A probability sample of teachers (n=480) responded to vignettes in which case characteristics were systematically manipulated. Case characteristics alone accounted for 50.30 percent of the variance in recognition and 51.08 percent of the variance in reporting; the strongest effects were from type and seriousness of abuse, positive behavior of the victim and positive psychology of the perpetrator. The inclusion of variables describing the teachers and the school explained only a very small additional proportion of the variance in teachers' responses. Teachers' responses to child abuse are relatively unbiased by either the extraneous characteristics of the perpetrator or victim, the responding teacher or the school setting. Findings of the study do not appear to support the problem of "overreporting." There is evidence for "underreporting," particularly in less serious cases involving physical and emotional abuse. Teachers are undeterred by the many problems and fears that may accompany a report of child abuse to Child Protective Services, and use discretion in reporting abuse they recognize. Table, references, appendixes