Clinical descriptions of families are often viewed as being too subjective for systematic scientific inquiry. This study examines the extent to which independent observers can recognize statements made about family interactions. These statements were of a type which clinicians dealing with families would make and comprised observations with varying levels of inference. A special method was devised to test a number of specific hypotheses concerning the processes of clinical observation. The results show that such descriptions are recognizable as being made about a particular family even when this involves discriminating between clinically similar families. Some of the implications of these results for observational research are discussed.