To date, researchers have assumed that continuums ranging from left to right (or top to bottom) and positive to negative (or negative to positive) are the optimal questionnaire format. This format is thought to encourage reading of all scale points and to promote comprehension of the words used for each response option. If verbal labels are constructed by researchers as arrayed in a continuum, then what respondents might actually be doing is ranking a construct according to its slot on a continuum relative to other constructs measured using the same type of scale.

This study compares the results yielded by two types of rating scales. One scale was designed to mimic the conventional scales widely used in survey research with response options fully labeled verbally. The second set of scales was presented one of four ways; 1) a slider with five verbal labels on it, exactly mimicking a traditional Likert scale, but with tick marks along the scale, 2) a slider with five tick marks on it, but only the end points labelled with the text anchors, 3) a slider with no tick marks and with only the end points labelled with the text anchors, 4) a slider with anchors but no tick marks but with the actual “numerical score” displayed as the slider is moved. Respondents in the United States, the United Kingdom, Germany, and China were members of an online panel and were randomly assigned to receive either the conventional, verbal, presentation or one of the four visual presentations.

Respondents report a preference for the slider approach because it allows for more accurate opinion reporting. The results suggest that using a slider rather than a traditional Likert scale will produce the same mean scores. It would therefore seem likely that the slider scales would also increase levels of engagement, which in turn could improve data quality by reducing satisficing. On the other hand, sliders take longer to complete, which may decrease quality (although completion time may decrease somewhat as respondents become used to using the slider). Additional analyses will illuminate which type of scale yields greater predictive validity.