In a variety of behavioral experiments, magnesium has effects that are similar to cocaine and other psychomotor stimulants. Of particular relevance to the present experiments is the recent finding that magnesium maintains responding in cocaine-trained rats. It would be expected, therefore, that injections of magnesium would alter the rate of responding maintained by self-administered cocaine in rats. Five experiments examined the specificity and selectivity of this interaction. Acute and multiple injections of MgCl2 (15-250 mg/kg) produced dose-dependent reductions in responding maintained by cocaine (0.1-2 mg/kg/infusion). Testing for acute injection effects occurred following injections, while testing for multiple injection effects occurred prior to daily injections. Doses of 30 and 125 mg/kg MgCl2 reduced responding maintained by doses of cocaine that were below the training dose of 0.75 mg/kg/infusion. MgCl2 in a dose of 250 mg/kg markedly suppressed responding maintained by each dose of cocaine. A magnesium-deficient diet produced a dose-dependent increase in responding maintained by 0.1 mg/kg/infusion cocaine. In order to determine the specificity and selectively of these effects, acute and multiple injections of MgCl2 were examined on glucose + saccharin- and food-maintained responding. The acute effects of MgCl2 injections were specific because food-maintained responding was not affected, except by the highest dose of 250 mg/kg. This demonstrates that lever pressing was not nonspecifically reduced by 30 and 125 mg/kg MgCl2 during cocaine availability. However, the effects on cocaine-maintained responding were not selective for cocaine because glucose + saccharin-maintained responding and responding during extinction from cocaine were affected by MgCl2 in a manner similar to cocaine-maintained responding.(ABSTRACT TRUNCATED AT 250 WORDS)