Although many web-based studies have investigated scale types using radio but­tons and numeric box entry, very few studies have examined the possibility of use of visual scales like a Visual Analog Scale (VAS) or Graphic Rating Scale (GRS). Funke (2005, GOR) conducted an initial study of respondent utilization of a VAS. As part of a larger study, we manipulated 4 factors for 6998 respondents (US 18+): topic - political or grocery store; element differentiation - high versus low differentiation; scale type - VAS with 5, 2 o~ no segments or a 5 segment GRS; banking - banked or not banked.

Respondents rated the importance of a set of 5 items (either political priorities or features of a grocery store). Immediately following, they rated their impressions concerning the scales (difficulty, accuracy, interest). Using a standard grid format,

they were then asked to rate how much each feature would affect their intention (political issues: likelihood to vote for a spending increase for the issue; grocery store: likelihood to shop at a store with the feature).

Results

Banking all 5 items on the same screen took about 11.75 seconds less for re­spondents to complete than presenting each scale one at a time. All scales were perceived as generally easy to do, though there was a significant ma in effect for scale type (F(7,6966)=2.18, p<.05), with the 2 segment VAS banked on a single screen being perceived as more difficult. For perceived accuracy, the no seg­ment VAS not banked came out as lower than the other scales (F(7,6966)=2.66, p<.01). We then examined the extent of criterion-related validity (the correlati­on of importance with likelihood), and found a significant effect for scale type (F(7,6966)=2.87, p<.01) with the 5 segment VAS and GRS not banked showing higher criterion-related validity than the banked versions (average proportion of variance = .492 and .501 versus .457 and .475, respectively). We will also present some distributional differences that we observed between the scale types. In ge­neral, the 5 segment GRS had lower means, whether banked or not.