The effect of the range of levels of the test stimuli on ratings of sweetness intensity is demonstrated for a lime drink. The results are consistent with Poulton's (1979) suggestion that range effects can be predicted from the deviation of the mean response in a session from the mid-point of the response dimension. It is demonstrated that this range bias can be avoided by adding a readily anchored mid-point (ideal) to the response dimension and selecting stimuli so that the mean response is close to this mid-point. This bias-reducing, relative-to-mid-point procedure also produces a stronger linear relationship between ratings and concentration ratios than does the conventional end-point-anchored intensity rating procedure. The similarity of the linear semi-log functions of response relative to extremely sweet and relative to ideally sweet is taken as evidence that the two types of responses are not independent.