Preprocessing Decisions Affect the Precision of P50 Estimates in the Paired-Click Paradigm

Abstract

Sensory gating is commonly indexed by suppression of the auditory P50 component in the paired-click paradigm. However, P50 findings have varied considerably across studies, laboratories, and populations, prompting calls for the standardisation of methodological practice. EEG preprocessing represents one source of methodological variation that requires standardisation. To date, filtering is the only preprocessing step to have been examined empirically. Segmentation and artifact handling have varied across paired-click studies without empirical evaluation, and, to our knowledge, the standardized measurement error (SME) has not been applied to P50 paired-click estimates. The present study addressed both gaps. Fifty-six neurotypical adults completed a paired-click paradigm consisting of 120 trials. Four preprocessing pipelines were constructed by crossing Segmentation (Short vs Long) with Artifact Handling (threshold-based rejection vs rejection of non-ocular artifacts followed by Gratton and Coles ocular correction). A bootstrapped SME (bSME) was used to quantify the precision of peak-to-peak P50 amplitudes at Cz and FCz, alongside an analysis of trial retention. Correction pipelines retained more trials and produced lower bSME values (higher precision) than Rejection pipelines, at both electrodes and across S1, S2, and suppression score outcomes. Segmentation and Artifact Handling did not operate independently: Segmentation had a negligible impact under Correction, whereas Long segmentation consistently produced the highest bSME values (lowest precision) under Rejection. A regression analysis showed that trial retention accounted for the large majority of the pipeline differences in bSME. This study therefore recommends that future paired-click studies correct, rather than reject, trials containing ocular artifacts, and adopt Short segmentation where matched S1-S2 averaging is not required. Impact statementInconsistent P50 findings have long prompted calls for the standardisation of paired-click methodology, yet preprocessing decisions are rarely justified and, with the exception of filtering, have never been evaluated empirically. By quantifying the precision of P50 estimates, we show that artifact handling and segmentation affect data quality almost entirely through the trials they discard. Measurement precision, not convention or preference, should decide how P50 data are preprocessed.