A large number of studies use interference effects to investigate the underlying mechanisms of numerical cognition. One of the most studied interference is called the SNARC effect which reveals the association between response side and numerosity. The classical method to measure interference effects implicitly assumes that most participants have the same associations between the interfering attributes. As an example, in SNARC effect, most of the people from the western culture associate LARGE numbers with RIGHT responses and SMALL numbers with LEFT responses producing a homogenous interference effect. However, previous studies showed that interferences are not necessarily homogeneous across participants. Iranian people, for example, have opposite associations than western people, namely associating LARGE numbers with LEFT responses and SMALL numbers with RIGHT responses. Here, we present a new method for detecting interferences in a population with heterogeneous associations. Considering the SNARC effect, the classical method calculates the linear regression slope for the difference of the right and left response side performance (number as independent and performance as dependent variable in the linear regression). In contrast, the new method computes two separate slopes for the performance on the right and the left sides and measures the correlation between them. We show that the classical method measures homogenous but not heterogenous, while the new method assesses heterogeneous but not homogenous interferences. Therefore, the classic and the new methods are complementary measurement of interferences. Finally, we demonstrate the importance of using the new method by revealing a new type of interference between parity and numerosity which, being heterogeneous across participants, would have been impossible to find with the classical method alone. Our work suggests that the usage of both the single and the dual indices is inevitable when investigating interference effects.