The reliability and responsiveness of the Barry-Albright Dystonia (BAD) Scale, a 5-point ordinal severity scale for secondary dystonia, was assessed. For interrater reliability, 13 raters scored 10 videotaped patients; for intrarater reliability, two raters rated the videotape again. For test-retest reliability, patients were rated on two occasions. Four inexperienced raters scored patients, received training, then scored additional patients. To assess responsiveness, we compared patient and physician global ratings of change (better, same, and worse) with BAD Scale score changes for 18 patients on intrathecal baclofen (ITB) trials. We assessed reliability with the intraclass correlation coefficient (ICC). The mean ICC for total BAD Scale scores were as follows: interrater reliability 0.866, intrarater reliability 0.967 and 0.978, test-retest reliability 0.978 (before training) and 0.967 (after training). We found the BAD Scale responsive to change, with most improved scores in patients rated by the patient, family, and neurosurgeon as 'better'. The total scores were reliable for experienced raters. We recommend training for clinicians interested in using the scale.