Using multidimensional scaling for the analysis of sexual offence behaviour: A replication and some cautionary notes

Abstract

Crime scene profiling has received considerable attention, particularly in popular media. First, since low inter-rater reliability would affect the validity of profiling, we tested inter-rater agreement for 33 variables often used in manuals for the coding of crime-related behaviour. Thirty cases were chosen randomly from 146 cases of assault rape with unknown male offenders in Sweden during 1990–1994. Second, we used multidimensional scaling (MDS) with all 146 cases to replicate earlier work by Canter and Heritage (Journal of Forensic Psychiatry, 1, 185–212, 1990) regarding motivational dimensions in unknown assault rapists. On average, inter-rater reliability was good; physical behaviours were more reliably coded than were verbal behaviours. However, a two-dimensional MDS solution with the motivational dimensions suggested by Canter and Heritage was not replicated. We argue for better empirical support for the inclusion of specific variables in coding formats for crime scene information. This is particularly important for analyses like the MDS, since the number, representativity, and intercorrelations of such variables could affect results considerably.