Variability of Single Trial ERP Measures Within a Session is Systematic and Nonlinear

Averaging multiple event-related potential (ERP) segments distorts the brain’s response to a stimulus given the false assumptions that the ERP signal is invariant and hidden by background noise. Our Single Trial Peaks (STP) procedure measures amplitude and latency of multiple peaks in each segment based on the peak latencies of the individual’s averaged ERP. This study examined correct trial data from 70 adults performing two repetitions of a speeded visual flanker task. STP peak data (P1, N1, P2, N2, and P3) were compared to single-trial averaged voltage in time windows (STW) and peaks of averaged ERPs. Results indicated that the STP approach had higher split-half reliability (odd-even, r =.70-.96) and test-retest reliability (r =.66-.97) than the STW approach. The mean signal/noise ratios for the amplitude measures for the STP ranged from 1.62 to 2.9 and were larger than the STW approach, although averaged ERPs had higher ratios than both. Coefficients of variation (CV) for the STP and STW approaches were similar, and both had smaller CVs than averaged ERPs. The validity of the STP approach was determined by regression analyses where mean trial-level peak amplitude, SDs across trials for amplitude and latency, and noise accounted for a significant amount of variance in corresponding peaks of the averaged ERPs (R2 =.73-.95). Curve fitting analyses showed nonlinear systematic patterns in amplitude over trials similar to changes in response time. These results support the validity of the STP approach which can be used to investigate systematic changes of brain activity across trial.