A multivariate brain signature for reward
Poster No:
1598
Submission Type:
Abstract Submission
Authors:
Judit Campdepadrós Barrios1, Sebastian Speer2, Christian Keysers3, Cas Teurlings3, Ale Smidts4, Maarten Boksem4, Tor Wager5, Valeria Gazzola3
Institutions:
1Netherlands Institute for Neuroscience, Amsterdam, NH, 2Princeton Neuroscience Institute, Princeton, NJ, 3Netherlands Institute for Neuroscience, Amsterdam, North Holland, 4Rotterdam School of Management, Rotterdam, South Holland, 5Department of Psychological and Brain Sciences, Hanover, NH
First Author:
Co-Author(s):
Introduction:
Reward and punishment processing is vital for adaptive behavior, with its dysregulation implicated in mental health and substance use disorders. Traditional univariate approaches to reward processing often yield small effect sizes and limited reliability. In contrast, multivariate methods decode distributed neural patterns, providing higher sensitivity and specificity (Reddan et al. 2017). In this study (Speer et al. 2023), we developed the Brain Reward Signature (BRS), a multivariate brain model (Kragel et al. 2018) to predict the magnitude of rewards and losses in decision-making tasks.
Methods:
The BRS was trained using data from five fMRI tasks across four independent samples (N = 1169). Initially, we trained a LASSO-PCR model on data from a Monetary Incentive Delay (MIDtrain) task (N = 39), predicting three levels of monetary reward. We selected voxels correlating more strongly with outcomes than salience to ensure specificity and used 5-fold cross-validation to prevent overfitting. The study used out-of-sample prediction to assess the brain regions' contributions to predicting monetary outcomes (Fig. 1A), with prediction accuracy evaluated using Spearman correlation at each fold. The resulting pattern maps were thresholded at p<0.001 through bootstrap analysis in the MIDtrain dataset (Fig. 1B). Validation tests included another MID task (MIDval; N = 12: Srirangarajan et al. 2021), a Gambling task (HCP, N = 1084: Van Essen et al., 2012), a Disgust-Delay Task (DDT, N = 39), and an Emotion Viewing Task (EVT, N = 27: manuscript in preparation).
Results:
The BRS successfully predicted outcomes in unseen subjects (RMSE = 2.89, r = 0.72, pperm < 0.001, BF10 > 1000) and achieved 92% accuracy in decoding rewards versus losses (Fig. 2A). It generalized to a different MID version and sample (RMSE = 2.97, r = 0.75, pperm < 0.001, BF10 > 1000), with 92% forced-choice accuracy in decoding high reward versus high loss (Fig. 2B). It also predicted outcomes in a gambling task (r = 0.21, pperm < 0.001, BF10 > 100), achieving 73% forced-choice accuracy in distinguishing high reward versus high loss (Fig. 2C). The BRS also predicted non-financial reinforcers in the DDT (r = 0.38, pperm < 0.001, BF10 > 100), by discriminating successful from unsuccessful trials in the feedback phase with 92% accuracy (Fig. 2E). However, it did not generalize to emotional salient outcomes, failing to distinguish between disgusting and neutral images (r = -0.13, pperm < 0.28, BF10 = 0.23; Fig. 2D). In the EVT, BRS values were positive across both positive and negative emotional facial expressions (all t > 4.5, p <0.001, BF10 > 100; Fig. 2F), suggesting it may capture processes like morbid curiosity. Supporting this hypothesis, an online sample (N = 200) rated high-intensity emotional videos (happy, angry, fearful, painful, and disgusting) as more interesting than neutral ones (all p < 0.001, all BF10 > 46). In addition, the BRS has already been applied successfully in a moral conflict learning task, loading positively on prediction errors for both money (receiving higher than expected monetary rewards) and painful shocks to another individual (for less intense than expected shocks; Fornari et al., 2023).
Conclusions:
The BRS robustly decodes the magnitude of reward and losses across tasks, generalizing to different samples and non-financial reinforcers, namely positive feedback and avoiding pain. Crucially, its specificity to rewarding outcomes, without extending to choice-dependent disgust outcomes, aligns with its intended purpose of capturing reward-related processes rather than unrelated affective responses. Additionally, its positive loading on non-action-contingent outcomes suggests sensitivity to broader processes, such as morbid curiosity (Oosterwijk et al., 2020). These findings underscore the utility and reliability of the BRS within its intended scope, while recognizing that any signature can be further refined and validated by testing across more diverse datasets.
Emotion, Motivation and Social Neuroscience:
Reward and Punishment 2
Modeling and Analysis Methods:
Multivariate Approaches 1
Keywords:
MRI
Multivariate
Other - reward, loss, neural signature, decoding, machine learning
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Kragel, P. A., Koban, L., Barrett, L. F., & Wager, T. D. (2018). Representation, pattern information, and brain signatures: From neurons to neuroimaging. Neuron, 99(2), 257–273. https://doi.org/10.1016/j.neuron.2018.06.009
Oosterwijk, S., Snoek, L., Tekoppele, J. et al. Choosing to view morbid information involves reward circuitry. Sci Rep 10, 15291 (2020). https://doi.org/10.1038/s41598-020-71662-y
Reddan, M. C., Lindquist, M. A., & Wager, T. D. (2017). Effect size estimation in neuroimaging. JAMA Psychiatry, 74(3), 207–208. https://doi.org/10.1001/jamapsychiatry.2016.3356
Speer, S. P. H., Keysers, C., Campdepadrós Barrios, J., Teurlings, C. J. S., Smidts, A., Boksem, M. A. S., Wager, T. D., & Gazzola, V. (2023). A multivariate brain signature for reward. NeuroImage, 119990. https://doi.org/10.1016/j.neuroimage.2023.119990
Srirangarajan, T., Mortazavi, L., Bortolini, T., Moll, J., & Knutson, B. (2021). Multi‐band fMRI compromises detection of mesolimbic reward responses. NeuroImage, 237, 118617. https://doi.org/10.1016/j.neuroimage.2021.118617
Van Essen, D. C., Ugurbil, K., Auerbach, E., Barch, D., Behrens, T. E. J., Bucholz, R., Chang, A., Chen, L., Corbetta, M., Curtiss, S. W., Della Penna, S., Feinberg, D., Glasser, M. F., Harel, N., Heath, A. C., Larson-Prior, L., Marcus, D., Michalareas, G., Moeller, S., Oostenveld, R., Yacoub, E. (2012). The Human Connectome Project: A data acquisition perspective. NeuroImage, 62(4), 2222–2231. https://doi.org/10.1016/j.neuroimage.2012.02.018