Abstract : Different methods of eliciting responses to the same question often produce different responses. In order to systematically study how response scales affect likelihood ratio judgments, two experiments were conducted. Experiment I manipulated two independent variables: the endpoints of the response scales (100:1, 1000:1, 10,000:1) and the spacing of the scales (logarithmic versus linear). Results compared the veridicality of responses on the six scales produced by crossing these factors plus another response mode in which subjects simply wrote their judgment in a blank (no scale). Logarithmic scales produced responses that were both more veridical and more consistent than responses on linear scales which were, in turn, better than simple written responses. Measures of the effect of the endpoints were somewhat inconsistent and probably interacted with the range of veridical likelihood ratios. Judgments of relatively small likelihood ratios were affected by the spacing; linear spacing caused overestimation. Experiment II manipulated two additional variables, data diagnosticity and the values of the true likelihood ratios. The results of Experiment I were confirmed while neither of the additional variables radically changed the effect of endpoints or spacing.