Variations in schedule from the usual 200 cGy daily, 5 days a week, for six or seven weeks, will be reviewed with both the radiobiological basis and clinical results so far. Hyperfractionated schedules (e.g., 1.15 Gy b.i.d. for 7 weeks) have given lo-15% better local control than the conventional 7-week schedule, with no extra late effects, both in Europe and USA. This is just as predicted in radiobiological grounds. Accelerated schedules (shorter overall times) are being tested in clinical trials (USA 6 weeks, Europe 5 weeks and UK 12 days). The effects of changing overall time or fraction size must be considered separately. Changes in overall time require changes in total (iso-effective) dose only if the tissue or tumour being considered is proliferating during the period of change. Therefore lengthening overall time allows both tumours and early-reacting tissues to protect themselves from damage by proliferating. There is little effect on late reactions from a change in overall time only. Shortening overall time will cause more damage to both tumour and early-reacting normal tissues because they proliferate less in the shorter overall time. Shortening overall time will have little effect on late reactions. The success of accelerated fractionation therefore depends upon the relative proliferation rates in tumours and early-reacting normal tissues. Recent advances in flow cytometry enable tumour proliferation to be measured within a day from biopsy. Accelerated fractionation will be necessary to achieve local control in some but not in all tumours. Within each tumour type there is a wide range of proliferation rates. Accelerated fractionation must not be achieved by using fewer and larger fractions than normal; this leads to excessive late damage. Tumour cells appear to respond as if they are rapidly proliferating. The relationship between total dose and dose-per-fraction can now be understood in terms of the greater curvature of the dose-response curve for late than for early reacting tissues. This is due to greater repair in those tissues for smaller doses per fraction, as if resting cells have more time to repair radiation injury. This is the rationale for hyperfractionation: a larger number of smaller dose fractions than normally. The late damage in normal tissues is spared more than the damage to early-reacting normal tissues and tumour cells. It is not essential to use the linear-quadratic formula to describe these relationships, but that formula is convenient and appears reliable between 1 and 10 Gy per fraction. The time factor must be considered separately from the dose-per-fraction factor. 'Juadratic formula as fo$lows. The surviving fraction S = e _b time factor can now be added to the Linear where: E = n(ad + Bd ) YT n = No of fractions d = size of each a/S = 2-5 Gy for late reactions = lo-20 Gy for tumor/early T = overall time Y = (log 2)/T where T = douglingPtime of cells T/T is the number of cell doublings. If proliferation doe!? not begin until the time Tk, replace T with (T-l?k). Flow cytometry suggests a range of 2-20 days for T , median 5 days. Some ways of combining the advantages of accelerafid and hyperfractionated schedules have been developed recently in clinical practice and will be discussed. Multiple fractions per day seem to be necessary for optimum schedules, but the intervals should be kept long, preferably 6 hours.