Abstract : The contribution of two substances may result in a chromatographic signal (in gas chromatography) whose shape is a single peak. This happens when the relative concentration of the two substances R and the difference between their retention times M are small enough. A criterion is proposed which, applied to a signal whose shape is a single peak, is able to detect the presence of two substances. This criterion is particularly efficient when R is of the order of a few percent, even in the presence of noise. Our method is limited to the detection problem, and does not aim to estimate the parameters of the possible second peak. The proposed detection method does not imply knowledge of the time origin and the width of a single peak either; it is based upon the comparison of two distribution functions; hence it is scarcely affected by noise: indeed, the information treatment is essentially carried out by means of the signal integral function. The study has been carried out by simulation on a computer and it uses a commonly accepted peak model: an exponentially modified Gaussian. To give an example: in a signal to noise ratio S/N = 100 (a value observed on real chromatograms) one detects with certainty a second substance of 2% of the first and for a distance between the retention times of 1.8 times the standard deviation of the Gaussian of the model. Last, the method is theoretically compatible with any mathematical model.