This study developed and evaluated a puppet interview that allows children to self-report on temperamental constructs. Structured child self-report measures are rarely utilised in clinical assessment of young children under the age of 7—8 years. Given that clinical assessment is often characterised by low convergence between raters, such a measure may offer important contributions. The present study developed and evaluated a measure based on items from the Colorado Childhood Temperament Inventory and reports two studies with child participants aged 4 to 5 years. Independent observations of the children were also made. Results showed moderate levels of internal consistency and stability, and convergence between child self-report and teacher/parent raters was low, but similar levels of agreement were achieved between adult informants. The puppet interview thus showed some potential but highlighted the difficulties of self-report in young children within a multiple informant framework in clinical assessment.