The model was based on a review of official statistics and use of controlled experimental methods. Most data were gathered in Great Britain. Results indicated that witness intervention is most likely in public places or in visible private places where the number of potential witnesses is likely to be larger, where the witnesses are isolated on-lookers rather than active participants, and where the criminal action is clearly recognizable. Witnesses are less likely to become involved in offenses taking place in private or where the criminal action is ambiguous. The probability for witness intervention is also determined by witnesses' evaluations of the seriousness of the offense. Witnesses are most likely to intervene in violent, breaking and entering, or vandalism incidents and least likely to intervene in assaults classified as domestic disputes. Witnesses may feel that the responsibility for intervening in criminal activity has been removed from their shoulders and assigned to the State. Witnesses are faced with four choices of action when confronted with an offense in progress: (1) inform the police or the victim that a crime is in progress, (2) take action to frighten off the offender, (3) catch the offender or provide information leading to the offender's identification or (4) do nothing. One of these alternatives is selected based on perceptions of potential cost to the witness. A figure, 51 references, and footnotes are provided.