The function of 121 challenging behaviours displayed by 20 people with severe intellectual disabilities was assessed using informant-based, experimental and descriptive methods. The results suggest that informant-based, descriptive and experimental approaches to assessment vary with regards to (1) the probability of ascribing functions across a range of challenging behaviours; (2) die types and characteristics of behaviours to which functions are more likely to be ascribed; (3) the types of functions which are likely to be identified; and (4) the probability with which they identify multiple functions. The results also suggest that the level of agreement between different approaches was disappointingly poor. It is suggested that, at present, the routine clinical application of functional assessment ought to be based on the construction of an individually-tailored multiple assessment format based on a pre-assessment of topography, environmental and person-related variables. Attention is also drawn to the importance of staff training to enhance the performance of informant-based methods and the further development of non-experimental approaches to functional assessment.