Problems with the measurement of illusory correlation

The illusory correlation eflect (ICE) refers to a tendency for people to over‐represent rare behaviours performed by members of an infrequently encountered (minority) group. On the basis of an empirtcal study ( N = 245) this paper examines three procedural features that are common to studies in this field: (a) excluding subjects who assign too many behaviours to the minority group, (b) asking subjects to estimate the incidence of distinctive behaviours, and (c) submitting phi coefficients to Z transformation. Results show that the ICE is significantly lower amongst subjects who (a) are excluded and (b) estimate non‐distinctive behaviours and (c) that Z transformation applied to phi coefficients renders the distribution of responses non‐normal and does not provide a more linear index of efect strength. Implications for future research are discussed