Reliability of gamma activity during semantic integration
The gamma band of the human EEG has become a point of increasing focus in studies on language; however, many of these effects remain comparatively untested in their reliability. More established previous research on event-related spectral perturbations (ERSPs) has been especially focused on the theta (Bastiaansen et al 2005, Heine et al 2006), alpha (Roehm et al., 2001), beta (Davidson & Indefrey 2007), and to a lesser extent delta bands (Roehm et al., 2004). Recently, a number of studies have reported an effect for semantic integration (related to cloze probability in classical N400 paradigms) in the gamma band (>30 Hz) (Hagoort et al., 2004, Penolazzi et al., 2009, Wang et al., 2012). Given recent concerns about reliability of effects in the brain sciences (Killner 2013, Simmons et al., 2012, Vul et al., 2009), the present study aimed to test the reliability of the gamma effect in scalp EEG during semantic integration.
We performed a reanalysis of data from ten existing studies involving two languages (ca. 300 subjects in total). Because previous findings are highly heterogenous in space, frequency and time, no specific prediction could be formulated, and we thus followed a bottom-up approach. In each study, for both semantically congruent sentences and the contrast congruent-incongruent, we identified the spatial-temporal feature with the strongest ERSP in the gamma band and then tested that effect in each of the other studies, so as to maintain statistical independence. Specifically, we found that the number of comparisons (5% of independent and 50% of circular analyses) reaching significance was equivalent to the number expected due to chance. In contrast, the same procedure reliably replicated results in the delta/theta, alpha and beta bands, confirming the sensitivity of our method.
We conclude that the heterogenity of reported gamma effects requires more experimental testing before gamma should be considered a reliable marker of semantic integration.