A psychometric evaluation of the Short Interpersonal Reactions Inventory (SIRI) in an Australian twin sample

The Short Interpersonal Reactions Inventory [SIRI (Grossarth-Maticek, Eysenck, Vetter & Schmidt, Topics in Health, 1988)] is a questionnaire designed to measure personality traits predisposing to disease, notably ischaemic heart disease and cancer. It has been criticized by several authors in terms of design, scoring and validity. In a sample of 762 Australian twins taking part in a genetic study of asthma, we have examined the psychometric properties of the SIRI, and its relationship with the Eysenck Personality Questionnaire. Asthma was not significantly associated with any of the items or subscales of the SIRI. Scores on the seven subscales as defined by the SIRI authors (Types 1–3, 4 + and 4−, 5–6) were heavily skewed or truncated in six cases. Disturbingly, the two alternative forms of the Type 4 subscale were poorly intercorrelated (r = 0.17). A factor analysis using polychoric correlations partially replicated these subscales, but with important differences, similar to those found by other authors (Ranchor, Sanderman & Bouma, Personality and Individual Differences, 14, 483–484 1992). Scores on the SIRI and EPQR/S personality types were found to correlate broadly, as predicted by Eysenck (Perceptual and Motor skills, 71, 216–218, 1990). Attempts to assign each subject to a single major type were unsuccessful, whether by maximum subscale score, or membership of the upper quartile of a subscale distribution. Finally, we found that the questionnaire could be shortened from 70 to 25 items with little loss of internal reliability. We conclude that the SIRI has a number of shortcomings as currently formulated, but that it may serve as the basis for shorter instruments with superior internal and construct validity.