Our aim was to characterise and classify structural changes in the alar ligaments in the late stage of whiplash injuries by use of a new MRI protocol, and to evaluate the reliability and the validity of this classification. We studied 92 whiplash-injured and 30 uninjured individuals who underwent proton density-weighted MRI of the craniovertebral junction in three orthogonal planes. Changes in the alar ligaments (grades 0-3) based on the ratio between the high signal area and the total cross-sectional area were rated twice at a 4-month interval, independently by three radiologists. Inter- and intraobserver statistics were calculated by ordinary and weighted kappa. Cases classified differently were reviewed to identify potential causes for disagreement. The alar ligaments were satisfactorily demonstrated in all cases (244 ligaments in 122 individuals). The lesions, 2-9 years after the injury, varied from small high-signal spots to high signal throughout the cross-sectional area. Signal was highest near the condylar insertion in 82 of 94 ligaments, indicating a lesion near that insertion, and near the dental insertion in eight, indicating a medial lesion. No grade 2 or 3 lesion was found in the control group. At least two observers assigned the same grade to 214 ligaments (87.7%) on the second occasion. In 30 ligaments (12.3%) this agreement was not obtained. Pair-wise interobserver agreement (weighted kappa) was fair to moderate (0.31-0.54) in the first grading, improving to moderate (0.49-0.57) in the second. Intraobserver agreement (weighted kappa) was moderate to good (0.43-0.70). Whiplash trauma can cause permanent damage to the alar ligaments, which can be shown by high-resolution proton density-weighted MRI. Reliability of classification of alar ligament lesions needs to be improved.