Construct Validity of the Circumplex Model of Marital and Family Systems: Results of a Preliminary Study and Some Methodological Suggestions

Abstract

　This paper reports the results of a preliminary experiment on the construct validity of the Circumplex model. This results were analyzed according to a Multitraits-Multimethods matrix. Two traits under examination are Adaptability and Cohesion. Two methods employed in this study are a paper and pencil test (Family Adaptability and Cohesion Evaluation Scale at Kwansei Gakuin of FACESKG) and a judge rating (Clinical Rating Scale or CRS). Ten clinic families responded to FACESKG. Their video-taped enactment interactions were observed and rated by independent judges using CRS.

　Convergent validation examines measures of the same trait assessed by different methods and is expected to be high. Cohesion demonstrated a moderate convergent validation (r=0.56) while adaptability showed a relatively poor result (r=0.19). Discriminant validation which is expected to be low, examines measures of different traits assessed by either the same or different methods. The results were also mixed. Very low to moderately low correlations were obtained from between-method comparisons of different traits; r=0.01 between CRS cohesion and FACESKG adaptability, and r=0.26 between CRS adaptability and FACESKG cohesion. However within-method comparisons of different traits were unexpectedly high; r=0.67 for a correlation of adaptability and cohesion within CRS, and r=0.45 for these traits within FACESKG.

　Inadequately low convergent validation for adaptability was discussed due to the difficulty of assessing this dimension whisin a family enactment situation. In order to induce both family adaptability and cohesion functions, this paper suggests more standardized, controlled and objective tasks such as the Simulated Family Activity Measurement (SIMFAM) for further study. Meanwhile, unexpectedly high correlation (i. e., poor discriminant varidation) of adaptability and cohesion measures within the same method, FACESKG or CRS, was attributed to a distorted sample distribution, the result of a small sample size. This paper suggests at least thirty or more families be involved in further experiments.

　Finally, it was argued that the current version of FACESKG mainly focusses on the concept of "boundary" for cohesion and that of "rule" and "role" for adaptability. Any other cohesiomn or adaptability variables are inadequately tapped by FACESKG. Further item construction for overlooked variables as well as item selection of overrepresented items were suggested for the next version of FACESKG.