The probability density functions (pdf's) for the longitudinal and vertical velocities, temperature, their derivatives, and momentum and sensible heat fluxes were measured in the atmospheric surface layer for a wide range of atmospheric stability conditions. The measured pdf's for both the velocity and the temperature fluctuations are near-Gaussian and consistent with corresponding laboratory measurements for near-neutral and stable stability conditions. Hence the first- and second-order moments are sufficient to predict the heat and momentum flux pdf's. The lower-order moments can be estimated from mean meteorological conditions using surface layer similarity theory. For unstable conditions the pdf for temperature is non-Gaussian and is strongly skewed due to local convective thermal plumes. For near-neutral and stable conditions the pdf's for the velocity and temperature longitudinal gradients have long exponential tails, in agreement with findings in laboratory experiments and numerical simulations.