Accuracy of MRI reports is taken for granted. In this paper the inter-observer reliability in the interpretation of meniscal lesions, degree of chondropathy, and integrity of the ACL was analyzed while taking the radiologist's experience and field strength into account. Fifty-two MRI studies of knees were interpreted by 11 radiologists independently. Twenty-two were acquired on 1.0-T, 20 on 1.5-T, and 10 on 3.0-T systems. Four of the radiologists had more than 5 years and seven had 3 to 5 years of experience in interpreting MRI studies. The findings were compared with the intra-operative findings. Inter-observer variance, specificity, and sensitivity were evaluated for each field strength. Inter-observer correlation ranged between 0.370 for cartilage lesions and 0.597 for meniscal tears. Correlation values did not increase with experience or field strength. The number of false reports was dependent on the observer, but not on field strength. The rate of false interpretations was significantly higher for most criteria in the less experienced group. In conclusion, inter-observer correlation was low, although the diagnostic criteria were defined. The use of the classification scheme should be standardized by uniform training. Radiologist experience seems to be more important than field strength.