Towards an individualised neural assessment of receptive language in children

Abstract

Keywords N400, EEG, language comprehension, sensitivity, Emotiv EPOC+, multivariate pattern analyses Highlights -New methods show hints that brain data can be used to infer language processing -A gaming headset records EEG waveforms comparable to a research grade system -50% of individual children show significant N400 effects in two paradigms -We can decode semantic context in up to 88% of children using multivariate analyses Abstract:Can we use electrophysiological data to assess language processing? Although we usually cannot infer cognition from brain data, there are some circumstances in which discriminative brain signals are enough to tell us that cognitive processing has occurred. For example, if we observe differential neural responses to identical stimuli that vary only in cognitive context, we can infer an influence of context on stimulus processing. Here, we used this logic to develop a test for receptive language ability in individual children. We were motivated by the suggestion that some non-verbal individuals, such as a subset of autistic children, may understand more language than they can demonstrate; for these people a neural test would be transformative. We developed two child-friendly paradigms in which typically developing children listened to identical spoken words in congruent and incongruent lexicalsemantic contexts. In the congruent condition, the target word was strongly predicted by the context, while in the incongruent condition, the target word violated lexical-semantic predictions. In both paradigms, we simultaneously measured EEG with a research-grade system, Neuroscan's SynAmps2, and an adapted gaming system, Emotiv's EPOC+. In both paradigms, at a group level, we found a statistically significant N400 effect to lexical-semantic violations, and we also detected significant effects in about half of our participants individually.The same effects were present, although with a numerically smaller amplitude, with EPOC+.Using multivariate analyses on individual children's EEG data increased our sensitivity to detect evidence of lexical-semantic processing, reaching an 88% detection rate in a paradigm using sentences. This provides a promising avenue to assess language comprehension in individuals, which could be used to identify language comprehension abilities in children who may otherwise struggle to demonstrate how much they understand.

Unconstrained multivariate EEG decoding can help detect lexical-semantic processing in individual children

“…Additionally, studies that do report individual-subject data are often limited by methodological or conceptual factors, such as a lack of proper statistical analyses, or lack of a control group. Recently, two studies have looked at the individual sensitivity of auditory N400 paradigms, in healthy adults (Cruse et al, 2014), and in healthy children (Petit et al, 2019). These studies each used two paradigms contrasting auditory words presented in congruent or incongruent semantic contexts.…”

“…These studies each used two paradigms contrasting auditory words presented in congruent or incongruent semantic contexts. Using a task in which participants were asked to make covert violation decisions, these studies reported a statistically significant N400 effect in a maximum of 58% of neurotypical adults (Cruse et al, 2014) and 56% of neurotypical children (Petit et al, 2019). This moderate sensitivity reflects the challenges associated with assessing ERPs at the individual level.…”

“…This moderate sensitivity reflects the challenges associated with assessing ERPs at the individual level. However, using multivariate pattern analyses (MVPA) on these EEG data, we recently increased the individual-subject detection rate to 88% (Petit et al, 2019). In the current study, we aimed to build on existing N400 paradigms by improving the stimulus delivery paradigm and by using a more robust and sensitive analysis framework.…”

“…We set out to extend the paradigm developed in Petit et al (2019), and to measure neural differential responses to auditory words presented in congruent and incongruent sentences frames (i.e., "The squirrel stored nuts in the tree/door"). In order to increase children's engagement to the task, and build up strong semantic contexts, the spoken sentences were accompanied by short animated cartoons (e.g., an animation of a squirrel storing nuts in a tree).…”

“…With traditional within-individual ERP analyses, it is necessary to choose a priori time windows, and electrodes, of interest, in order to reduce the number of comparisons and thus increase the statistical power. However, we have previously demonstrated that for individual subject analysis, a priori assumptions about spatio-temporal location should be avoided as there is substantial inter-individual variability in the location and timing of N400-like effects (Petit et al, 2019). For clinical populations, for whom inter-individual variability may be even higher, it is essential to allow some inter-individual variability in the location and timing of the effect of interest.…”