1. Mag-fura-2 fluorescence microscopy and whole-cell patch-clamp recordings were used to measure glutamate-induced changes in the intracellular free Mg2+ concentration ([Mg2+]i) and Mg2+ currents, respectively, in cultured forebrain neurones from fetal rats in the absence of extracellular Na+ (Nao+) and Ca2+ (Cao2+). 2. Increasing the extracellular Mg2+ concentration ([Mg2+]o) from 9 to 70 mM significantly enhanced the maximum [Mg2+]i induced by a 5 min 100 microM glutamate plus 1 microM glycine stimulation ([Mg2+]i,5 min) from 2.04 +/- 0.07 to 2.98 +/- 0.20 mM. Increasing [Mg2+]o from 9 to 70 mM also significantly enhanced the initial rate of rise in [Mg2+]i upon glutamate stimulation from 0.41 +/- 0.02 to 0.81 +/- 0.08 mM min-1. 3. The glutamate-stimulated increase in [Mg2+]i was not altered by prior depletion of intracellular free Na+ (Nai+). For paired stimulations in single neurones, the mean [Mg2+]i,5 min was 1.95 +/- 0.17 mM under Na(+)-depleted conditions and 1.94 +/- 0.16 mM under control conditions. 4. The glutamate-stimulated increase in [Mg2+]i was significantly reduced when NMDA channel-permeant Cs+ or K+ ions were used as the Na+ substitute instead of the presumably NMDA channel-impermeant ions N-methyl-D-glucamine (NMDG), Tris or sucrose. The mean [Mg2+]i,5 min was 0.56 +/- 0.06 and 0.74 +/- 0.08 mM in the presence of Cs+ or K+, respectively, compared with 2.13 +/- 0.10, 1.93 +/- 0.11 and 2.07 +/- 0.22 mM in the presence of NMDG, Tris or sucrose, respectively. 5. In whole-cell recordings performed with Cs+ as the primary intracellular cation, application of 100 microM NMDA plus 10 microM glycine induced inward currents that reversed around -55 mV in an extracellular solution containing 70 mM Mg2+ and 31 mM NMDG as the only cations. The currents were reversibly inhibited by DL-2-amino-5-phosphonovaleric acid (APV). In an extracellular solution containing 2 mM Mg2+ and 140 mM NMDG, NMDA plus glycine activated outward currents at potentials more depolarized than -90 mV. 6. In whole-cell recordings made with NMDG as the principal cation in the patch pipette, application of NMDA plus glycine in the 70 mM Mg2+ extracellular solution induced inward currents at voltages more negative than +15 mV. The ratio of the current measured under these conditions to the current measured in an extracellular solution containing Na+ as the principal cation (0.073:1) was nearly constant from cell to cell. 7. Following a 5 min glutamate stimulation in the presence of 9 mM Mg2+, [Mg2+]i returned to basal levels at a mean rate of 58.1 +/- 2.1 microM min-1. Complete removal of Nao+ significantly inhibited the rate of recovery to 31% of control. Raising [Mg2+]o to 30 mM in combination with removal of Nao+ did not inhibit recovery significantly more than either manipulation alone (28% of control). 8. These results suggest that glutamate-stimulated increases in [Mg2+]i that occur in the absence of Nao+ and Cao2+ result from Mg2+ entry through NMDA-activated ion channels. Furthermore, recovery from a glutamate-induced Mg2+ load appears to be primarily due to Mg2+ efflux via a mechanism whose characteristics are consistent with Na(+)-Mg2+ exchange.