Given three person families the Subjective Family Image Test assesses the six dyadic relations on a valence and potency dimensions. Besides the total sum score (developmental conditions in the family) other aggregate scores are unsettled.Data of 461 patients with eating disorders was reanalysed in order to clarify which aggregate scores are comprehensive. Structural equation analysis followed the family social relations model.Eight further latent variables with sufficient variance in the lay psychology of the patients were identified: at the valence dimension the three reciprocal dyads and the parents as two dispositional sources, at the potency dimension a dispositional source for each family member with complementarity for the parents (e. g. the dominant father to whom all others submit themselves vs. the submissive father who is dominated by the others). Because of the found complementarity a potency total score is inadmissible. For the same reason, the potency dimension is interpreted as dominance according to the Interpersonal Circumplex. The differences between parental and child roles in the lay psychology of the patients can be explained by attribution theory.Future analyses of the subjective family image test will be guided by the 9 variables found. They allow interpreting the lay family psychology of patients with eating disorders in the terminology of the Interpersonal Circumplex. Beyond, the empirically found typology in combination with anamnestic data and observed family interaction allow to focus problematic areas of family relationships in family oriented interventions.