A measurement scale should be short and quick to complete if it is to be practically useful. Drawing on data from a community-based survey of 2,178 people in Hong Kong, we compared five short forms (5- to 10-item) and the original version (20-item) of the Center for Epidemiologic Studies-Depression Scale (CES-D; Radloff, 1977) in predicting suicidal attempts and suicidal thoughts. Short forms with as few as nine items performed in ways very similar to the full version; a version with only five items had a detectable difference from the full version. Sensitivity, specificity, and predictive values in differentiating people with and without suicidal thought or attempt change almost linearly with the cut-offs.