IT HAS become customary for the police to treat family fights, even where physical assaults occur, as semi-private matters to be settled informally and within the context of the family. Police have been reluctant to interfere in the relationship between a husband and wife. There has also been a fear that harsh action, such as an arrest, would so provoke a violent husband that he might be more likely to retaliate later against his wife and children.

Gradually in the last 15 years, a literature of conventional wisdom has developed emphasizing this approach and encouraging training police as mediators in handling domestic disputes. Most big-city departments provide such courses to officers. Now a study released by the Police Foundation has called into question the underlying premise of this approach.

In an experiment funded by the Justice Department's National Institute of Justice, social scientists in Minneapolis monitored 252 cases of domestic conflict for a six-month period after police intervention in each case. What they learned was this: those men who were arrested were least likely to assault their wives thereafter. Conciliation at the scene or forced separation were far less successful techniques.

Researchers who conducted the study stress that it does not show that all cases of domestic disturbance should result in arrest. Where there is no physical abuse or threat of violence, conciliation is clearly the appropriate first step. But what the experiment does indicate is that police need not fear that arrest-- when called for--will necessarily result in retaliation, making the wife's position more difficult and vulnerable than before the assault. In fact, in the Minneapolis cases quite the reverse was true. Surely this is a finding that should be brought to the attention of all who train police personnel and those law enforcement officers who cope with domestic disturbances on a regular basis. The best way to handle violence in the home may be the most obvious and familiar response after all.