The rate of transient beta frequency events predicts behavior across tasks and species
- Brown University, United States
- Providence VA Medical Center, United States
Figures
Figure 1 with 1 supplement
Average 1 s prestimulus beta (15–29 Hz) power is higher on non-detected/attend out trials in SI.
(i) Mean prestimulus beta power as a function of time relative to stimulus onset in (A) human MEG during a tactile detection task (source localized data from hand area of SI; N = 10 subjects, 100 …
Figure 1—figure supplement 1
Evoked response to suprathreshold sensory input and beta power on catch trials.
(A) (i) Human MEG evoked response to a suprathreshold tap to the fingertip of the third digit of the right hand. (reproduced from Jones et al., 2007) (ii) Mouse LFP evoked response to a …
Figure 2
Beta emerges as brief events in non-averaged spectrograms.
(A) Top panel shows averaged spectrogram (1–40 Hz) in the 1 s prestimulus period from 100 trials, from an example subject in the human detection dataset (average across 100 non-detected trials). The …
Figure 3
Schematic illustration of possible features underlying differences in averaged prestimulus beta power.
Given that surges in beta power in non-averaged data occur as transient events, higher trial mean beta power could be due to an increase in (A) event number (i.e. rate), (B) event power, (C) event …
Figure 4
Beta events defined by a 6X median power cutoff show consistently high correlation with average prestimulus beta.
(i) Pearson’s correlation coefficient between mean prestimulus beta power and the percent area (i.e. percentage of pixels in the spectrogram) above cutoff in the non-averaged spectrogram, for …
Figure 5 with 1 supplement
Beta event features are highly conserved across tasks and species.
Probability histogram for each prestimulus beta event feature; (i) event number, (ii) event power, (iii) event duration, (iv) event frequency span; defined at 6X median cutoff in each dataset (A–C). …
Figure 5—figure supplement 1
Beta event duration and frequency span is stereotyped when maxima power is higher than 6X median cutoff.
Scatter plots of all local maxima aggregated across all subjects/sessions in each dataset.
Figure 6 with 2 supplements
The number of beta events has a higher correlation with trial mean prestimulus beta power than event power, duration or frequency span.
Box and whisker plots over subjects/sessions depicting the Pearson’s correlation coefficients () between the trial summary of each beta event feature and trial mean prestimulus beta power. Note, …
Figure 6—figure supplement 1
Example scatter plots showing relationship between trial mean prestimulus beta power and corresponding beta event number, trial mean event power, duration and frequency span.
Scatter plot showing trial-by-trial relationship between trial mean prestimulus beta power and trial summary of each event feature. Data shown for one representative subject (A, C)/session (B) from …
Figure 6—figure supplement 2
Dependence of correlation with mean power on choice of cutoff.
Pearson’s correlation with trial mean prestimulus power for event number per trial (blue), trial mean event power (red), trial mean event duration (yellow) and trial mean event frequency span …
Figure 7 with 2 supplements
Beta event number per trial (i.e. rate) consistently predicts detection/attention across tasks and species.
(i) Probability time histogram of beta event occurrence in the 1 s prestimulus window. Data binned in 50 ms windows, sliding in 1 ms steps. All other notations are same as Figure 1i. (ii-iv) …
Figure 7—figure supplement 1
Dependence of relationship between beta event features and behavior on choice of power cutoff.
Percentage of subjects with DP/AP significantly less than 0.5 (i.e. beta is miss predictive) for trial summary of each beta event feature, over a range of cutoffs (significance determined by 95% …
Figure 7—figure supplement 2
Optimal criterion for classifying behavior based on event number is two events per trial.
A-C (i) Probability histogram of event number across conditions (blue: detected/attend in; red: non-detected/attend out; mean ± SEM across subjects/sessions). (ii) % subjects/sessions with optimal …
Figure 8 with 1 supplement
Non-detected trials are more likely to have a beta event closer to the time of the stimulus.
For each trial, the event occuring closest to the stimulus onset was termed the ‘most recent beta event’. All analyses are analogous to Figure 7, in the human (A) and mouse (B) detection datasets, …
Figure 8—figure supplement 1
Dependence of relationship between ‘most recent beta event’ features and behavior on choice of cutoff.
Same as in Figure 7—figure supplement 1, but for ‘most recent beta event’ features.
Figure 9
Independent influence of rate and timing on detection.
(i)Left panel: Event number histogram in detected (blue, mean ± SEM), non-detected (red, mean ± SEM), and event number matched (black, mean across subjects/sessions) trials via a random trial …
Figure 10
Inter-event-interval (lEI) analyses suggest that beta events are not rhythmic or a renewal process.
(i) IEI distributions (20 ms bins) on detected/attend in (blue) and non-detected/attend out (red) conditions; mean across subjects with ± SEM as error bars. (ii) Fano factor and for each subject …
Figure 11
Beta event findings are robust under redefinition of events as ‘non-overlapping beta events’.
(i) Probability histograms for number of local maxima per supra-cutoff region (i.e. a ‘non-overlapping beta event’), in detected (blue) and non-detected (red) trials; mean ± SEM across subjects. …
Figure 12
Beta event emergence is consistent with a bursty generator mechanisms, as opposed to dynamic amplitude modulation of a sustained beta oscillation.
(A) Schematic illustration of two alternative mechanisms that could underlie transient surges in beta power in the spectrogram: (i) Dynamic Amplitude Modulation mechanism; versus, (ii) Bursty …
Author response image 1
Hilbert-transformed amplitude histograms do not consistently show bimodality.
Hilbert-transformed amplitude histograms of band-pass (14~29Hz) filtered data for each subject / session in A. Human Detection, B. Mouse Detection, C. Human Attention datasets. The amplitude values …
Author response image 2
Spearman’s correlation with trial mean prestimulus beta power.
All trials were included for all features, and trial mean event power, duration and frequency span as zero for zero event trials.
Author response image 3
Multiple linear regression slopes, with trial mean prestimulus beta power as the response variable and the four trial summary event features as regressors.
All trials were included for all four features; trial mean event power, duration and frequency span were defined as zero for zero event trials.
Additional files
-
Transparent reporting form
- https://doi.org/10.7554/eLife.29086.021
