How robustly do multivariate EEG patterns track individual-subject lexico-semantic processing of visual stimuli?

Abstract

Electroencephalography may be a valuable tool for assessing lexico-semantic processing in conditions where behavioural measures are unreliable. Detecting and quantifying effects in individuals is crucial for clinical applications, but individual-subject analyses are frequently not reported, and are hampered by low signal-to-noise. Multivariate analyses (MVPA) may be more sensitive than traditional approaches, so we asked how robustly they could detect differential neural responses to semantically matched and mismatched word/picture pairs in individuals. With clinical application in mind, we compared data from a research-grade EEG system to concurrently recorded data from the wireless Emotiv EPOC +. In both EEG systems, despite robust group-level effects, we only detected statistically significant processing of lexico-semantic condition in 50% of individuals. Surprisingly, detection rates were similar for MVPA and univariate analyses. MVPA may be advantageous when individual responses are heterogeneous, but in this simple paradigm, lexico-semantic processing could not be reliably detected at the individual level.