Discrepancies with theory revealed by recent nucleation experiments have been examined. It is suggested that a minor deviation between experimental nucleation rate and the atomistic theoretical model, may be ascribed to a value between 1 and 5×1011sec−1(instead of 1013sec−1) for the desorption rate pre‐exponential. The much larger deviation with the spherical cap classical model is shown to be due to inability of the model to adequately evaluate the Gibb's free energy of formation from the monomer, &Dgr;Gn′, for nuclei in the rangen= 1–3 atoms, which is the actual critical size in these experiments. If, for Bi on C, the surface energy term is chosen to give the observed values ofn*, then evaluating &Dgr;Gn′ forn= 1 gives &Dgr;G1′ = −20 kcal/mole, whereas it should be zero. If the surface energy is chosen to make &Dgr;G1′ = 0, then the calculated values ofn* and the nucleation rate do not agree with experiments. A modified form of the monolayer disc classical model gives good agreement between theory and experiment but is less satisfactory than the atomistic model. The experimental temperature variation of the maximum number density of nuclei disagrees with that expected as a result of depletion of single atoms. The results can be explained by depletion of mobile nuclei by mutual capture.