Directly measuring emotional state using latency difference of millisecond responses to depressive (D) vs. non-depressive (ND) words can overcome self-report limitations. Cross-sectional (n=682) data show two factors (D=15 items σ²≈11%; ND=12 items σ²≈11%), sensitivity .73, specificity .91, predictive and concurrent validity with PHQ-9/CES-D (p<.001). A longitudinal sample (n=97): replicates and extends validity; inversely tracks wellbeing; has good reliability (test-retest=.6; Cronbach's α≈.7).
