Earliest childhood memories (EMs) have been utilized as expressions of relationship paradigms, but few empirical studies have been conducted. This study outlines the construction and development of an EM relationship scoring system and scale utilizing nonclinical and clinical samples. It was hypothesized that relationship episodes could be reliably coded using EMs and that they would demonstrate convergent validity with measures of attachment style (Separation-Individuation Test of Adolescence), mood (Profile of Mood States), and clinical symptomatology (Symptom Checklist 90-Revised [SCL-90-R] and the Minnesota Multiphasic Personality Inventory [MMPI]). The hypotheses received broad support. Findings indicate that relationship episodes may be reliably coded from EMs. Associations between individual EM variables, the EM relationship scale, and objective measures suggest that the quality of relationships expressed in EMs is associated with degree of general maladjustment. Suggestions for further research are discussed.