One such issue is the presence of non-sinusoidal sawtooth-like oscillations in electrophysiological data, which can result in spurious PAC (Lozano-Soldevilla et al., 2016), especially when phase is obtained with wide band-pass filters. By computing the ratio between rise-time and decay-time of alpha oscillations within area V1, we showed that non-sinusoidal oscillations did not differ between baseline and grating periods, and are unlikely to account for our results. Another issue in trial-based PAC analysis is data length, with some previous reports suggesting that 10 s or more is required for detecting theta-gamma coupling (Dvorak and Fenton, 2014; Aru et al., 2015). However, using simulated alpha-gamma PAC we determined that 1 s of data was sufficient to obtain stable estimates. We encourage the reader to run similar follow-up analyses after finding significant PAC to check for spurious coupling caused by, for example, non-sinusoidal oscillations (Jensen et al., 2016; Lozano-Soldevilla et al., 2016) and/or insufficiently long trials (Dvorak and Fenton, 2014).
