1.0 Introduction
For over seven decades, electroencephalography (EEG) has been used to
study brain activity elicited by the presentation of sensory stimuli, or
during performance of motor or cognitive tasks (Kappenman & Luck,
2016). A cornerstone of the general methodology used to obtain measures
of brain activity has been the processing step of averaging multiple
segments of EEG signals elicited by the same stimulus to create averaged
event-related potentials (ERPs). Then, the resulting peaks and valleys
of the averaged ERP are quantified for a specified time window as
either: 1) the average of the voltage values of the data points in the
window or 2) as the amplitude and latency of the largest peak found in
the time window.
Justification for the processing step of averaging multiple segments is
based on three assumptions. Firstly, EEG obtained at any scalp site is a
continuous mixture of electrical activity from multiple brain regions,
where some of the activity is not related to the stimulus. This
unrelated activity is often referred to as background (brain) noise.
Secondly, the brain’s electrical signal elicited by any stimulus (or
response) event is difficult to measure as it is assumed to be smaller
than, and mixed with, the continuous background noise. Thirdly, the
brain’s signal in response to an event is invariant in amplitude and
time for each occurrence of the event (Dawson, 1954; Hu et al., 2011;
Quian Quiroga & Garcia, 2003).
While the validity of the second and third assumptions have received
considerable discussion over the history of ERP-based research, some
researchers propose that the best approach to address these assumptions
is using the average voltage in a window measurement (e.g., De Vos,
Thorne, Yovel, & Debener, 2012; Jongsma et al., 2006; Kappenman &
Luck, 2016; Kosciessa, Grandy, Garrett, & Werkle-Bergner, 2020; Luck,
Stewart, Simmons, Rhemtulla., 2021). Yet even this measurement approach
relies on the third assumption, signal invariance, being true or at
least having minimal variations in the latencies of the various positive
and negative peaks across segments being averaged. While latency
variability is often minimal for adults, research in children has shown
that greater latency variability of peaks leads to decreased amplitude
measures, which in turn leads to erroneous conclusions about brain
development (e.g., DuPuis et al., 2015; Gavin, Lin, & Davies, 2019).
The present study investigates
the feasibility of using a simple procedure, the Single Trial Peak
approach (STP), for obtaining reliable and valid peak amplitude and
latency measures at the trial (i.e., segment) level. This procedure does
not depend on the validity of the second and third assumptions
traditionally presumed in research using averaged ERPs; that is, it does
not require a signal-to-noise ratio (SNR) of any value, nor does it
assume signal invariance. In contrast, the STP procedure does assume,
allows for, and quantifies not only the variability between participants
representing individual differences in processing abilities but also the
variability across trials within a paradigm/task completed by the
participant. Consequently, the STP procedure enables researchers to test
the hypothesis that some of the variability in amplitude and latency of
peaks across trials may be systematic rather than random. More
importantly, the STP procedure can produce measures of brain activity at
the single trial level with psychometric properties comparable to or
exceeding the commonly used measure of average voltage in defined time
windows. The rationale for proposing the STP procedure is presented in
the following sections.