The aim was to investigate consensus between speech and language therapists (SLT) during the initial assessment of preschool children. A multimethod design was used. The qualitative component provided SLT's own views of the process as the starting point for the study. Data were generated from semistructured interviews and focus group discussions. The interviews investigated factors considered during an initial assessment of a preschool child. The focus groups discussed the process and content of a videotaped initial assessment carried out by the author. Quantitative data was generated from six case analyses and from a card sorting exercise. In the former, SLT were asked to record important features of newly referred cases. In the latter, SLT sorted statements about preschool children according to a simple concern/not concerned decision. Data collection tasks and analysis took place successively to build a composite and triangulated picture. The main subjects were 11 SLT, experienced in childrens' work with a further six acting as a broader reference group. The results show an interacting pattern of consensus and variation. SLT agreed about the relative priority of children, categories that informed the assessment; they made similar interpretations of incoming information. There was variation in their preferred assessment methods, in the thresholds of concern and in the interventions offered. Sources of variation that were identified included individual confidence in and access to assessments, personal levels of concern and in the approach to the SLT's role, as well as the influence of workplace policies. The study concluded that, consensus and variation both exist within the context of SLT's initial assessment of preschool children, but that there are discernible patterns to the variation. While such variation may be inevitable, it is suggested that any variation should be made explicit and used to challenge and develop the shared knowledge base of the speech and language therapy profession.