Identifying lame cows and quantifying the prevalence of lameness are important elements of cattle welfare assessment that are generally achieved by methods involving observations of each animal walking. There is no published method for assessing lameness in cows confined in tie-stalls. The objective of this study (carried out within the European Commission"s Welfare Quality(R) project) was to develop a suitable method and validate it for lameness detection against a published locomotion score. A series of indicators of lameness visible in tied cows was formalized into a stall assessment protocol. This was validated against a traditional locomotion score and tested for repeatability between 2 observers. A total of 98 cows on 4 farms were assessed. Overall interobserver agreement was 91%. Sensitivity compared with locomotion scoring was 0.54 to 0.77, dependent on observer and threshold definition. Assessment in the stall underestimated the herd prevalence of lameness revealed by locomotion scoring by 11 to 37% (mean 27%). The discrepancy between herd lameness prevalence assessed in the stall and by locomotion scoring was not affected significantly by farm or observer. The cases of lameness that were not detected in the stall tended to be the least severe. The proposed method for lameness detection in tie-stalls could be used for herd-level assessment of lameness and detection of individual lame animals by farmers and their advisors, but it is important to remember that it is less sensitive than locomotion scoring.