Assumptions of reliability and consistency of self-report of pain by patients using visual analogue scales (VAS) and numerical rating scales (NRS) are based on narrow considerations of possible sources of error. This study examined patients' use of VASs and NRSs, by their own description, with particular attention to rating of multiple pains, of different dimensions of pain, and of interpretation and use of lower and upper endpoints and increments on the scales. These have implications for the approximation of the scales to psychometric requirements. An interview developed from a small pilot project was given to 78 volunteer chronic pain patients embarking on a pain management course, and consisted of both forced choice questions and free response. Data are described with reference to lack of concordance between patients and of consistency within patients; responses suggested that ratings incorporate multiple partially differentiated dimensions of pain, with particular importance placed on function or mobility. Labels assigned to scale endpoints by researchers, whether lexical or numerical, appeared to affect their use; however, covert relabelling of scale points was revealed in free response. The action of arriving at a rating is better conceptualised as an attempt to construct meaning, influenced by and with reference to a range of internal and external factors and private meanings, rather than as a task of matching a distance or number to a discrete internal stimulus.