The effects of context on the correspondence between observational ratings and questionnaire reports of hostile behavior: a multitrait, multimethod approach.

Scholars who study close relationships often measure the same concept with both observational ratings and questionnaire items. In studies where this has been done, the correlations between the two measures have often been low, and one reason may be a "context by measurement" confound. Observational ratings are usually based on behaviors during a specific task, whereas most questionnaire responses are based on a broader context such as "during the past month." In 2001, a total of 292 young couples were videotaped during a discussion task and asked to report on their own behavior and that of their partner during the discussion task and during the past month. The data were configured as a multitrait, multimethod matrix, and the magnitudes of the correlations were estimated with confirmatory factor analysis. The resulting correlations between observer ratings and questionnaire reports of hostile behavior during the videotaped discussion were roughly twice the magnitude of the correlations between observer rating and questionnaire reports during the past month. This suggests that the correspondence between behavioral categories and the content of questionnaire items is greater than previous studies have indicated.