Mixed evidence for the rhythmicity of auditory perceptual judgements in humans
- Department for Cognitive Neuroscience, Faculty of Biology, Bielefeld University, Germany
Figures
Figure 1
Experimental design and rationale.
(A) Rhythmicity in behavioural data could have several origins. On the one extreme are rhythmic processes in monaural auditory processes that operate at different frequencies for sounds in each ear (left scheme). As a result, one would observe signatures of rhythmicity in behavioural data for monaural targets at different time scales, and a yet different pattern when presenting binaural stimuli (orange). On the other extreme are high-level processes possibility more tied to cognition than sensation, that operate in the same manner on any sound and which give rise to the same rhythmicity in behavioural data following monaural or binaural stimuli (right scheme). (B) General design of the four experiments. In each experiment, independent white noise was presented to each ear over a period of 1.8 s after a pseudo-random fixation period (0.3–0.5 s). The task-relevant sounds were presented monaurally and at random time points between 0.3 and 1.5 s following the noise onset and either comprised one (40 ms) or two (30 ms) sounds. Experiments 1, 2, 4 featured a tone discrimination task with a decision criterion fixed across trials (tones categorised as ‘low’ or ‘high’). Experiment 3 featured a within-trial discrimination of two subsequent tones. Experiments 1 and 2 differ in that sound presentation was automatically paced or required manual initialisation by the participant. Experiment 1 and 4 differed in that the latter also required participants to perform a dual task on in the visual fixation dot. This was intended to divert attention across sensory modalities.
Figure 2
Calibration of analysis approaches on simulated data.
We simulated data with and without rhythmicity to calibrate the statistical specificity (false positive rate) of each approach. The approaches are based on the spectra derived from delay-binned data (‘Spectra’), the vector strength of a rhythmic component in linear models applied to delay-binned data (‘Binned’) and the vector strength of a rhythmic component in linear models applied to single-trial data (‘Trials’). For each, we determined the respective first-level threshold (i.e. method-specific p-value) that results in a false-positive rate of about 0.05 when probing for rhythmicity at any frequency. (A) Sensitivity of each approach in detecting a rhythmic effect for data generated with a rhythmic effect and different signal-to-noise ratios (SNR). (B) Specificity, shown here as false positive rate in detecting a rhythmic effect in data generated without such an effect (calibrated to about 0.05). (C) Illustration of the (log-transformed) first-level p-values for simulated data with an effect at 4 Hz and different SNRs, showing the frequency specificity of each approach in detecting an effect.
Figure 3
Illustration of the data for experiments 1–3.
(A) Sensitivity (d-prime), bias, and reaction times (rt) regardless of target delay. These metrics are shown separately for each ear (R, L) and reaction times (in seconds) are shown separately for each stimulus condition (f1, f2; corresponding to the two target frequencies in experiments 1 and 2 or the two orders of pitch in experiment 3). (B) The same metrics for trials with targets presented at specific delays within the 1.2 s range of target uncertainty. Grey dots and lines denote the individual participant data, thick dots and error bars denote the group average and standard deviation.
Figure 4
Significance of rhythmicity in the behavioural data.
The individual panels show the group-level (first-level) p-values for each approach (A–C) and experiment together with the statistical cut-off used to determine significance (thick grey line; corresponding to p<0.05 corrected for multiple tests across frequencies, calibrating false-positive rate across analyses). For comparison, the dashed grey line also shows the cut-off at p<0.01. The coloured shadings indicate significant effects to facilitate their comparison across panels. The precise frequencies with significant effects were: Exp 1: Binned/d-prime: 1.8–2.0 Hz; Exp 2: Binned/bias 6.4–6.6 Hz; Exp 3: Binned/d-prime 2.4–2.8 Hz.
Figure 5
Rhythmic effects in the binned analysis for experiment 1 in the actual data (coloured) and the surrogate data.
The rhythmic effect here is the combined vector strength of the sine and cosine predictors in the binned analysis, averaged across participants and shown for each metric and ear (colour-coded). For the surrogate data, we picked the percentile corresponding to the significance level of p=0.05 and showed the respective group-average surrogate data. In experiment 1, only d-prime for the right ear data was significant around 2 Hz. The figure also illustrates how the effective strength of a rhythmic predictor required to achieve significance varies between metrics, frequencies, and ear combinations.
Figure 6
Prevalence of significant effects across random variations in the participant sample.
Using a resampling approach, we determined the percentage of participant samples that yield a significant effect when repeating the analysis. For each simulated participant sample, we determined significant effects at Pp<0.05 (as in Figure 4) and then counted the number of simulations with effects at each frequency. A value of above e.g., 50% (dashed line) indicates that when randomly sampling from within the collected participants more than 50% of such simulated experiments yield a significant effect. The coloured shadings highlight the same effects as shown in Figure 4.
Figure 7
Averaged prevalence across experiments 1–3.
We focus on the two approaches and metrics yielding the strongest effects in the data for individual experiments. To allow better visibility of the differences between individual ears, the combined-ear data are shown dashed.
Figure 8
Influence of dual task and oculomotor behaviour on rhythmicity in experiment 4.
Evidence for rhythmicity was obtained as the group-average vector strength in the model-based analysis of delay-binned data. This evidence was compared between conditions with the graphs showing the difference (mean, s.e.m.). (A) Comparison of blocks with dual task those without (n=36). (B) Comparison of trials with large pupil diameter minus trials with small diameter (n=24). (C) Comparison of trials with less fixation stability (more eye mobility) minus trials with more stability (less eye mobility, n=24). The left panels show the group-level difference (mean, s.e.m.) across the collected participant sample for the combined-ear data. Dots indicate significant effects (paired t-tests, corrected for multiple tests across frequencies at p<0.05; for d-prime in panel C, these are 3.4–3.5 Hz). The right panels show the prevalence of effects across random variations in the participant sample for both the combined-ear and individual data.
Tables
Key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Software, algorithm | Matlab | Mathworks | Version R2022a |
Additional files
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)
Mixed evidence for the rhythmicity of auditory perceptual judgements in humans
eLife 14:RP105734.
https://doi.org/10.7554/eLife.105734.3
