Using EEG to detect lapses in sustained attention to moving stimuli.

Sustaining attention is effortful but crucial for daily life. Despite this, attentional lapses are common and can have fatal consequences (e.g., when driving). The spontaneous nature of these lapses makes studying their underlying phenomena elusive. As such, methods capable of determining when lapses have occurred may be fruitful research tools, with the potential to save lives if implemented within real-world settings. Here, we capitalised on a recent hierarchical classification method, which uses multivariate decoding to index how well human observers sustain their attention within a dynamic visual environment. We asked whether this method could be used to anticipate behavioural errors based on neural activity measured with electroencephalography (EEG, N = 25). We first decoded how long until a stimulus would reach a task-critical point within a Multiple Object Monitoring (MOM) task from multivariate patterns of EEG amplitudes. The extent to which we could decode this information depended on whether the stimulus was relevant for behaviour, and was lower before participants failed to detect (or 'missed') target stimuli relative to hits, presumably due to attentional lapses. We then exploited this drop in neural decodability to predict whether errors were about to occur on each trial. The results form a foundation for sensitive and specific methods to objectively detect errors in sustained attention tasks based on patterns of brain activity.