There are two major problems with the study of Gudjonsson et al (Reference Gudjonsson, Rabe-Hesketh and Szmukler2004) which render interpretation of the results problematic. First, violent incidents were identified from untoward incident reporting forms. Formally evaluated scales measuring violent behaviour were not used. It is well recognised that nursing staff may underreport violent behaviour on incident forms (Reference Shah, Fineberg and JamesShah et al, 1991). Moreover, there is no mention of incident forms other than those for untoward incidents. In some hospitals there are different types of incident forms (including that for untoward incidents) depending how the incident is classified. Furthermore, no data are provided on the exact number of staff completing these forms and reliability between different raters in the reporting of violent incidents. The second concern is with the definition of ethnicity. The authors provide no information on how ethnicity was defined. It was simply ascertained from the record of ethnicity on the untoward incident form. There are many problems with the definition of ethnicity. Unless ethnicity is clearly defined and all those completing the incident forms use the same definition, this is likely to introduce bias in the findings. Again, no data are provided on how many staff completed the incident forms and the reliability in the reporting of ethnicity between different staff members. These issues are important because findings on psychiatric issues and ethnicity are often considered to be controversial and emotive to all sectors of society.