We studied 20 patients with cervical dystonia who had started to respond poorly to botulinum toxin A (BTXA) injections after an initial good response. All patients had extensor digitorum brevis (EDB) tests performed in addition to BTXA immunoprecipition assay (IPA) and mouse bioassay (MBA) antibody testing. The patients were reexamined and then treated with carefully placed electromyogram (EMG)-guided BTXA. Nine patients had a good clinical response to EMG-guided injections and all of these patients showed an obvious decrement on the EDB test. All were BTXA blocking antibodies (Abs)-negative via IPA and MBA (apart from one patient who had low BTXA antibodies titers using IPA but no antibodies by MBA). In the other 11 patients, there was a poor clinical response to EMG-guided BTXA injections. Seven of these 11 had small EDB decrement and BTXA antibodies using IPA, suggesting resistance to BTXA. Of the remaining four patients, two had obvious EDB decrement and low antibody titers via IPA (one of them had no antibodies via MBA), while the other two patients showed obvious decrement on the EDB test and no antibodies via IPA. This study shows that the EDB test correlates better with the clinical response than the antibody assays and that EDB decrement does not always correlate quantitatively with the BTXA antibody titers. In patients with secondary nonresponsiveness, it is recommended that an EDB test is the initial investigation of choice. In those patients where the EDB test does not demonstrate resistance to BTXA, a reexamination of the patients and carefully placed injections under EMG guidance may improve results.