Correlated ab initio quantum chemical methods based on second-order perturbation theory and density functional theory (DFT) together with large atomic orbital (AO) basis sets are used to calculate the structures of four cyclophanes with two aromatic rings and one sulphur-containing phane with one aromatic ring. The calculated geometrical data for [2.2]paracyclophane, cyclophane (superphane), 8,16-dimethyl[2.2]metacyclophane, 16-methyl[2.2]metaparacyclophane, and 2,6,15-trithia[3(4,10)][7]metacyclophane are compared to experimental data from X-ray crystal structure determinations. In all cases, very accurate theoretical predictions are obtained from the recently developed spin-component-scaled MP2 (SCS-MP2) method, in which the deviations are within the experimental accuracy and expected crystal-packing or vibrational effects. Especially the inter-ring distances, which are determined by a detailed balance between attractive van der Waals (dispersive) and repulsive (Pauli) contributions, are very sensitive to the level of theory employed. While standard MP2 theory in general overestimates the dispersive interactions (pi-pi correlations) between the two aromatic rings leading to too short distances (between 3 and 8 pm), the opposite is observed for DFT methods (errors up to 15 pm). This indicates that an explicit account of dispersive-type electron correlation effects between the clamped aromatic units is essential for a quantitative description of cyclophane structures. In order to distinguish these effects from "normal" van der Waals interactions, the term "overlap-dispersive" interaction may be employed. The popular B3 LYP hybrid density functional offers no advantage over the pure PBE functional that at least qualitatively accounts for some of the dispersive effects. The use of properly polarized AO basis sets of at least valence-triple-zeta quality is strongly recommended to obtain quantitative predictions with traditional wave function methods.