The purpose was to test the reliability of short samples of parent/child interaction for use in single-subject research. Four variable pairs of mother/child behaviour were coded for seven mother/child play sessions. Each session lasted 20 min and 18 min of the session was behaviourally coded using frame-by-frame analysis. The co-occurrence of the mother/child behaviours within a given time window was computed and an odds ratio was calculated for the co-occurrence of the targeted behaviours. The play session was divided into shorter segments (3, 6 and 9 min) and odds ratios of the variable pairs from the shorter segments were compared to the odds ratios from the entire session. Segments of 3 and 6 min did not yield the same pattern of results as the entire session. In single-subject research, evidence of the reliability of the time segment for behavioural coding should be reported in the methods section of original research manuscripts.