Seven T 2 maps (multiecho (ME) sequence: 3000 ms, eight echoes with 13.2 ms of echo spacing, 20 sections) and T 1 -weighted (T 1 -w) fast low-angle shot (FLASH-water excitation (WE)) data sets from four imaging sessions (right patellae of 10 healthy volunteers) were obtained. A segmentation of cartilage (WE sequence) was overlaid on the ME data and T 2 values were calculated for total cartilage, three layers, three facets (global), and 240 ROIs (regional). Reproducibility (precision error) was calculated as the root mean square average (RMSA) of the individual coefficients of variation (COVs, %) and standard deviations (SDs, ms) for intra-and intersession reproducibility. The precision error was 3-7% and 6 -29% for global and regional T 2 , respectively. There was no difference between intraand intersession reproducibility, but there was worse reproducibility in the superficial layers compared to the deeper layers. Peripheral ROI reproducibility (mean ؍ 13%) was worse than in the central portions (mean ؍ 11%), but omission of the periphery did not positively affect the globally calculated T 2 reproducibility. The precision errors were small compared to reported changes in diseased cartilage, suggesting good discriminatory power of the technique. Our data provide a first estimate of global and regional reproducibility errors of T 2 in healthy cartilage, and may serve as a basis for sample size calculations and aid study designs for longitudinal and cross-sectional trials in osteoarthritis (OA). Magn Reson Med 56:527-534, 2006.