Radiobiological models for estimating normal tissue complication probability (NTCP) are increasingly used in order to quantify or optimize the clinical outcome of radiation therapy. A good NTCP model should fulfill at least the following two requirements: (a) it should predict the sigmoid shape of the corresponding dose-response curve and (b) it should accurately describe the probability of a specified response for arbitrary non-uniform dose delivery for a given endpoint as accurately as possible, i.e. predict the volume dependence. In recent studies of the volume effect of a rat spinal cord after irradiation with narrow and broad proton beams the authors claim that none of the existing NTCP models is able to describe their results. Published experimental data have been used here to try to quantify the change in the effective dose (D(50)) causing 50% response for different field sizes. The present study was initiated to describe the induction of white matter necrosis in a rat spinal cord after irradiation with narrow proton beams in terms of the mean dose to the effective volume of the functional subunit (FSU). The physically delivered dose distribution was convolved with a function describing the effective size or, more accurately, the sensitivity distribution of the FSU to obtain the effective mean dose deposited in it. This procedure allows the determination of the mean D(50) value of the FSUs of a certain size which is of interest for example if the cell nucleus of the oligodendrocyte is the sensitive target. Using the least-squares method to compare the effective doses for different sizes of the functional subunits with the experimental data the best fit was obtained with a length of about 9 mm. For the non-uniform dose distributions an effective FSU length of 8 mm gave the optimal fit with the probit dose-response model. The method could also be used to interpret the so-called bath and shower experiments where the heterogeneous dose delivery was used in the convolution process. The assumption of an effective FSU size is consistent with most of the effects seen when different portions of the rat spinal cord are irradiated to different doses. The effective FSU length from these experiments is about 8.5 +/- 0.5 mm. This length could be interpreted as an effective size of the functional subunits in a rat spinal cord, where multiple myelin sheaths are connected by a single oligodendrocyte and repair is limited by the range of oligodendrocyte progenitor cell diffusion. It was even possible to suggest a more likely than uniform effective FSU sensitivity distribution from the experimental data.