Among the most commonly used measurement instruments in social and psychological research is the response scale – a question or statement with an accompanying set of response alternatives arranged on a numeric or verbal scale. Research has accumulated which demonstrates that the construction of the response scale may influence substantially the way in which respondents answer questions. We examine the effects on the responses of (i) numeric and verbal anchors – labels at the endpoints of the scale – and (ii) numeric labelling of scale positions. Two experiments, both conducted in large-scale surveys, are described. In the first we compare two sets of numeric labels (the ranges – 5 to 5 and 0 to 10) for 11-point scales, and test whether explicit mention of the numeric anchors in the question stem modifies the effect of the labels. We find a clear effect for the scales, and a possible effect for mentioning the numeric anchors. In the second we vary both the numeric and the verbal anchors and partition the total effect according to source. We find significant effects for each factor but no interaction between them. For both experiments we observe that the characteristics of the scale (the numeric scale, uni- or bipolar anchors) lead to differences in the use of the midpoint and endpoints of the scales.