For many, the course of major depressive disorder (MDD) is recurrent, with periods of remission between major depressive episodes (MDEs); those in remission are known to be at elevated risk of future MDEs. A common and distressing symptom of MDD is overgeneralised self-blame, and this also persists into remission. In order to study the involvement of self-blame in vulnerability to MDD, a large cohort of participants was recruited: a group with remitted MDD (rMDD) and a matched healthy control (HC) group with no personal or family history of MDD. Participants completed electrophysiological and neuropsychological tasks. The rMDD group also completed a 14-month follow-up period, during which symptoms were monitored at intervals; this was to study the predictive effects of electrophysiological and neuropsychological variables, with a view to development of a biomarker with predictive value. The main method was electroencephalography (EEG), chosen for its high temporal resolution in comparison to a commonly used technique, functional magnetic resonance imaging (fMRI). On a practical level, EEG is also more cost effective and widely available, making it more suitable for future clinical transfer of any biomarker developed. A task previously used in fMRI was adapted for EEG; in this task, short sentences designed to evoke negative feelings related to the self and others were presented. The theta signal was abnormally sustained over time during self-blame in the rMDD group relative to the HC group. Given the involvement of theta in temporal binding, this may represent a correlate of dysfunction within the neural network underpinning self-blaming emotions. Correlation of sustained theta with separately collected fMRI data indicated the dorsolateral prefrontal cortex (dlPFC) was involved in this network. In a source analysis of the EEG data, the dlPFC was identified again; it showed reduced activation in the rMDD group relative to the HC group during other-blame. In summary, activation of the dlPFC appears to be adaptive in both self- and other-blame, as the HC group showed higher activation than the rMDD group; further work is required to confirm the clinical relevance of this. For a separate study of memory overgeneralisation, a known feature of MDD, a novel associative memory task was designed. A loss of bias towards remembering positive memories was found in a subgroup of the rMDD cohort with early life stress (ELS). This reduced positive bias correlated with the number of past MDEs, indicating that the cumulative effect of MDEs reactivating early traumatic memories leads to selective loss of positive memory bias. In summary, although no electrophysiological or neurocognitive predictive markers of recurrence risk were found, clear effects were seen in the cross-sectional results. Importantly, EEG was also validated as a technique for detecting self-blame-selective neural correlates of depression vulnerability. There were clear effects in the temporal domain, which highlight the benefits of EEG above other imaging techniques. However, the sources identified did not correlate with parallel fMRI work, so further work is required to understand the temporal dynamics of these sources. This research provides a platform from which future EEG investigations can develop.

Previous studies support that the relative width of the upper face (facial width-to-height ratio; fWHR) has evolved to signal threat, but these studies rely greatly on subjective facial ratings and measurements prone to confounds. The present study objectively quantifies threat perception to the magnitude of the observers’ electrophysiological reaction, specifically the event-related potential (ERP) called the late positive potential (LPP), and investigate if brow height and jaw width could have confounded previous fWHR studies. Swedish and international students (N = 30, females = 11, Mage = 24 years, SDage = 2.9) were shown computer-generated neutral faces created with the underlying skeletal morphology varying in brow ridge height, cheekbone width and jaw width. Participants first rated how threatening each face was and then viewed 12 blocks of 64 faces while their electroencephalography (EEG) was recorded. The results supported that the LPP could be used to index threat perception and showed that only brow height significantly affected both facial ratings (p < .001, ɳp2 = .698) and magnitude of the LPP within the 400 to 800 ms latency (p = .02, d = .542). Hence, brow height, not facial width, could explain previous findings. The results contradict the hypothesis that fWHR is an evolved cue of threat and instead support the overgeneralisation hypothesis in that faces with similar features to anger will be perceived as more threatening.