Background. The relative merit of operation in the treatment of Graves' disease has been questioned, and the extent of surgical resection is still a matter of debate. Methods. We have analyzed retrospectively the incidence of recurrent hyperthyroidism (frequency and time point) in 215 consecutive patients subjected sequentially to subtotal thyroidectomy (n = 63; remnant mass 6 to 8 g, based on surgeons' estimates and dimensions measured during operation), extensive subtotal thyroidectomy (n = 106; remnant mass similar to 4 g) and near-total (n = 27; unilateral capsular remnant of <2 g) or total thyroidectomy (n = 19). In addition we have evaluated the postoperative kinetics of thyroid hormone elimination (free triiodothyronine and free thyroxine) in 14 selected patients with hyperthyroidism who underwent operation under P-adrenergic blockade but without any thyrostatic pretreatment. Results. The size of the remnant significantly (P < .05) affected the relapse rate (23.8%, 9.4%, and 0% in subtotal, extensive subtotal, and near-total/total thyroidectomy, respectively). However, the time point at which the relapse occurred did not differ in subtotal and extensive subtotal thyroidectomy. All relapses occurred within the first 70 weeks. The incidence of complications (permanent recurrent nerve paresis and persistent hypocalcemia) was comparable in all groups. The elimination of fT3 was biphasic and rapid such that the levels were within the normal range on the second day. In contrast, 15 days were required until the fT4 level had declined below the upper limit in all patients. Conclusions. We propose that the therapeutic goal in thyroid operations is to avoid recurrent hyperthyroidism. This is not reliably achieved by subtotal thyroidectomy; in contrast, near-total and total thyroidectomy are effective and safe. On the basis of the postoperative elimination kinetics, hormone replacement is to be instituted within 2 weeks after operation.