How Respondents Use Verbal and Numeric Rating Scales: A Case for Rescaling