In COPD, several walking tests have been proposed to measure exercise tolerance but their relative merits are uncertain. We studied 57 moderate-to-severe, stable COPD patients (mean FEV1 35 +/- 12% predicted). Within a 2-month period, we compared reliability (inter-subject variability) and repeatability (intra-subject variability) of the most widely used 6-min walks (6MWD), with self-paced 2-min walks (2MWD) and externally paced, incremental shuttles (Shuttle). On 9 separate days either of the three 6MWD, 2MWD or Shuttles were performed (nine walks of each type). Then, each walk was performed before and after bronchodilators (BD) to assess sensitivity to change (mean change/standard deviation of change (sensitivity index--SI)). For all exercise tests, reliability was >90% (2MWD 95%, 6MWD 92% and Shuttle 91%). Repeatability was excellent (overall <10% intra-subject variation; for 2MWD 5%, 6MWD 8% and Shuttle 9%). On average, the first walking distance was significantly shorter, but there were no significant differences between second and third walks. Dyspnoea scores were much less reproducible. BD produced highly significant improvements in Shuttle (pre-BD 27 SD=11 --> post-BD 30 SD=11), 6MWD (424 m SD=115 --> 462 m SD=106) and 2MWD (153 m SD=35 --> 162 m SD=34), (P < 0.0001). SI was similar for all walks (6MWD 0.84, 2MWD 0.75 and Shuttle 0.76). In moderate-to-severe COPD, 2MWDs are as informative as 6MWDs without their disadvantages. Self-paced walks are as useful as externally paced Shuttles.