Improving the timeline for intervention in Alzheimer's disease (AD) has considerable potential to delay and mitigate disability and suffering. Neuropsychological assessment is useful for distinguishing AD from normal aging and other dementias but is less useful in preclinical detection due to its limited sensitivity. The N400 (N4), a language-based EEG event-related potential (ERP) related to semantic functioning, is a promising candidate marker of AD with potential to improve early detection and monitoring of AD. For example, studies have shown that individuals with AD show a reduced N4 "effect"—a smaller difference in the size of the N4 to semantically congruent vs. incongruent word-pairs. The goal of this study is to assess the presence of the N4 effect in healthy seniors, and those with amnestic mild cognitive impairment (MCI) or mild AD, and to evaluate associations between performance on the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) and the N4 across these samples.

Fifty older adults (intact=27, combined MCI/mild AD group=23; "impaired") completed neuropsychological testing, including the RBANS, as part of a larger study. Participants were re-contacted and returned for EEG assessment between several weeks to one year later. During EEG recording, participants completed a word-pair judgement paradigm, which involved distinguishing between semantically congruent and incongruent word-pairs. Data was collected and analyzed according to customized N4 analysis scripts provided as part of ERPCORE, an online resource for acquiring and analyzing common ERP components (Kappenman et al., 2021; https://osf.io/thsqg/). The change in N4 amplitude between congruent and incongruent trials (the N4 effect) was used as an index of participants' semantic functioning. Participants' N4 effect was quantified using the mean amplitude from 300-550 milliseconds poststimulus at electrode Cz.

Repeated measures ANOVAs indicated a significant effect of trial type on the N400 amplitude in the intact individuals (F(1, 26)=77.66, p<.001), which remained significant in the sample as a whole (F(1, 48)=65.18, p<.001). Although intact participants numerically showed a larger N4 effect (intact: M=-4.02, SD=2.37; impaired: M=-2.60, SD=3.40), the expected group-by-trial interaction was not significant (F(1, 48)=3.01, p=.089). Correlational analyses revealed no significant associations between the N4 effect and the RBANS Total Scale scores (r=-.14, p=.32), nor for the Immediate Memory (r=-.002, p=.99), Visuospatial/Constructional (r=-.069, p=.63), Language (r=-.15, p=.30) Attention (r=-.21, p=.14), or Delayed Memory (r=-.18 p=.58) indexes.

Results confirmed the presence of the N4 effect in intact participants and in the sample as a whole. Although the N4 effect was numerically smaller in the impaired group as expected, this difference was not significant in the present sample. Likewise, we observed no evidence for associations between the size of N4 effect and performance on RBANS indexes. Overall, the present study provides mixed evidence for the utility of the N4 as a biomarker in mild AD. Factors that may have contributed to the lack of associations between the N4 effect and the RBANS include the limited sample size and variable lengths of time between participants' initial cognitive assessments and EEG testing.