ABSTRACT
Naturalistic assessments are concerned with noting a piece of ‘real life’ behaviour, in an unstructured and naturally occurring context, and coding it against a predetermined analysis framework. Even the most accurate, life-like assessment of a child’s expressive language can only give data on which inferences about a child’s language ability can be made. There are difficulties inherent in linguistic and psychological measurements that must try to assess an assumed underlying language ability using examples of language performance. The problem of collecting ‘representative’ examples of child-language output has always been of concern to clinicians, and indeed to child-language research in general, and is in the end a problem of validity. As well as considering the validity of the language database obtained, assessment procedures, whether naturalistic or standardized, should consider several types of validity. One of the most important factors in making a valid assessment is that the procedures adopted should be reliable.
