The link between stress and health has not received strong empirical support, possibly due to problems in the stress measures used. Here, the first wholly empirical development of a new "Stress Overload Scale" is described. A pool of 150 items was formed to reflect "overload", a common denominator in stress theories. Then, the results of five sequenced studies, conducted in heterogeneous community samples, were used to pare the item pool. Exploratory (n = 431) and confirmatory (n = 433) analyses revealed two factors (Event Load and Personal Vulnerability) corresponding to theoretical constructs; only the best factor markers were submitted to further construct validity (n = 310) and reliability tests (n = 342). The 24 strongest items were selected for the SOS, which demonstrated criterion validity in predicting who (n = 285) would become sick following a common stressor. The SOS is (1) psychometrically strong, especially in its validity relative to popular measures; (2) appropriate to community research, due to its brevity and fit to a broad demographic spectrum; (3) unique in its ability to cross-section individuals into risk categories. It should prove useful to community psychologists in determining etiology, diagnosing risk for pathology, and evaluating the efficacy of interventions.