**3. Results**

Observers' capability of detecting video speed modulation is shown in Figure 2. With weak signals (10% amplitude of video speed modulation), perceptual sensitivity to speed modulation was practically null with both Ripples and RDK video clips (confidence intervals crossed zero in all cases), but increased significantly with speed modulation amplitude. Neither speed modulation frequency nor the interaction frequency × amplitude reached statistical significance (Table 1).

**Figure 2.** Perceptual sensitivity to video speed modulation as a function of video speed modulation amplitude (10%, 30%, and 50%) and frequency (1, 2, and 4 Hz), shown separately for Ripples and RDK videos. Error bars are 95% confidence intervals.

**Table 1.** Results of the generalized linear mixed models (GLMM) analysis of perceptual sensitivity.


Video speed modulation amplitude, but not frequency, or their interaction, was statistically significant. Abbreviations: e—coe fficient estimate, t—t-statistics (Wald t-test), d—degrees of freedom, *p*—*p*-value, l—lower confidence bound, u—upper confidence bound.

By contrast, cortical oscillatory responses (steady-state visual evoked potentials, SSVEPs) were found at all video speed modulation amplitudes, with mean adjusted R<sup>2</sup> for the sinusoidal fittings ranging from 13% to 36%, which indicated that a relevant component of cortical activity was pulsating at the stimulus frequency. Although an evident oscillation could not be discerned in all traces (e.g., the 4 Hz, 10% condition of Figure 3), in many cases the cortical potentials followed the rhythm of the video speed (e.g., the 2 Hz, 10% condition of Figure 3). SSVEP amplitude, as computed through sinusoidal fitting, was significantly above zero at all amplitudes and frequencies of speed modulation, as shown by confidence intervals (Figure 4). Also, SSVEP amplitude tended to moderately but significantly increase as speed modulation amplitude increased, though not as steeply as perceptual sensitivity. Speed modulation frequency, but not the interaction frequency × amplitude, also reached statistical significance. The results of these analyses are reported in Table 2. Note also that Figure 3 suggests little SSVEP di fferentiation across the cortex: indeed, there was a large overlap among the random coe fficients of the eight recording channels, with the highest contribution to SSVEP variability coming from F4 (Table 3).

**Figure 3.** Examples of electroencephalographic (EEG) recordings from one participant showing the traces from all channels (shown in different colors, see legend), averaged across a 1-s time window and superimposed. The sinusoidal best-fitting curves are also shown (black lines; for graphical simplicity, here only a single curve averaged across channels is shown). Also reported are the values of stimulus modulation frequency, stimulus modulation strength, and adjusted R<sup>2</sup> of the fitted functions.

**Figure 4.** Steady-state visual evoked potential (SSVEP) amplitude as a function of speed modulation amplitude and frequency, shown separately for Ripples and RDK videos. Data have been averaged across participants. Error bars are 95% confidence intervals.

**Table 2.** Results of the GLMM analysis of SSVEP amplitude.


Video speed modulation amplitude and frequency, but not their interaction, were statistically significant. See Table 1 for abbreviations.


**Table 3.** Coefficient estimates resulting from the GLMM analysis of SSVEP amplitude for the eight recording channels. See text and Table 1 for abbreviations.

In Table 4, the effect sizes for both perceptual and cortical responses are reported. Perceptual responses tended to be more affected by speed modulation amplitude than speed modulation frequency, whereas the opposite held for cortical responses.

**Table 4.** Estimates of effect size (partial η2) of video speed modulation amplitude and frequency on perceptual and cortical responses, shown separately for Ripples and RDK stimuli.

