The aim of my thesis was to develop a method for estimating the actual stress level, based on short ECG recordings. During the 4 semesters, I tried several different methods in the laboratory, while I found what I thought to be the most appropriate. The measurements were processed in MATLAB environment, and for the characterization we introduced a parameter set, pNNtri, which was not used elsewhere yet. The differences between the adjacent cardiac cycle lengths were divided into three ranges, – under 20 msec (pNN20), between 20-50 msec, and above 50 msec (pNN50) – and I used the ratio of these intervals for the quantitative analysis. The measurement protocol consisted of 6 consecutive measurements: 1: at rest 2: at rest 3: under blood pressure measurement 4: control (at rest) 5: under psychic Stress 6: after physical stress. I examined the short and long term reproducibility, and the mean deviation of the pNNtri between the values belonging to the suspected stress condition and at the rest state. Although in most of the cases, – except in the post physical stress measurements –, I did not observe significant changes, but in many cases the tendency what we thought logic, was justified. I examined the change of pNNtri in time domain while sliding a 1 minute average-window along the ECG signal. I have also made some interesting conclusions about these results. Ultimately I found that in order to determine the actual stress level, we need more measurements from the same person in different times, and with other additional methods the process probably could be more accurate.