Reinforcement biases subsequent perceptual decisions when confidence is low: a widespread behavioral phenomenon

  1. Armin Lak  Is a corresponding author
  2. Emily Hueske
  3. Junya Hirokawa
  4. Paul Masset
  5. Torben Ott
  6. Anne E Urai
  7. Tobias H Donner
  8. Matteo Carandini
  9. Susumu Tonegawa
  10. Naoshige Uchida
  11. Adam Kepecs  Is a corresponding author
  1. University of Oxford, United Kingdom
  2. MIT, United States
  3. Doshisha University, Japan
  4. Cold Spring Harbor Laboratory, United States
  5. University Medical Center Hamburg-Eppendorf, Germany
  6. University College London, United Kingdom
  7. Massachusetts Institute of Technology, United States
  8. Harvard University, United States

Abstract

Learning from successes and failures often improves the quality of subsequent decisions. Past outcomes, however, should not influence purely perceptual decisions after task acquisition is complete since these are designed so that only sensory evidence determines the correct choice. Yet, numerous studies report that outcomes can bias perceptual decisions, causing spurious changes in choice behavior without improving accuracy. Here we show that the effects of reward on perceptual decisions are principled: past rewards bias future choices specifically when previous choice was difficult and hence decision confidence was low. We identified this phenomenon in six datasets from four laboratories, across mice, rats, and humans, and sensory modalities from olfaction and audition to vision. We show that this choice-updating strategy can be explained by reinforcement learning models incorporating statistical decision confidence into their teaching signals. Thus, despite being suboptimal from the experimenter’s perspective, confidence-guided reinforcement learning optimizes behavior in uncertain, real-world situations.

Data availability

The data used in this study is available at http://dx.doi.org/10.6084/m9.figshare.4300043

The following previously published data sets were used

Article and author information

Author details

  1. Armin Lak

    Department of Physiology, Anatomy and Genetics, University of Oxford, Oxford, United Kingdom
    For correspondence
    armin.lak@dpag.ox.ac.uk
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1926-5458
  2. Emily Hueske

    Picower Institute, MIT, Cambridge, United States
    Competing interests
    No competing interests declared.
  3. Junya Hirokawa

    Graduate School of Brain Science, Doshisha University, Kyotanabe, Japan
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1238-5713
  4. Paul Masset

    Cold Spring Harbor Laboratory, Cold Spring Harbor, United States
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2001-7515
  5. Torben Ott

    Cold Spring Harbor Laboratory, Cold Spring Harbor, United States
    Competing interests
    No competing interests declared.
  6. Anne E Urai

    Cold Spring Harbor Laboratory, Cold Spring Harbor, United States
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5270-6513
  7. Tobias H Donner

    Department of Neurophysiology and Pathophysiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
    Competing interests
    Tobias H Donner, Reviewing editor, eLife.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7559-6019
  8. Matteo Carandini

    UCL Institute of Ophthalmology, University College London, London, United Kingdom
    Competing interests
    No competing interests declared.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4880-7682
  9. Susumu Tonegawa

    Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, United States
    Competing interests
    No competing interests declared.
  10. Naoshige Uchida

    Center for Brain Science, Harvard University, Cambridge, United States
    Competing interests
    Naoshige Uchida, Reviewing editor, eLife.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5755-9409
  11. Adam Kepecs

    Cold Spring Harbor Laboratory, Cold Spring Harbor, United States
    For correspondence
    akepecs@wustl.edu
    Competing interests
    No competing interests declared.

Funding

Wellcome (106101)

  • Armin Lak

Wellcome (213465)

  • Armin Lak

National Institutes of Health (R01 MH110404)

  • Naoshige Uchida

National Institutes of Health (R01MH097061 and R01DA038209)

  • Naoshige Uchida

Wellcome (205093)

  • Matteo Carandini

Deutsche Forschungsgemeinschaft (DO 1240/2-1 and DO 1240/3-1)

  • Tobias H Donner

RIKEN-CBS

  • Emily Hueske
  • Susumu Tonegawa

JPB Foundation

  • Emily Hueske
  • Susumu Tonegawa

Howard Hughes Medical Institute

  • Emily Hueske
  • Susumu Tonegawa

German Academic Exchange Service

  • Anne E Urai

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Reviewing Editor

  1. Emilio Salinas, Wake Forest School of Medicine, United States

Ethics

Animal experimentation: The experimental procedures were approved by Institutional committees at Cold Spring Harbor Laboratory (for experiments on rats), MIT and Harvard University (for mice auditory experiments) and were in accordance with National Institute of Health standards (project ID: 18-14-11-08-1). Experiments on mice visual decisions were approved by the home Office of the United Kingdom (license 70/8021). Experiments in humans were approved by the ethics committee at the University of Amsterdam (project ID: 2014­-BC­-3376).

Human subjects: The ethics committee at the University of Amsterdam approved the study, and all observers gave their informed consent.project ID: 2014-BC-3376

Version history

  1. Received: July 1, 2019
  2. Accepted: April 9, 2020
  3. Accepted Manuscript published: April 14, 2020 (version 1)
  4. Accepted Manuscript updated: April 15, 2020 (version 2)
  5. Version of Record published: May 11, 2020 (version 3)

Copyright

© 2020, Lak et al.

This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 7,058
    Page views
  • 1,086
    Downloads
  • 46
    Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Armin Lak
  2. Emily Hueske
  3. Junya Hirokawa
  4. Paul Masset
  5. Torben Ott
  6. Anne E Urai
  7. Tobias H Donner
  8. Matteo Carandini
  9. Susumu Tonegawa
  10. Naoshige Uchida
  11. Adam Kepecs
(2020)
Reinforcement biases subsequent perceptual decisions when confidence is low: a widespread behavioral phenomenon
eLife 9:e49834.
https://doi.org/10.7554/eLife.49834

Share this article

https://doi.org/10.7554/eLife.49834

Further reading

    1. Neuroscience
    Kiwamu Kudo, Kamalini G Ranasinghe ... Srikantan S Nagarajan
    Research Article

    Alzheimer’s disease (AD) is characterized by the accumulation of amyloid-β and misfolded tau proteins causing synaptic dysfunction, and progressive neurodegeneration and cognitive decline. Altered neural oscillations have been consistently demonstrated in AD. However, the trajectories of abnormal neural oscillations in AD progression and their relationship to neurodegeneration and cognitive decline are unknown. Here, we deployed robust event-based sequencing models (EBMs) to investigate the trajectories of long-range and local neural synchrony across AD stages, estimated from resting-state magnetoencephalography. The increases in neural synchrony in the delta-theta band and the decreases in the alpha and beta bands showed progressive changes throughout the stages of the EBM. Decreases in alpha and beta band synchrony preceded both neurodegeneration and cognitive decline, indicating that frequency-specific neuronal synchrony abnormalities are early manifestations of AD pathophysiology. The long-range synchrony effects were greater than the local synchrony, indicating a greater sensitivity of connectivity metrics involving multiple regions of the brain. These results demonstrate the evolution of functional neuronal deficits along the sequence of AD progression.

    1. Cell Biology
    2. Neuroscience
    Zhenyong Wu, Grant F Kusick ... Shigeki Watanabe
    Research Article

    Despite decades of intense study, the molecular basis of asynchronous neurotransmitter release remains enigmatic. Synaptotagmin (syt) 7 and Doc2 have both been proposed as Ca2+ sensors that trigger this mode of exocytosis, but conflicting findings have led to controversy. Here, we demonstrate that at excitatory mouse hippocampal synapses, Doc2α is the major Ca2+ sensor for asynchronous release, while syt7 supports this process through activity-dependent docking of synaptic vesicles. In synapses lacking Doc2α, asynchronous release after single action potentials is strongly reduced, while deleting syt7 has no effect. However, in the absence of syt7, docked vesicles cannot be replenished on millisecond timescales. Consequently, both synchronous and asynchronous release depress from the second pulse onward during repetitive activity. By contrast, synapses lacking Doc2α have normal activity-dependent docking, but continue to exhibit decreased asynchronous release after multiple stimuli. Moreover, disruption of both Ca2+ sensors is non-additive. These findings result in a new model whereby syt7 drives activity-dependent docking, thus providing synaptic vesicles for synchronous (syt1) and asynchronous (Doc2 and other unidentified sensors) release during ongoing transmission.