The efficiency with which coefficients in probit mod- els are estimated is improved by exploiting data on continuous ancillary variates. In this paper the resulting gains in efficiency are examined and illustrative calculations are provided. Extra precision is achieved at the cost of making an extra assumption but this assumption can be tested. It is shown that fully efficient maximum likelihood estimation of the probit model with a continuous ancillary variate can be achieved by a simple two step procedure involving an ordinary least squares and a probit estimation.