Auditory cue-induced prediction templates fluctuate at alpha frequencies. A, Top, The power spectrum of pre-stimulus (−1,250 to 0 ms) shape decoding shows significant deviations from an empirical null distribution at 10 and 11 Hz (***p = 0.003). The baseline power spectrum (dark blue line) was obtained by bootstrapping (n = 10,000) shuffled label decoding data (n = 25 per participant). Mean and shaded regions indicate SD. Solid orange line indicates the 95th percentile of the null distribution. Error bars indicate SEM. A, Bottom, Pre-stimulus (−1,250 to 0 ms) MEG data revealed significantly higher 10–11 Hz power for shape A versus D decoding than for shape B versus C decoding (*p < 0.05). Error bars indicate power of shape A versus D decoding; dark purple line indicates power of shape B versus C decoding (applied to identical pre-stimulus prediction data). Shaded regions and error bars indicate SEM. B, Top, Topography of sensor-level alpha (10 Hz) coherence between shape-specific signals (i.e., decoder output) and pre-stimulus sensor-level raw MEG signals (−1,250 to 0 ms). B, Bottom, Topography of raw pre-stimulus sensor-level alpha power (10 Hz, −1,250 to 0 ms). C, Top, Source-localized alpha coherence between shape-specific and pre-stimulus raw MEG signals (−1,250 to 0 ms). C, Bottom, Source-localized raw alpha power (10 Hz with 2 Hz spectral smoothing, −1,250 to 0 ms).