In the present empirical study, subjects without legal training were presented with a set of homicide cases and asked to assign them to offence categories, as well as to rate the seriousness of each case and their confidence in their assignments. Prior criminal records of offenders and information about penalties were also manipulated. Results showed good agreement among subjects overall, but there were differences among the categories. There was most agreement and greatest confidence in the assignment of first degree murder cases while agreement in second degree cases was much lower. Information about prior records or penalties had very little effect on assignments. The pattern of the data suggests that subjects are able to reliably identify first degree murder cases on the basis of specific features, and that manslaughter cases were discriminated because of their lesser seriousness, but that neither of these criteria applies well for second degree murder cases. This model may aid in attempts to improve the definitions of the homicide offence categories. (Author abstract)