Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2017:1:0139.
doi: 10.1038/s41562-017-0139. Epub 2017 Jul 10.

Perceptual confidence neglects decision-incongruent evidence in the brain

Megan A K Peters  1 , Thomas Thesen  2   3   4 , Yoshiaki D Ko  5 , Brian Maniscalco  6 , Chad Carlson  2 , Matt Davidson  5 , Werner Doyle  2 , Ruben Kuzniecky  2 , Orrin Devinsky  2 , Eric Halgren  3 , Hakwan Lau  1   7
Affiliations

Perceptual confidence neglects decision-incongruent evidence in the brain

Megan A K Peters et al. Nat Hum Behav. 2017.
No abstract available

PubMed Disclaimer

Conflict of interest statement

Competing Interests The authors declare no competing interests.

Figures

Figure 1
Figure 1
Behavioral task and results. (a) Subjects discriminated noisy stimuli as faces/houses and indicated their confidence (high vs low) with a single button press; responses were all made with one hand. (b) As expected, subjects showed higher accuracy for high versus low contrast stimuli, and for high confidence versus low confidence responses (2 (contrast: high/low) x 2 (confidence: high/low) repeated measures ANOVA: F(1,5)confidence = 8.418, p = .034; F(1,5)contrast = 1.783, p = .239; F(1,5)confidenceXcontrast = 0.502, p = .10), but showed negligible bias to respond ‘face’ more often than ‘house’ (t(5) = 0.316, p = 0.765). Error bars represent the standard error of the mean across subjects.
Figure 2
Figure 2
Spatiotemporal dissociation between Decision and Confidence decoding. (a) Decoder accuracy for both estimators (Decision and Confidence) rises just around 200ms post stimulus onset. However, decodability for Decision rises more quickly and peaks earlier than for Confidence. Shaded regions indicate the standard error of the mean. Lower bars denote 50ms post-stimulus time bins in which decodability was above chance for some proportion of participants. (b) To localize factors contributing to decoding performance, we projected each electrode’s contribution index C (see Supplementary Methods: Neuroanatomical localization of representations) onto its MNI coordinates across all subjects, averaged across coarse time bins of 200ms. C < 0 (dark blue) indicates the electrode contributed very little, whereas C > 0 (red) indicates the electrode contributes more to decoding. (c) We calculated average C within four broadly-defined regions of interests by lobe, and plotted it as a function of time after stimulus onset. Decision shows strong contributions from occipital electrodes around 200–700ms, while Confidence occupies a more distributed spatial representation.
Figure 3
Figure 3
Choice probability analyses show Confidence computations were insensitive to Decision-Incongruent Evidence. (a) Differences in Decision versus Confidence representations mapped onto differential use of Decision-Congruent Evidence versus Decision-Incongruent Evidence for Decision and Confidence computations. (b) Decision and Confidence were predicted differentially by the Balance-Of-Evidence rule than by the Decision-Congruent-Only rule: Decision was significantly better predicted by Balance-Of-Evidence, but Confidence showed no difference between Balance-Of-Evidence and Decision-Congruent-Only computation rules. This indicates that the computation of Confidence overly relied on the magnitude of Decision-Congruent Evidence, and did not appear to utilize Decision-Incongruent Evidence.
Figure 4
Figure 4
Violations of the normative model for Confidence but not Accuracy. (a) In signal detection theory, on a given trial the internal evidence available to a system can be represented as x, a sample drawn from one of two distributions representing stimulus categories in a discrimination task. The sign of x dictates which category an unbiased observer will choose, such that positive x (above the decision criterion at zero) leads to a ‘face’ Decision and negative x (below the decision criterion) to a ‘house’ Decision. Likewise, x’s magnitude, or its distance from the decision criterion at zero, indicates how strongly it indicates a ‘face’ or ‘house’ choice: the farther x is from zero, the more likely observers are to be correct, and so the more confident they should be in their categorization choices. Thus, the absolute value of x predicts both the trial-by-trial Accuracy (trial-by-trial correct choices/errors) and Confidence in a Decision. (b) We fitted the assumed decoding noise in the signal detection theoretic model, αdecoding, to each subject by degrading the predicted Decision decodability (based on subjects’ performance and the stimulus decoder; see Methods: Signal detection theoretic forward model) to match the observed Decision decodability. Incorporating this noise, we then used the model to predict the theoretical maximum for Accuracy and Confidence decodability for each subject. (c) Given the presence of observed decoding noise, the model predicts that the theoretically expected maximal level of decodability for Confidence will be above that for Accuracy. (d) We compared the actual Accuracy and Confidence decodability achieved via the model to the theoretical maxima predicted by the model. While mean Accuracy decodability reached the theoretical maximum (t(5) = 1.58, p = .173), Confidence decodability was significantly worse (t(5) = 2.868, p = .035). This indicates that Confidence cannot depend purely on the same internal information as Decision and Accuracy. Shaded regions indicate the standard error of the mean.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Charles L, King JR, Dehaene S. Decoding the dynamics of action, intention, and error detection for conscious and subliminal stimuli. J Neurosci. 2014;34:1158–1170. - PMC - PubMed
    1. Fleming SM, Huijgen J, Dolan RJ. Prefrontal contributions to metacognition in perceptual decision making. J Neurosci. 2012;32:6117–6125. - PMC - PubMed
    1. Fleming SM, Weil RS, Nagy Z, Dolan R, Rees G. Relating Introspective Accuracy to Individual Differences in Brain Structure. Science. 2010;329 - PMC - PubMed
    1. Kepecs A, Uchida N, Zariwala HA, Mainen ZF. Neural correlates, computation and behavioural impact of decision confidence. Nature. 2008;455:227–231. - PubMed
    1. Kiani R, Corthell L, Shadlen MN. Choice Certainty Is Informed by Both Evidence and Decision Time. Neuron. 2014;84:1329–1342. - PMC - PubMed

LinkOut - more resources