Coordinates of gamma-band activity
Study Space Frequency Time Condition --------------------- ------ ---------- --------- ------------------------------------------------------------------------------- Hagoort et al 2004 Fz 35-45 300-800 World Knowledge Violation, but not Control (50% cloze) or Semantic Violation Hald et al 2006 F4? 35-45 0-600 Control (50% cloze), but not Semantic Violation Penolazzi etal 2009 ? 30-100 0-512? Control > Semantic Violation Urrutia et al 2012 F3 80-90 300-800 Factual, but not counterfactual Rommers et al 2013 CPz? 55-70 200-1000 Control, but not (unrelated) Semantic Violations Wang et al 2012 Pz 40-50 0-1200 Control (68% cloze), but not Control (? < 68%) or Semantic Violation
Hagoort et al. 2004 and Hald et al. 2006 may be the same experiment, with the exclusion of a condition by Hald et al. 2006. Bastiaansen et al. 2012 write that the stimuli in this experiment, not differentiating between Hagoort et al 2004 and Hald et al 2006, lead to an increase in the gamma band for Control, but not Semantic Violation. However, Hagoort et al. 2004 explicitly state that no difference was observed in Control.
Possible methodological differences - larger Baseline (350 msec average for Bastiaansen, over 1 sec for us), which if at all provides a more robust estimate - Low-pass filter (we used no high-pass filter and a 125 hz sampling rate; though note that Hagoort et al. 2004 and Hald et al. 2006 apply a 70 hz low pass! Wang et al. and Sommers et al. 2012 use 100 and 125 hz respectively) - ICA artifact correction (we rejected non-dipolar ICs, which is assumed to attenuate muscle artifacts) - We use Wavelet filtering, some studies used multi-taper/FFT - We do not apply a cluster technique to control type I errors, but rather derive hypotheses from independent data sets - Our protocol features substantially higher power (due to lack of multiple comparison control and higher n, < 200 vs. 24+30+24) - We do not believe methodological differences are substantial
Possible differences in material - our sentences span the whole range of observed gamma band differences, including multiple experiments including high- and low-cloze contexts and violations in high- and low-cloze contexts. Every experiment features at least a congruent control and an incongruent comparison <<<<<<< HEAD:übersicht.txt
Cluster Correction may be inappropriate for cortical gamma-band activity in general since gamma is known to be synchronized only very focally {vonStein:2000wf}{DuunHenriksen:2013gn}, and Cluster Correction is highly sensitive to distributed effects, while being insensitive to focal effects (Groppe et al., 2011)
>>>>>>> 9b8cefe3234a6bd6213b1190ae8ea17c11613f1a:übersicht.rst