This paper presents results of a generalizability assessment of two sets of scales designed to evaluate the quality of verbal and nonverbal interactions between parent and child. The Nursing Child Assessment Feeding Scale (NCAFS) and Nursing Child Assessment Teaching Scales (NCATS) were developed as part of a longitudinal study aimed at determining early indicators of child development. Designed for use by trained observers, both sets of scales are Likert-type and both provide individual scale scores plus cluster scores reflecting the measure of constructs believed important to parent-child interactions. Both scales were administered to children and parents who were members of a health care cooperative in Seattle. The Feeding Scales were given four times during the child's first year and the Teaching Scales four times during the child's first two years. Results of generalizability analyses suggested further research with the instrument with more heterogeneous samples. Also, findings indicated that increasing the number of administrations across time could best decrease measurement error. Detailed technical reports on the reliability of the scales are included. (Author/RH)