The objectives of this study were to determine the test-retest reliability of two self-perceived recovery questions in patients with recent whiplash-associated disorders (WAD), and to assess whether remembering previous answers influences reliability. The self-perceived general recovery and self-perceived change in neck pain questions were administered to 46 patients with recent WAD 6 weeks after recruitment and again 3-5 days later. At follow-up, we also asked participants if they remembered their previous answers. We used the intra-class correlation coefficients (ICC) to measure the reliability of the original ordinal response structure and kappa statistics for dichotomized responses. The ICC [95% confidence intervals (CI)] for the general recovery and for the change in neck pain questions were 0.70 (0.60-0.80) and 0.80 (0.72-0.87), respectively. The kappa statistic (95% CI) for the general recovery question was 0.81 (0.64-0.99) when recovery was defined as "completely better" or "much improved". The kappa statistic (95% CI) for the change in neck pain question was 0.80 (0.62-0.99) when recovery was defined as "very much better" or "better". Our analysis suggests that the test-retest reliability may be higher for participants who remembered their previous responses. In conclusion, our results suggest that self-perceived recovery questions have adequate reliability for use in epidemiological research of WAD.