A clear understanding of the nucleation mechanisms is important to understand the lifecycle of atmospheric particles and to provide reliable predictions of climate change associated with aerosols. On the basis of the classical ternary homogeneous nucleation (THN) theory developed in recent years, the presence of ppt level NH3 concentration significantly enhances nucleation rates. Here we compare the predicted NH3 enhancement with that derived from several reported experiments. We find that the differences between results predicted and observed are very large. Compared to the binary homogeneous nucleation of H2SO4-H2O (no NH3), the classical THN theory predicts up to ~30 orders of magnitude enhancement in the nucleation rates in the presence of ppt level NH3, while the laboratory measurements show only one to two orders of magnitude enhancement in the presence of several ppt to several ppm level of NH3. Furthermore, the classical THN theory predicts a decrease in nucleation rates as RH increases, but the measurements indicate an opposite trend. Since experimental results are probably more reliable than the theoretical model, we developed a kinetic THN model so that the nucleation enhancement effect of NH3 derived from these experimental results could be included. The kinetic model assumes that H2SO4 is the principal specie dominating the cluster growth and nucleation, while H2O and NH3 are secondary species influencing the cluster compositions and hence the evaporation coefficient of H2SO4 from the clusters. The kinetic model is a new approach to simulating THN, and it can take into account the thermodynamic data of molecular clusters derived from density functional theory study and prenucleation cluster measurements. The preliminary simulations using the kinetic THN model constrained by the experimental results indicate a negligible contribution of THN to new particle formation in the boundary atmosphere. Our analysis suggests that the kinetic THN model is unlikely to underpredict the real ternary nucleation rates by much as long as there is no significant error in the experimental results. The implication of this finding and areas for further investigations are discussed.