Theoretical studies are presented for 4,-N,N-dimethylaminobenzonitrile (DMABN) by using the semiempirical Austin model 1 (AM1) and ab initio Hartree-Fock (HF) methodology for optimization of the electronic ground and AM1/configuration interaction with both single and double excitations (CISD) and HF/configuration interaction with single excitation (CIS) for the lowest excited states. For a correct description of the ground-state structure, additional polarization functions and at least a split-valence double-C basis set have to be used. For both the ground and excited states of DMABN, the relative orientation of the two methyl groups is important: AM1/CISD predicts both the first (L-1(b) character) and second excited state (L-1(a) character) to be of untwisted and slightly pyramidalized structure with the methyl groups oriented in a staggered conformation. HF/CIS computes the L-a state at lower energy than the L-b state in contrast to experimental data. This incorrect state ordering represents a serious problem for geometry optimization as only the lowest excited state of a given symmetry can be optimized because of root flipping. The HF/CIS L, optimized geometry is twisted by about 30degrees yielding the methyl groups in an eclipsed conformation. Optimization of the twisted intramolecular charge-transfer state (TICT) yields different geometries for both methods. Both methods calculate the dimethylamino group for a 90degrees-fixed twist angle to be of sp(2)-hybridization (i.e., without pyramidalization). The AM1/CISD-optimized structure, however, has a widened amino-carbon bond length and aromatic (nearly equal) benzene bonds, whereas the HF/CIS-optimized structure yields a shortened amino-carbon bond and alternating benzene bond lengths. The results of AM1/CISD, HF/CIS, complete active space self-consistent field (CASSCF), and second-order perturbation theory (CASPT2), time-dependent density functional theory (TDDFT), density functional theory/single-excitation configuration interaction (DFT/SCI) and multireference configuration interaction (DFT/MRCI) single-point calculations are compared by using both the AM1/CISD- and HF/CIS-optimized geometries for the calculation of absorption and emission energies. The results of both the CASPT2 and all DFT-based methods are in qualitatively good agreement with experimentally obtained absorption energies. A comparison of calculated emission energies by using excited-state geometries with data using ground-state optimized geometries shows the necessity to use optimized excited-state geometries for computation of emission energies. The first excited-state energy surface pathway corresponding to the photoreaction from the planar L-1(b) to the (TICT)-T-1 state can only be obtained with AM1/CISD geometries. A strongly endothermic reaction is predicted by AM1/CISD, HF/CIS, and CASSCF, a slightly exothermic reaction by CASPT2 and the DFT/configurations interaction methods, and a strongly exothermic reaction by the time-dependent DFT methodology. Experimentally, a slight increase in energy is found.