科技科学

Neural basis of compositional control

Nature作者：Assia Chericoni2026年8月12日正文已收录本站

Data availability

Code availability

References

  1. Gordon, J. et al. The road towards understanding embodied decisions. Neurosci. Biobehav. Rev. 131, 722–736 (2021).

    Article  PubMed  PubMed Central  Google Scholar 

  2. Maselli, A. et al. Beyond simple laboratory studies: developing sophisticated models to study rich behavior. Phys. Life Rev. 46, 220–244 (2023).

    Article  ADS  PubMed  Google Scholar 

  3. Yoo, S. B. M., Hayden, B. Y. & Pearson, J. M. Continuous decisions. Philos. Trans. R. Soc. B 376, 20190664 (2021).

    Article  Google Scholar 

  4. Merel, J., Botvinick, M. & Wayne, G. Hierarchical motor control in mammals and machines. Nat. Commun. 10, 5489 (2019).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  5. Cisek, P. Making decisions through a distributed consensus. Curr. Opin. Neurobiol. 22, 927–936 (2012).

    Article  CAS  PubMed  Google Scholar 

  6. Gallivan, J. P., Chapman, C. S., Wolpert, D. M. & Flanagan, J. R. Decision-making in sensorimotor control. Nat. Rev. Neurosci. 19, 519–534 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Yoo, S. B. M., Tu, J. C., Piantadosi, S. T. & Hayden, B. Y. The neural basis of predictive pursuit. Nat. Neurosci. 23, 252–259 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Fabian, S. T., Sumner, M. E., Wardill, T. J., Rossoni, S. & Gonzalez-Bellido, P. T. Interception by two predatory fly species is explained by a proportional navigation feedback controller. J. R. Soc. Interface. 15, 20180466 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

  9. Sridhar, V. H. et al. The geometry of decision-making in individuals and collectives. Proc. Natl Acad. Sci. USA 118, e2102157118 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  10. Yang, Q. et al. Monkey plays Pac-Man with compositional strategies and hierarchical decision-making. eLife 11, e74500 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Bertsekas, D. P. in Encyclopedia of Optimization (eds Pardalos, P. M. & Prokopyev, O. A.) 1–6 (Springer, 2025).

  12. Sutton, R. S. & Barto, A. G. Reinforcement Learning: An Introduction (MIT Press, 1998).

  13. Theodorou, E. A., Buchli, J. & Schaal, S. A generalized path integral control approach to reinforcement learning. J. Mach. Learn. Res. 11, 3137–3181 (2010).

    MathSciNet  Google Scholar 

  14. Dvijotham, K. & Todorov, E. in Reinforcement Learning and Approximate Dynamic Programming for Feedback Control (eds Lewis. F. L. & Liu, D.) 119–141 (Wiley, 2012).

  15. Wolpert, D. M. & Kawato, M. Multiple paired forward and inverse models for motor control. Neural Netw. 11, 1317–1329 (1998).

    Article  CAS  PubMed  Google Scholar 

  16. Lake, B. & Baroni, M. Generalization without systematicity: on the compositional skills of sequence-to-sequence recurrent networks. In Proc. 35th International Conference on Machine Learning (eds. Dy, J. & Krause, A.) 2873–2882 (PMLR, 2018).

  17. Todorov, E. Compositionality of optimal control laws. In Advances in Neural Information Processing Systems 22 (eds Bengio, Y. et al.) (NeurIPS, 2009).

  18. Kurth-Nelson, Z. et al. Replay and compositional computation. Neuron 111, 454–469 (2023).

    Article  CAS  PubMed  Google Scholar 

  19. Whittington, J. C. R., McCaffary, D., Bakermans, J. J. W. & Behrens, T. E. J. How to build a cognitive map. Nat. Neurosci. 25, 1257–1272 (2022).

    Article  CAS  PubMed  Google Scholar 

  20. Eichenbaum, H. & Cohen, N. J. Can we reconcile the declarative memory and spatial navigation views on hippocampal function? Neuron 83, 764–770 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Behrens, T. E. J. et al. What is a cognitive map? Organizing knowledge for flexible behavior. Neuron 100, 490–509 (2018).

    Article  CAS  PubMed  Google Scholar 

  22. Kay, K. et al. Constant sub-second cycling between representations of possible futures in the hippocampus. Cell 180, 552–567.e25 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Park, S. A., Miller, D. S., Nili, H., Ranganath, C. & Boorman, E. D. Map making: constructing, combining, and inferring on abstract cognitive maps. Neuron 107, 1226–1238.e8 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  24. Sanders, H., Wilson, M. A. & Gershman, S. J. Hippocampal remapping as hidden state inference. eLife 9, e51140 (2020).

    Article  PubMed  PubMed Central  Google Scholar 

  25. Rushworth, M. F. S., Noonan, M. P., Boorman, E. D., Walton, M. E. & Behrens, T. E. Frontal cortex and reward-guided learning and decision-making. Neuron 70, 1054–1069 (2011).

    Article  CAS  PubMed  Google Scholar 

  26. Wikenheiser, A. M. & Schoenbaum, G. Over the river, through the woods: cognitive maps in the hippocampus and orbitofrontal cortex. Nat. Rev. Neurosci. 17, 513–523 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Yeung, N. & Summerfield, C. Metacognition in human decision-making: confidence and error monitoring. Philos. Trans. R. Soc. B 367, 1310–1321 (2012).

    Article  Google Scholar 

  28. Eppinger, B., Goschke, T. & Musslick, S. Meta-control: From psychology to computational neuroscience. Cogn. Affect. Behav. Neurosci. 21, 447–452 (2021).

    Article  PubMed  Google Scholar 

  29. Musslick, S., Cohen, J. D. & Goschke, T. in Encyclopedia of the Human Brain (ed. Grafman, J. H.) 269–285 (Elsevier, 2025).

  30. Alexander, W. H. & Brown, J. W. Medial prefrontal cortex as an action-outcome predictor. Nat. Neurosci. 14, 1338–1344 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Kennerley, S. W., Walton, M. E., Behrens, T. E. J., Buckley, M. J. & Rushworth, M. F. S. Optimal decision making and the anterior cingulate cortex. Nat. Neurosci. 9, 940–947 (2006).

    Article  CAS  PubMed  Google Scholar 

  32. Kolling, N., Behrens, T., Wittmann, M. & Rushworth, M. Multiple signals in anterior cingulate cortex. Curr. Opin. Neurobiol. 37, 36–43 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Shenhav, A., Botvinick, M. M. & Cohen, J. D. The expected value of control: an integrative theory of anterior cingulate cortex function. Neuron 79, 217–240 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Akam, T. et al. The anterior cingulate cortex predicts future states to mediate model-based action selection. Neuron 109, 149–163.e7 (2021).

    Article  CAS  PubMed  Google Scholar 

  35. Heilbronner, S. R. & Hayden, B. Y. Dorsal anterior cingulate cortex: a bottom-up view. Annu. Rev. Neurosci. 39, 149–170 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Sarafyazd, M. & Jazayeri, M. Hierarchical reasoning by neural circuits in the frontal cortex. Science 364, eaav8911 (2019).

    Article  CAS  PubMed  Google Scholar 

  37. Padoa-Schioppa, C. Neurobiology of economic choice: a good-based model. Annu. Rev. Neurosci. 34, 333–359 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Hunt, L. T. et al. Triple dissociation of attention and decision computations across prefrontal cortex. Nat. Neurosci. 21, 1471–1481 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  39. Wilson, R. C., Takahashi, Y. K., Schoenbaum, G. & Niv, Y. Orbitofrontal cortex as a cognitive map of task space. Neuron 81, 267–279 (2014).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  40. Elston, T. W. & Wallis, J. D. Context-dependent decision-making in the primate hippocampal–prefrontal circuit. Nat. Neurosci. 28, 374–382 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Yoo, S. B. M., Tu, J. C. & Hayden, B. Y. Multicentric tracking of multiple agents by anterior cingulate cortex during pursuit and evasion. Nat. Commun. 12, 1985 (2021).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

  42. Wilson, R. C. & Collins, A. G. E. Ten simple rules for the computational modeling of behavioral data. eLife 8, e49547 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

  43. Khona, M. & Fiete, I. R. Attractor and integrator networks in the brain. Nat. Rev. Neurosci. 23, 744–766 (2022).

    Article  CAS  PubMed  Google Scholar 

  44. Hayden, B. Y., Pearson, J. M. & Platt, M. L. Neuronal basis of sequential foraging decisions in a patchy environment. Nat. Neurosci. 14, 933–939 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Anderson, D. J. & Perona, P. Toward a science of computational ethology. Neuron 84, 18–31 (2014).

    Article  CAS  PubMed  Google Scholar 

  46. Brown, A. E. X. & de Bivort, B. Ethology as a physical science. Nat. Phys. 14, 653–657 (2018).

    Article  CAS  Google Scholar 

  47. Milner, D. & Goodale, M. The Visual Brain in Action (Oxford Univ. Press, 2006).

  48. Gershman, S. J. & Niv, Y. Learning latent structure: carving nature at its joints. Curr. Opin. Neurobiol. 20, 251–256 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  49. Zutshi, I. et al. Hippocampal neuronal activity is aligned with action plans. Nature 639, 153–161 (2025).

    Article  ADS  CAS  PubMed  Google Scholar 

  50. Shadmehr, R. & Krakauer, J. W. A computational neuroanatomy for motor control. Exp. Brain Res. 185, 359–381 (2008).

    Article  PubMed  PubMed Central  Google Scholar 

  51. Bakkour, A. et al. The hippocampus supports deliberation during value-based decisions. eLife 8, e46080 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

  52. Stachenfeld, K. L., Botvinick, M. M. & Gershman, S. J. The hippocampus as a predictive map. Nat. Neurosci. 20, 1643–1653 (2017).

    Article  CAS  PubMed  Google Scholar 

  53. Vikbladh, O. M. et al. Hippocampal contributions to model-based planning and spatial memory. Neuron 102, 683–693.e4 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  54. Edelson, M. G. & Hare, T. A. Goal-dependent hippocampal representations facilitate self-control. J. Neurosci. 43, 7822–7830 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  55. Cohen, J. D., Botvinick, M. & Carter, C. S. Anterior cingulate and prefrontal cortex: who’s in control?. Nat. Neurosci. 3, 421–423 (2000).

    Article  CAS  PubMed  Google Scholar 

  56. Verguts, T. Binding by random bursts: a computational model of cognitive control. J. Cogn. Neurosci. 29, 1103–1118 (2017).

    Article  PubMed  Google Scholar 

  57. Silvetti, M., Vassena, E., Abrahamse, E. & Verguts, T. Dorsal anterior cingulate-brainstem ensemble as a reinforcement meta-learner. PLoS Comput. Biol. 14, e1006370 (2018).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

  58. Friedman, A. et al. A corticostriatal path targeting striosomes controls decision-making under conflict. Cell 161, 1320–1333 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  59. Kane, G. A. et al. Rat anterior cingulate cortex continuously signals decision variables in a patch foraging task. J. Neurosci. 42, 5730–5744 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  60. Botvinick, M. M., Niv, Y. & Barto, A. G. Hierarchically organized behavior and its neural foundations: A reinforcement learning perspective. Cognition 113, 262–280 (2009).

    Article  PubMed  Google Scholar 

  61. Franch, M. et al. A population code for semantics in human hippocampus. Preprint at bioRxiv https://doi.org/10.1101/2025.02.21.639601 (2025).

  62. Chaure, F. J., Rey, H. G. & Quian Quiroga, R. A novel and fully automatic spike-sorting implementation with variable number of features. J. Neurophysiol. 120, 1859–1871 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  63. Groppe, D. M. et al. iELVis: An open source MATLAB toolbox for localizing and visualizing human intracranial electrode data. J. Neurosci. Methods 281, 40–48 (2017).

    Article  CAS  PubMed  Google Scholar 

  64. Jenkinson, M. & Smith, S. A global optimisation method for robust affine registration of brain images. Med. Image Anal. 5, 143–156 (2001).

    Article  CAS  PubMed  Google Scholar 

  65. Jenkinson, M., Bannister, P., Brady, M. & Smith, S. Improved optimization for the robust and accurate linear registration and motion correction of brain images. Neuroimage 17, 825–841 (2002).

    Article  PubMed  Google Scholar 

  66. Joshi, A. et al. Unified framework for development, deployment and robust testing of neuroimaging algorithms. Neuroinformatics 9, 69–84 (2011).

    Article  PubMed  PubMed Central  Google Scholar 

  67. Dale, A. M., Fischl, B. & Sereno, M. I. Cortical surface-based analysis. Neuroimage 9, 179–194 (1999).

    Article  CAS  PubMed  Google Scholar 

  68. Yang, A. I. et al. Localization of dense intracranial electrode arrays using magnetic resonance imaging. Neuroimage 63, 157–165 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

  69. Magnotti, J. F., Wang, Z. & Beauchamp, M. S. RAVE: Comprehensive open-source software for reproducible analysis and visualization of intracranial EEG data. Neuroimage 223, 117341 (2020).

    Article  PubMed  PubMed Central  Google Scholar 

  70. Wang, Z., Magnotti, J. F., Zhang, X. & Beauchamp, M. S. YAEL: your advanced electrode localizer. eNeuro 10, ENEURO.0328-23.2023 (2023).

    Article  PubMed  PubMed Central  Google Scholar 

  71. Chericoni, A. et al. Neural geometry in the human hippocampus enables generalization across spatial position and gaze. Preprint at arXiv https://doi.org/10.48550/arXiv.2603.04747 (2026).

  72. Gómez, V., Kappen, H. J., Peters, J. & Neumann, G. in Machine Learning and Knowledge Discovery in Databases (eds Calders, T. et al.) 482–497 (2014).

  73. Peng, X. Bin, Chang, M., Zhang, G., Abbeel, P. & Levine, S. MCP: learning composable hierarchical control with multiplicative compositional policies. In Advances in Neural Information Processing Systems 32 (eds Wallach, H. et al.) (NeurIPS, 2019).

  74. Matsuo, Y. et al. Deep learning, reinforcement learning, and world models. Neural Netw. 152, 267–275 (2022).

    Article  ADS  PubMed  Google Scholar 

  75. Murphy, K. P. Probabilistic Machine Learning: An Introduction (The MIT Press, 2022).

  76. Wood, S. N. Generalized additive models. Annu. Rev. Stat. Appl. 12, 497–526 (2025).

    Article  MathSciNet  Google Scholar 

  77. Seabold, S. & Perktold, J. Statsmodels: econometric and statistical modeling with Python. In Proc. 9th Python in Science Conference (SciPy 2010) https://doi.org/10.25080/Majora-92bf1922-011 (SciPy, 2010).

  78. Balzani, E., Lakshminarasimhan, K., Angelaki, D. & Savin, C. Efficient estimation of neural tuning during naturalistic behavior. In Advances in Neural Information Processing Systems 33 (eds. Larochelle, H. et al.) (NeurIPS, 2020).

  79. Wood, S. N. Generalized Additive Models: An Introduction with R (CRC Press/Taylor & Francis Group, 2017).

  80. Gelman, A., Hwang, J. & Vehtari, A. Understanding predictive information criteria for Bayesian models. Stat. Comput. 24, 997–1016 (2014).

    Article  MathSciNet  Google Scholar 

  81. Kriegeskorte, N. & Wei, X.-X. Neural tuning and representational geometry. Nat. Rev. Neurosci. 22, 703–718 (2021).

    Article  CAS  PubMed  Google Scholar 

  82. Kobak, D. et al. Demixed principal component analysis of neural population data. eLife 5, e10989 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

  83. Chericoni, A. Dataset for: Neural basis of compositional control. Figshare https://doi.org/10.6084/m9.figshare.32572764.v3 (2026).

Download references

Acknowledgements

We thank P. Schrater, X. Pitkow, J. Pearson, S. Linderman, H. Ritz and A. Shenhav for invaluable discussions and insights on solving the controller problems; J. Johnston for discussion on neural analysis; J. Adkinson, R. Mathura and V. Pirtle for invaluable assistance; and A. Thome for early and visionary efforts in developing this task and core questions.

Funding

This project was supported by NIH grants R01 DA038615, R01 MH125377, U01 NS121472, and R01 MH129439, and by the McNair foundation.

Author information

Author notes

  1. These authors contributed equally: Assia Chericoni, Justin M. Fine

  2. These authors jointly supervised this work: Sameer A. Sheth, Benjamin Y. Hayden

Authors and Affiliations

  1. Department of Neurosurgery, Baylor College of Medicine, Houston, TX, USA

    Assia Chericoni, Justin M. Fine, Taha S. Ismail, Melissa C. Franch, Elizabeth Mickiewicz, Ana G. Chavez, Eleonora Bartoli, Danika L. Paulo, Garrett P. Banks, Nisha Giridharan, Nicole R. Provenza, Andrew Watrous, Sameer A. Sheth & Benjamin Y. Hayden

  2. Department of Neuroscience, University of Minnesota, Minneapolis, MN, USA

    Gabriela Delgado Salazar

  3. Department of Electrical and Computer Engineering, Rice University, Houston, TX, USA

    Vaishnav Krishnan, Nicole R. Provenza, Sameer A. Sheth & Benjamin Y. Hayden

  4. Department of Neuroscience, Baylor College of Medicine, Houston, TX, USA

    Eleonora Bartoli, Vaishnav Krishnan, Nicole R. Provenza, Sameer A. Sheth & Benjamin Y. Hayden

  5. Department of Neurology, Baylor College of Medicine, Houston, TX, USA

    Vaishnav Krishnan, Mohamed Hegazy, Alica M. Goldman & Lu Lin

  6. Department of Psychiatry and Behavioral Sciences, Baylor College of Medicine, Houston, TX, USA

    Vaishnav Krishnan & Sameer A. Sheth

  7. Department of Neurosurgery, Michael E. DeBakey VA Medical Center, Houston, TX, USA

    Garrett P. Banks

  8. Department of Neurosurgery, College of Medicine, Imam Abdulrahman Bin Faisal University, Dammam, Saudi Arabia

    Mohammed Hasen

  9. Department of Bioengineering, Rice University, Houston, TX, USA

    Nicole R. Provenza

  10. Neuroengineering Initiative, Rice University, Houston, TX, USA

    Nicole R. Provenza, Sameer A. Sheth & Benjamin Y. Hayden

  11. Center for Neuroscience Imaging Research, Sungkyunkwan University, Suwon, South Korea

    Seng Bum Michael Yoo

  12. Department of Biomedical Engineering, Sungkyunkwan University, Suwon, South Korea

    Seng Bum Michael Yoo

  13. Department of Intelligence Health Precision Convergence, Sungkyunkwan University, Suwon, South Korea

    Seng Bum Michael Yoo

  14. Gordon and Mary Cain Pediatric Neurology Research Foundation Laboratories, Jan and Dan Duncan Neurological Research Institute, Texas Children’s Hospital, Houston, TX, USA

    Sameer A. Sheth

Authors

  1. Assia Chericoni
  2. Justin M. Fine
  3. Taha S. Ismail
  4. Gabriela Delgado Salazar
  5. Melissa C. Franch
  6. Elizabeth Mickiewicz
  7. Ana G. Chavez
  8. Eleonora Bartoli
  9. Danika L. Paulo
  10. Vaishnav Krishnan
  11. Mohamed Hegazy
  12. Alica M. Goldman
  13. Lu Lin
  14. Garrett P. Banks
  15. Nisha Giridharan
  16. Mohammed Hasen
  17. Nicole R. Provenza
  18. Andrew Watrous
  19. Seng Bum Michael Yoo
  20. Sameer A. Sheth
  21. Benjamin Y. Hayden

Contributions

J.M.F. and A.C. devised and further developed the modelling approach, respectively. J.M.F., A.C. and T.S.I. validated the modelling approach. J.M.F., A.C. and T.S.I. implemented the analysis codes. A.C., J.M.F. and B.Y.H. designed the experimental procedure and downstream analysis. D.L.P., G.P.B., N.G., M. Hasen and S.A.S. performed the intracranial stereotactic surgery. V.K., M. Hegazy, A.M.G. and L.L. provided neurological care for the study participants. J.M.F., A.C. and B.Y.H. drafted the manuscript. J.M.F. and A.C. devised and further developed the neural analyses, respectively. J.M.F., A.C. and T.S.I. implemented them. A.C., T.S.I., M.C.F., E.A.M. and A.G.C. collected the sEEG data. G.D.S., A.C. and E.A.M. collected the behavioural data. E.B., N.R.P. and A.W. contributed to data collection. E.B., N.R.P., A.W. and S.B.M.Y. contributed to data interpretation.

Corresponding author

Correspondence to Benjamin Y. Hayden.

Ethics declarations

Competing interests

S.A.S. is a consultant for Boston Scientific, Abbott, Koh Young, Neuropace, Zimmer Biomet and co-founder of Motif Neurotech. The other authors declare no competing interests.

Peer review

Peer review information

Nature thanks Sebastian Musslick and the other, anonymous, reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Fig. 1 Model recovery of wt value and time at switch point (wt = 0.5, t = 0 ms).

(A). Empirical simulation regime (n = 30 simulations × 30 trials × 2 prey-controllers): distribution of recovered wt values at the true switch point (wt = 0.5), quantifying bias in switch magnitude recovery across controller models. Red line indicates the median recovered value; dashed lines indicate the central 95% interval of the recovered distribution. (B). Distribution of temporal lag between recovered and true switch times for wt = 0.5, quantifying timing accuracy of switch detection across controller models. Red line indicates the median lag; dashed lines indicate the 95% CI. (C–D). Same as (A–B), but for the random 1-5 simulation regime. (E–F). Same as (A–B), but for the random 1-40 simulation regime.

Extended Data Fig. 2 Model recovery of controller gains across controller classes.

(A). Empirical simulation regime (n = 30 simulations × 30 trials × 2 prey-controllers): scatter plot showing the relationship between the simulated gains and the gains recovered from the model. Each point represents one recovered gain form a simulated trial. Each column represents a controller class. (B). Same as panel A, but for the random 1-5 simulation regime. (C). Same as panel A, but for the random 1-40 simulation regime.

Extended Data Fig. 3 Relationship between recovered controller gains and recovered wt trajectories across controller classes.

(A). Empirical simulation regime (30 simulations × 30 trials × 2 prey-specific controllers): scatter plots showing the relationship between recovered controller gains and recovered wt trajectories at the start, middle, and end of the trial. Each point represents one recovered gain estimate from one prey-specific policy in a simulated trial. Each column represents a controller class. Pearson’s correlation coefficient (r) is shown for each trial epoch. (B). Same as panel A, but for the random 1-5 simulation regime. (C). Same as panel A, but for the random 1-40 simulation regime.

Extended Data Fig. 4 Pairwise relationships between recovered controller gains across controller classes.

(A). Empirical simulation regime (30 simulations × 30 trials × 2 prey-specific controllers): scatter plots showing pairwise relationships between recovered controller gains across controller classes. Each point represents one recovered gain estimate from a simulated trial. Pearson’s correlation coefficient (r) is shown for each gain pair. (B). Same as panel A, but for the random 1-5 simulation regime. (C). Same as panel A, but for the random 1-40 simulation regime.

Extended Data Fig. 5 Model recovery of controller class and blending parameter wt.

We simulated ground-truth trajectories using known controller class (p) and known wt profiles, considering five types of wt. Simulations were generated by sampling 20 trials across sessions, with trial durations matched to the empirical distribution (approximately 0.91, 1.45, 2.83, and 4.1 s). To fix model complexity, simulations were fit using a fixed number of basis functions, rather than the model-averaging procedure used in the main text. (A). Posterior probability of recovering the true generative controller model relative to a baseline position-only controller. Across all controller classes, the true generative model was recovered above baseline (all probabilities > 0.5). Controller classes are defined in Extended Data Table 3. (B). Correlation between true and predicted position trajectories for each generative controller class, averaged across simulated trials, showing good recovery across classes (mean r = 0.83). (C–H). Example simulations showing recovery of the ground-truth blending parameter wt for each controller class. Blue lines indicate true wt, orange lines indicate recovered wt. Left panels (scaling = 1) show low-complexity trajectories, and right panels (scaling = 5) show higher-complexity trajectories. Complexity was controlled by the scale parameter of the Gaussian process used to generate wt. Across models, the mean recovery of wt was r = 0.87.

Extended Data Fig. 6 Full model recovery confusion matrix.

To assess overall model identifiability, we performed a full confusion analysis across three simulation regimes. For each generative model corresponding to a controller class (Extended Data Table 3), we simulated datasets with known parameters (30 datasets with 30 trials each), fit all candidate models to each dataset, and selected the best-fitting model using the evidence lower bound (ELBO). (A–C). Confusion matrices quantifying how reliably each generative controller class could be distinguished from the others under the (A). empirical simulation regime, (B). random 1-5, and (C). random 1-40 simulation regimes. Recoverability was significantly above chance for controller classes p, pv and pvi (binomial test, all p < 0.0001), whereas the remaining classes showed weaker identifiability. The pv model, which provided the best fit to participant behavior, was significantly recovered (43% correct; 95% CI [0.40, 0.47], p < 0.0001), but was frequently confounded with pvi and pvf. This overlap likely reflects the nested structure of the controller classes, indicating that interpretation of fitted model identities should account for partial non-identifiability between related controller families.

Extended Data Fig. 7 Behavioral predictors of target switching and average controller dynamics.

(A). Target switch GLM restricted to sEEG subjects (n = 19), showing the contribution of behavioral predictors at two time windows before the switch onset. Left: predictors computed as the mean value within the 100 ms preceding switch onset (−100 to 0 ms). Right: predictors computed as the mean value from an earlier window preceding switch onset (−400 to −300ms). Bars indicate the mean posterior effect size (percent change), and error bars indicate the 95% credible interval, as in Fig. 2h. Asterisks indicate predictors whose 95% credible interval excluded zero. (C). Averaged wt trajectory across all the participants (n = 69).

Extended Data Table 1 Experimental cohorts and participants demographics

Full size table

Extended Data Table 2 Number of recorded neurons across participants and brain regions

Full size table

Extended Data Table 3 Controller types

Full size table

Supplementary information

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Chericoni, A., Fine, J.M., Ismail, T.S. et al. Neural basis of compositional control. Nature (2026). https://doi.org/10.1038/s41586-026-10896-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • Version of record:

  • DOI: https://doi.org/10.1038/s41586-026-10896-8