Adults, 9-year-old children, and 5-year-old children were trained on multiple A-B and A-C matching tasks. Then they received a series of tests, first symmetry (B-A, C-A), then equivalence (B-C, C-B), and finally equivalence-equivalence tests (BC-BC). The latter tests assessed whether the subjects matched BC compounds with equivalent elements with one another and BC compounds with nonequivalent elements with one another; for example, B1C1-B3C3 (equivalence-equivalence) and B1C2-B2C3 (nonequivalence-nonequivalence). Most adults and 9-year-old children demonstrated equivalence-equivalence and nonequivalence-nonequivalence (Experiments 1 and 2). These performances were not seen with any of the 5-year-old children (Experiments 1 - 3) without first having the opportunity to match compounds with trained correct relations between elements (e.g., A1B1-A3B3) and compounds with trained incorrect relations between elements (e.g., A3C1-A3C2) (baseline-baseline, Experiment 4). Present findings suggest a developmental divide similar to that reported in earlier developmental research on analogical reasoning for which equivalence-equivalence has been used as a model. Yet, they should be taken only as tentative. Although equivalence-equivalence and classical analogies (a:b::c:?) require subjects to match functionally same relations, the procedures for measuring equivalence-equivalence are suffiently different from those used in classical analogy tests, not to permit any direct comparisons.