The comparison of numbers of successes and failures in 2 treatment groups by the usual chi-squared criterion with 1 degree of freedom becomes invalid if the groups are heterogeneous, falling into categories with different success rates in propor- tions which differ systematically for the treatments. If these numbers are adjusted beforehand by a simple procedure of the type used in standardising mortality rates the validity of the criterion is restored, subject to multiplication by an inflating factor which exceeds unity by an amount which is usually so small that it may be neglected. To make the technique accessible to non-statistical workers it is presented in non-mathematical language in the 1st half of the paper, discussion of its statistical basis and efficiency being left for the 2nd half.