Bernstein's languages of description: Some generative principles

There are numerous ways of displaying Likert‐type scales but only a few investigators have investigated these differences systematically. In this study we report the results that we found when we compared four different layouts: scales that went numerically from ‘0’ to ‘10’, or from ‘10’ to ‘0’, and scales that went verbally from ‘clear’ to ‘unclear’, or ‘unclear’ to ‘clear’. Over 450 participants rated each of seven aspects of a structured abstract in a web‐based study, with each one using only one of the four scale formats listed above. The resulting data showed that the scale ‘Clear – 10 … 0 – Unclear’ consistently led to significantly higher ratings in all seven cases. Such findings have implications for the design of Likert‐type scales and for the data that are gathered from them.