The reproducibility of the 6-min walking test (6MWT) needs to be more solidly studied. This study aimed to investigate the reproducibility of two 6MWTs performed on subsequent days in a large and representative sample of patients with chronic obstructive pulmonary disease (COPD), and to quantify the learning effect between the two tests, as well as its determinants. In a retrospective observational study, 1,514 patients with COPD performed two 6MWTs on subsequent days. Other measurements included body composition (dual X-ray absorptiometry), dyspnoea (Medical Research Council scale) and comorbidity (Charlson index). Although the 6MWT was reproducible (intraclass correlation coefficient = 0.93), patients walked farther in the second test (391 m, 95% CI 155-585 m versus 418 m, 95% CI 185-605 m; p<0.0001). On average, the second 6MWT increased by 27 m (or 7%), and 82% of patients improved in the second test. Determinants of improvement ≥ 42 m in the second test (upper limit of the clinically important change) were as follows: first 6MWT <350 m, Charlson index <2 and body mass index <30 kg · m(-2) (OR 2.49, 0.76 and 0.60, respectively). The 6MWT was statistically reproducible in a representative sample of patients with COPD. However, the vast majority of patients improved significantly in the second test by an average learning effect of 27 m.