Data were drawn from 262 official police reports concerning domestic disturbances. The propensity of police to make arrests is examined through a variety of exogenous variables, such as whether the disturbance occurred between legally married spouses, the race of the male partner, whether the female called the police, whether the incident occurred on a Saturday night, whether both principals were present with police, whether only the female alleged violence, whether the male had been drinking, and whether there were signs of property damage or injury. These recorded incident characteristics were subjected to multivariate analysis, which revealed four variables to be significant to police arrest decisions. Significant positive influence was exerted by the female victim's refusal on agreement to sign a citizen's arrest warrant. When both principals were present at the point of police intervention, the female's allegation of violence and drinking by the male also increased the likelihood of arrest. The probability of arrest decreased if the female had made the initial call for help. These results indicate that when the police arrive at the scene of a domestic disturbance, they have to construct a 'story' of what happened from the immediate characteristics of the encounter. Officers' interpretation of these events, their assessment of them, and their decision to act upon them is a process more complex than previously thought, involving legal considerations, personal or occupational attitudes and experience, and situation-specific rationales. In this context, discretionary powers are not an abuse but part of the normal exercise of duty. Tabular data, footnotes, and over 50 references are given.