Cardiac arrest etiology is often assigned according to the Utstein template, which differentiates medical (formerly "presumed cardiac") from other causes. These categories are poorly defined, contain within them many clinically distinct etiologies, and are rarely based on diagnostic testing. Optimal clinical care and research require more rigorous characterization of arrest etiology. We developed a novel system to classify arrest etiology using a structured chart review of consecutive patients treated at a single center after in- or out-of-hospital cardiac arrest over four years. Two reviewers independently reviewed a random subset of 20% of cases to calculate inter-rater reliability. We used X2 and Kruskal-Wallis tests to compare baseline clinical characteristics and outcomes across etiologies. We identified 14 principal arrest etiologies, and developed objective diagnostic criteria for each. Inter-rater reliability was high (kappa = 0.80). Median age of 986 included patients was 60 years, 43% were female and 71% arrested out-of-hospital. The most common etiology was respiratory failure (148 (15%)). A minority (255 (26%)) arrested due to cardiac causes. Only nine (1%) underwent a diagnostic workup that was unrevealing of etiology. Rates of awakening and survival to hospital discharge both differed across arrest etiologies, with survival ranging from 6% to 60% (both P < 0.001), and rates of favorable outcome ranging from 0% to 40% (P < 0.001). Timing and mechanism of death (e.g. multisystem organ failure or brain death) also differed significantly across etiologies. Arrest etiology was identifiable in the majority cases via systematic chart review. "Cardiac" etiologies may be less common than previously thought. Substantial clinical heterogeneity exists across etiologies, suggesting previous classification systems may be insufficient.