Semantic differentials are used to assess different dimensions of a single construct. In contrast to unrelated items – where it is advantageous to present each item on a single screen to minimize anchoring and contrast effects – all dimensions are presented on a single (Web) page. Due to the relatively small number of categories in comparison to the number of items, it can be quite burdensome for the respondent to judge all dimensions that are presented. In a Web experiment we replaced categorical scales by visual analogue scales (VAS) created with http://vasgenerator.net. VAS are continuous graphical rating scales. The obvious advantage over discrete scales is that – as each pixel can be clicked and corresponds to a possible value – very fine graduations can be measured. An additional advantage of VAS is that data can be recoded to a rank order, as the probability of getting the same value is quite small. On two Web pages we presented 26 semantic differentials, 13 on each page. In a between subjects design we modified the scale type used in the semantic differential: either 5 categories or a VAS with 250 possible values. In addition to the usual indicators of data quality (item nonresponse, breakoffs) we examined the variance of each given answer and the correlation between all dimensions. Thanks to computerized assessment of data, we were also able to record the response time and the number of clicks needed to answer the questions. From these data we were able to infer how often judgements were modified.