Summary.A moderately speeded, moderately difficult multiple‐choice mathematics test was set under three different instructions to three random samples of 15‐year‐old pupils of average and above average ability. The aim of the instructions was to dispose the pupil so as to inhibit his propensity to guess (A), to encourage it (C), and to be neutral in respect to guessing (B).Analysis of variance of both raw and corrected scores showed there to be no significant effect attributable to instructions either in the whole test or in the earlier or later part of the test. The instructions affected the pupils' item response pattern. There were fewer omits and unfinished' items than would be expected by chance arising from the inhibiting instructions (A), more from the instruction encouraging guessing (C) and the level expected by chance from the neutral instructions (B). The concurrent validity was not significantly affected by the instruc