The phase of ongoing oscillations mediates the causal relation between brain excitation and visual perception

Abstract

Why does neuronal activity in sensory brain areas sometimes give rise to perception, and sometimes not? Although neuronal noise is often invoked as the key factor, a portion of this variability could also be due to the history and current state of the brain affecting cortical excitability. Here we directly test this idea by examining whether the phase of prestimulus oscillatory activity is causally linked with modulations of cortical excitability and with visual perception. Transcranial magnetic stimulation (TMS) was applied over human visual cortex to induce illusory perceptions (phosphenes) while electroencephalograms (EEGs) were simultaneously recorded. Subjects reported the presence or absence of an induced phosphene following a single pulse of TMS at perceptual threshold. The phase of ongoing alpha (∼10 Hz) oscillations within 400 ms before the pulse significantly covaried with the perceptual outcome. This effect was observed in occipital regions around the site of TMS, as well as in a distant frontocentral region. In both regions, we found a systematic relationship between prepulse EEG phase and perceptual performance: phosphene probability changed by ∼15% between opposite phases. In summary, we provide direct evidence for a chain of causal relations between the phase of ongoing oscillations, neuronal excitability, and visual perception: ongoing oscillations create periodic "windows of excitability," with sensory perception being more likely to occur at specific phases.

Figure 1.

Prepulse phase predicts phosphene perception. A, Statistical comparison between phase-locking values for trials grouped according to perception (phosphene/no-phosphene, averaged over 64 electrodes and 9 subjects) and surrogate phase-locking values computed over random subsets of trials. A significant effect indicates that ongoing phase differs statistically for phosphene and no-phosphene trials. The color map represents uncorrected p values (similar in B and C) and the white outline indicates significant effects corrected for multiple comparisons using the FDR method. There is a strongly significant phase effect (FDR = 10⁻⁵; corresponding to a p value threshold of 6.4 × 10⁻⁷) in the last 400 ms preceding the pulse, ranging in frequency from 7 to 17 Hz. The topography shows the scalp distribution of phase-locking values in the time–frequency range of 7 to 17 Hz and −400 to −50 ms. Further analyses are performed on two regions of interest: a frontal and an occipital one. B, Statistical significance of the phase difference between phosphene and no-phosphene trials, in the frontal region of interest at each time–frequency point. There is a strongly significant phase effect (FDR = 10⁻³; corresponding to a p value threshold of 6.2 × 10⁻⁵) in the 400 ms preceding the pulse and in the alpha frequency range (∼10 Hz). C, Same as B, but for the occipital region of interest.

Figure 2.

Differential activity (top), and periodic fluctuations of phosphene probability (bottom). A, Difference of ERPs between phosphene trials and no-phosphene trials for an occipital electrode (PO3). B, Difference of ERPs for a frontal electrode (AFz). The shaded area around the ERP difference corresponds to the SEM across nine subjects (SEM). The gray rectangle from −1 ms to 150 ms masks the duration of electrical artifacts induced by the TMS pulse. The black line at 0 ms corresponds to the pulse onset and at 600 ms to the onset of the response screen, asking subjects to report the presence or absence of a phosphene. The ERP difference diverges from zero between 250 and 600 ms after the pulse, supporting the notion that the two conditions differ at the perceptual level. In addition, the difference oscillates during the 400 ms before the pulse onset, suggesting that the two conditions correspond to different prepulse phases. C, Relationship between the prepulse occipital EEG phase (expressed in radians) at −77 ms and 10.7 Hz (time–frequency point of maximal significance), and phosphene perception (expressed as a modulation of average phosphene probability). Single trials were sorted in 10 phase bins. Phosphene probability was computed for each phase bin, then averaged over all electrodes in the occipital region of interest and over all subjects (error bars represent SEM across subjects). For enhanced readability, the curve is plotted over two consecutive cycles (phase values that were replicated are represented over shaded backgrounds). The curve demonstrates that phosphene report probability significantly oscillates along with ongoing EEG phase (one-way ANOVA over the 10 phase values, F_(9,89) = 2.35, p = 0.021). The magnitude of this periodic modulation (measured between optimal and opposite phase values) is ∼13%. D, Relationship between the prepulse EEG phase at −40 ms and 12.6 Hz (time–frequency point of maximal significance) and phosphene perception, averaged over all electrodes of the frontal region of interest. Plotting conventions are similar to those in C. As previously, phosphene probability significantly oscillates along with ongoing EEG phase (one-way ANOVA, F_(9,89) = 2.72, p = 0.008). The magnitude of this periodic modulation is ∼15%.