It is shown that, in the case of asymmetric nuclear interactions, the application of the traditional dihadron correlation approach to determining a two-particle correlation function C may lead to a form distorted in relation to the canonical pair correlation function C2. This result was obtained both by means of exact analytic calculations of correlation functions within a simple string model for proton–nucleus and deuteron–nucleus collisions and by means of Monte Carlo simulations based on employing the HIJING event generator. It is also shown that the method based on studying multiplicity correlations in two narrow observation windows separated in rapidity makes it possible to determine correctly the canonical pair correlation function C2 for all cases, including the case where the rapidity distribution of product particles is not uniform.