Very little is known about the early stages of self-injurious behaviour (SIB) in young children with developmental disabilities, even though there has been a great deal of research into the prevalence, assessment and treatment of well-established SIB in older individuals. In the present initial study, teachers in special schools for children under II years of age with severe intellectual disability and/or autism were asked to identify children who were beginning to show early self-injury (the index group). These children were then matched to classroom controls (of the same ability level and mobility), and teachers were interviewed about the children's behaviours and skills. The index children showed significantly more potential SIB than the control group children, but there was overlap between the groups in terms of percentage duration of potential SIB, suggesting that teachers do not find it easy to identify children with 'early' SIB. The index children's skills and problem behaviours, their sensory impairments and degree of autism did not differ significantly from those of the control group. When all the children showing any potential SIB were pooled together, it transpired that developmental age and degree of mobility were significantly correlated with percentage duration of SIB, suggesting that these characteristics may be important risk markers. The index children were also observed at 3-month intervals at school over the following 18 months and self-injury clearly escalated for some of the index children, while it did not do so for others. Using regression analysis, increases in SIB were shown to be associated only with the degree of concern expressed about the child's behaviour at time I by the teacher, no other variables predicting increases in SIB.