[Construct validation study of the Relationship Scales Questionnaire (RSQ) on an adult sample].

INTRODUCTION: The authors present the construct validity of the French version of the Relationship Scales Questionnaire (RSQ) designed by Griffin and Bartholomew (1994), which is the most widely-used self-report concerning adult attachment.
OBJECTIVES: To improve knowledge on the psychometric properties of the RSQ by studying the construct validity on a sample of adults.
METHOD: Subjects were recruited in a primary social care setting in Paris, France, and asked to fill in the RSQ twice: the first time just before meeting the social worker in charge, the second time, at home, with a prestamped form, two days later. A questionnaire on sociodemographic and economic variables was also filled in during the first time. Statistical analyses were conducted using Exploratory Factor Analysis with orthogonal rotation. The reliability was studied using Cronbach's coefficient for each new scale from the factor analysis. The test-retest reliability was studied for the prototypic scales and for the scales from the factor analysis, using the intra-class correlation method.
RESULTS: One hundred and twenty-six subjects were recruited and completed the two forms (mean interval: 2.16 days). The factor analysis gives a three-factor structure explaining 48% of the total variance. The three factors were: factor 1 "Avoidance" with seven items explaining 21% of the total variance; factor 2 "Anxiety in the relationships" (five items) explaining 14% of total variance and factor 3 "Security" (five items) explaining 11% of the total variance. Cronbach's coefficient was low for the prototypical scales (0.41 for "secure", 0.54 for "fearful", 0.22 for "preoccupied", and moderate for "dismissive" (0.64). It was moderate for the scales designed from the factor analysis (0.66 for F1, 0.69 for F2 and 0 .60 for F3). The Intraclass Coefficients (ICC) were modest for the four prototypical scales (ICC<0.70) and were good for the scales designed from the factor analysis (F1: ICC=0.80; F2: ICC=0.85 and F3: ICC=0.78). DISCUSSION: The construct validity studied on an adult sample confirms the good psychometric properties of the RSQ considering the factor analysis, the test-retest short time reliability and the internal consistency. The factor analysis with three factors provides a different structure of classical descriptions with only two factors, but confirms the most recent results on Attachment Self-Reports that find a factor concerning security and two factors concerning management of insecurity (avoidance and anxiety in relationships). To be confirmed, the results require further research (confirmatory factor analysis, larger sample, other type of population). Copyright (c) 2009 L'Encéphale, Paris. Published by Elsevier Masson SAS. All rights reserved.