Brain-based markers of psychopathology reflect risk factors for future mental illness or indicators of current disease states. One solution to differentiating trait-like risk factors from indicators of disease states is trait-state-occasion (TSO) modeling, a novel structural equation model that uses repeated observations to parse variance due to stable factors (i.e., trait) from that due to momentary changes (i.e., state). To date, TSO models have largely been applied to self-report data, with only a handful of studies applying TSO models to psychophysiological markers. Importantly, these psychophysiological studies have only applied TSO models to resting-state activity, making this the first study to model psychophysiological responses to stimuli in this way. This study conducted a "proof-of-concept" to examine trait- and state-variance in event-related potential (ERP) responses (specifically, startle-elicited N1 and P3 ERPs) to unpredictable threat in 83 adults across three time-points. TSO models were applied for the following condition contrasts: unpredictable shock>
no shock and unpredictable shock>
predictable shock. TSO models fit well for the N1 and P3 for both condition contrasts. In comparison to responses to no shock and predictable shock, respectively, the N1 and P3 to unpredictable threat showed substantial trait variance (N1 = 66 % & 84 %, P3 = 69 % & 71 %), less state residual variance (N1 = 32 % & 15 %, P3 = 28 % & 25 %) variance, and little autoregressive variance (N1 = 3 % & 2 %, P3 = 4 % & 6 %). Longitudinal modeling of task-based brain data can elucidate novel findings regarding the relative contribution of trait-/state-factors of biomarkers reflecting responses to stimuli.