An Intrinsic Role of Beta Oscillations in Memory for Time Estimation

Abstract

The neural mechanisms underlying time perception are of vital importance to a comprehensive understanding of behavior and cognition. Recent work has suggested a supramodal role for beta oscillations in measuring temporal intervals. However, the precise function of beta oscillations and whether their manipulation alters timing has yet to be determined. To accomplish this, we first re-analyzed two, separate EEG datasets and demonstrate that beta oscillations are associated with the retention and comparison of a memory standard for duration. We next conducted a study of 20 human participants using transcranial alternating current stimulation (tACS), over frontocentral cortex, at alpha and beta frequencies, during a visual temporal bisection task, finding that beta stimulation exclusively shifts the perception of time such that stimuli are reported as longer in duration. Finally, we decomposed trialwise choice data with a drift diffusion model of timing, revealing that the shift in timing is caused by a change in the starting point of accumulation, rather than the drift rate or threshold. Our results provide evidence for the intrinsic involvement of beta oscillations in the perception of time, and point to a specific role for beta oscillations in the encoding and retention of memory for temporal intervals.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1

EEG Re-analysis of Wiener and Thompson. (A) Time/frequency plot of slope values for a linear regression of power with the stimulus duration (electrode FCz). Notably, no effect was found for the present trial duration (Direct Effect), only the previous trial (Carryover Effect). A significant cluster within the beta frequency range (17–23 Hz) was observed, peaking at approximately 600 ms. (B) Scalp distribution of mean slope values from within the white box inset, demonstrating a frontocentral peak at FCz. Significant electrodes are highlighted in green (cluster corrected p < 0.05). (C) Time-course of mean beta power in the 17–23 Hz range for both direct (bottom) and carryover (top) effects, demonstrating the staggering of late beta power with duration for the prior but not present trial duration.

Figure 2

EEG Re-analysis of Wiener, et al.. (A) Time/frequency representation of the difference between rSMG and Mid-Occ stimulation revealed an increase in high beta power following rSMG stimulation (electrode Cz; highlighted regions are significant at p < 0.05, uncorrected for visualization purposes). (B) Scalp distribution of beta power differences from within the white box at inset, representing the cluster-corrected region of signifiance. (C) The time-course of mean beta power from electrode Cz in the high beta range.

Figure 3

(A) Task design for visual temporal bisection. Subjects viewed a fixation point for 500 ms, and where then presented with a visual stimulus (Gaussian luminance blur) for one of seven possible durations, logarithmically-spaced between 300 and 900 ms. At offset, subjects were required to classify the presented stimulus into “long” or “short” duration categories. (B) Experimental setup. All subjects started each session by performing a baseline version of the temporal bisection task. Following this, tACS electrodes were administered on the subject and stimulation was initiated; all subjects performed a baseline questionnaire at this time. In the stimulation run, subjects again performed the bisection task, while concurrently receiving stimulation (10 Hz or 20 Hz, separate days). Following stimulation, subjects again filled out a questionnaire. (C) At top, the electrode montage used for tACS, with electrodes placed over FC1 and FC2 in the international 10–20 system. At bottom, the results of electric field modeling for our chosen montage, demonstrating a maximal electric field generation within the supplementary motor area.

Figure 4

Results of tACS on temporal bisection. Top and middle graphs feature psychometric data, with the average proportion at which each interval was classified as “long”. Top graphs display the mean and spread of bisection points; inner bars and outer bars represent 95% confidence intervals and the standard deviation, respectively. bottom graphs feature chronometric functions, with the average reaction time for each interval, regardless of choice. Gumbel distribution curves for psychometric data are fit to the average proportions for visualization purposes only. Individual open data points on top panels represent the bisection point for all subjects. Stimulation was found to reduce reaction time for both frequencies, whereas beta stimulation alone significantly shifted the bisection point leftward, characterized by a greater propensity to classify stimuli as “long”. Error bars represent +/− standard error.

Figure 5

Example of the TopDDM model for temporal bisection (cf.). On a given trial, at the onset of a to-be-timed interval, a first-stage drift process is initiated that accumulates at a particular rate (A) until duration offset occurs. At offset, a second-stage is initiated, with the ending accumulated time estimate of the first-stage serving as the starting point (z). From here, a decision variable accumulates towards one of two decision boundaries (a) at a particular rate (v). The direction of the drift depends on where the starting point lies relative to a categorical boundary that is used for classifying stimuli. In the example above, the same interval may be categorized as short or long, depending on the first-stage drift rate.

Figure 6

Fits of model parameters for the hierarchical drift diffusion model. Insets represent model parameters collapsed across duration for each stimulation condition. Shaded regions and error bars represent standard deviation across subjects. Model parameters confirm earlier findings and show that stimulation decreases the threshold and increases the drift rate at both frequency ranges, but exclusively increases the starting point for beta stimulation. No impact on non-decision time was found.

Figure 7

Sliding window analysis of within-session effects. A sliding window of 131 trials was ran across choice data for each session and the bisection point was calculated within each window as described in the methods. This window was chosen as the minimum size needed for each of the seven durations to have the minimum number of trials needed to fit a psychometric function (n = 8). Data displayed above are smoothed by a 10-point moving average for the visualization of within-session trends. At right, frequency distributions of the bisection point values for each trace. The overall finding of above is that the shift in the bisection point values for beta stimulation are largely consistent across the session, rather than larger at one point in time over another. Shaded regions represent +/− standard error.