PURPOSE: To assess the usefulness of three-dimensional Fourier transformation constructive interference in steady state (CISS) for the evaluation of chondromalacia. MATERIALS AND METHODS: In 110 knee joints which underwent both MR imaging and arthroscopy, the findings were retrospectively reviewed. MR imaging sequences included two-dimensional dual-echo turbo spin-echo imaging along the sagittal and coronal planes, two-dimensional fast low-angle shot (FLASH) with magnetization transfer along the axial plane, and three-dimensional CISS along the sagittal plane. After the cartilage surfaces of each joint were divided into eight areas (each medial and lateral area of patellar facets, trochlear surfaces, femoral condyles, and tibial plateaux), a total of 880 areas were assessed. Using both combined two-dimensional (2-D turbo spin-echo and FLASH) and CISS imaging during different sessions, each chondromalacia case was assigned one of five grades. RESULTS: Arthroscopy revealed the presence of chondromalacia in 162 areas. This was first grade in 77 areas, second grade in 38, third grade in 21, and fourth grade in 26. The sensitivity, specificity, and accuracy of 2-D and CISS imaging were 48.1%, 93.7% and 85.3%, and 45.7%, 95.3% and 86.1%, respectively. Agreement between MR and arthroscopic staging occurred in 81.48% of 2-D imaging procedures and 82.16% of CISS procedures. If a difference of one grade was accepted, these proportions rose to 84.32% and 85.22%, respectively, though this increase was statistically insignificant. CONCLUSION: Though CISS imaging was less sensitive than 2-D imaging in the grading of chondromalacia, additional CISS imaging can help improve the accuracy of this grading.