In this paper, we develop semiparametric inference theory to disentangle and identify the probability weighting and utility functions implicit in option prices. Our approach relies on kernel estimation based on suitable probability integral transforms and profile maximum likelihood. We establish the asymptotic properties of our estimation procedure. Monte Carlo simulations show the good finite sample performance of our approach. Empirical results reveal the relevance of probability weighting. The shape of the probability weighting function implicit in S&P 500 index option prices over the period 1996–2023 is found to be inverse S-shaped and appears to be robust to the specification of the utility function of wealth.