Antipyrylazo III was introduced into frog cut twitch fibers (17-19~ by diffusion. After action potential stimulation, the change in indicator absorbance could be resolved into two components that had different time courses and wavelength dependences. The first component was early and transient and due to an increase in myoplasmic free [Ca] (Maylie, J., M. Irving, N.L. Sizto, and W. [C Chandler, 1987,Journal of General Physiology, 89:83-143). The second component, usually measured at 590 nm (near the isosbestic wavelength for Ca), developed later than the Ca transient and returned towards baseline about 100 times more slowly. Although the wavelength dependence of this component is consistent with an increase in either free [Mg] or pH, its time course is clearly different from that of the signals obtained with the pH indicators phenol red and 4',5'-dimethyl-5-(and -6-) carboxyfluorescein, suggesting that it is mainly due to an increase in free [Mg]. After a single action potential in freshly prepared cut fibers that contained 0.3 mM antipyrylazo III, the mean peak amplitude of &A(590) would correspond to an increase in free [Mg] of 47 #M if all the signal were due to a change in [Mg] and all the intracellular indicator reacted with Mg as in cuvette calibrations. With either repetitive action potential stimulation or voltage-clamp depolarization, the AA(590) signal continued to develop throughout the period when free [Ca] was elevated and then recovered to within 40-90% of the prestimulus baseline with an average rate constant between 0.5 and 1.0 s -1. With prolonged voltage-clamp depolarization, both the amplitude and rate of development of the &A(590) signal increased with the amplitude of the depolarization and appeared to saturate at levels corresponding to an increase in free [Mg] of 0.8-1.4 mM and a maximum rate constant of 3-4 s -1, respectively. These results are consistent with the idea Address reprint requests to Dr.