Measured production rates of tropospheric ultrafine particles (similar to 3 nm diameter) are reported for the first time and are shown to be orders of magnitude greater than nucleation rates predicted by the binary theory of homogeneous nucleation for sulfuric acid and water. Furthermore, the functional dependence of observed particle formation rates on sulfuric acid vapor concentrations is much weaker than predicted by binary theory. We present arguments to show that these discrepancies might be due to the participation of a species such as ammonia which could stabilize subcritical clusters, thereby enhancing nucleation rates. The data suggest that atmospheric nucleation may occur by a collision-limited process, rather than by a condensation/evaporation controlled process as is assumed in the classical theory.