FIRST BAYES is a Microsoft Windows based package to assist in teaching elementary Bayesian statistics. As a Windows program, it is particularly user-friendly, and students can pick up quickly how to use it. First Bayes includes analysis of simple one-parameter models (binomial or Poisson data or normal data with known variance), using essentially arbitrary priors. A range of summaries and plots of prior, posterior and predictive distributions are available. First Bayes also includes analysis of simple linear models (one or more normal samples with common unknown variance, simple linear regression), using conventional improper priors. Summaries as above for posterior marginal densities or predictive densities are available. The regression analysis includes scatter and residual plots. All plots can be resized or printed. They and other summaries are automatically updated if any details of prior or data are changed. Some simple datasets are provided and others may be created in First Bayes. The author is Tony O'Hagan, University of Nottingham, UK.