Report on the results of regression analysis studies concerning median nerve somatosensory evoked potentials (SEPs) latencies, as dependent variables, and subject height, limb length and nerve conduction velocity (NCV), as independent variables. The tests were performed on 23 normal volunteers. Absolute SEP latencies could be predicted by a linear regression model when the independent variable was arm length; when it was subject height, however, both exponential and polynomial models proved better, the Latter showing the best coefficients of determination, R 2. Multiple linear regression with two independent variables (arm length and NCV) was found to be better than simple linear regression for predicting P/N13 latency. The regression line for EP-P/N13 latency on height was found to be a polynomial curve; although the regression was found to be significant by the «F» test (alpha= 1%), the model had a low R2 value (0.41). The same applies to the P/N13-N19 interpeak latency regression curve, but the regression was significant for alpha =5% in that case. Although interwave latencies are the most useful parameters for clinical interpretation of median SEPs, absolute latencies may occasionally be important, and should be corrected for body size; in unusually tall subjects, it might be useful to double-check EP-P/N13 interwave latency prolongation by estimating the maximum expected P/N13 latency, using a model that takes into account both limb length and NCV.