A probit analysis has been made of data from the literature on local control of tumours and injury to normal tissue as a function of dose of radiation. Fifteen series were analysed for local tumour control in man and ten series for complications. The analysis yielded values for the D50 dose (50% incidence of effect) and the probit width (K), a measure of the steepness of the dose-incidence curve. The same analyses were made of data for rodents. Broadly, K was proportional to D50 in the ratio 1:7, with no major differences between tumours and reported complications. D50 was plotted as a function of dose per fraction for four normal tissues and two tumours in rodents. D50 decreased more rapidly with increasing dose per fraction for the normal tissues than for tumours. The probit width, K, varied inversely with increasing dose per fraction for normal tissues and this contrasted with the tumour response. Thus with increasing dose per fraction, the threshold for effect decreased and the steepness of the ensuing dose-incidence curve increased, relatively more rapidly for normal tissue than for tumour. These curves of gross response have been analysed also by the double negative log method of Gilbert [23], in an attempt to estimate the number and survival characteristics of "tissue-rescuing cells". These were calculated to be less than 1 in 10(4) of the numbers of clonogenic cells measured by excision assays. The D0 values of the derived survival curves for these tissue-rescuing cells were higher than those measured by excision assays.