To investigate whether local anesthetic neurotoxicity results from sodium channel blockade, we compared the effects of intrathecally administered lidocaine, bupivacaine, and tetrodotoxin (TTX), the latter a highly selective sodium channel blocker, on sensory function and spinal cord morphology in a rat model. First, to determine relative anesthetic potency, 25 rats implanted with intrathecal catheters were subjected to infusions of lidocaine (n = 8), bupivacaine (n = 8), or TTX (n = 9). The three drugs produced parallel dose-effect curves that differed significantly from one another: the EC50 values for lidocaine, bupivacaine, and TTX were 28.2 mM (0.66%), 6.6 mM (0.19%), and 462 nM, respectively. Twenty-five additional rats were then given intrathecal lidocaine (n = 8), bupivacaine (n = 8), or TTX (n = 9) at concentrations 10 times the calculated EC50 for sensory block. Lidocaine and bupivacaine induced persistent sensory impairment, whereas TTX did not. Finally, 28 rats were given either intrathecal bupivacaine (n = 10) or TTX (n = 9) at 10 times the EC50, or normal saline (n = 9). Significant sensory impairment again occurred after infusion of bupivacaine, but not after infusion of TTX or saline. Neuropathologic evaluation revealed moderate to severe nerve root injury in bupivacaine-treated animals; histologic changes in TTX- and saline-treated animals were minimal, similar, and restricted to the area adjacent to the catheter. These results indicate that local anesthetic neurotoxicity does not result from blockade of the sodium channel, and suggest that development of a safer anesthetic is a realistic goal.