A theoretical analysis has been carried out of the structure of metalliferous epoxy–chelate polymers (MECP) based on diglycidyl ether of bisphenol-A (DGEBA) hardened with metal complexes of the formula [M(L)n(X)p], where M is the cation of the transition metal; R, a nitrogen-containing ligand; X, the anion of an organic acid; n, the number of the ligands in the complex molecule (n = 1 or 2), and p, the metal valency (p = 2 or 3). On the basis of the correlations between the tensile strength (σt) and tensile modulus (Et), and flexural strength (σf) and flexural modulus (Ef), of MECP, σt = f(Et)and σf = f(Ef), and supposing that when the condition equation image is fulfilled, where A and B are complex hardeners of different structures but of the same class, the epoxy–chelate matrices have similar structures. The influence of the structural fragments of the hardener molecule (the metal, ligand, and anion) on the polymer properties was evaluated and it was found out that the biggest contribution to these properties belongs to the metal, the alteration of which changes the thermal stability (ΔM is the polymer mass loss after thermal treatment in air), deformability (ε), σf, Ef, and deflection temperature (DT) significantly. By this, the effect of the hardener structure change on the alteration of the MECP properties is maximal for ΔM, is minimal for the compressive strength (σc), and decreases in the series: ΔM > ε > DT > σf > Ef > σc. The type of the anion affects σc significantly, but the ligand type contributes the least to the polymer properties. The obtained dependencies of the MECP properties on the structural fragments of the complex hardeners allow preliminary evaluation of the structure of the chelates and epoxy–chelate compositions necessary to produce epoxy polymers with required properties. The new method of the theoretical investigation of the effect of the structural fragments (method of TIESF) of the polymer matrix on the polymer properties can be used to analyze the structures of the polymers of other classes and to predict the optimal structures, promising the production of the materials with the optimal properties. © 1993 John Wiley & Sons, Inc.