Scope: This study focuses on measurement error, one component of error of observation in the framework of the total survey error (Groves et al., 2009). More precisely, this research is about is formatting error that occurs if a rating scale does not provide a perfectly matching response option (see Schwarz & Oyserman, 2001). Therefore data collected with two different closed-ended rating scales – conventional 5-point scales and graphical visual analogue scales (VASs) – were checked against each other. About VASs: The general advantages of VASs are (1) great sensitivity because of a great range, (2) data are less affected by error, leading to more statistical power (see Funke, 2010), and (3) there are far more possibilities for data analysis (e.g., recoding into odd and even number of categories, as well as into any empirical quantile). Study Design: Respondents (N = 460) completed a 40 item Big Five personality test (found at http://ipip.ori.org/ipip/). In a between-subjects design participants were randomly assigned to a questionnaire that was either made of 5-point scales or of VASs. The VASs used in the study were plain, horizontal lines with only the ends verbally anchored. They had a range of 250 values and they were generated with the free Web service VAS Generator (see http://vasgenerator.net).

Results: Overall, higher loadings on predicted factor and lower loadings on unpredicted factors lead to more explained variance with VASs in comparison to 5-point scales. The expected factor structure was considerably clearer with VASs than with 5-point scales.

Conclusion: This study adds further evidence that VASs can have a beneficial effect on data quality and that one should think about the general reluctance to use this rating scale. Overall, VASs’ positive scale characteristics should be taken advantage of in computerized data collection.