Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Prediction errors bidirectionally bias time perception

Abstract

Time perception and prediction errors are essential for everyday life. We hypothesized that their putative shared circuitry in the striatum might enable these two functions to interact. We show that positive and negative prediction errors bias time perception by increasing and decreasing perceived time, respectively. Imaging and behavioral modeling identify this interaction to occur in the putamen. Depending on context, this interaction may have beneficial or adverse effects.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Fig. 1: PEs bidirectionally bias time perception.
Fig. 2: Modeling the PE–time bias.
Fig. 3: Putamen activity underlies PE–time interaction and behavioral bias.

Data availability

All data supporting the findings of this study are available from the corresponding author upon reasonable request.

Code availability

Custom code for behavioral and imaging tests is available from the corresponding author upon reasonable request.

References

  1. 1.

    Mauk, M. D. & Buonomano, D. V. The neural basis of temporal processing. Annu. Rev. Neurosci. 27, 307–340 (2004).

    CAS  Article  Google Scholar 

  2. 2.

    Dirnberger, G. et al. Give it time: neural evidence for distorted time perception and enhanced memory encoding in emotional situations. Neuroimage 63, 591–599 (2012).

    Article  Google Scholar 

  3. 3.

    Droit-Volet, S. & Meck, W. H. How emotions colour our perception of time. Trends Cogn. Sci. 11, 504–513 (2007).

    Article  Google Scholar 

  4. 4.

    Schultz, W., Dayan, P. & Montague, P. R. A neural substrate of prediction and reward. Science 275, 1593–1599 (1997).

    CAS  Article  Google Scholar 

  5. 5.

    Niv, Y. & Schoenbaum, G. Dialogues on prediction errors. Trends Cogn. Sci. 12, 265–272 (2008).

    Article  Google Scholar 

  6. 6.

    Petter, E. A., Gershman, S. J. & Meck, W. H. Integrating models of interval timing and reinforcement learning. Trends Cogn. Sci. 22, 911–922 (2018).

    Article  Google Scholar 

  7. 7.

    Mikhael, J. G. & Gershman, S. J. Adapting the flow of time with dopamine. J. Neurophysiol. 121, 1748–1760 (2019).

    CAS  Article  Google Scholar 

  8. 8.

    Coull, J. & Nobre, A. Dissociating explicit timing from temporal expectation with fMRI. Curr. Opin. Neurobiol. 18, 137–144 (2008).

    CAS  Article  Google Scholar 

  9. 9.

    Jahanshahi, M., Jones, C. R., Dirnberger, G. & Frith, C. D. The substantia nigra pars compacta and temporal processing. J. Neurosci. 26, 12266–12273 (2006).

    CAS  Article  Google Scholar 

  10. 10.

    D'ardenne, K., McClure, S. M., Nystrom, L. E. & Cohen, J. D. BOLD responses reflecting dopaminergic signals in the human ventral tegmental area. Science 319, 1264–1267 (2008).

    CAS  Article  Google Scholar 

  11. 11.

    Soares, S., Atallah, B. V. & Paton, J. J. Midbrain dopamine neurons control judgment of time. Science 354, 1273–1277 (2016).

    CAS  Article  Google Scholar 

  12. 12.

    Allman, M. J. & Meck, W. H. Pathophysiological distortions in time perception and timed performance. Brain 135, 656–677 (2012).

    Article  Google Scholar 

  13. 13.

    Jones, C. R. & Jahanshahi, M. Motor and perceptual timing in Parkinson’s disease. Adv. Exp. Med. Biol. 829, 265–290 (2014).

    Article  Google Scholar 

  14. 14.

    Bavard, S., Lebreton, M., Khamassi, M., Coricelli, G. & Palminteri, S. Reference-point centering and range-adaptation enhance human reinforcement learning at the cost of irrational preferences. Nat. Commun. 9, 1–12 (2018).

    CAS  Article  Google Scholar 

  15. 15.

    Klein, T. A., Ullsperger, M. & Jocham, G. Learning relative values in the striatum induces violations of normative decision making. Nat. Commun. 8, 16033 (2017).

    CAS  Article  Google Scholar 

  16. 16.

    Fayolle, S., Gil, S. & Droit-Volet, S. Fear and time: fear speeds up the internal clock. Behav. Processes 120, 135–140 (2015).

    Article  Google Scholar 

  17. 17.

    Smith, S. D., McIver, T. A., Di Nella, M. S. & Crease, M. L. The effects of valence and arousal on the emotional modulation of time perception: evidence for multiple stages of processing. Emotion 11, 1305–1313 (2011).

    Article  Google Scholar 

  18. 18.

    Eagleman, D. M. Human time perception and its illusions. Curr. Opin. Neurobiol. 18, 131–136 (2008).

    CAS  Article  Google Scholar 

  19. 19.

    Failing, M. & Theeuwes, J. Reward alters the perception of time. Cognition 148, 19–26 (2016).

    Article  Google Scholar 

  20. 20.

    Tse, P. U. Attention underlies subjective temporal expansion. in Attention and Time (eds Nobre, A. C. & Coull, J. T.) 137–150 (Oxford University Press, 2010).

  21. 21.

    Brainard, D. H. & Vision, S. The psychophysics toolbox. Spat. Vis. 10, 433–436 (1997).

    CAS  Article  Google Scholar 

  22. 22.

    Pelli, D. G. The VideoToolbox software for visual psychophysics: transforming numbers into movies. Spat. Vis. 10, 437–442 (1997).

    CAS  Article  Google Scholar 

  23. 23.

    Rammsayer, T. & Ulrich, R. The greater temporal acuity in the reminder task than in the 2AFC task is independent of standard duration and sensory modality. Can. J. exp. Psychol. 66, 26–31 (2012).

    Article  Google Scholar 

  24. 24.

    Ulrich, R. & Vorberg, D. Estimating the difference limen in 2AFC tasks: pitfalls and improved estimators. Atten. Percept. Psychophys. 71, 1219–1227 (2009).

    Article  Google Scholar 

  25. 25.

    Kahneman, D. Prospect theory: an analysis of decisions under risk. Econometrica 47, 263–292 (1979).

    Article  Google Scholar 

  26. 26.

    Tom, S. M., Fox, C. R., Trepel, C. & Poldrack, R. A. The neural basis of loss aversion in decision-making under risk. Science 315, 515–518 (2007).

    CAS  Article  Google Scholar 

  27. 27.

    Levitt, H. Transformed up‐down methods in psychoacoustics. J. Acoust. Soc. Am. 49, 467–477 (1971).

    Article  Google Scholar 

  28. 28.

    Allan, L. G. The time-order error in judgments of duration. Can. J. Psychol. 31, 24–31 (1977).

    Article  Google Scholar 

  29. 29.

    Hellström, Å. The time-order error and its relatives: mirrors of cognitive processes in comparing. Psychol. Bull. 97, 35–61 (1985).

    Article  Google Scholar 

  30. 30.

    Laufer, O. & Paz, R. Monetary loss alters perceptual thresholds and compromises future decisions via amygdala and prefrontal networks. J. Neurosci. 32, 6304–6311 (2012).

    CAS  Article  Google Scholar 

  31. 31.

    Taatgen, N. A., Van Rijn, H. & Anderson, J. An integrated theory of prospective time interval estimation: the role of cognition, attention and learning. Psychol. Rev. 114, 577–598 (2007).

    Article  Google Scholar 

  32. 32.

    Deichmann, R., Gottfried, J. A., Hutton, C. & Turner, R. Optimized EPI for fMRI studies of the orbitofrontal cortex. Neuroimage 19, 430–441 (2003).

    CAS  Article  Google Scholar 

  33. 33.

    Dale, A. M. Optimal experimental design for event-related fMRI. Hum. Brain Mapp. 8, 109–114 (1999).

    CAS  Article  Google Scholar 

  34. 34.

    Goebel, R., Esposito, F. & Formisano, E. Analysis of functional image analysis contest (FIAC) data with brainvoyager QX: from single‐subject to cortically aligned group general linear model analysis and self‐organizing group independent component analysis. Hum. Brain Mapp. 27, 392–401 (2006).

    Article  Google Scholar 

  35. 35.

    Talairach, J. & Tournoux, P. Co-planar Stereotaxic Atlas of the Human Brain (Thieme Medical Publishers, 1988).

  36. 36.

    Eklund, A., Nichols, T. E. & Knutsson, H. Cluster failure: why fMRI inferences for spatial extent have inflated false-positive rates. Proc. Natl Acad. Sci. USA 113, 7900–7905 (2016).

    CAS  Article  Google Scholar 

  37. 37.

    Woo, C.-W., Krishnan, A. & Wager, T. D. Cluster-extent based thresholding in fMRI analyses: pitfalls and recommendations. Neuroimage 91, 412–419 (2014).

    Article  Google Scholar 

  38. 38.

    Stephan, K. E., Penny, W. D., Daunizeau, J., Moran, R. J. & Friston, K. J. Bayesian model selection for group studies. Neuroimage 46, 1004–1017 (2009).

    Article  Google Scholar 

Download references

Acknowledgements

We thank E. Furman-Haran and F. Attar for MRI procedures. This work was supported by a Joy-Ventures grant, Israel Science Foundation grant no.2352/19 and a European Research Council grant (ERC-2016-CoG no. 724910) to R.P.

Author information

Affiliations

Authors

Contributions

I.T. and R.P. designed the study. I.T. performed the experiments and analyzed the data. K.C.A. contributed ideas for analysis. I.T., K.C.A. and R.P. wrote the manuscript.

Corresponding author

Correspondence to Rony Paz.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

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

Extended data

Extended Data Fig. 1 Behavior-only group’s performance, separated into five levels of PE.

Data presented is similar to main Fig. 1, but only for participants in the behavior-only group, when separating trials into 5 levels of PE (PE0, small (<|3|) and large (>=|3|) PE+/PE-). a. Proportion of discrimination errors as a function of objective time difference between the two images (Δt), averaged across all trials separately for each PE-type. Shown is the main effect of Δt (F4,68 = 26.9, p <10−12, η2 = 0.61), but no difference between PE types (F4,68 = 0.7, p = 0.59). b. Proportion of discrimination errors as a function of PE-type and Trial-type. An interaction between PE-type and Trial-type is evident as PE+ and PE- bias performance in opposite directions relative to PE0 trials (F4,68 = 9.79, p < 10−5, η2 = 0.36). Inset: after correction for individual Time-Order Error (TOE). c. Proportion of discrimination errors as a function of time difference between images (Δt), separately for LS and SL trials. PE+ and PE- bias performance in opposite directions relative to PE0, for all values of Δt. Different outcome magnitudes are not significantly different (p>0.12), and no main effect of Trial-type (F1,17 = 4.1, p = 0.06). d. Proportion of discrimination errors as a function of time difference between images (Δt), separately for LS and SL trials and corrected for TOE. PE+ and PE- bias performance in opposite directions relative to PE0, for all values of Δt. e. Proportion of discrimination errors as a function of individual Just-Noticeable-Difference (JND), corrected for TOE, fitted to a logistic function after JND normalization. Error bars represent SEM (n=18).

Extended Data Fig. 2 Behavior-only group’s performance, separated into three levels of PE.

Data is presented for participants in the behavior-only group, when separating trials into three levels of PE. a. Proportion of discrimination errors as a function of objective time difference between the two images (Δt), averaged across all trials separately for each PE-type. No difference between PE-types (F2,34 = 1.5, p = 0.24). b. Proportion of discrimination errors as a function of PE-type and Trial-type. An interaction between PE-type and Trial-type (F2,34 = 10.89, p = 0.0002, η2 = 0.39). SL trials: mean PE+ = 0.3, mean PE- = 0.49, mean PE0 = 0.37. Better performance in PE+ vs. PE- (p=0.008, Cohen’s d = 0.81), worse performance in PE- compared to PE0 (p = 0.05, Cohen’s d = 0.6). LS trials: mean PE+ = 0.39, mean PE- = 0.21, mean PE0 = 0.275. Worse performance during PE+ vs. PE- (p = 0.005, Cohen’s d = 0.86) and PE0 (p = 0.02, Cohen’s d = 0.68). There was no significant difference between PE- and PE0 trials (p = 0.11). Inset: Performance after correcting for individual Time-Order Error (TOE). c. Proportion of discrimination errors as a function of the time difference between images (Δt), separately for LS and SL trials. PE+ and PE- bias performance in opposite directions relative to PE0 for all values of Δt. No main effect of Trial-type (F1,17 = 4.03, p = 0.06). d. Proportion of discrimination errors as a function of the time difference between images (Δt), separately for LS and SL trials and corrected for TOE. PE+ and PE- bias performance in opposite directions relative to PE0 for all values of Δt. e. Proportion of discrimination errors as a function of the individual Just-Noticeable-Difference (JND), corrected for TOE, fitted to a logistic function after JND normalization. f. Model fit to individual behavioral data, averaged over subjects. Error bars and error bands represent SEM (n=18).

Extended Data Fig. 3 Imaging-group’s performance.

Data is presented for participants in the imaging (fMRI) group, similar to main Fig. 1 and Supplementary Fig. 1, 2. a. Proportion of discrimination errors as a function of objective time difference between the two images (Δt), averaged across all trials separately for each PE-type. Shown is the main effect of Δt (F4,136 = 59.05, p < 10−5, η2 = 0.63), as well as significant interaction between Trial type and Δt (F4,136 = 6.6, p < 10−5, η2 = 0.16). b. Proportion of discrimination errors as a function of PE-type and Trial-type. An interaction between PE-type and Trial-type (F2,68 = 9.1, p = 0.003, η2 = 0.21). SL trials: mean PE+ = 0.33, mean PE- = 0.5, mean PE0 = 0.46. Better performance in PE+ vs. PE- (p<0.001, Cohen’s d = 0.77) and PE0 (p<0.001, Cohen’s d = 0.64), no difference between PE- and PE0. LS trials: mean PE+ = 0.31, mean PE- = 0.23, mean PE0 = 0.28, worse performance in for PE+ vs. PE- (p=0.05, Cohen’s d = 0.4), no difference in performance between PE+ or PE- and PE0. Inset: Performance after correcting for individual Time-Order Error (TOE). c. Proportion of discrimination errors as a function of time difference between images (Δt), separately for LS and SL trials. PE+ and PE- bias performance in opposite directions relative to PE0, for all values of Δt. d. Proportion of discrimination errors as a function of time difference between images (Δt), separately for LS and SL trials and corrected for TOE. PE+ and PE- bias performance in opposite directions relative to PE0, for all values of Δt. e. Proportion of discrimination errors as a function of individual Just-Noticeable-Difference (JND), corrected for TOE, fitted to a logistic function after JND normalization. f. Model fit to individual behavioral data, averaged over subjects. Error bars and error bands represent SEM (n=35). Notice that in this group (imaging), PE0 induced a slight time perception bias, because the distance between PE+ and PE0 is not similar to the distance between PE- and PE0 (see b, c, d, e). This is explained by the different values used in this experiment (−2,4), that create an expected value of 0.2 on average, and hence PE0 also has a small negative PE. For a full explanation and direct experimental evidence, see Supplementary Note subsection “A behavioral difference between the imaging and the behavior-only groups and its explanation (or: why does PE0 also have a slight bias in the fMRI group)”.

Extended Data Fig. 4 Comparing performance between the behavior-only and the imaging group.

a. Time order error (TOE) as a function of Δt, measured from the behavior. TOE is positive in both groups across all Δt, indicating that the first stimulus is perceived as longer. b. Mean expected value from the model in the last 50 trials. Overall, a higher expected value can be seen in the imaging group (two-samples t-test; t49 = 2.4, p = 0.02). This difference is caused by a larger difference in magnitude between the positive and negative outcomes used for the imaging group (+4/−2), as compared to the behavior-only group (+2.5/−2.5). c. Proportion of discrimination errors grouped by PE-type. No difference in performance is seen between the groups for PE+/- (t51 = 0.8/0.9, p = 0.4/0.37, Bayes-Factor >2.5). However, higher proportion of errors in PE0 was found in the imaging group (t51 = 2.52, p = 0.01, Cohen’s d = 0.185), suggesting a non-zero value assigned to these trials during the main paradigm. Right: Proportion of discrimination errors for PE0 trials in the first and second half of main paradigm in the imaging group. The shallower slope (that is indicating worse discrimination) for later trials may result from the fact that the expected value of a trial becomes non-zero as a result of the imbalance between positive and negative outcomes (+4/−2, see b and c, same Figure). Error bars and error bands represent SEM (n=18, 35 for behavior-only and imaging groups, respectively).

Extended Data Fig. 5 Proportion of discrimination errors and TOE correction.

Proportion of discrimination errors as a function of trial type (Long-Short; LS or Short-Long; SL) and (Δt). a. Left: For the raw data, there is a larger difference in performance between LS and SL trials for small Δt (“hard” trials, Δt < 67 ms) compared with large Δt (“easy” trials; mean difference = 0.2 in “hard” and 0.08 in “easy” trials), resulted from significant interaction between Trial-type and Δt (F4,208 = 11.3, p < 10−7, η2 = 0.175). Right: Data corrected for TOE shows similar performance differences between the different levels of Δt for LS and SL trials (F4,208 = 0.75, p = 0.55). b. Separated according to group (before correction to TOE). TOE occurred in all groups, and was therefore corrected for all (see top-right). Error bars represent SEM (n=18, 35 for behavior-only and imaging groups, respectively).

Extended Data Fig. 6 Stability of perceptual thresholds and value control.

a. Comparison of four different methods for computing individual Just-Noticeable-Difference (JND), estimated on the behavior-only group, indicating no difference between the approaches (see Methods for the different approaches; F2,46 = 0.7, p = 0.5, Bayes-Factor > 1.5) which validates the use of staircase JND to normalize the results (n=18). b. Comparison between JND before and after main paradigm. Left plot demonstrates no difference between different measures of JND after main paradigm, whereas right plot demonstrates that perceptual threshold did not change (after minus before) due to main paradigm and was identical across different images (n=47 for first three violins, n=22 for last two violins). The two sets of left violins measure JND when the two consecutive images were overlaid with the same number (positive or negative), to control for value and visual differences. The middle (PE0) and right set of violins (PE+/PE-) measure JND when the two consecutive images are either [0 0],[0 +4],[0–2], correspondingly. Therefore, although we denote it as PE here for clarity, there is no real prediction-error because the outcome is fully predicted within the block. Before the paradigm numbers did not have any meaning (that is did not provide reward/loss), while at the end they did (like in the main paradigm). Overall, the different conditions show that perceptual thresholds did not change: before to after the main paradigm, with or w/o real outcome, and with outcome that is predicted (that is no PE). Error bars represent SEM.

Extended Data Fig. 7 Performance in the Gain-only group.

a. Proportion of discrimination errors as a function of objective time difference between the two images (Δt), averaged across all trials separately for each PE-type. Shown is main effect of Δt (F5,90 = 21.1, p <10−13, η2 = 0.5) but there was no difference between PE-types (F2,36 = 0.59, p = 0.55) or Trial-type (F1,18 = 3.4, p = 0.078), as seen in the main experiment. b. Proportion of discrimination errors as a function of PE-type and Trial-type. An interaction between PE-type and Trial-type (PE-Type x Trial-type: F2,36 = 6.77, p = 0.003, η2 = 0.273). SL trials: mean PE+ = 0.38, mean PE- = 0.48, mean PE0 = 0.41. Better performance in PE+ vs. PE- (p=0.03, Cohen’s d = 0.64). LS trials: mean PE+ = 0.4, mean PE- = 0.3, mean PE0 = 0.32. Worse performance in PE+ vs. PE- (p=0.07, Cohen’s d = 0.54). Inset: Performance after correcting for individual Time-Order Error (TOE). c. Proportion of discrimination errors as a function of time difference between images (Δt), separately for LS and SL trials. PE+ and PE- bias performance in opposite directions. d. Proportion of discrimination errors as a function of individual Just-Noticeable-Difference (JND), fitted to a logistic function after JND normalization. Error bars represent SEM (n=19).

Extended Data Fig. 8 ROIs where activation correlates with trial-to-trial model-derived PE+ signals.

Shown are activations in the dACC and PCC (n=35). Displayed on average brain. Time courses represent mean % signal change extracted from the ROIs ± SEM. Black vertical lines represent trial onset, trial offset and average onset of next trial, respectively. Activation was set to statistical threshold of q = 0.055 for visualization.

Supplementary information

Supplementary Information

Supplementary Note and Supplementary Tables 1–4.

Reporting Summary

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Toren, I., Aberg, K.C. & Paz, R. Prediction errors bidirectionally bias time perception. Nat Neurosci 23, 1198–1202 (2020). https://doi.org/10.1038/s41593-020-0698-3

Download citation

Further reading

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing