Corrosion was studied in vitro in Nor's solution (CO2 - bicarbonate buffered Hank's solution) at 37 degrees C, and in vivo implanted in the lower back muscle of rats. Nor's solution is a good model for HP Mg and WZ21, because (i) the pH is maintained by the same buffer as in blood and (ii) concentrations of corrosive chloride ions, and other inorganic constituents, are similar to those in blood. The higher in vitro corrosion rate of AZ91 was caused by micro-galvanic from second phases. The lower in vivo corrosion rate of AZ91 was tentatively attributed to suppression of micro-galvanic corrosion by tissue encapsulation.