Abstract The comparison of scaling methods used to value health states sometimes rests upon an analysis of aggregate scores. This analysis is usually undertaken once “inconsistent’ respondents have been excluded from the data. However, it is important to be able to judge the extent to which respondents as a whole are logically consistent when assigning values to health states. The degree of inconsistency will depend on how the health states are described, how the questionnaire is administered and who the respondents are. This paper analyses the inconsistency rates from two studies in which valuations for EuroQol health states were elicited using the visual analogue scale (VAS) method. The studies differed in design and incorporated several different variants of the standard EuroQol questionnaire, thus providing an opportunity to examine the relative importance of the different factors that were thought to affect inconsistency rates. Our general conclusions are that inconsistency rates are higher for interviewer-based than for postal surveys, possibly due to response bias, and that inconsistency rates are positively related to age and negatively related to educational attainment.