Abstract : The “Distribution Function Method”, which makes possible the measurement of systematic variations in the shape of signals, has been experimentally tested on the detection of two visually confounded gas chromatographic peaks, irrespective of the estimation of the resolution or the concentration ratio. Two detection procedures are suggested. The first compares directly, under the same experimental conditions, the shape of a pure peak and that of a peak suspected of contamination. The second makes use of variations of column temperature and gas flow rates to distinguish the behaviour of a pure substance from that of a contaminated substance. A marking procedure for the shape of a pure peak is then set up to dispense with the need of a reference standard. In practice, the two procedures make possible the detection of heptane as an impurity in n-hexane at levels of one percent and under experimental conditions artificially chosen so as to make the resolution inferior to 0.25.