The Objective of this study was to test the hypothesis that presenting risk information using numbers rather than words is a more effective way of communicating the residual risk inherent in a screen negative test result. We used a randomised controlled trial in a large UK teaching hospital. Two hundred and twenty pregnant women who received negative results on serum screening for Down syndrome participated. Presentation of screen negative test results were given either as a numerical probability (e.g. you have a 1:650 chance of having a baby with Down syndrome) or as a verbal probability (your chance of having a baby with Down syndrome is low). In both interventions the verbal anchor 'it is unlikely that your baby has Down syndrome' was used. Our aims were to measure the understanding of the residual risk in a screen negative result, and anxiety. Immediately after receipt of the results, 97% of those receiving their results in the form of a numerical probability and 91% of those receiving their results in the form of a verbal probability correctly understood that their baby probably did not have Down syndrome (95% CI for difference: 0% to 12%; p=0.04). All those who were incorrect believed that their baby definitely did not have Down syndrome. Subgroup analysis showed that this effect was confined to those with lower levels of education (i.e. those without a university degree), amongst whom understanding was poorer. There was no difference between intervention groups in understanding the results at four months. There were no differences between intervention groups in the levels of anxiety at one week or four months after receiving their results. In conclusion, communicating residual risks using numbers rather than words has a small beneficial effect of increasing awareness of residual risks without raising anxiety. Further work is needed to estimate the size of this effect in less well-informed and less highly educated populations. Copyright 2000 John Wiley & Sons, Ltd.