplementary Figure 1). In the best-fitting model, the subjective value V of each offer was computed as: Here, the subjective value of an offer was modelled as a quadratic function of the proposed EFFORT LEVEL , and a linear function of the REWARD LEVEL , weighted by individual parameters β Effort and β Reward , reflecting effort and reward sensitivity, respectively (see below for more details). To account for potential time-dependent effects on acceptance rates across repeated trials, the model also incorporated a β Time parameter, which linearly modulated the cost component as a function of trial number ( Le Heron et al., 2018 ). This parameter thus captured time-on-task effects on acceptance rates, whether reflecting fatigue (progressive decline in acceptance across trials) or habituation and rising motivation (progressive increase). The intercept β₀ captured baseline choice bias, reflecting the propensity to accept an offer with zero reward. After model fitting, we extracted individual β Effort and β Reward parameters as proxies for effort and reward sensitivity, respectively. Parameter estimation was performed using constrained nonlinear optimization (fmincon function in MATLAB 2022b, MathWorks) to minimize the negative log-likelihood of observed choices. Model- derived choice probabilities (range: 0 – 1) were compared against actual behavior. In this framework, more negative β Effort values indicated stronger effort sensitivity — i.e. , a steeper relative decline in acceptance rates with increasing effort. Of note, due to the quadratic nature of the cost function in the winning model ( i.e. , β Effort × EFFORT LEVEL ²), this effect is particularly pronounced at high effort levels: the more a participant’s acceptance rate drops for high efforts ( e.g. , 80% MVC), the steeper the curvature of the effort-sensitivity function and the more negative the β Effort . For ease of interpretation in subsequent analyses, β Effort values were multiplied by −1, such that higher (more positive) values consistently reflected stronger effort sensitivity. Conversely, more positive β Reward values corresponded to stronger reward sensitivity, as reflected by steeper relative increase in acceptance with higher rewards. β Reward values were transformed using a log function to normalize the distribution, in accordance with former studies ( Le Heron et al., 2018 ; Derosiere et al., 2024 ). Subject- specific β Effort and β Reward values were used as dependent variables in voxel-wise, cluster-based permutation analyses (see Statistical Analysis section). MRI data analyses Diffusion data were preprocessed using the Elikopy pipeline ( Dessain et al., 2024 ), including brain extraction ( Hoopes et al., 2022 ), thermal noise removal ( Veraart et al., 2016 ), and corrections for susceptibility-induced distortions, eddy currents, and head motion using FSL (v6.0.7.8). As reversed phase-encoding b0 images were not acquired, susceptibility distortion correction was performed using Synb0-DISCO, which generates a synthetic distortion-free b0 image from the T1-weighted anatomical scan. Post-processing was also performed using Elikopy, which included the mathematical reconstruction of the diffusion-weighted images (64 directions) to derive the diffusion tensor imaging (DTI) model. From this, volume-weighted microstructural maps of FA and MD were generated for each participant in MNI space ( Figure 1 .C), providing a voxel-wise measure of microstructure across the entire brain. Statistical analyses As mentioned in the Introduction , we used a two-stage strategy to first determine white matter circuits underlying inter-individual differences in effort and reward sensitivity and to then test whether these sensitivities could be decoded from microstructure metrics. First, four whole-brain cluster-based analyses were performed to detect clusters where FA or MD covaried with β Effort or β Reward . Second, the FA and MD values from these clusters were used as features in supervised machine learning classifiers to assess whether they could decode individual differences in sensitivity in out-of-sample predictions. These two stages are presented in detail below. Whole-brain cluster-based analysis Voxel-based correlation analyses were conducted using BrainVoyager™ (Brain Innovation, version 23.2), examining the associations between FA and MD maps and β Effort or β Reward ( Figure 1 .C). These analyses controlled for potential confounding factors by including them as covariates, specifically gender, age, intracranial volume, and other variables that could influence effort and reward sensitivity, such as self-reported depression and anhedonia (assessed using the EDAS and SHAPS scales; ( Bonnelle et al., 2016 )). Statistical significance was set at a voxel-wise p-value of .005, and correction for multiple comparisons was performed using cluster-size thresholding (threshold: 44 mm 3 ; Forman et al., 1995 ). This approach revealed clusters where white matter microstructure significantly covaried with β Effort or β Reward . To interpret these findings, we use the term white matter integrity to describe FA and MD values in our healthy cohort. This term reflects microstructural organization and does not imply pathology. FA and MD provide complementary information: higher FA indicates greater directional coherence of water diffusion, often linked to axonal alignment, density, or myelination, whereas higher MD reflects greater overall diffusivity, potentially due to increased extracellular space or reduced barriers. Because associations may arise in one metric, the other, or both ( Hayakawa et al., 2014 ) analyzing both metrics offers a more comprehensive view of white matter organization. Effects observed across both FA and MD ( i.e. , with FA and MD clusters overlapping), when replicating in direction, provide stronger evidence for a robust microstructural-behavioral relationship. To align their interpretability, we computed the additive inverse of MD (-MD), allowing FA and -MD to be interpreted in the same direction (higher values indicating greater integrity). For visualization, clusters showing significant negative associations between integrity (FA or -MD) and β Effort or β Reward are highlighted in red, and positive associations in green. To assign tract labels, we used the XTRACT HCP Probabilistic Tract Atlas ( Warrington et al., 2020 ), which reliably identifies the major canonical white matter bundles, including deep, long-range pathways such as the cingulum, corticospinal tract, and corpus callosum. These tracts, typically located deep, form the principal structural connections within and between hemispheres and are thus the primary focus of our study. A minority of clusters did not overlap with these canonical bundles, instead lying within more superficial white matter (i.e., closer to the cortex) that supports short-range connections between adjacent cortical areas. Because such clusters fall outside the coverage of canonical atlases like XTRACT, we used the EBRAINS Human Connectome Project superficial white matter atlas (Labra-Avila et al., 2020) to facilitate their labelling and describe them in the Supplementary Materials (Supplementary Figures 2 and 3). All tract visualizations presented in this study were derived from white matter bundle definitions and population-averaged templates in MNI ICBM 2009a space, based on previously published work ( Garyfallidis et al., 2018 ; Yeh et al., 2018 ). Finally, to characterize the relationship between white matter microstructure and individual differences in effort and reward sensitivity, we conducted partial Pearson correlation analyses. For each significant cluster identified in the voxel-wise analyses, we extracted the FA and -MD values from all voxels within the cluster and computed a mean value per cluster for each participant. These averaged microstructural metrics served as independent variables in separate partial correlation analyses with β Effort and β Reward , controlling for the potential confounding factors mentioned above, including age, gender, intra-cranial volume and self-reported measures of depression and anhedonia. The resulting correlation R coefficients, p-values, and corresponding regression plots are presented in the Results section for each cluster. In these plots, the y-axis depicts residualized β Effort or β Reward values (after controlling for covariates), while the x-axis represents the mean FA or -MD values for each cluster, allowing for direct visualization of the strength and direction of the adjusted associations. As before, significant negative and positive partial correlations are shown in red and green, respectively, to facilitate visualization throughout the Results section. Decoding effort and reward sensitivity using machine learning As a second stage of our approach, we used the clusters to train supervised machine learning models, testing whether β Effort and β Reward could be decoded from their microstructural properties. Such predictive analyses are critical, as they provide a principled framework to estimate how focal white matter alterations (e.g., following stroke, surgery, or stimulation) might affect sensitivity to effort and reward, and ultimately disrupt goal-directed behavior. Here, we performed two supervised machine learning analyses, one for β Effort and one for β Reward . In each case, we used averaged FA and -MD values of each voxel extracted from the significant clusters identified within major, canonical bundles in the voxel-wise FA and -MD analyses as input features. Participants were assigned to either a low or high sensitivity class based on a median split of β Effort or β Reward values, enabling binary classification. To avoid bias associated with choosing a single classifier and to ensure robust performance across data distributions, we conducted an algorithmic classifier selection. Specifically, we evaluated 12 classifiers: logistic regression, support vector machines, random forests, gradient boosting, extra trees, adaptive boosting, decision trees, multilayer perceptron, k-nearest neighbors, Gaussian naive Bayes, linear discriminant analysis, and quadratic discriminant analysis. To estimate classifier performance while avoiding overfitting, we used a nested cross-validation procedure. The outer loop consisted of a stratified 5-fold cross- validation that provided an unbiased estimate of generalization performance. Within each outer fold, a second, inner stratified cross-validation was performed on the training data to conduct a grid search over hyperparameters. The best model identified in the inner loop was then evaluated on the outer test fold. This procedure was repeated 1,000 times, with random re-sampling at each iteration. For each iteration, we computed two key performance metrics: classification accuracy and the area under the curve (AUC) of the receiver operating characteristic curve (ROC). Classification accuracy reflects the proportion of correctly predicted labels, while the ROC plots the true positive rate against the false positive rate across all decision thresholds. The AUC gives a synthetic and quantitative measure of this curve, with 0.5 indicating chance-level performance and 1.0 representing perfect classification. For each target variable (β Effort and β Reward ), the classifier achieving the highest mean AUC across the 1,000 iterations was selected for further analysis. The best- performing classifier was extra trees for β Effort and logistic regression for β Reward , although all classifiers performed above chance as described in the Results section. To assess whether our classification accuracy and AUC was significantly above chance level, we repeated the entire classification procedure using the best- performing classifiers 1,000 times with randomly permuted class labels, generating a null distribution of accuracy and AUC under the assumption of no association between white matter microstructure and β Effort and β Reward . We then evaluated whether classification accuracy and AUC in the true-label data exceeded that expected by chance using three complementary criteria: (1) a Monte Carlo p-value, defined as the proportion of classification yielding higher accuracy or AUC with the permuted-label data than that obtained with the true-label data; (2) one-sided t-tests comparing the distributions of accuracy or AUC for true-label versus permuted-label data; and (3) one-sided t-tests testing whether accuracy or AUC for true-label and permuted-label data were significantly higher from the chance level of 0.5. Statistical significance was set at p < .05. Further, to determine whether the observed classification performance was specific to the white matter clusters identified in the voxel-wise analysis, we conducted a control analysis using randomly selected clusters. For each of the two target variables (β Effort and β Reward ), we randomly sampled clusters from the whole-brain white matter mask, matching both the number and size of the original significant clusters. We then repeated the full classification pipeline, including evaluation across 12 classifiers, nested cross-validation, and performance estimation based on accuracy and AUC, for these randomly located clusters. As with the primary analyses, we assessed whether AUC exceeded chance (0.5) using one-sample t-tests; p-values were FDR-corrected. This control analysis allowed us to test whether decoding performance was specific to the significant clusters identified in the voxel-wise analyses, rather than reflecting generic variability of white matter microstructure across subjects. Finally, we conducted feature importance and recursive feature elimination analyses to identify which clusters contributed most to classification performance. In this context, the term “feature” refers specifically to the individual white matter clusters included as input variables in the model, a standard term in machine learning used to denote each parameter or predictor considered during the analysis. For feature importance, we used the best-performing classifier for each target variable (extra trees for β Effort , logistic regression for β Reward ) trained on true-label data. We extracted built- in feature weights (i.e., impurity-based importance scores for tree-based models and absolute coefficient weights for logistic regression) to rank all identified clusters’ microstructural measures (FA and -MD) according to their predictive contribution. Using these rankings, we implemented recursive feature elimination by iteratively removing the least informative cluster and repeating the full classification analysis, recalculating AUC at each iteration. This process was repeated 100 times with random re-sampling to ensure stability of the performance estimates. For each number of retained clusters (from one up to the full set), we computed the mean AUC and tested whether AUC significantly exceeded chance level (0.5) using one-sided t-tests across iterations. This analysis provided a quantitative estimate of the optimal subset of clusters required to decode β Effort and β Reward above chance, while identifying the specific clusters driving classification performance. RESULTS Effort and reward sensitivity show substantial inter-individual variability across healthy subjects Behavioral data revealed strong inter-individual differences in how participants modulated their acceptance rates based on effort and reward magnitudes, reflecting differences in effort and reward sensitivity ( Figure 2 .A). This variability was quantitatively captured by the density distribution of the β Effort and β Reward parameters, which showed broad dispersion across individuals ( Figure 2 .B; β Effort : mean = 16.92, SD = 11.10; β Reward : mean = -0.30, SD = 0.98). The coefficient of variation reached 65.6% for β Effort and 318.1% for β Reward , confirming substantial heterogeneity in sensitivity to effort and reward. To illustrate this diversity of profiles, Figure 2 .C presents individual examples of acceptance behavior in participants with low vs. high β Effort and β Reward values. These findings confirm the presence of heterogenous sensitivity profiles in the healthy population. Download figure Open in new tab Figure 2: Effort and reward sensitivity show substantial inter-individual variability across healthy subjects. A. Individual acceptance rate maps. Each map represents the acceptance rate of a given subject as a function of effort (in % MVC; y-axes) and reward (x-axes). These maps show the strong inter-individual differences in how participants modulated their acceptance rates based on effort and reward magnitudes. B. Distributions of β Effort and β Reward . Parameters β Effort and β Reward were estimated from the modelling of acceptance rates presented in A using the equation presented in section Computational modelling of acceptance rates . Here again, the density distributions depict the high inter-individual variability in β Effort and β Reward values among our group of 45 subjects. As described in the section Computational modelling of acceptance rates , for ease of interpretation, β Effort values initially obtained following our modelling procedure were multiplied by −1, such that higher (more positive) values reflect stronger effort sensitivity in the left panel. Further, more positive β Reward values correspond to stronger reward sensitivity in the right panel. C. Representative acceptance curves from individuals with divergent β Effort and β Reward values . Left panel: Compared to subject A, subject B showed a stronger decrease in acceptance rates with increasing effort, reflected by a higher β Effort (24.44 vs. 6.06). Right panel : Compared to subject C, subject D showed a stronger increase in acceptance rates with increasing reward, reflected by a higher β Reward (0.46 vs. –1.51). These examples illustrate how modulation of decision behavior by effort and reward varies across healthy individuals. Key findings on the microstructural correlates of effort and reward sensitivity As detailed in the Methods , in the first stage of our statistical approach, we conducted four cluster-based analyses to examine the microstructural correlates of individual differences in β Effort and β Reward , using FA and the additive inverse of MD (- MD, with higher values reflecting greater integrity) to capture complementary aspects of white matter integrity (see section Whole-brain cluster-based analysis , for more details). Across these analyses, we identified 12 significant clusters located within major, canonical white matter bundles: 5 covarying with effort sensitivity and 7 with reward sensitivity. All of the 5 effort-related clusters were situated within tracts connected to frontal valuation regions, such as SMA, dACC, and OFC ( e.g ., within the anterior cingulum tract). In contrast, the 7 reward-related clusters were more widely distributed: 3 were located within tracts linked to frontal valuation regions, while 4 lay outside this network, encompassing fronto-parietal and sensorimotor pathways. The most robust associations included one cluster shared across effort and reward sensitivity analyses and another consistent across FA and -MD metrics, both within SMA-connected tracts. Machine learning analyses further showed that microstructural metrics from these 5 effort-related and 7 reward-related clusters reliably predicted effort and reward sensitivity, respectively, whereas randomly positioned clusters did not. Table 1 summarizes all clusters identified across the four analyses (i.e., FA and - MD analyses for β Effort and β Reward ), their tract labels and associated partial correlation results, which are described individually in the following sections. View this table: View inline View popup Download powerpoint Table 1. White matter clusters showing significant associations with β Effort and β Reward . The table reports 12 significant clusters (5 associated with β Effort and 7 with β Reward ). For each cluster, MNI coordinates correspond to the cluster center, and volume reflects cluster size (mm 3 ). Tract labels were derived using the XTRACT HCP Probabilistic Tract Atlas and, when necessary, refined with anatomical masks from prior studies. The two rightmost columns report partial correlation coefficients (R) and p-values between β Effort or β Reward and the average FA or -MD values extracted from voxels within each cluster, controlling for age, gender, and other covariates (see Methods). Clusters located within tracts directly connected to frontal valuation regions ( e.g. , projecting to the SMA or OFC) are highlighted in black, while clusters belonging to fronto-parietal or sensorimotor pathways are shown in gray. Microstructure in tracts connected to key frontal valuation regions is associated with effort sensitivity Among the 5 significant clusters associated with β Effort , 4 showed negative associations, where reduced FA or -MD ( i.e. , lower microstructural integrity) was linked to higher β Effort ( i.e. , stronger effort sensitivity), while 1 showed a positive association. Below, we detail these clusters. The first negative association with β Effort was found for FA in a cluster located within the SMA portion of the left corticospinal tract (MNI: -13.95, -10.01, 58.02; 94 mm 3 ; R = -.697; p < .001; Figure 3A ). Probabilistic atlas labelling identified the corticospinal tract, and anatomical masks ( Bonnelle et al., 2016 ; Beckmann et al., 2009) showed it was located within SMA proper. As typical in such labelling, the cluster extended beyond the tract (46% of voxels in the tract), likely involving other SMA- related projections, including the SMA-NAcc pathway, which showed in prior work a negative association between effort sensitivity and macroscopic structural connectivity (i.e., streamline count; Derosiere et al., 2024 ). Importantly, the same cluster also emerged as a key correlate of β Reward (see below), underscoring its importance as a shared microstructural substrate of both effort and reward sensitivity. Download figure Open in new tab Figure 3: Microstructure in tracts connected to frontal valuation regions showing negative associations with effort sensitivity (β Effort ). A. Cluster in the SMA portion of the left corticospinal tract. (i) A significant cluster (red) showed a negative correlation between FA and β Effort (MNI: -13.95, -10.01, 58.02). (ii) Probabilistic tract labelling identified the corticospinal tract; anatomical masks ( Bonnelle et al., 2016 ; Beckmann et al., 2009) confirmed the cluster’s location within SMA proper. The cluster extended beyond the tract (46% of its voxels), likely encompassing SMA-related projections such as the SMA–NAcc pathway, previously linked to β Effort via macroscopic connectivity measures ( Derosiere et al., 2024 ). (iii) Partial correlation confirmed reduced FA in this cluster was significantly correlated with higher effort sensitivity (R = -.697, p < .001). Across all panels, residualized β values (adjusted for age, gender, intracranial volume, depression, and anhedonia) are plotted against mean FA or -MD per cluster to visualize adjusted associations. B. Clusters in the mid-anterior corpus callosum. (i) Three significant clusters (red) exhibited consistent negative correlations between microstructural integrity and β Effort . The FA cluster (top row, left) overlapped with one -MD cluster (middle row, left), while another -MD cluster was located symmetrically in the right hemisphere (bottom row). (ii) Probabilistic tract labelling localized all clusters to the mid-anterior corpus callosum, specifically in its premotor portion containing interhemispheric fibers connecting bilateral SMAs ( Xiong et al., 2024 ). (iii) Partial correlation analyses confirmed strong negative associations for each cluster (R = -.556, R = -.641, and R = -.575, all p < .001). The replication across hemispheres and both FA and -MD provides particularly strong evidence for a robust link between effort sensitivity and compromised integrity in SMA-related interhemispheric pathways. A second robust pattern involved the mid-anterior corpus callosum. FA analyses revealed a cluster in its left portion (MNI: -10.93, 15.96, 19.02; 56 mm 3 ; R = -.556; p < .001; Figure 3B ), in a locus containing interhemispheric fibers predominantly connecting bilateral SMAs ( Xiong et al., 2024 ). Supporting this, -MD analyses revealed two additional bilateral clusters with overlapping spatial distributions with the FA cluster (left: -7.55, 14.03, 20.11; 322 mm 3 ; R = -.641; p < .001; right: 8.71, 15.55, 20.32; 139 mm 3 ; R = -.575; p < .001; Figure 3B ). The fact that these effects replicate bilaterally and across both FA and -MD, consistently showing decreased integrity with higher β Effort , provides particularly strong evidence for a robust relationship. These results suggest that higher effort sensitivity is linked to reduced directional integrity (lower FA, reflecting reduced axonal alignment or myelination) and increased water diffusion (lower -MD, suggesting greater extracellular space), which may disrupt interhemispheric communication between bilateral SMAs and other medial frontal regions critical for effort processing. In contrast, one cluster showed a positive association between FA and β Effort . This cluster was located in the left anterior cingulum bundle (MNI: -9.95, 34.7, 11.22; 209 mm 3 ; R = .623; p < .001; Figure 4 ), a tract that links medial frontal regions. To identify which medial frontal regions were specifically concerned by this cluster, we projected it onto a high-resolution atlas (Labra-Avila et al., 2020), which revealed it was located in fibers connecting the dACC with SMA and medial OFC, key frontal valuation hubs. No additional clusters emerged in major tracts, and the -MD analysis revealed none with positive associations. Download figure Open in new tab Figure 4: Microstructure in a tract connected to frontal valuation regions showing a positive association with effort sensitivity (β Effort ). (i) A significant cluster (green) exhibited a positive correlation between FA and β Effort (MNI: -9.95, 34.7, 11.22). (ii) Probabilistic tract labelling identified the anterior cingulum. To clarify which medial frontal regions were specifically connected by this cluster, we projected it onto a high-resolution atlas (Labra-Avila et al., 2020), revealing fibers linking the dACC with SMA and medial OFC, core valuation hubs. (iii) Partial correlation confirmed that greater FA in this cluster was associated with higher effort sensitivity (R = .623, p < .001). No additional clusters were identified in major tracts, and the - MD analysis did not reveal any clusters with positive associations. Together, these findings indicate that individual differences in effort sensitivity are anchored in several white matter pathways connected to frontal valuation regions. Reduced microstructural integrity in SMA-connected clusters, including clusters identified in the corticospinal tract and interhemispheric fibers of the mid-anterior corpus callosum, is consistently associated with stronger effort sensitivity, representing some of the most robust effects in the present study (consistent across metrics and, for one cluster, across valuation dimensions [see below]). Conversely, increased integrity of the left anterior cingulum bundle, encompassing connections between SMA, dACC, and medial OFC, is associated heightened effort sensitivity, potentially reflecting a distinct mechanism whereby enhanced communication among these valuation hubs amplifies effort sensitivity (see Discussion section). Microstructure in tracts connected to frontal valuation regions, as well as fronto- parietal and sensorimotor structures, is associated with reward sensitivity As mentioned earlier, the FA and -MD analyses revealed 7 significant clusters, more widely distributed than the effort-related clusters: 3 were located within tracts linked to frontal valuation regions, while 4 lay outside this network, encompassing fronto-parietal and sensorimotor pathways. 2 clusters showed negative associations, where lower FA or -MD ( i.e. , lower microstructural integrity) was linked to higher β Reward ( i.e. , stronger reward sensitivity), while 5 clusters showed positive associations. Below, we detail these clusters, beginning again with the negative associations. The first negative association with β Reward was for FA in a cluster located within the SMA portion of the left corticospinal tract (MNI: -13.81, -10.44, 59.95; 79 mm 3 ; R = -.614; p < .001; Figure 5A ), as identified with probabilistic atlas labelling and anatomical masks ( Bonnelle et al., 2016 ; Beckmann et al., 2009) and overlapped extensively with the cluster identified for β Effort (MNI: -13.95, -10.01, 58.02; 94 mm 3 ; Figure 6 ). Importantly, this shared cluster represents the sole locus where microstructural integrity covaries with both β Effort and β Reward , suggesting a shared white matter substrate where reduced integrity may amplify both effort and reward sensitivity. To confirm that this dual association was not trivially driven by the fact that β Effort and β Reward were derived from the same acceptance rates, we tested whether FA values in these clusters correlated with the two other model-derived parameters: β Time and β₀. No significant correlations were observed either with the effort-related FA values (β Time : R = -.180, p = .237; β₀: R = -.163, p = .284) or with the reward-related FA values (β Time : R = -.111, p = .466; β₀: R = -.148, p = .332; Supplementary Figure 4), underscoring the specificity of this SMA white matter locus for β Effort and β Reward . Download figure Open in new tab Figure 5: Microstructure in tracts connected to frontal valuation regions and sensorimotor structures and showing negative associations with reward sensitivity (β Reward ). A. Cluster in the SMA portion of the left corticospinal tract. (i) A significant cluster (red) exhibited a negative correlation between FA and β Reward (MNI: -13.81, -10.44, 59.95). (ii) Probabilistic tract labelling identified the corticospinal tract; anatomical masks ( Bonnelle et al., 2016 ; Beckmann et al., 2009) confirmed the cluster’s location within SMA proper. The cluster overlapped substantially with the β Effort cluster (see Figure 6 ), suggesting a shared white matter substrate where reduced integrity may amplify both effort and reward sensitivity. (iii) Partial correlation confirmed that reduced FA in this region was significantly correlated with greater reward sensitivity (R = -.614, p < .001). B. Cluster in the right middle cerebellar peduncle, extending to the left cerebello-ponto-cortical tract. (i) A significant cluster (red) exhibited a negative correlation between -MD and β Reward (MNI: 11.05, -29.94, -34.32). (ii) Probabilistic tract labelling localized the cluster right middle cerebellar peduncle, extending to the left cerebello-ponto-cortical tract. (iii) Partial correlation confirmed that reduced integrity in this region was significantly correlated with greater reward sensitivity (R = -.547, p < .001). Across all panels, residualized β values (adjusted for age, gender, intracranial volume, depression, and anhedonia) are plotted against mean FA or -MD per cluster to visualize adjusted associations. Download figure Open in new tab Figure 6. Exclusive overlap of white matter clusters associated with effort and reward sensitivity within the SMA portion of the corticospinal tract. FA analyses identified two significant clusters within the SMA portion of the left corticospinal tract. The cluster associated with β Effort (red) was centered at MNI: X = –13.95, Y = –10.01, Z = 58.02 (94 mm 3 ), while the cluster associated with β Reward (green) was centered at MNI: X = –13.81, Y = –10.44, Z = 59.95. These clusters overlapped extensively (yellow), forming the only locus across the entire FA and -MD maps where microstructural integrity covaried with both β Effort and β Reward . This convergence underscores the specificity of this SMA corticospinal segment as a shared structural substrate for individual differences in sensitivity to both effort and reward. The second negative association was located in a tract connected to sensorimotor structures, specifically the right middle cerebellar peduncle, with part of the cluster extending into the left cortico-ponto-cerebellar tract, which projects to the middle cerebellar peduncle, as identified by probabilistic tract labelling (MNI: 11.05, –29.94, –34.32; 84 mm 3 ; R = –.547; p < .001; Figure 5B ). Here, lower -MD values, indicative of greater extracellular space, was associated with higher β Reward . Although less expected, this result suggests that disrupted cortical input to the cerebellum ( i.e. , since the middle cerebellar peduncle conveys afferences from cortex to cerebellum) may promote action initiation when high rewards are at stakeThis finding illustrates how a whole-brain, cluster-based approach applied to FA and -MD maps can uncover functionally relevant pathways beyond those typically targeted by tract-of-interest analyses. 5 clusters showed positive associations between FA and β Reward . Two bilateral clusters were located in the anterior portion of the forceps minor (left: MNI: -17.82, 53.12, -2.3; 73 mm 3 ; R = .572; p < .001; right: 17.9, 53.81, 4.19; 67 mm 3 ; R = .561; p < .001; Figure 7A ), a tract interconnecting bilateral OFCs ( Filbey et al., 2014 ), key hubs for reward valuation. Two additional clusters were found in a fronto-parietal pathway, namely along the right superior longitudinal fasciculus, one in its frontal portion (MNI: 39.85, 15.76, 17.7; 126 mm 3 ; R = .579; p < .001) and one in its parietal portion (MNI: 36.22, -24.6, 33.78; 86 mm 3 ; R = .561; p < .001; Figure 7B ). Finally, one other cluster was located in a tract connected to motor structures, namely in the left mid-body of the corpus callosum (MNI: -9.61, 5.02, 26.19; 137 mm 3 ; R = .566; p < .001; Figure 7C ), a locus of fibers primarily connecting bilateral M1s ( Hofer & Frahm, 2006 ; Tarumi et al., 2022 ; Wahl et al., 2007 ). No additional clusters emerged in major tracts, and the -MD analysis revealed none with positive associations. Download figure Open in new tab Figure 7: Microstructure in tracts connected to frontal valuation regions, fronto-parietal and sensorimotor structures showing positive associations with reward sensitivity (β Reward ). A. Bilateral clusters in the forceps minor. (i) Two significant clusters (green) exhibited a positive correlation between FA and β Reward (MNI: left: -17.82, 53.12, -2.3; right: 17.9, 53.81, 4.19). (ii) Probabilistic tract labelling localized both clusters to the forceps minor, interconnecting bilateral OFCs, key hubs for reward valuation. (iii) Partial correlation confirmed significant positive associations with reward sensitivity (R = .572 and R = .561, both p < .001). B. Clusters in the right superior longitudinal fasciculus. (i) Two significant clusters (green) exhibited a positive correlation between FA and β Reward (MNI: anterior: 39.85, 15.76, 17.7; posterior: 36.22, -24.6, 33.78). (ii) Probabilistic tract labelling indicated both clusters were located within the superior longitudinal fasciculus. (iii) Partial correlation confirmed significant associations with reward sensitivity (R = .579 and R = .561, both p < .001). C. Cluster in the left mid-body of the corpus callosum (motor portion). (i) One significant cluster (green) exhibited a positive correlation between FA and β Reward (MNI: -9.61, 5.02, 26.19). (ii) Probabilistic tract labelling indicated the cluster was located in callosal fibers connecting bilateral motor cortices. (iii) Partial correlation confirmed a significant association (R = .566, p < .001). Collectively, these results show that differences in reward sensitivity are associated with white matter integrity across distributed systems, encompassing circuits directly connected to frontal valuation regions as well as fronto-parietal and sensorimotor pathways. Regarding frontal valuation regions, heightened sensitivity was linked to reduced integrity within an SMA cluster, comprising SMA-connected corticospinal fibers, and increased integrity within the OFC-connected anterior forceps minor. In addition, reduced integrity in cerebellum-connected tracts and greater integrity within a major fronto-parietal bundle (i.e., the superior longitudinal fasciculus) and motor callosal fibers was linked to stronger reward sensitivity, highlighting additional contributions from fronto-parietal and sensorimotor pathways. This pattern suggests that reward sensitivity reflects the integrated influence of frontal valuation circuits together with fronto-parietal and motor-related networks, rather than any single network. Decoding effort and reward sensitivity from brain microstructure using machine learning classifiers We next tested whether β Effort and β Reward could be predicted from the microstructural properties of the clusters identified above. Our machine learning models used as input the averaged FA and -MD values extracted from the significant clusters identified in voxel-wise analyses and were trained to classify participants into low versus high β Effort and β Reward groups. Performance metrics (accuracy and AUC) were estimated using nested cross-validation on true labels and compared against null distributions generated by 1,000 label permutations. For β Effort , the classifier achieved a mean accuracy of 0.648 ± 0.053 on the true- label data, well above the 0.5 chance level for binary classification. In contrast, accuracy on the permuted-label data averaged 0.506 ± 0.086, with only 37 out of 1,000 permutations reaching or exceeding the accuracy obtained with true-label data, yielding a Monte Carlo p-value of .037 ( Figure 8 .A). This difference in accuracy between true-label and permuted-label data was statistically significant (t 998 = 45.57, p 2.5), and accuracy on true-label data was significantly above chance (t 999 = 88.19, p < .001), unlike the permuted-label data (t 999 = -1.466, p = .93). To complement these results, we also examined the AUC, which provides a threshold- independent measure of classification performance. The AUC obtained on true-label data was 0.719 ± 0.050, higher than the AUC from permuted-label data (0.498 ± 0.118), with only 28 out of 1,000 permutations matching or exceeding it (Monte Carlo p = .028). The difference between true-label and permuted-label data was also statistically significant (t 998 = 54.40, p 2.5), and only the AUC obtained on the true-label data was significantly greater than chance (t 999 = 137.3, p < .001; permuted-label data: t 999 = -0.553, p = .580). Download figure Open in new tab Figure 8: White matter microstructure predicts individual differences in effort and reward sensitivity. (A) β Effort decoding. (i) Classifier accuracy on true-label data (blue line) compared to the null distribution obtained from 1,000 label permutations (pink); only 37/1,000 permutations matched or exceeded true-label accuracy (Monte Carlo p = 0.037). (ii) Accuracy was significantly higher for true-label data compared to permuted-label and chance-level data (true vs. permuted: t 998 = 45.57, P < 0.001; true vs. chance: t 999 = 88.19, P < 0.001; permuted vs. chance: t 999 = –1.466, P = 0.93). (iii) ROC curve showing classifier performance (mean AUC = 0.720 ± 0.039; 95% CI shown in shading). (iv) True-label AUC (blue line) compared to null distribution from permuted-label data (pink); 28/1,000 permutations matched or exceeded true-label AUC (Monte Carlo p = 0.028). (v) AUC was significantly higher for true-label data (true vs. permuted: t 998 = 54.40, P < 0.001; true vs. chance: t 999 = 137.3, P < 0.001; permuted vs. chance: t 999 = –0.553, P = 0.580). (B) β Reward decoding . (i) True-label accuracy (blue line) exceeded that of 1,000 permutations (pink); only 2 permutations matched or exceeded it (Monte Carlo P = 0.002). (ii) Accuracy was significantly higher for true-label data (true vs. permuted: t 998 = 88.68, P < 0.001; true vs. chance: t 999 = 183.2, P < 0.001; permuted vs. chance: t 999 = –4.95, P = 0.99). (iii) ROC curve (mean AUC = 0.829 ± 0.022). (iv) AUC distribution from permuted-label data (pink) with true-label AUC (blue line); no permutations matched or exceeded the true-label AUC (Monte Carlo P < 0.001). (v) AUC was significantly higher for true-label data (true vs. permuted: t 998 = 112.7, P < 0.001; true vs. chance: t 999 = 314.5, P < 0.001; permuted vs. chance: t 999 = –1.698, p = 0.955). Boxplots show median, interquartile range, and full range. Symbols including # and * indicate statistical significance: ### p < .001, ***p < .001; n.s. = not significant) Similarly, for β Reward , the classifier achieved a mean accuracy of 0.718 ± 0.038 on the true-label data. In contrast, accuracy on the permuted-label data averaged 0.489 ± 0.073, with only 2 out of 1,000 permutations reaching or exceeding the accuracy obtained with true-label data, yielding a Monte Carlo p-value of .002 ( Figure 8 .B). This difference between true-label and permuted-label data was statistically significant (t 998 = 88.68, p 2.5), and accuracy on true-label data was significantly above chance (t 999 = 183.2, p < .001), unlike the permuted-label data (t 999 = -4.95, p = .99). The AUC from true-label data was 0.837 ± 0.033, while the one from permuted data-label was 0.495 ± 0.090, with 0 out of 1,000 permutations matching or exceeding it (Monte Carlo p < .001). This difference was also statistically significant (t 998 = 112.7, p 2.5), and only the AUC from true-label data was significantly greater than chance (t 999 = 314.5, p < .001; permuted-label data: t 999 = -1.698, p = .955). To further establish the robustness and specificity of our machine learning findings, we conducted two additional analyses. First, we verified that decoding performance was not dependent on the choice of the best classifier by repeating the t-tests against chance level across the AUC values obtained for all 12 tested classifiers (i.e., with p values FDR-corrected). AUC was significantly above chance level across nearly all classifiers for β Effort (all p-values < .001, except SVM: p = .999) and across all classifiers for β Reward (all p-values < .001; see Figure 9 & Supplementary Tables 1-2), underscoring that predictive performance was classifier-independent and generalizable. Second, we tested whether decoding performance was specific to the white matter clusters identified in our voxel-wise analyses. For each target variable, we randomly sampled clusters from the whole-brain white matter mask, matched in number and size to the original clusters, and re-ran the full classification pipeline, including evaluation across all classifiers and nested cross-validation. In contrast to the primary analyses, AUC for these randomly located clusters was consistently near chance (β Effort prediction: AUC values = [.364 to .506], all p-values = [.360 to 1.00]; β Reward prediction: AUC values = [.401 to .534], all p-values = [.368 to 1.00]; Figure 9 , pink). In fact, the AUC obtained from the original, significant clusters with the best- performing classifier was significantly higher than the AUC obtained from the randomly located clusters (t 999 = 192.4, p < .001). This second analysis demonstrates that the decoding of β Effort and β Reward is driven by microstructural features of the specific clusters identified in the voxel-wise analyses, rather than reflecting generic inter- individual variability in white matter. Download figure Open in new tab Figure 9: Decoding effort and reward sensitivity is classifier-independent and cluster-specific. A. Classifier-independent and cluster-specific decoding of β Effort . Left panel (blue) : Area under the curve (AUC) values for 12 machine learning classifiers trained on microstructural measures from the 5 clusters significantly associated with β Effort . Decoding performance was significantly above chance for nearly all classifiers (all FDR-corrected p < .001), except for SVM (FDR-corrected p = .99), demonstrating classifier-independent decoding. Classifier names: LR: logistic regression; SVM: support vector machine; RF: random forest; GB: gradient boosting; ET: ExtraTrees; AdaB: AdaBoost; DT: decision tree; MLP: multilayer perceptron; KNN: k-nearest neighbors; GNB: Gaussian naive Bayes; LDA: linear discriminant analysis; QDA: quadratic discriminant analysis. Middle panel (pink) : AUC values for the same classifiers trained on microstructural measures from randomly sampled white matter clusters, matched in size and number to the original clusters, showing performance near chance. Right panel : Statistical comparison of AUC values from the best-performing classifier (ExtraTrees) trained on the 5 significant clusters (blue) versus 5 random clusters (pink). Performance with significant clusters was significantly higher than with random clusters (***FDR-corrected p < .001) and above chance ( ### FDR-corrected p < .001), whereas performance with random clusters did not differ from chance (n.s.). Box plots indicate median and interquartile range; whiskers denote data within 1.5 × IQR. B. Classifier-independent and cluster-specific decoding of β Reward . The panels follow the same organization as A. but for β Reward . To delineate which white matter clusters most strongly drove the decoding of β Effort and β Reward , we conducted feature importance and recursive feature elimination analyses (see Methods section). In this context, the term “feature” refers specifically to the individual white matter clusters included as input variables in the classifier, a standard term in machine learning used to denote each parameter or predictor considered during the analysis. Feature importance rankings (derived from the best-performing classifiers), identified the cluster located within the SMA portion of corticospinal tract as the most predictive feature for both β Effort and β Reward ( Figure 10 A and B). This finding aligns with the overlap of this cluster across the β Effort and β Reward cluster-based analyses and its lack of correlation with other computational parameters ( i.e. , both β Time and β₀), underscoring its specificity to effort and reward sensitivities. Recursive feature elimination indicated that high decoding accuracy could be achieved without the full cluster set. For β Effort , classification performance peaked when restricted to the four most informative clusters (AUC = .806 ± .147; Figure 10 ), three of which were connected to SMA and one with the anterior cingulum. For β Reward , classification performance also reached maximal performance with the four most predictive clusters (AUC = .840 ± .136). Of these, two clusters belonged to tracts directly connected to frontal valuation regions (SMA corticospinal tract and OFC-connected forceps minor), while the remaining two involved fronto-parietal and sensorimotor pathways (right superior longitudinal fasciculus and right middle cerebellar peduncle/left cortico-ponto- cerebellar tract). Hence, these analyses indicate that while SMA-connected pathways are the most consistent predictors of both β Effort and β Reward , optimal decoding of reward sensitivity additionally relies on contributions from fronto-parietal and motor pathways, reinforcing the idea that inter-individual differences in reward sensitivity emerge from distributed and functionally diverse white matter circuits. Download figure Open in new tab Figure 10: SMA-connected clusters dominate in terms of predictive power, with distributed circuits contributing to reward sensitivity decoding. A. Cluster contributions and optimal cluster set for decoding β Effort . Left panel : Feature importance analysis, where “features” denote the individual white matter clusters used as input variables in the model (a standard machine learning term for predictors). The most predictive cluster was the left corticospinal tract, SMA portion [FA], followed by the right and left premotor portions of the corpus callosum [-MD and FA], and the left anterior cingulum [FA]. Right panel : Recursive feature elimination analysis. Classification performance (AUC) peaked when restricted to the four most informative clusters (green diamond), which are highlighted by the green dashed box in the left panel. A. Cluster contributions and optimal cluster set for decoding β Reward . The panels follow the same organization as A. but for β Reward . Here, the left corticospinal tract, SMA portion [FA], was again the most predictive feature, followed by clusters in the cortico-ponto-cerebellar tract [-MD], superior lateral fasciculus [FA], and forceps minor [FA]. Maximal classification performance (AUC) was also reached with four features, indicating that reward sensitivity is best predicted by a combination of SMA-connected, fronto- parietal, and cerebellar pathways. Altogether, these results demonstrate that white matter microstructure within the clusters identified through voxel-wise analyses carries sufficient information to reliably decode inter-individual differences in both effort and reward sensitivity. Predictions were consistently above chance across complementary metrics (accuracy and AUC) and robust across all tested classifiers. Importantly, classification performance did not exceed chance when using randomly sampled white matter clusters of matched size and number, confirming that predictive information is specific to the loci uncovered by the voxel-wise analyses rather than reflecting generic microstructural variability. Feature importance and recursive elimination further revealed that decoding performance is driven by a subset of clusters, most prominently those connected to SMA, while additional contributions from fronto-parietal and motor pathways are required for optimal prediction of reward sensitivity. DISCUSSION White matter pathways are increasingly recognized as active modulators of neural function, capable of amplifying or attenuating electrophysiological signals ( Innocenti et al., 2022 ; Thiebaut de Schotten and Forkel, 2022 ), thereby shaping brain activity and behavior ( López-Barroso et al., 2013 ). Variability in white matter microstructure is therefore thought to drive individual differences in cognition and behavior (Forkel et al., 2022), and mapping these structure-behavior relationships can help identify biomarkers for neurological and psychiatric conditions ( Thiebaut de Schotten and Forkel, 2022 ) and predict the behavioral impact of structural lesions ( Koch et al., 2021 ). Recent work highlights the clinical potential of such predictive modelling, from forecasting disease progression ( e.g. , Parkinson’s, catatonia; Huang et al., 2024 , Peretzke et al., 2025 ) to anticipating behavioral outcomes after neurosurgery ( Aylmore et al., 2025 ; Essayed et al., 2017 ; Ordonez-Rubiano et al., 2023 ), suggesting that extending these approaches to motivational constructs could improve risk assessment for apathy and impulsivity and guide personalized interventions. Here, using a whole- brain, data-driven approach combining DWI with computational modelling of decision- making, we show that white matter integrity covaries with individual differences in effort and reward sensitivity, two key determinants of goal-directed behavior. Twelve clusters emerged: 5 linked to effort sensitivity, all within tracts connected to frontal valuation hubs (SMA, dACC, OFC), and 7 linked to reward sensitivity, spanning frontal valuation, fronto-parietal, and sensorimotor pathways. The most robust effects localized to two SMA-connected clusters, one common to effort and reward sensitivity and another converging across FA and -MD metrics and related to effort sensitivity. FA and -MD metrics from the 5 effort-related and 7 reward-related clusters reliably predicted effort and reward sensitivity in out-of-sample machine learning analyses, respectively, whereas randomly sampled clusters did not. SMA-connected tracts dominated decoding in these analyses, but fronto-parietal and sensorimotor pathways also strongly contributed to the decoding of reward sensitivity. We found both negative and positive associations between white matter integrity and effort/reward sensitivity, supporting the view that white matter modulates neural signals ( Innocenti et al., 2022 ) and that changes in its integrity, whether reductions or increases, alter signal modulation between gray matter regions and influence behavior. Similar bidirectional patterns have been reported across networks in disorders affecting effort and reward sensitivity, including apathy ( Baggio et al., 2015 ; Tay et al., 2019 ), depression ( Leaver et al., 2016 ; Lynch et al., 2024 ; Oestreich et al., 2022 ), and addictions ( Tolomeo and Yu, 2022 ). These observations suggest that both hypo- and hyperconnectivity can heighten sensitivity to effort and reward, depending on the circuit and the stage of dysfunction. As mentioned above, one of the strongest effects of our data-driven analyses localized to SMA-connected pathways, with a shared cluster in the SMA segment of the corticospinal tract covarying with both effort and reward sensitivity, suggesting a generic role in action valuation. This dual association was construct-specific, as FA values in this cluster did not correlate with β Time and β₀, which index sensitivity to time- on-task and overall baseline acceptance, respectively. Machine learning feature importance analyses independently ranked these SMA clusters as dominant predictors, reinforcing their central role in predicting both effort and reward sensitivity. Although SMA activity has been consistently linked to value computation, particularly in relation to effort processing ( Zénon et al., 2015 ; Bonnelle et al., 2016 ; Husain and Roiser, 2018 ; Heron et al., 2019 ), the prevailing view attributes these changes primarily to interactions with other fronto-striatal valuation hubs, such as the dACC and NAcc (see Husain and Roiser, 2018 , for review). Our findings suggest a potentially complementary mechanism: changes in SMA activity may not only reflect interactions with other fronto-striatal valuation regions, but also indicate SMA’s direct influence on behavioral engagement via its corticospinal projections, a pathway largely overlooked in current models. Future studies assessing corticospinal excitability via transcranial magnetic stimulation of the SMA ( Entakli et al., 2014 ; Spieser et al., 2013 ; Neige et al., 2023 ) during effort-reward decisions could further characterize the contribution of SMA corticospinal projections to effort and reward sensitivity. Probabilistic tract labelling further showed that, while the shared cluster indeed mainly overlapped with the SMA corticospinal tract, it also extended beyond this tract, likely engaging additional, more classical valuation-related pathways, such as the SMA-NAcc tract, where reduced macroscopic connectivity has likewise been linked to heightened effort sensitivity ( Derosiere et al., 2024 ). Altogether, reduced integrity in these different SMA-originating circuits may disrupt SMA’s dual roles: that is, invigorating action initiation when movement execution is costly ( Fried et al., 1991 ; Potgieser et al., 2014 ; Zimnik et al., 2019 ) and exerting inhibitory control to suppress action initiation when reward incentives are high ( Chen et al., 2010 ), thereby amplifying sensitivity to both effort and reward. A second major effect involved reduced integrity in the mid-anterior corpus callosum, a key pathway supporting interhemispheric communication, in part between bilateral SMAs, which was robustly associated with heightened effort sensitivity. This effect was strong and spatially consistent, spanning both FA and -MD metrics and forming symmetrical clusters across hemispheres. Extending the view that SMA contributions are not limited to interactions with other frontal valuation nodes (such as the OFC, dACC or NAcc), these findings point to interhemispheric communication as another critical substrate for individual differences in effort computation. Notably, our forceps minor results suggest a parallel role for commissural fibers linking bilateral OFCs in reward sensitivity (see below), underscoring the broader importance of interhemispheric integration in value computation. In addition to these SMA-connected pathways, additional clusters were identified in classical frontal valuation circuits. As described above, reward sensitivity was positively associated with integrity in the forceps minor, which interconnects bilateral OFCs, regions central to reward processing. Increased connectivity in this tract has also been linked to impulsivity ( Jeong et al., 2016 ), suggesting that stronger OFC communication may enhance reward signal integration and drive greater behavioral engagement when incentives are high. Similarly, higher integrity in the anterior cingulum bundle, connecting SMA and dACC, both key to effort processing, was associated with heightened effort sensitivity. While this might seem at odds with prior findings of reduced cingulum integrity in apathy ( Bonnelle et al., 2016 ), it is consistent with clinical evidence from anterior cingulotomy in obsessive-compulsive disorder, where lesions to this tract reduce action initiation ( Bubb et al., 2018 ). This may suggest that increased integrity in the anterior cingulum might enhance interactions between effort processing regions such as the SMA and dACC ( Innocenti et al., 2022 ; Thiebaut de Schotten and Forkel, 2022 ), amplifying the influence of high effort costs on behavioral disengagement. Notably, beyond the canonical frontal valuation regions, our data-driven analysis uncovered less expected white matter pathways linked to reward sensitivity, including the right superior longitudinal fasciculus, cerebellar connections, and the motor segment of the corpus callosum. These associations, unlikely to emerge from hypothesis-driven tractography restricted to predefined valuation circuits, highlight the capacity of whole-brain approaches to reveal non-canonical circuits. While the functional implications remain uncertain, these findings raise the possibility that reward-related signals are transmitted and modulated not only through core valuation hubs but also via broader fronto-parietal and motor networks, potentially influencing how individuals adjust behavioral engagement as rewards increase. Such a role would align with emerging evidence implicating fronto-parietal regions ( Etzel et al., 2016 ), the cerebellum ( Kostadinov and Häusser, 2022 ) and M1 ( Derosiere et al., 2017a , 2017b , 2025 ; Prévost et al., 2010 ; Pessiglione et al., 2007 ) in reward-related computations during decision-making. A key contribution of the present study is the demonstration that individual differences in effort and reward sensitivity can be accurately predicted from white matter microstructure using machine learning. By moving beyond correlational analyses and using a decoding framework, we show that DWI metrics within specific clusters carry sufficient information to decode effort and reward sensitivity across multiple classifiers, highlighting the robustness of these structure-behavior associations. This decoding framework builds on recent advances applying machine learning on DWI metrics in clinical settings (e.g., to predict Parkinson’s progression, Huang et al., 2024 or catatonia, Peretzke et al., 2025 ) and extends them to dimensional constructs relevant to motivation. Crucially, our findings suggest that microstructure metrics could be leveraged to anticipate the impact of white matter alterations, whether pathological ( e.g. , following stroke, multiple sclerosis, tumor infiltration, etc ) or iatrogenic ( e.g. , following surgical tumor resection, callosotomy, etc ), on goal-directed behavior. Such predictive modelling may support personalized risk assessment and pre-operative planning in the context of surgery ( Aylmore et al., 2025 ; Essayed et al., 2017 ; Ordonez-Rubiano et al., 2023 ), as well as inform targeted rehabilitation strategies (e.g., reward-based interventions, Vassiliadis et al., 2021 , 2022 ), and guide the development of neurotechnological interventions targeting key hubs of the valuation network ( Vassiliadis et al., 2024 ) particularly in cases of motivational impairments such as apathy or impulsivity, which remain challenging to anticipate. Together, these findings provide a robust anatomical mapping of individual differences in effort and reward sensitivity. By integrating data-driven tract identification with predictive modelling, we isolate specific white matter pathways, particularly SMA-connected tracts, that reliably account for variability in motivational processes. This framework offers a concrete basis for anticipating how structural disruptions may affect goal-directed behavior, with potential relevance for clinical applications in neurology and psychiatry. Conflict of interest statement The authors declare no conflicts of interest AUTHOR CONTRIBUTIONS NT: Machine learning analysis, interpretation of the results, data visualisation, article writing. LD, QD: MRI data processing, article writing. PV: computational modelling, interpretation of the results, article writing. JD: study conception, interpretation of the results, article writing. TW: Machine learning analysis, interpretation of the results. GD: study conception, data acquisition, interpretation of the results, article writing from the first draft to the final version. Acknowledgement Nam Trinh was supported by the financial grant from Science Foundation Ireland (SFI) Centre for Research Training in Machine Learning at Dublin City University under Grant number 18/CRT/6183. Gerard Derosiere was supported by a FNRS Research associate grant and by an ANR-JCJC grant (ANR-24-CE37-6839-01). Tomás Ward is supported by Research Ireland under grant agreement 12/RC/2289_P2. Pierre Vassiliadis was supported by a Swiss National Science Foundation (SNSF) fellowship (P500PB_230720). Julie Duque was supported by three FNRS research grants (T008219F, J005921F and T007023F). We would like to thank Matthieu Boisgontier, Ahmad Noureddine and Valentin Touzé for their participation in data acquisition. Funder Information Declared Science Foundation Ireland, https://ror.org/0271asj38 , 18/CRT/6183 FNRS , ANR-24-CE37-6839-01 Research Ireland , 12/RC/2289_P2 Swiss National Science Foundation, https://ror.org/00yjd3n13 , P500PB_230720 FNRS , T008219F , J005921F , T007023F REFERENCES Labra-Avila , N. , Lebenberg , J. , Uszynski , I. , Rivière , D. , Auzias , G. , Fischer , C. , Poupon , F. , Guevara , P. , Poupon , C. , & Mangin , J.-F . ( 2023 ). Atlas of the short fiber bundles inferred from the Human Connectome Project diffusion MRI dataset (v1.1) [Data set] . EBRAINS . doi: 10.25493/MK9H-ZE1 OpenUrl CrossRef ↵ Aylmore H , Young F , Aquilina K , Clark CA , Clayden JD ( 2025 ) The use of intraoperative tractography in brain tumor and epilepsy surgery: a systematic review and meta-analysis . Front Neuroimaging 4 Available at: https://www.frontiersin.org/journals/neuroimaging/articles/10.3389/fnimg.2025.1563996/full [Accessed July 30, 2025]. ↵ Baggio HC , Segura B , Garrido-Millan JL , Marti M-J , Compta Y , Valldeoriola F , Tolosa E , Junque C ( 2015 ) Resting-state frontostriatal functional connectivity in Parkinson’s disease-related apathy . Mov Disord Off J Mov Disord Soc 30 : 671 – 679 . OpenUrl ↵ Beck D , de Lange A-MG , Maximov II , Richard G , Andreassen OA , Nordvik JE , Westlye LT ( 2021 ) White matter microstructure across the adult lifespan: A mixed longitudinal and cross-sectional study using advanced diffusion models and brain- age prediction . NeuroImage 224 : 117441 . OpenUrl CrossRef PubMed ↵ Bonnelle V , Manohar S , Behrens T , Husain M ( 2016 ) Individual Differences in Premotor Brain Systems Underlie Behavioral Apathy . Cereb Cortex 26 : 807 – 819 . OpenUrl CrossRef PubMed ↵ Brainard DH ( 1997 ) The Psychophysics Toolbox . Spat Vis 10 : 433 – 436 . OpenUrl CrossRef PubMed Web of Science ↵ Bubb EJ , Metzler-Baddeley C , Aggleton JP ( 2018 ) The cingulum bundle: Anatomy, function, and dysfunction . Neurosci Biobehav Rev 92 : 104 – 127 . OpenUrl CrossRef PubMed ↵ Chen X , Scangos KW , Stuphorn V ( 2010 ) Supplementary Motor Area Exerts Proactive and Reactive Control of Arm Movements . J Neurosci 30 : 14657 – 14675 . OpenUrl Abstract / FREE Full Text ↵ Costello H , Husain M , Roiser JP ( 2024 ) Apathy and Motivation: Biological Basis and Drug Treatment . Annu Rev Pharmacol Toxicol 64 : 313 – 338 . OpenUrl CrossRef PubMed ↵ Derosiere G , Shokur S , Vassiliadis P ( 2025 ) Reward signals in the motor cortex: from biology to neurotechnology . Nat Commun 16 : 1307 . OpenUrl CrossRef PubMed ↵ Derosiere G , Vassiliadis P , Demaret S , Zénon A , Duque J ( 2017a ) Learning stage- dependent effect of M1 disruption on value-based motor decisions . NeuroImage 162 : 173 – 185 . OpenUrl CrossRef PubMed ↵ Derosiere G , Vassiliadis P , Dricot L , Dessain Q , Delinte N , Zénon A , Duque J ( 2024 ) Fronto-motor circuits linked to effort-based decision-making and apathy in healthy subjects . :2024.10.07.616861 Available at: https://www.biorxiv.org/content/10.1101/2024.10.07.616861v3 [Accessed May 29, 2025]. ↵ Derosiere G , Zénon A , Alamia A , Duque J ( 2017b ) Primary motor cortex contributes to the implementation of implicit value-based rules during motor decisions . NeuroImage 146 : 1115 – 1127 . OpenUrl CrossRef PubMed ↵ Dessain Q , Fuchs C , Macq B , Rensonnet G ( 2024 ) Fast multi-compartment Microstructure Fingerprinting in brain white matter . Front Neurosci 18 Available at: https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2024.14 00499/full [Accessed May 31, 2025]. ↵ Entakli J , Bonnard M , Chen S , Berton E , De Graaf JB ( 2014 ) TMS reveals a direct influence of spinal projections from human SMAp on precise force production . Eur J Neurosci 39 : 132 – 140 . OpenUrl CrossRef PubMed ↵ Essayed WI , Zhang F , Unadkat P , Cosgrove GR , Golby AJ , O’Donnell LJ ( 2017 ) White matter tractography for neurosurgical planning: A topography-based review of the current state of the art . NeuroImage Clin 15 : 659 – 672 . OpenUrl PubMed ↵ Etzel JA , Cole MW , Zacks JM , Kay KN , Braver TS ( 2016 ) Reward Motivation Enhances Task Coding in Frontoparietal Cortex . Cereb Cortex 26 : 1647 – 1659 . OpenUrl CrossRef PubMed ↵ Filbey FM , Aslan S , Calhoun VD , Spence JS , Damaraju E , Caprihan A , Segall J ( 2014 ) Long-term effects of marijuana use on the brain . Proc Natl Acad Sci 111 : 16913 – 16918 . OpenUrl Abstract / FREE Full Text ↵ Forman SD , Cohen JD , Fitzgerald M , Eddy WF , Mintun MA , Noll DC ( 1995 ) Improved Assessment of Significant Activation in Functional Magnetic Resonance Imaging (fMRI): Use of a Cluster-Size Threshold . Magn Reson Med 33 : 636 – 647 . OpenUrl CrossRef PubMed Web of Science ↵ Fried I , Katz A , McCarthy G , Sass KJ , Williamson P , Spencer SS , Spencer DD ( 1991 ) Functional organization of human supplementary motor cortex studied by electrical stimulation . J Neurosci 11 : 3656 – 3666 . OpenUrl Abstract / FREE Full Text ↵ Fuentes-Claramonte P , Ávila C , Rodríguez-Pujadas A , Costumero V , Ventura-Campos N , Bustamante JC , Rosell-Negre P , Barrós-Loscertales A ( 2016 ) Characterizing individual differences in reward sensitivity from the brain networks involved in response inhibition . NeuroImage 124 : 287 – 299 . OpenUrl CrossRef PubMed ↵ Garyfallidis E , Côté M-A , Rheault F , Sidhu J , Hau J , Petit L , Fortin D , Cunanne S , Descoteaux M ( 2018 ) Recognition of white matter bundles using local and global streamline-based registration and clustering . NeuroImage 170 : 283 – 295 . OpenUrl CrossRef PubMed ↵ Gilmour GS , Langer LK , Lang AE , MacGillivray L , Lidstone SC ( 2023 ) Neuropsychiatric phenotypes in functional movement disorder . CNS Spectr 28 : 747 – 755 . OpenUrl PubMed ↵ Hayakawa YK , Sasaki H , Takao H , Hayashi N , Kunimatsu A , Ohtomo K , Aoki S ( 2014 ) Depressive symptoms and neuroanatomical structures in community-dwelling women: A combined voxel-based morphometry and diffusion tensor imaging study with tract-based spatial statistics . NeuroImage Clin 4 : 481 – 487 . OpenUrl PubMed ↵ Heron CL , Holroyd CB , Salamone J , Husain M ( 2019 ) Brain mechanisms underlying apathy . J Neurol Neurosurg Psychiatry 90 : 302 – 312 . OpenUrl Abstract / FREE Full Text ↵ Hofer S , Frahm J ( 2006 ) Topography of the human corpus callosum revisited— Comprehensive fiber tractography using diffusion tensor magnetic resonance imaging . NeuroImage 32 : 989 – 994 . OpenUrl CrossRef PubMed Web of Science ↵ Hogan PS , Chen SX , Teh WW , Chib VS ( 2020 ) Neural mechanisms underlying the effects of physical fatigue on effort-based choice . Nat Commun 11 : 4026 . OpenUrl CrossRef PubMed ↵ Hoopes A , Mora JS , Dalca AV , Fischl B , Hoffmann M ( 2022 ) SynthStrip: skull-stripping for any brain image . NeuroImage 260 : 119474 . OpenUrl CrossRef PubMed ↵ Huang X , He Q , Ruan X , Li Y , Kuang Z , Wang M , Guo R , Bu S , Wang Z , Yu S , Chen A , Wei X ( 2024 ) Structural connectivity from DTI to predict mild cognitive impairment in de novo Parkinson’s disease . NeuroImage Clin 41 : 103548 . OpenUrl PubMed ↵ Husain M , Roiser JP ( 2018 ) Neuroscience of apathy and anhedonia: a transdiagnostic approach . Nat Rev Neurosci 19 : 470 – 484 . OpenUrl CrossRef PubMed ↵ Innocenti GM , Schmidt K , Milleret C , Fabri M , Knyazeva MG , Battaglia-Mayer A , Aboitiz F , Ptito M , Caleo M , Marzi CA , Barakovic M , Lepore F , Caminiti R ( 2022 ) The functional characterization of callosal connections . Prog Neurobiol 208 : 102186 . OpenUrl CrossRef PubMed ↵ Jeong BS , Han DH , Kim SM , Lee SW , Renshaw PF ( 2016 ) White matter connectivity and Internet gaming disorder . Addict Biol 21 : 732 – 742 . OpenUrl PubMed ↵ Klein-Flügge MC , Bongioanni A , Rushworth MFS ( 2022 ) Medial and orbital frontal cortex in decision-making and flexible behavior . Neuron 110 : 2743 – 2770 . OpenUrl CrossRef PubMed ↵ Koch PJ , Park C-H , Girard G , Beanato E , Egger P , Evangelista GG , Lee J , Wessel MJ , Morishita T , Koch G , Thiran J-P , Guggisberg AG , Rosso C , Kim Y-H , Hummel FC ( 2021 ) The structural connectome and motor recovery after stroke: predicting natural recovery . Brain 144 : 2107 – 2119 . OpenUrl CrossRef PubMed ↵ Kostadinov D , Häusser M ( 2022 ) Reward signals in the cerebellum: Origins, targets, and functional implications . Neuron 110 : 1290 – 1303 . OpenUrl CrossRef PubMed ↵ Le Heron C , Manohar S , Plant O , Muhammed K , Griffanti L , Nemeth A , Douaud G , Markus HS , Husain M ( 2018 ) Dysfunctional effort-based decision-making underlies apathy in genetic cerebral small vessel disease . Brain 141 : 3193 – 3210 . OpenUrl CrossRef PubMed ↵ Leaver AM , Espinoza R , Joshi SH , Vasavada M , Njau S , Woods RP , Narr KL ( 2016 ) Desynchronization and Plasticity of Striato-frontal Connectivity in Major Depressive Disorder . Cereb Cortex 26 : 4337 – 4346 . OpenUrl CrossRef PubMed ↵ Long X , Wang X , Tian F , Cao Y , Xie H , Jia Z ( 2022 ) Altered brain activation during reward anticipation in bipolar disorder . Transl Psychiatry 12 : 300 . OpenUrl PubMed ↵ López-Barroso D , Catani M , Ripollés P , Dell’Acqua F , Rodríguez-Fornells A , de Diego-Balaguer R ( 2013 ) Word learning is mediated by the left arcuate fasciculus . Proc Natl Acad Sci 110 : 13168 – 13173 . OpenUrl Abstract / FREE Full Text ↵ Luijten M , Schellekens AF , Kühn S , Machielse MWJ , Sescousse G ( 2017 ) Disruption of Reward Processing in Addiction : An Image-Based Meta-analysis of Functional Magnetic Resonance Imaging Studies . JAMA Psychiatry 74 : 387 – 398 . OpenUrl PubMed ↵ Lynch CJ et al. ( 2024 ) Frontostriatal salience network expansion in individuals in depression . Nature 633 : 624 – 633 . OpenUrl CrossRef PubMed ↵ Morris L-A , Horne K-L , Manohar S , Paermentier L , Buchanan CM , MacAskill MR , Myall DJ , Apps M , Roxburgh R , Anderson TJ , Husain M , Le Heron CJ ( 2025 ) Decision cost hypersensitivity underlies Huntington’s disease apathy . Brain 148 : 861 – 874 . OpenUrl PubMed ↵ Neige C , Vassiliadis P , Ali Zazou A , Dricot L , Lebon F , Brees T , Derosiere G ( 2023 ) Connecting the dots: harnessing dual-site transcranial magnetic stimulation to quantify the causal influence of medial frontal areas on the motor cortex . Cereb Cortex 33 : 11339 – 11353 . OpenUrl PubMed ↵ Oestreich LKL , Wright P , O’Sullivan MJ ( 2022 ) Hyperconnectivity and altered interactions of a nucleus accumbens network in post-stroke depression . Brain Commun 4 :fcac281. ↵ Oldfield RC ( 1971 ) The assessment and analysis of handedness: The Edinburgh inventory . Neuropsychologia 9 : 97 – 113 . OpenUrl CrossRef PubMed Web of Science ↵ Ordonez-Rubiano EG , Johnson JM , Abdalá-Vargas N , Zorro OF , Marin-Munoz JH , Álvarez-Tobián R , Forlizzi V , Rangel CC , Luzzi S , Campero A , Patiño-Gómez JG , Baldoncini M ( 2023 ) Preoperative tractography algorithm for safe resection of tumors located in the descending motor pathways zone . Surg Neurol Int 14 : 255 . OpenUrl PubMed ↵ Peretzke R , Neher PF , Brandt GA , Fritze S , Volkmer S , Daub J , Northoff G , Bohn J , Kirchhoff Y , Roy S , Maier-Hein KH , Meyer-Lindenberg A , Hirjak D ( 2025 ) Deciphering white matter microstructural alterations in catatonia according to ICD- 11: replication and machine learning analysis . Mol Psychiatry 30 : 2095 – 2107 . OpenUrl PubMed ↵ Pessiglione M , Schmidt L , Draganski B , Kalisch R , Lau H , Dolan RJ , Frith CD ( 2007 ) How the brain translates money into force: a neuroimaging study of subliminal motivation . Science 316 : 904 – 906 . OpenUrl Abstract / FREE Full Text ↵ Pessiglione M , Vinckier F , Bouret S , Daunizeau J , Le Bouc R ( 2018 ) Why not try harder? Computational approach to motivation deficits in neuro-psychiatric diseases . Brain 141 : 629 – 650 . OpenUrl CrossRef PubMed ↵ Potgieser ARE , de Jong BM , Wagemakers M , Hoving EW , Groen RJM ( 2014 ) Insights from the supplementary motor area syndrome in balancing movement initiation and inhibition . Front Hum Neurosci 8 Available at: https://www.frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2014.00960/full [Accessed July 11, 2025]. ↵ Prévost C , Pessiglione M , Météreau E , Cléry-Melin M-L , Dreher J-C ( 2010 ) Separate Valuation Subsystems for Delay and Effort Decision Costs . J Neurosci 30 : 14080 – 14090 . OpenUrl Abstract / FREE Full Text ↵ Schurr R , Reznik D , Hillman H , Bhui R , Gershman SJ ( 2024 ) Dynamic computational phenotyping of human cognition . Nat Hum Behav 8 : 917 – 931 . OpenUrl PubMed ↵ Shenhav A , Botvinick MM , Cohen JD ( 2013 ) The Expected Value of Control: An Integrative Theory of Anterior Cingulate Cortex Function . Neuron 79 : 217 – 240 . OpenUrl CrossRef PubMed Web of Science ↵ Song JY , Fleysher R , Ye K , Kim M , Zimmerman ME , Lipton RB , Lipton ML ( 2025 ) Characterizing the microstructural transition at the gray matter-white matter interface: Implementation and demonstration of age-associated differences . NeuroImage 306 : 121019 . OpenUrl PubMed ↵ Spieser L , Aubert S , Bonnard M ( 2013 ) Involvement of SMAp in the intention-related long latency stretch reflex modulation: a TMS study . Neuroscience 246 : 329 – 341 . OpenUrl CrossRef PubMed Web of Science ↵ Suzuki S , Lawlor VM , Cooper JA , Arulpragasam AR , Treadway MT ( 2021 ) Distinct regions of the striatum underlying effort, movement initiation and effort discounting . Nat Hum Behav 5 : 378 – 388 . OpenUrl PubMed ↵ Tarumi T , Fukuie M , Yamabe T , Kimura R , Zhu DC , Ohyama-Byun K , Maeda S , Sugawara J ( 2022 ) Microstructural organization of the corpus callosum in young endurance athletes: A global tractography study . Front Neurosci 16 Available at: https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2022.1042426/full [Accessed July 8, 2025]. ↵ Tay J , Tuladhar AM , Hollocks MJ , Brookes RL , Tozer DJ , Barrick TR , Husain M , de Leeuw F-E , Markus HS ( 2019 ) Apathy is associated with large-scale white matter network disruption in small vessel disease . Neurology 92 : e1157 – e1167 . OpenUrl CrossRef PubMed ↵ Thiebaut de Schotten M , Forkel SJ ( 2022 ) The emergent properties of the connected brain . Science 378 : 505 – 510 . OpenUrl CrossRef PubMed ↵ Tolomeo S , Yu R ( 2022 ) Brain network dysfunctions in addiction: a meta-analysis of resting-state functional connectivity . Transl Psychiatry 12 : 41 . OpenUrl PubMed ↵ Vassiliadis P , Beanato E , Popa T , Windel F , Morishita T , Neufeld E , Duque J , Derosiere G , Wessel MJ , Hummel FC ( 2024 ) Non-invasive stimulation of the human striatum disrupts reinforcement learning of motor skills . Nat Hum Behav 8 : 1581 – 1598 . OpenUrl PubMed ↵ Vassiliadis P , Derosiere G , Dubuc C , Lete A , Crevecoeur F , Hummel FC , Duque J ( 2021 ) Reward boosts reinforcement-based motor learning . iScience 24 Available at: https://www.cell.com/iscience/abstract/S2589-0042(21)00789-6 [Accessed July 30, 2025]. ↵ Vassiliadis P , Lete A , Duque J , Derosiere G ( 2022 ) Reward timing matters in motor learning . iScience 25 Available at: https://www.cell.com/iscience/abstract/S2589-0042(22)00560-0 [Accessed July 30, 2025]. ↵ Veraart J , Novikov DS , Christiaens D , Ades-aron B , Sijbers J , Fieremans E ( 2016 ) Denoising of diffusion MRI using random matrix theory . NeuroImage 142 : 394 – 406 . OpenUrl CrossRef PubMed ↵ Wahl M , Lauterbach-Soon B , Hattingen E , Jung P , Singer O , Volz S , Klein JC , Steinmetz H , Ziemann U ( 2007 ) Human Motor Corpus Callosum: Topography, Somatotopy, and Link between Microstructure and Function . J Neurosci 27 : 12132 – 12138 . OpenUrl Abstract / FREE Full Text ↵ Warrington S , Bryant KL , Khrapitchev AA , Sallet J , Charquero-Ballester M , Douaud G , Jbabdi S , Mars RB , Sotiropoulos SN ( 2020 ) XTRACT - Standardised protocols for automated tractography in the human and macaque brain . NeuroImage 217 : 116923 . OpenUrl CrossRef PubMed ↵ Xiong Y , Yang L , Wang C , Zhao C , Luo J , Wu D , Ouyang Y , Thiebaut de Schotten M , Gong G ( 2024 ) Cortical mapping of callosal connections in healthy young adults . Hum Brain Mapp 45 : e26629 . OpenUrl PubMed ↵ Yeh F-C , Panesar S , Fernandes D , Meola A , Yoshino M , Fernandez-Miranda JC , Vettel JM , Verstynen T ( 2018 ) Population-averaged atlas of the macroscale human structural connectome and its network topology . NeuroImage 178 : 57 – 68 . OpenUrl CrossRef PubMed ↵ Zénon A , Sidibé M , Olivier E ( 2015 ) Disrupting the supplementary motor area makes physical effort appear less effortful . J Neurosci Off J Soc Neurosci 35 : 8737 – 8744 . OpenUrl Abstract / FREE Full Text ↵ Zimnik AJ , Lara AH , Churchland MM ( 2019 ) Perturbation of Macaque Supplementary Motor Area Produces Context-Independent Changes in the Probability of Movement Initiation . J Neurosci 39 : 3217 – 3233 . OpenUrl Abstract / FREE Full Text View the discussion thread. Back to top Previous Next Posted August 19, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following White matter microstructure predicts effort and reward sensitivity Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share White matter microstructure predicts effort and reward sensitivity Nam Trinh , Laurence Dricot , Pierre Vassiliadis , Quentin Dessain , Julie Duque , Tomas Ward , Gerard Derosiere bioRxiv 2025.08.19.671080; doi: https://doi.org/10.1101/2025.08.19.671080 Share This Article: Copy Citation Tools White matter microstructure predicts effort and reward sensitivity Nam Trinh , Laurence Dricot , Pierre Vassiliadis , Quentin Dessain , Julie Duque , Tomas Ward , Gerard Derosiere bioRxiv 2025.08.19.671080; doi: https://doi.org/10.1101/2025.08.19.671080 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neuroscience Subject Areas All Articles Animal Behavior and Cognition (7622) Biochemistry (17648) Bioengineering (13871) Bioinformatics (41880) Biophysics (21423) Cancer Biology (18558) Cell Biology (25460) Clinical Trials (138) Developmental Biology (13364) Ecology (19866) Epidemiology (2067) Evolutionary Biology (24290) Genetics (15589) Genomics (22475) Immunology (17711) Microbiology (40327) Molecular Biology (17145) Neuroscience (88473) Paleontology (666) Pathology (2827) Pharmacology and Toxicology (4816) Physiology (7635) Plant Biology (15114) Scientific Communication and Education (2044) Synthetic Biology (4286) Systems Biology (9815) Zoology (2268)