Serial viable counts on Mycobacterium tuberculosis exposed in vitro to isoniazid, rifampicin and ethambutol in Tween-albumin liquid medium showed no bactericidal synergism between isoniazid and rifampicin and no influence of ethambutol on the bactericidal activity of isoniazid or isoniazid plus rifampicin. Guinea-pigs with moderately advanced experimental tuberculosis were treated fro 11 weeks with either (1) ethambutol, (2) isoniazid, (3) isoniazid plus ethambutol, (4) isoniazid plus rifampicin, (5) isonaiazid plus rifampic in plus ethambutol, or (6) no chemotherapy. The amount of tuberculous disease was scored and the spleen cultured in groups killed at intervals from 0--7 1/2 months after the end of chemotherapy. The regimens containing rifampicin were no more bactericidal during treatment than the corresponding regimens without rifampicin, but the onset of relapse after chemotherapy was delayed for at least 2 months following the rifampicin-containing regimens. Ethambutol alone did not protect guinea-pigs, nor did it influence the response to isoniazid or to isoniazid plus rifampicin. It was concluded that rifampicin may act selectively on a small proportion of the bacterial population and that it may be unnecessary to prescribe it for long periods in short course chemotherapy in man. Ethambutol does not appear likely to contribute to the sterilizing activity of short course regimens though it may prevent the emergence of drug resistance.