In aqueous acidic solutions, a higher peak current and a more positive peak potential for the first oxidation peak of polyaniline (PAN) compared with those obtained in a multicycle voltammogram was observed after holding at a sufficiently negative potential. This is referred to as a memory effect in PAN. An exception to this behavior was observed at low scan rate, when the peak potential in the first voltammogram can be less positive than that in the multicycle voltammogram. When the solvent was substituted by a mixture of water and some organic solvents, the first oxidation peak obtained in aqueous solution split into two subpeaks. The first subpeak, with a less positive peak potential, contributed strongly to the memory effect observed in PAN, but the second subpeak, with a more positive peak potential, was little influenced by holding in the reduced state. The exception mentioned above could be explained in terms of the existence of these two subpeaks if it could be shown that they were present for PAN in any solvent. These two subpeaks may reflect different structures in reduced PAN or be due to solvent effects on PAN intermediates during cycling.