To determine the false-positive rate of anesthetic blocks of the medial branches of the cervical dorsal rami in the diagnosis of cervical zygapophysial joint pain. Comparison between single diagnostic blocks, and a criterion standard of double-blind, controlled, differential anesthetic blocks. Tertiary referral center. The first 55 consecutive patients with neck pain for >3 months after and attributable to a motor vehicle accident, and who had completed a second diagnostic block after an initial positive response. A total of 60 joints was studied, with five patients providing two joints each. The mean age was 41 years: 61% were female. Each patient had been investigated with radiologically controlled blocks of the medial branches of the cervical dorsal rami to anesthetize the target cervical zygapophysial joint. The initial block was performed using either 0.5% bupivacaine or 2% lignocaine, randomly selected. The duration of pain relief was assessed in a double-blind fashion. The procedure was repeated with the complementary anesthetic. Only patients experiencing a longer period of pain relief from bupivacaine were considered to have true-positive responses. The second block failed to relieve pain in two of the tested joints. In a further 14 joints, the control blocks relieved pain, but the patient failed to correctly discriminate the longer acting anesthetic. The remaining 44 joints met the criteria for true-positive responses. The false-positive rate of single blocks was 16 of 60 or 27% (95% confidence interval 15%, 38%). Uncontrolled diagnostic blocks are compromised by a significant false-positive rate that seriously detracts from the specificity of the test.