Achieving reliability in observational research: lessons from a study of newborn temperament

Observational studies of human responses frequently yield only tenuous or no support for hypothesized relationships. These null findings often may be due to reliance on single observations rather than on multiple measures aggregated over contexts or occasions. This article discusses the principle of aggregation and describes the development and evaluation of aggregated measures in a study of the relationship of nursery nurse's behavior to emerging newborn temperament.