Dipolar-coupled spectra of mols. dissolved in liq. cryst. solvents were analyzed by applying pattern recognition procedures to 2-dimensional z-filtered correlation spectra (z-COSY). Transitions that share common common energy levels can be recognized as such by recording spectra with different flip angles. This allows one to obtain a picture of the energy-level diagram without knowledge of the Hamiltonian, although in the current preliminary version of the program, the no. of spins and their symmetry must be known prior to anal. By adapting the programs NMREN and NMRIT, the energies of the eigenstates derived from the spectra are assigned to the diagonal elements of the Hamiltonian in the product base. These assignments can be made without guesswork, except for states with the same quantum nos., for which it is necessary to permute assignments by trial and error. The algorithm is iterated for different tentative assignments of the eigenstates to yield the shifts and couplings. The best Hamiltonian (best assignment) is chosen on the grounds of a comparison between simulated and exptl. spectra. Examples of entirely automated anal. are shown for A3B and A2BC systems. [on SciFinder (R)]