As other neurophysiological studies, ours has some methodological shortcomings. For instance, the investigators were blinded during off-line analyses of VEP data, as applied in previous studies by independent groups [15, 73], but not to diagnosis during the recording session, although this is probably only a minor risk for bias. As a matter of fact, in a clinical setting it is quasi impossible to totally blind a neurophysiological study. Even in VEP studies that found no abnormalities in migraineurs and were claimed to be blinded to diagnosis [12, 74, 75], blinding was not perfect for various reasons according to the reported methodology: 1) the neurophysiologist knew which set of responses belonged to each of the 6 blocks of averagings [74], which allows a selection bias in favour of low amplitude responses and thus normal habituation [76]; 2) the neurophysiologist was not blinded to check size [74], to which VEP amplitudes are quite sensitive [77]; 3) after each recording, the investigators guessed the correct diagnosis in more than half of subjects [12]; 4) although the investigators were blinded to diagnosis during the first examination, they were not during the 3 subsequent recording sessions [75].