In-vivo ³¹P-nuclear magnetic resonance (NMR) studies were undertaken with anesthetized rats to determine: a. whether systemic administration of MgCl2 could protect animals against cocaine-induced hemorrhagic stroke, and b. whether a relationship exists between basal levels of brain intracellular free magnesium ions ([Mg²⁺]_i), phosphometabolites, and stroke risk. Repeat ³¹P-NMR spectra were obtained at various intervals of time (3-120 min, or up until death) after administration of cocaine (5 + 30 mg/kg). Ion selective electrodes were used to measure plasma Mg²⁺, K⁺, Na⁺ and Ca²⁺. Forty percent of animals died in the absence of Mg²⁺ infusion following high dosage of cocaine. Only 13% died with cocaine following Mg²⁺ infusion (p <0.005). In the Mg²⁺-protected animals, neither brain [Mg²⁺]_i,intracellular pH (pHi), [phosphocreatine-PCr]/[ATP], nor brain [inorganic phosphate-Pi]/[ATP] fell when toxic and lethal doses of cocaine were given. Low basal brain [Mg²⁺]_i (275 +/- 24 vs. 466 +/- 35 microM, p <0.01) and low basal brain [PCr] (3.36 +/- 0.35 vs. 4.26 +/- 0.24 mM, p <0.01) were found to be associated with a 3-fold increased incidence of stroke. A positive correlation (r = 0.31, p <0.03) between brain [Mg²⁺]₁ and [PCr]/[ATP] was found. It is possible that both brain [Mg²⁺]_i and [PCr] may be useful as important predictors of susceptibility to cocaine-induced hemorrhagic stroke.