Reply: Replicability and impact of statistics in the detection of neural responses of consciousness

Sir,The investigation of preserved neural functions in comatose patients and their link to long-term outcome is largely based on the analysis of electrophysiological measurements at the scalp (Morlet and Fischer, 2014; Juan et al. , 2015). Because the analysis and interpretation of these data can influence treatments and the patients’ recovery, the introduction of quantitative methods for the analysis of EEG has been increasingly recognized as a major advancement for a systematic use of evoked activity measurements in the clinical domain (Lodder and van Putten, 2013; Noirhomme et al. , 2014, 2015; Rossetti et al. , 2014; Hermans et al. , 2016). In this context, the letter by Gabriel et al. , (2016) raises an important point of discussion about the reliability and replicability of the methods for measuring the neural correlates of violation detection in mismatch negativity paradigms (Garrido et al. , 2009). In particular the letter highlights the inconsistency between different approaches in detecting such neural correlates at the single subject level using the same dataset. These results question the reliability of previous studies in detecting differential brain activity in response to different sensory stimuli and stimulate a debate around the ‘best’ analysis method for EEG recordings, particularly in a clinical setting. In our reply we would like to contribute to this debate by drawing attention to the possible sources of such discrepancies and how they could be taken into account for allowing a fair comparison. First, these various analyses stem from different hypotheses about the source of the neural response to sensory stimuli. These hypotheses lead to different choices of the EEG features undergoing the statistical analysis and may impact the final results. Within the methods selected in the letter by Gabriel and co-authors, the vast majority rely on the analysis …