Pith. sign in

REVIEW 3 major objections 4 minor 60 references

A modular state-space model of human perception, cognition, and decision dynamics

T0 review · 3 major / 4 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A modular state-space model aims to make human perception, cognition, and decision dynamics jointly analyzable and controllable.

desk verdict A well-built modular state-space framework with honest stability analysis, but the psychological labels are a labeling convention until human data appears. read the letter →

arxiv 2607.14078 v1 pith:HXHVJYTA submitted 2026-07-15 eess.SY cs.SYq-bio.NC

Sven Schoonebeek , Anahita Jamshidnejad , Carlo Cenedese This is my paper · ORCID
classification eess.SYcs.SYq-bio.NC
keywords humancognitionmodularstate-spacemodellatentinternaldynamicsperception-cognition-decisionpipelinestabilityanalysisinput-to-statepredictivecontrolrehabilitationhuman-robotinteraction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that human behavior can be written as a coupled perception–cognition–decision pipeline in discrete-time state-space form, with latent states for attention, perceptual estimates, beliefs, goals, emotions, intentions, and actions. It proves sufficient conditions for these states to stay bounded, invariant, and stable under admissible sensory inputs, so the model is not just interpretable but also analyzable with tools from nonlinear control. The authors further claim that a receding-horizon controller built on this model can adapt task difficulty in a simulated rehabilitation setting and achieves lower realized cumulative cost than target-following and random baselines. A sympathetic reader would care because this promises an alternative to black-box behavior predictors: a white-box dynamical structure that supports estimation, validation, and feedback control in human-centered systems.

What carries the argument

The load-bearing object is the coupled state-space recursion: the LPE update (equation 7) rewrites predictive inference as a convex combination, guaranteeing boundedness; the cognition update (equation 8) combines an exponential self-inhibition term (9), a DCM-style coupling matrix (10) with baseline, input-gated, and state-gated components, and an additive perceptual drive; and the decision module (equations 14–16) maps goals and beliefs to intentions and then to actions through a sigmoid gate with competition and thresholding. The stability proofs rest on a norm-ball dissipativity certificate and a contraction condition α_R < 1 that quantifies when self-inhibition dominates recurrent ampli

What would settle it

Fit the model to human behavioral data and check two things: whether the estimated latent 'threat' and 'fatigue' states track independent physiological or self-report measures, and whether the forward-invariance and ISS conditions hold on real trajectories; failure on either would undercut the central claim, given the paper's own acknowledgment that the rehabilitation showcase is model-class-matched and not an empirical validation.

Watch

Extended reading notes

Core claim

The central claim is that a modular state-space model—composed of divisive-normalization attentional selection, a convex-combination predictive-inference update for a latent perceptual estimate, DCM-style self-inhibition and bilinear couplings among cognitive states, and gated intention/action selection—simultaneously achieves psychological interpretability and mathematical tractability. Under stated regularity, boundedness, and dissipativity assumptions, the paper establishes forward invariance of the perceptual-estimate set, exponential contraction of perceptual inference under constant input, and input-to-state stability of the cognitive-state dynamics on a bounded operating set. The clos

Load-bearing premise

The paper's interpretability and personalization claims stand or fall on whether the psychological labels attached to its latent states (goals, beliefs, emotions) correspond to measurable, validated constructs—Remark 2 states these labels are interpretive, not implied by the equations—and the stability analysis assumes a fixed-step forward-Euler discretization faithfully approximates human cognitive dynamics.

Editorial extensions

If this is right

  • If the central claim holds, latent cognitive states such as goals, beliefs, and emotions can be estimated and constrained inside a bounded operating set, making them usable in feedback control without divergence.
  • The model's parameters have interpretable directional effects: increasing prior precision smooths perceptual estimates but worsens tracking, while stronger self-inhibition stabilizes cognition and reduces state excursions.
  • A receding-horizon controller using the model can keep a simulated patient in the 'perform' regime longer than feedback-free baselines, suggesting a path toward difficulty adaptation in rehabilitation.
  • The stability conditions identify parameter regimes in which the model is safe for prediction and control, while deliberately excluding regimes intended to represent dysregulated or strongly conflicted internal states.
  • The modular structure supports compositional analysis and future extensions to observer-based state estimation and robust or stochastic cognitive MPC.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the model's qualitative predictions generalize to humans, they imply that simple mechanisms—divisive normalization, leaky integrators, and gated thresholds—may be sufficient to reproduce observed trade-offs between perceptual tracking and smoothing in sensorimotor tasks.
  • The stability certificates could be used to design 'safe' personalization algorithms that keep intervention parameters within a region where the model's latent states remain bounded, avoiding experimentally induced dysregulation.
  • A testable extension suggested by the sensitivity analysis is that increasing attention gain should improve perceptual tracking while increasing reactivity, a prediction that could be checked in controlled human psychophysics experiments.
  • Because the controller relies on exhaustive enumeration, scaling to richer action spaces would require approximate optimization; a natural next test is whether a learned residual model can correct structural mismatch between the model class and real human behavior.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a modular discrete-time state-space model of the perception–cognition–decision loop, with explicit attentional selection, predictive inference, DCM-inspired cognitive-state dynamics, intention formation, and action selection. Section 4 derives sufficient conditions for boundedness, Lipschitz regularity, forward invariance, contraction of the perceptual inference under constant input, and regional ISS of the cognitive dynamics. Section 5 reports one-at-a-time parameter sensitivity analyses with synthetic inputs and a closed-loop rehabilitation showcase in which a receding-horizon controller uses a perturbed copy of the same model to adapt movement difficulty for a simulated, model-class-matched patient. The mathematical development is self-contained and detailed, and the authors candidly state several limitations, including the absence of human-participant validation, the model-class-matched nature of the rehabilitation study, and the lack of formal identifiability results.

Significance. The main strength is the formal, modular dynamical formulation: the constituent mappings are explicit, the assumptions are stated, and the stability analysis in Appendix A is substantive rather than ornamental. The paper also ships public code and reports precise sensitivity metrics, which supports reproducibility. If the framework is later connected to measured human behavior, it could provide a useful white-box template for estimation and control in human-centered systems. However, the significance is currently prospective: the psychological interpretability rests on latent-state labels that the authors themselves state are not implied by the equations, and the numerical/control results do not involve human data or structural model mismatch. The paper would be strengthened by either adding an empirical/identifiability component or explicitly reframing the contribution as a formal modeling framework whose psychological semantics are hypotheses to be tested, not established properties.

major comments (3)
  1. [§3.2, Remark 2; §5.5] The central value proposition is a psychologically interpretable latent-state model, but Remark 2 concedes that the labels 'beliefs', 'goals', and 'emotions' are not properties implied by the state equations. Section 5.5 further states that the sensitivity analysis is 'not a formal identifiability result'. Without construct-relevant measurements or an identifiability study, the interpretability claim is a labeling convention rather than an established property. This is load-bearing because the paper's contribution is framed as white-box and interpretable. I recommend either adding a validation or identifiability component, or substantially reframing the contribution as a mathematically analyzable mechanistic model whose psychological semantics remain to be established.
  2. [§5.6 and Conclusion] The rehabilitation showcase uses a simulated patient generated by the same model class as the robot's prediction model. The authors acknowledge in the conclusion that the experiment 'does not test robustness to structural mismatch with real human behavior', but the abstract states without qualification that the controller 'achieves lower realized cumulative cost' than baselines. The reported advantage is a self-consistency check, not evidence of control efficacy for real patients. The abstract and Section 5.6 should carry the same caveat as the conclusion, and the claim should be explicitly scoped to the model-class-matched simulation setting.
  3. [§5.1 and §5.4, Table 4] Several of the 'working hypotheses' are effectively enforced by the model assumptions, so the sensitivity analyses do not provide independent tests of the underlying mechanisms. For example, H2 (precision modulates update gain) follows from the monotonicity conditions imposed on F_pi in Definition 3, and H5 (supportive/suppressive beliefs modulate intention gain) follows from the coordinate-wise monotonicity of G_if in Definition 6. The numerical sweeps are consistency checks of the chosen parameterization, not validations of the neuro-cognitive mechanisms. This should be stated more prominently in Section 5.1 and the abstract, otherwise readers may over-interpret the sensitivity results as empirical support.
minor comments (4)
  1. [§4.2, Lemma 5 and Proposition 2] The notation for the lower and upper self-inhibition bounds is hard to distinguish in the text: both appear as 'Λ' in some passages. Please use a clearly different symbol or explicit under/overbar throughout, including in equations (21) and (A2).
  2. [Appendix B.2, Table B2] The row 'd 10.0' appears to be a unit/formatting inconsistency; if d is the maximum difficulty, state it as an integer or clarify the normalization.
  3. [§5.6] The claim that the model-based controller 'sustains simulated task participation' is based on a representative run in Figure 12a, while the cumulative-cost comparison is over 50 runs. Please clarify whether task participation (margin staying positive) holds across all repeated simulations or only in the illustrated run.
  4. [§3.2] The forward-Euler discretization of the DCM-style ODE with fixed step size Δt is introduced as a modeling choice, but the empirical appropriateness of this discretization for human cognitive dynamics is not discussed. A sentence noting the range of Δt used and its relation to the stability conditions would help.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found: the formal stability and regularity results are self-contained, and the interpretability and model-matched simulation issues are disclosed validation limitations rather than circular reasoning.

full rationale

The paper's analytical claims in Section 4 are derived from explicit assumptions (Assumptions 1–3) and proved directly in Appendix A; the LPE invariance/contraction, cognition-module forward invariance, and ISS results are standard consequences of the model's defining bounds and are stated as sufficient, feasibility-type conditions (e.g., Eq. (21), Corollary 1) rather than empirical predictions. The self-citations [6], [7], and [12] are motivational or future-extension references and are not load-bearing for the formal derivation. Remark 2 explicitly disclaims that the latent-state labels are not implied by the equations, so the interpretability claim is an interpretive/validation limitation, not a circular derivation. Similarly, the sensitivity hypotheses in Section 5.1 are explicitly said not to constitute independent empirical validation, and the closed-loop rehabilitation showcase in Section 5.6 is explicitly called a model-class-matched proof of concept, with the Conclusion acknowledging that the simulated patient is generated by the same model class used by the controller. These passages show that the authors do not present the self-consistency checks as independent predictions. Because the actual formal results are self-contained and the limitations are disclosed, no circularity is present.

Assumptions & free parameters 9 free parameters · 5 assumptions · 1 invented entities

The model introduces many hand-specified parameters that define operating regimes and simulation scenarios. The stability proofs are conditional on dissipation inequalities and a forward-Euler discretization assumption. The latent states are interpretive constructs without direct empirical grounding.

free parameters (9)
  • Attentional Naka–Rushton parameters η1, η2 = 1.80, 0.60 (sensitivity); 1.80, 0.25 (rehab)
    Hand-selected to define drive steepness and semi-saturation; not estimated from data.
  • Normalization offset β_ats and pooling weights γ_ats = 0.65; 0.70
    Control divisive suppression; chosen to stabilize attention behavior.
  • Precision parameters φ_pi, θ_pi, χ_pi = 1.0, 1.0, 1.0
    For predictive-inference update gain; not fit to human data.
  • Self-inhibition gain κ and log-leak vector γ = 0.90 (sensitivity), 0.55 (rehab); γ vector in Table B3/B2
    Crucial for the stability inequalities in Lemma 5 and Proposition 2.
  • Coupling matrices Φ_base, Ψ_base, Θ_base = Matrices in Tables B3/B2
    Hand-specified directed couplings; stability depends on their induced norms.
  • State-gated couplings Ξ = Zero in sensitivity; sparse entries in rehab
    Enables endogenous modulation; values chosen by hand.
  • Decision function parameters η8–η12, w±, sigmoid slopes = Various values in Table B3/B2
    Shape goal-salience and belief-gate curves; chosen for the simulated regimes.
  • MPC cost weights λ1–λ7 and target schedule = λ = (2.0, 1.5, 20, 1, 3, 1, 0.1); d_target pattern
    Define clinical preferences in the simulation; not estimated from clinical outcomes.
  • Robot perturb/model mismatch and noise settings = σ_par=0.10, σ_ω=0.008, σ_comfort=0.03
    Simulation-only values controlling the closed-loop proof of concept.
assumptions (5)
  • standard math Standard real analysis: mean value theorem, compactness, induced matrix norm sub-multiplicativity.
    Used throughout Appendix A proofs (e.g., Lemma 1, Proposition 2).
  • domain assumption Input sets U_ℓ are compact and convex and contain zero; attentional drive functions are continuously differentiable and bounded.
    Assumption 1 and Definition 1; needed for Lipschitz and well-posedness.
  • domain assumption Cognitive dynamics are a forward-Euler discretization of DCM-style ODEs with fixed step size Δt, and Δt ≤ 1/Λ in stability results.
    Section 3.2; the discrete-time structure is a modeling choice, and Proposition 2 requires the Δt bound.
  • domain assumption Sufficient dissipation condition (21) and σ<0 in Corollary 1: self-inhibition dominates coupling.
    The stability certificates are conditional on such inequalities holding; these are not guaranteed by the model alone.
  • ad hoc to paper Goal/belief partition of cognitive states and product-form intention mapping (goal salience × belief gain).
    Remark 3 and Definition 6; a design choice not derived from data.
invented entities (1)
  • Interpretable latent cognitive state vector x (beliefs, goals, emotions)
    purpose: Provide white-box internal states for estimation and control; connect to neuro-cognitive mechanisms.
    These states are not directly measured; the paper itself (Remark 2) states their psychological labels are interpretations rather than properties implied by the equations, so there is no falsifiable handle outside the model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A modular state-space model of human perception, cognition, and decision dynamics." pith.science (2026). https://pith.science/paper/HXHVJYTA

@misc{pith2026260714078,
  author       = {Pith},
  title        = {Pith review of: A modular state-space model of human perception, cognition, and decision dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HXHVJYTA}},
  note         = {Machine review of arXiv:2607.14078}
}
read the original abstract

Human-centered adaptive systems require behavioral models that are both psychologically interpretable and mathematically analyzable. Many existing predictors either operate as black-box input-output mappings or provide limited access to latent internal dynamics. This paper addresses this gap by modeling behavior as a perception-cognition-decision pipeline. We propose a modular state-space model in which attentional selection, predictive inference, cognitive-state evolution, intention formation, and action selection are represented by coupled mathematical mappings. The model links sensory inputs to observable behavior through latent internal states while retaining interpretable connections to neuro-cognitive mechanisms. We establish sufficient conditions for boundedness, Lipschitz regularity, forward invariance, contraction of perceptual inference under constant input, and input-to-state stability of the cognitive state dynamics. Numerical sensitivity analyses show that the model yields interpretable changes in perceptual tracking, cognitive amplification, intention expression, and action decisiveness. We further demonstrate a closed-loop rehabilitation case study in which a receding-horizon controller uses the model to adapt movement difficulty from partial feedback. In this proof-of-concept setting, the model-based controller sustains simulated task participation and achieves lower realized cumulative cost than target-following and random baselines. Overall, the framework provides a white-box dynamical structure for estimation, validation, and model-based control in human-centered settings.

Figures

Figures reproduced from arXiv: 2607.14078 by the authors.

Figure 1
Figure 1. High-level modular architecture of the proposed closed-loop human behavior model as a perception-cognition-decision pipeline. The environment provides [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Block diagram of the perception module. upstream preprocessing. The stacked sensory input is denoted by u(k) ∈ U B Q ℓ∈L Uℓ , where Q ℓ∈L denotes the Cartesian product. Each admissible set Uℓ is compact and convex (imply￾ing the same for U), and represents the finite operating range of the feature score extracted for channel ℓ. The range reflects sensory saturation and normalization applied during upstream pre-proce… view at source ↗
Figure 3
Figure 3. Block diagram of the cognition module. guarantee F c ii(·) < 0 on Xlpe without imposing hard inequality constraints during estimation. The resulting parameterization is stated next. Definition 4. Let x lpe(k) ∈ Xlpe ⊆ R |L| be the exogenous per￾ceptual input at time step k. For each i ∈ I, we parameterize the self-inhibition as follows: F c ii  x lpe(k)  B −κ exp   γi + X ℓ∈L x lpe ℓ (k)Λiℓ   , (9)… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Block diagram of the decision-making module. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Sensitivity analysis of attentional selection. (a) Endpoint-change heatmap for one-at-a-time parameter sweeps across input families. Colors encode [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Sensitivity analysis of predictive inference. (a) Endpoint-change heatmap for one-at-a-time parameter sweeps across input families. Colors encode [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]
Figure 7
Figure 7. Figure 7: Sensitivity and constant-input local stability analysis of the cognition module. (a) Endpoint-change heatmap for maximum state norm, mean step [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Two-dimensional constant-input sweep of the cognition module over recurrent coupling scale [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]
Figure 9
Figure 9. Figure 9: Sensitivity analysis and belief-gate visualization of intention formation. (a) Endpoint-change heatmap for sampled I/ [PITH_FULL_IMAGE:figures/full_fig_p017_9.png]
Figure 10
Figure 10. Figure 10: Sensitivity analysis of the action selection stage. (a) Endpoint-change heatmap for sampled I/ [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: Block diagram of the closed-loop rehabilitation showcase. [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Closed-loop rehabilitation showcase. (a) Closed-loop rehabilitation case-study results. The top three panels show one representative closed-loop run: the [PITH_FULL_IMAGE:figures/full_fig_p020_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 15 canonical work pages

  1. [1]

    Schürmann, P

    T. Schürmann, P. Beckerle, Personalizing human-agent interaction through cognitive models, Frontiers in Psy- chology V olume 11 - 2020 (2020). doi:10.3389/fpsyg. 2020.561510

    arXiv 2020

  2. [2]

    Leite, C

    I. Leite, C. Martinho, A. Paiva, Social robots for long-term interaction: A survey, International Jour- nal of Social Robotics 5 (2013) 291–308. doi:10.1007/ s12369-013-0178-y

    2013

  3. [3]

    M. K. Ho, T. L. Griffiths, Cognitive science as a source of forward and inverse models of human 20 decisions for robotics and control, Annual Re- view of Control, Robotics, and Autonomous Sys- tems 5 (2022) 33–53. doi:https://doi.org/10.1146/ annurev-control-042920-015547

    2022

  4. [4]

    Huang, B

    Y . Huang, B. Yang, T. W.-L. Wong, S. S. M. Ng, X. Hu, Personalized robots for long-term telerehabilitation af- ter stroke: a perspective on technological readiness and clinical translation, Frontiers in Rehabilitation Sciences V olume 4 - 2023 (2024). doi:10.3389/fresc.2023. 1329927

    work page doi:10.3389/fresc.2023 2023

  5. [5]

    W. Zu, X. Huang, T. Xu, L. Du, Y . Wang, L. Wang, W. Nie, Machine learning in predicting outcomes for stroke patients following rehabilitation treatment: A sys- tematic review, PLOS ONE 18 (2023) 1–14. doi:10. 1371/journal.pone.0287308

    2023

  6. [6]

    M. L. Morão Patrício, A. Jamshidnejad, Leveraging sys- tems and control theory for social robotics: a model- based behavioral control approach to human-robot inter- action, Applied Intelligence 56 (2026). doi:10.1007/ s10489-026-07100-9

    2026

  7. [7]

    M. L. M. Patrício, A. Jamshidnejad, Dynamic math- ematical models of theory of mind for socially assis- tive robots, IEEE Access 11 (2023) 103956–103975. doi:10.1109/ACCESS.2023.3316603

    arXiv 2023

  8. [8]

    Powers, Quantitative analysis of purposive systems: Some spadework at the foundations of scientific psychol- ogy, Psychological Review 85 (1978) 417–435

    W. Powers, Quantitative analysis of purposive systems: Some spadework at the foundations of scientific psychol- ogy, Psychological Review 85 (1978) 417–435. doi:10. 1037/0033-295X.85.5.417

    1978

Show all 60 references
  1. [9]

    M. Han, L. Zhang, J. Wang, W. Pan, Actor-critic rein- forcement learning for control with stability guarantee, IEEE Robotics and Automation Letters 5 (2020) 6217–

    2020

  2. [10]

    Z. C. Lipton, The mythos of model interpretability, Com- mun. ACM 61 (2018) 36–43. doi:10.1145/3233231

    2018 doi

  3. [11]

    A. S. Rao, M. P. Georgeff, Bdi agents: From theory to practice, in: International Conference on Multiagent Sys- tems, 1995, pp. 312–319

    1995

  4. [12]

    M. L. M. Patrício, A. Jamshidnejad, A systems-theoretic approach to mental state estimation for theory-of-mind- aware social robots, IEEE Access 13 (2025) 158467– 158482. doi:10.1109/ACCESS.2025.3607165

    2025

  5. [13]

    Feldman, K

    H. Feldman, K. Friston, Attention, uncertainty, and free- energy, Frontiers in Human Neuroscience V olume 4 - 2010 (2010). doi:10.3389/fnhum.2010.00215

    2010 arXiv

  6. [14]

    Friston, A theory of cortical responses, Philosophical Transactions of the Royal Society B: Biological Sciences 360 (2005) 815–836

    K. Friston, A theory of cortical responses, Philosophical Transactions of the Royal Society B: Biological Sciences 360 (2005) 815–836. doi:10.1098/rstb.2005.1622

    2005

  7. [15]

    T. Parr, K. J. Friston, Generalised free energy and active inference, Biological Cybernetics 113 (2019) 495–513. doi:10.1007/s00422-019-00805-w

    2019 doi

  8. [16]

    Friston, L

    K. Friston, L. Harrison, W. Penny, Dynamic causal mod- elling, NeuroImage 19 (2003) 1273–1302. doi:10.1016/ S1053-8119(03)00202-7

    2003

  9. [17]

    K. E. Stephan, L. Kasper, L. M. Harrison, J. Daunizeau, H. E. den Ouden, M. Breakspear, K. J. Friston, Nonlinear dynamic causal models for fmri, NeuroImage 42 (2008) 649–662. doi:10.1016/j.neuroimage.2008.04.262

    2008 doi

  10. [18]

    Ratcliff, G

    R. Ratcliff, G. McKoon, The diffusion decision model: Theory and data for two-choice decision tasks, Neu- ral Computation 20 (2008) 873–922. doi:10.1162/neco. 2008.12-06-420

    2008 doi

  11. [19]

    M. A. Bertolero, B. T. T. Yeo, D. S. Bassett, M. D’Esposito, A mechanistic model of connector hubs, modularity and cognition, Nature Human Behaviour 2 (2018) 765–777. doi:10.1038/s41562-018-0420-6

    2018 doi

  12. [20]

    Kotseruba, J

    I. Kotseruba, J. K. Tsotsos, 40 years of cognitive ar- chitectures: core cognitive abilities and practical appli- cations, Artificial Intelligence Review 53 (2020) 17–94. doi:10.1007/s10462-018-9646-y

    2020 doi

  13. [21]

    J. I. Gold, M. N. Shadlen, The neural basis of decision making, Annual Review of Neuroscience 30 (2007) 535–574. doi:10.1146/annurev.neuro.29. 051605.113038

    2007

  14. [22]

    Rauss, G

    K. Rauss, G. Pourtois, What is bottom-up and what is top- down in predictive coding?, Frontiers in Psychology V ol- ume 4 - 2013 (2013). doi:10.3389/fpsyg.2013.00276

    2013

  15. [23]

    Dijkstra, P

    N. Dijkstra, P. Zeidman, S. Ondobaka, M. A. J. van Gerven, K. Friston, Distinct top-down and bottom- up brain connectivity during visual perception and im- agery, Scientific Reports 7 (2017) 5677. doi:10.1038/ s41598-017-05888-8

    2017

  16. [24]

    A. M. Treisman, G. Gelade, A feature-integration the- ory of attention, Cognitive Psychology 12 (1980) 97–136. doi:10.1016/0010-0285(80)90005-5

    1980 doi

  17. [25]

    Carandini, D

    M. Carandini, D. J. Heeger, Normalization as a canoni- cal neural computation, Nature Reviews Neuroscience 13 (2012) 51–62. doi:10.1038/nrn3136

    2012 doi

  18. [26]

    J. H. Reynolds, D. J. Heeger, The normal- ization model of attention, Neuron 61 (2009) 168–185. doi:10.1016/j.neuron.2009.01.002, doi: 10.1016/j.neuron.2009.01.002

    2009 doi

  19. [27]

    Moore, M

    T. Moore, M. Zirnsak, Neural mechanisms of selective visual attention, Annual Review of Psychology 68 (2017) 47–72. doi:10.1146/ annurev-psych-122414-033400

    2017

  20. [28]

    D. J. Heeger, Normalization of cell responses in cat striate cortex, Visual Neuroscience 9 (1992) 181–197. doi:10. 1017/S0952523800009640. 21

    1992

  21. [29]

    Sajid, P

    N. Sajid, P. J. Ball, T. Parr, K. J. Friston, Active infer- ence: Demystified and compared, Neural Computation 33 (2021) 674–712. doi:10.1162/neco_a_01357

    2021 doi

  22. [30]

    D. C. Knill, A. Pouget, The bayesian brain: the role of uncertainty in neural coding and computation, Trends in Neurosciences 27 (2004) 712–719. doi:https://doi. org/10.1016/j.tins.2004.10.007

    2004 doi

  23. [31]

    M. R. Nassar, K. M. Rumsey, R. C. Wilson, K. Parikh, B. Heasly, J. I. Gold, Rational regulation of learning dy- namics by pupil-linked arousal systems, Nature Neuro- science 15 (2012) 1040–1046. doi:10.1038/nn.3130

    2012 doi

  24. [32]

    Friston, K

    K. Friston, K. H. Preller, C. Mathys, H. Cagnan, J. Hein- zle, A. Razi, P. Zeidman, Dynamic causal modelling re- visited, NeuroImage 199 (2019) 730–744. doi:10.1016/ j.neuroimage.2017.02.045

    2019

  25. [33]

    L. S. Petro, A. T. Paton, L. Muckli, Contextual modulation of primary visual cortex by auditory signals, Philosoph- ical Transactions of the Royal Society B: Biological Sci- ences 372 (2017) 20160104. doi:10.1098/rstb.2016. 0104

    2017 doi

  26. [35]

    P. R. Cohen, H. J. Levesque, Intention is choice with commitment, Artificial Intelligence 42 (1990) 213–261. doi:10.1016/0004-3702(90)90055-5

    1990 doi

  27. [36]

    Cisek, J

    P. Cisek, J. F. Kalaska, Neural mechanisms for interact- ing with a world full of action choices, Annual Review of Neuroscience 33 (2010) 269–298. doi:https://doi. org/10.1146/annurev.neuro.051508.135409

    2010

  28. [38]

    R. C. O’Reilly, M. J. Frank, Making working memory work: A computational model of learning in the prefrontal cortex and basal ganglia, Neural Comput. 18 (2006) 283–328. doi:10.1162/089976606775093909

    2006 doi

  29. [39]

    Ajzen, The theory of planned behavior, Organizational Behavior and Human Decision Processes 50 (1991) 179–

    I. Ajzen, The theory of planned behavior, Organizational Behavior and Human Decision Processes 50 (1991) 179–

    1991

  30. [40]

    R. C. Wilson, Y . K. Takahashi, G. Schoenbaum, Y . Niv, Orbitofrontal cortex as a cognitive map of task space, Neuron 81 (2014) 267–279. doi:10.1016/j.neuron. 2013.11.005

    2014 doi

  31. [41]

    R. L. Albin, A. B. Young, J. B. Penney, The func- tional anatomy of basal ganglia disorders, Trends in Neurosciences 12 (1989) 366–375. doi:10.1016/ 0166-2236(89)90074-X

    1989

  32. [42]

    J. W. Mink, The basal ganglia: Focused selection and inhibition of competing motor programs, Progress in Neurobiology 50 (1996) 381–425. doi:10.1016/ S0301-0082(96)00042-1

    1996

  33. [43]

    Nambu, H

    A. Nambu, H. Tokuno, M. Takada, Functional signif- icance of the cortico–subthalamo–pallidal ‘hyperdirect’ pathway, Neuroscience Research 43 (2002) 111–117. doi:10.1016/S0168-0102(02)00027-5

    2002 doi

  34. [45]

    Gurney, T

    K. Gurney, T. J. Prescott, P. Redgrave, A computational model of action selection in the basal ganglia. i. a new functional anatomy, Biological Cybernetics 84 (2001) 401–410. doi:10.1007/PL00007984

    2001 doi

  35. [46]

    Gurney, T

    K. Gurney, T. J. Prescott, P. Redgrave, A computational model of action selection in the basal ganglia. ii. analysis and simulation of behaviour, Biological Cybernetics 84 (2001) 411–423. doi:10.1007/PL00007985

    2001 doi

  36. [47]

    Schwedhelm, B

    P. Schwedhelm, B. S. Krishna, S. Treue, An extended nor- malization model of attention accounts for feature-based attentional enhancement of both response and coherence gain, PLOS Computational Biology 12 (2016) 1–22. doi:10.1371/journal.pcbi.1005225

    2016 doi

  37. [48]

    A. J. Yu, P. Dayan, Uncertainty, neuromodulation, and attention, Neuron 46 (2005) 681–692. doi:10.1016/j. neuron.2005.04.026

    2005 doi

  38. [49]

    Zeidman, A

    P. Zeidman, A. Jafarian, N. Corbin, M. L. Seghier, A. Razi, C. J. Price, K. J. Friston, A guide to group effective connectivity analysis, part 1: First level analy- sis with dcm for fmri, NeuroImage 200 (2019) 174–190. doi:10.1016/j.neuroimage.2019.06.031

    2019 doi

  39. [50]

    Bartra, J

    O. Bartra, J. T. McGuire, J. W. Kable, The valua- tion system: A coordinate-based meta-analysis of bold fmri experiments examining neural correlates of subjec- tive value, NeuroImage 76 (2013) 412–427. doi:10. 1016/j.neuroimage.2013.02.063

    2013

  40. [51]

    E. K. Miller, J. D. Cohen, An integrative theory of pre- frontal cortex function, Annual Review of Neuroscience 24 (2001) 167–202. doi:https://doi.org/10.1146/ annurev.neuro.24.1.167

    2001

  41. [52]

    D. W. D. , J. J. A. , Statistics of natural time-varying im- ages, Network: Computation in Neural Systems 6 (1995)

    1995

  42. [53]

    Haegens, E

    S. Haegens, E. Zion Golumbic, Rhythmic facilitation of sensory processing: A critical review, Neuroscience & Biobehavioral Reviews 86 (2018) 150–165. doi:10. 1016/j.neubiorev.2017.12.002

    2018

  43. [54]

    J. H. McDermott, E. P. Simoncelli, Sound texture percep- tion via statistics of the auditory periphery: Evidence from sound synthesis, Neuron 71 (2011) 926–940. doi:10. 1016/j.neuron.2011.06.032

    2011

  44. [55]

    Poeppel, M

    D. Poeppel, M. F. Assaneo, Speech rhythms and their neu- ral foundations, Nature Reviews Neuroscience 21 (2020) 322–334. doi:10.1038/s41583-020-0304-4

    2020 doi

  45. [56]

    E. P. Simoncelli, B. A. Olshausen, Natural image statis- tics and neural representation, Annual Review of Neu- roscience 24 (2001) 1193–1216. doi:10.1146/annurev. neuro.24.1.1193

    2001 doi

  46. [57]

    L. N. Vinke, I. M. Bloem, S. Ling, Saturating nonlinear- ities of contrast response in human visual cortex, Jour- nal of Neuroscience 42 (2022) 1292–1302. doi:10.1523/ JNEUROSCI.0106-21.2021

    2022

  47. [58]

    D. J. Heeger, Half-squaring in responses of cat striate cells, Visual Neuroscience 9 (1992) 427–443. doi:10. 1017/S095252380001124X

    1992

  48. [59]

    C. R. Fetsch, G. C. DeAngelis, D. E. Angelaki, Bridg- ing the gap between theories of sensory cue integration and the physiology of multisensory neurons, Nature Re- views Neuroscience 14 (2013) 429–442. doi:10.1038/ nrn3503

    2013

  49. [60]

    1.2 0.00.0 1.2 # ,

    C. R. Fetsch, A. Pouget, G. C. DeAngelis, D. E. Ange- laki, Neural correlates of reliability-based cue weighting during multisensory integration, Nature Neuroscience 15 (2012) 146–154. doi:10.1038/nn.2983. Appendix A. Proof of analytical results This appendix contains the proo...

    2012 doi

  50. [211]

    doi:10.1016/0749-5978(91)90020-T, theories of Cognitive Self-Regulation

    doi

  51. [345]

    doi:10.1088/0954-898X/6/3/003. 22

    doi

  52. [6224]

    doi:10.1109/LRA.2020.3011351

    2020

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.