department of psychology, university of …...the final version may differ from this version....

Accepted manuscripts are peer-reviewed but have not been through the copyediting, formatting, or proofreadingprocess.

Copyright © 2018 the authors

This Accepted Manuscript has not been copyedited and formatted. The final version may differ from this version.

Research Articles: Behavioral/Cognitive

Attractor-like dynamics in belief updating in schizophrenia

Rick A Adams1,2, Gary Napier1, Jonathan P Roiser1, Christoph Mathys3,4,5 and James Gilleen6,7

1Institute of Cognitive Neuroscience, UCL, 17 Queen Square, London, WC1N 3AZ, UK2Division of Psychiatry, UCL, 6th floor, 149 Tottenham Court Road, London, W1T 7NF, UK3Scuola Internazionale Superiore di Studi Avanzati (SISSA), Via Bonomea 265, 34136 Trieste, Italy4Translational Neuromodeling Unit (TNU), Institute for Biomedical Engineering, University of Zurich and ETHZurich, Wilfriedstrasse 6, 8032 Zurich, Switzerland5Max Planck UCL Centre for Computational Psychiatry and Ageing Research, 10-12 Russell Square, London,WC1B 5EH, UK6Department of Psychology, University of Roehampton, London, SE15 4JD.7Department of Psychosis Studies; Institute of Psychiatry, Psychology and Neuroscience, Kings CollegeLondon, London, SE5 8AF.

DOI: 10.1523/JNEUROSCI.3163-17.2018

Received: 2 November 2017

Revised: 3 May 2018

Accepted: 27 June 2018

Published: 5 September 2018

Author contributions: R.A.A., J.R., C.M., and J.G. designed research; R.A.A., G.N., and J.G. performedresearch; R.A.A. and G.N. analyzed data; R.A.A. wrote the first draft of the paper; R.A.A., G.N., J.R., C.M., andJ.G. wrote the paper; C.M. contributed unpublished reagents/analytic tools.

Conflict of Interest: The authors declare no competing financial interests.

The authors are very grateful to Dr Emmanuelle Peters for providing them with dataset 1. Dr Rick Adamsis funded by the Academy of Medical Sciences (AMS-SGCL13-Adams) and the National Institute of HealthResearch (CL-2013-18-003). JG was supported in his contribution to this project by the British Academy

Correspondence should be addressed to corresponding author: [email protected]

Cite as: J. Neurosci ; 10.1523/JNEUROSCI.3163-17.2018

Alerts: Sign up at www.jneurosci.org/cgi/alerts to receive customized email alerts when the fully formattedversion of this article is published.

Attractor-like dynamics in belief updating in schizophrenia

Abstract

N

Significance Statement

Introduction

more

less consistent

N

Methods and Materials

Subject characteristics

Experimental design

Computational modelling

ω

φ κ1 σ2(0) φ κ1

x2

ω

φ

κ1

x2

m

x2 φ x2 m ω

x2

m

φ

φ

μ2

ω κ1

κ1

κ1 u

κ1

κ1

κ1

κ1

μ2 σ2(0)

κ1

φ

κ1

φ

more

ω ν σ2(0) φ κ1

.

μ2(0)

σ2(0)

σ2(0)

κ1

Model fitting and statistical analysis

ω ν φ κ1

σ2(0)

κ1

σ2(0)

σ2(0)

σ2(0)

Results

Behavioural results: dataset 1

p(adj) p post hoc

F

p

p(adj)

p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F

F p

p(adj) p(adj)

F p

F p

p(adj)

ρ p

ρ p

p

Behavioural results: dataset 2

t p t

t p d t

t p t p

t

ρ p ρ p

p

Modelling results: dataset 1

p(adj) p post hoc

κ1

χ2 p η

ν χ2 p η

σ2(0) ω κ1

p(adj)

p(adj)

p(adj)

ν p(adj)

p(adj)

p(adj)

κ1 χ2

p η ν

χ2 p η σ2(0) ω

κ1 p(adj)

p(adj)

ν

p(adj)

p(adj)

p(adj)

κ1 ν

p κ1

r p

κ1 ν

ν F

p κ1 F p

κ1 ν

κ1 ν

ν ν ν

p

κ1

p ν

ρ p

ρ κ1 ν

ρ p ρ p

κ1 ω ρ p ρ

p simulated

κ1 ω r

κ1 ω

σ2(0)

ν κ1

κ1

ρ ω κ1

Modelling results: dataset 2

κ1

Z p

r ν Z

p r σ2(0)

Z p r

ω

κ1

κ1

ω

κ1

ν Z p Z p

σ2(0) Z

p

κ1 ν

p κ1 ν

ν r p

p

κ1 ν

p

κ1 ν

κ1 ν

ν ν ν κ1

F p ν

F p ν t p

κ1 ν ρ p

κ1

σ2(0) ρ p r κ1 ν

ρ ρ ρ

σ2(0) κ1

r

Discussion

κ1 ν

ν κ1

ν

κ1

κ1

κ1 ν

κ1

ν

Parameter relationships with cognition and symptoms

κ1 ν

ν

κ1 ν

κ1 ν

ν

ν

against

and back towards

κ1

Related modelling studies

Limitations

ν

Conclusion

Acknowledgements

Conflicts of Interest

References

Figure Legends

Figure 1: Effects of attractor network dynamics on belief updating

Figure 2: Beads task schematic and group average confidence ratings in

Datasets 1 and 2.

Figure 3: The structure of the Hierarchical Gaussian Filter (Model 6) and

some simulated data

μ2 x2

k

ω

ω

u(k)

x1 s

ν

k

Figure 4: Simulated data illustrating the effects of φ (Models 3 and 4)

and κ1 (Model 5 and 6) on inference

φ κ1

σ2(0) ω

y

φ k

. μ2 m

σ μ2 φ

φ σ μ2

κ1

κ1

κ1

κ1

κ1

κ1 φ

κ1

σ2(0) ω

κ1

κ1

u

κ1

κ1

Figure 5: Recovery of model parameters from simulated data

σ2(0) ω κ1 ν

κ1 ν

Figure 6: Bayesian model selection results for both datasets.

Figure 7: Probability density plots for Model 6 parameters in dataset 1.

σ2(0) ω ν κ1

Figure 8: Model 6 parameters in dataset 2 – distributions and correlation

σ2(0) ω ν κ1

σ2(0)

p

ν κ1

Figure 9: Responses and model fits for two control subjects

u(k)

k y

–

σ2(0)

ν κ1

ω

κ1

κ1

ν

Figure 10: Responses and model fits for two Scz subjects

ω

κ1

ω

Dataset 1 Dataset 2

Non-

clinical

controls

t1

Non-

clinical

controls

t2

Clinical

controls t1

Clinical

controls t2

Psychotic

t1

Psychotic

t2

Controls

(all) Scz

Controls

(subset)

Cognitive

measures

Delusion

proneness

Schizotypy

Diagnosis/

Symptoms

Beads task

Table 1: Demographic, psychological and behavioural details of both datasets

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

F p

p(adj) p(adj)

p(adj)

t p d

t p d

F p

p(adj) p

p(adj)

F p

F p

p(adj) p

p(adj)

F p

p(adj) p

p(adj)

F p

p(adj) p

p(adj)

F p

p(adj) p

p(adj)

F p

F p

p(adj) p

p(adj)

ρ p

ρ p ρ p

ρ p ρ

ρ p

p

t p d t p

d t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

t p d

ω νω σ2(0) ν ω φ ν ω σ2(0) φ ν ω κ1 ν ω σ2(0) κ1 ν

Table 2: Models, parameters and their prior distributions.

σ2(0) ω ν (κ1)

Dataset 1 (baseline,

n=80)

p

η2

p=

η2

p

η2

p

η2

p(adj) p(adj) p(adj) p(adj)



Dataset 1 (follow-up,

n=55)

p

η2 p

η2

p

η2

p

η2




Dataset 2 (n=167)

Z

p

r

Z

p

r

Z

p

r

Z

p

r

Dataset 2

(better-matched

controls, n=116)

Z

p

r

Z

p

r

Z

p

r

Z

p

r

Table 3: Parameter distributions and statistical tests in Datasets 1 and 2

department of psychology, university of …...the final version may differ from this version....

Documents