Previous work (Ahn and Liu (1990) J. Aerosol. Sci. 21, 249-261; Brockmann (1981) Ph.D. Thesis, University of Minnesota; Rebours et al. (1992) J. Aerosol. Sci. 23, S189-S192; Stolzenburg (1988) Ph.D. thesis, University of Minnesota) has shown that for particles smaller than about 15 nm, pulse heights produced by the optical detector in a white-light ultrafine condensation nucleus counter (UCNC; Stolzenburg and McMurry (1991) Aerosol. Sci. ＇Technol. 14, 48-65) decrease with initial particle size. We have previously reported on the use of pulse heights from this instrument to determine the concentrations of freshly nucleated atmospheric nanoparticles in the 3-4 nm diameter range (Weber et al. (1995) J. Atm. Sci. 52, 2242-2257) Weber et al. (1997) J. Geophys. Res. 102, 4375-4385). In this paper we report on the inversion of measured pulse-height distributions to obtain size distributions of particles in the 3-10 nm diameter range. Using methods developed by Stolzenburg (Stolzenburg (1988) Ph.D. Thesis, University of Minnesota) the effect of diffusional broadening is taken into account so as to obtain monodisperse kernel functions from measured pulse-height distributions produced by DMA-generated calibration aerosols in the 3-50 nm diameter range. These kernel functions are then used with the MICRON algorithm described by Wolfenbarger and Seinfeld (1990, J. Aerosol. Sci. 21, 227-247) to obtain size distributions of nanoparticle aerosols from measured pulse height distributions. Calculations were done to ensure that assumed pulse-height data generated from selected known size distributions can be inverted to recover the original size distribution. Results from these validation studies are discussed. Applications of the inversion algorithm to data acquired in studies of homogeneous nucleation in the atmosphere are also presented.