Accessibility research studies often gather subjective responses to technology using Likert-type items, where participants respond to a prompt statement by selecting a position on a labeled response scale. We analyzed recent ASSETS papers, and found that participants in non-anonymous accessibility research studies gave more positive average ratings than those in typical usability studies, especially when responding to questions about a proposed innovation. We further explored potential positive response bias in an experimental study of two telephone information systems, one more usable than the other. We found that participants with visual impairment were less sensitive to usability problems than participants in a typical student sample, and that their subjective ratings didn't correlate as strongly with objective measures of performance. A deeper understanding of the mechanism behind this effect would help researchers to design better accessibility studies, and to interpret subjective ratings with more accuracy.