Using trial‐level data and multilevel modeling to investigate within‐task change in event‐related potentials

Abstract

EEG data, and specifically the ERP, provide psychologists with the power to examine quickly occurring cognitive processes at the native temporal resolution at which they occur. Despite the advantages conferred by ERPs to examine processes at different points in time, ERP researchers commonly ignore the trial-to-trial temporal dimension by collapsing across trials of similar types (i.e., the signal averaging approach) because of constraints imposed by repeated measures ANOVA. Here, we present the advantages of using multilevel modeling (MLM) to examine trial-level data to investigate change in neurocognitive processes across the course of an experiment. Two examples are presented to illustrate the usefulness of this technique. The first demonstrates decreasing differentiation in N170 amplitude to faces of different races across the course of a race categorization task. The second demonstrates attenuation of the ERN as participants commit more errors within a task designed to measure implicit racial bias. Although the examples presented here are within the realm of social psychology, the use of MLM to analyze trial-level EEG data has the potential to contribute to a number of different theoretical domains within psychology.

Relation of depression symptoms to sustained reward and loss sensitivity

“…All MLM analyses were carried out using R packages lme4 (Bates, Mächler, Bolker, & Walker, 2014), lmerTest (Kuznetsova, Brockhoff, & Christensen, 2017), and jtools (Long, 2018). We used R model specification procedures similar to Volpert-Esmond et al (2017) and referenced the code that these authors made available online (https://github. com/hiv8r3/MLM-ERP/) to assist with our analyses.…”

“…We selected these electrodes because they correspond to the frontal and central sites on the 64‐channel cap and encompass the region where the FN has previously been shown to be maximal (Foti et al, ). We selected multiple frontocentral sites (i.e., not only FCz and Cz) to account for individual‐level heterogeneity in RewP/FN responding and improve power to model variance between electrode sites, consistent with prior studies (Volpert‐Esmond et al, ). In contrast to previous studies that have analyzed the RewP/FN as an average difference score between reward and loss trials (e.g., Foti & Hajcak, ), the current study calculated the RewP and FN trial by trial for each participant individually, consistent with other time‐based ERP analyses (Von Gunten, Volpert‐Esmond, & Bartholow, ; Volpert‐Esmond et al, ).…”

Moderators of the internal consistency of error‐related negativity scores: A meta‐analysis of internal consistency estimates

“…When internal consistency was estimated using the first subset of error trials (α = .71; sensitivity analysis: α = .70), internal consistency was higher than when it was estimated using a random subset of all error trials (α = .63; sensitivity analysis: α = .62). In a study that used multilevel modeling to look at the relationship between error trials across time and ERN amplitude, ERN decreased as participants committed more errors within a task (Volpert-Esmond, Merkle, Levsen, Ito, & Bartholow, 2017). It seems likely that ERN from the beginning error trials would be more similar in amplitude than ERN trials randomly sampled from the entire task.…”

“…It is possible that once all error trials are included in analysis that internal consistency estimates for any single study would be higher. However, this assumption only holds if ERN item covariance is constant across the entire task (Cronbach, Gleser, Nanda, & Rajaratnum, 1972), which does not appear to be the case (Volpert-Esmond et al, 2017). Second, internal consistency estimates from large samples are weighted more heavily than those from small samples in the meta-analysis.…”

Data Quality and Reliability Metrics for Event-Related Potentials (ERPs): The Utility of Subject-Level Reliability

“…However, changes in cognitive or affective state over the course of a task can impact ERP signals, and such within-person changes might be of interest. Some recent RewP and ERN studies used mixed-effects models to explain changes in trial-to-trial scores (e.g., Brush, Ehmann, Hajcak, Selby, & Alderman, 2018;Volpert-Esmond, Merkle, Levsen, Ito, & Bartholow, 2017), and such mixedeffects models have the statistical capability to separate within-person variance into true-score variance and measurement error (e.g., within-person changes in ERP amplitudes across the course of a task). These models can examine factors that impact between-trial variance within and across participants.…”