Stellar distances from spectroscopic observations: a new technique

A Bayesian approach to the determination of stellar distances from photometric and spectroscopic data is presented and tested both on pseudo-data, designed to mimic data for stars observed by the Radial Velocity Experiment survey, and on the real stars from the Geneva-Copenhagen survey. It is argued that this method is optimal in the sense that it brings to bear all available information and that its results are limited only by observational errors and the underlying physics of stars. The method simultaneously returns the metallicities, ages and masses of programme stars. Remarkably, the uncertainty in the output metallicity is typically 44 per cent smaller than the uncertainty in the input metallicity.