While research has shown that both black and white children can clearly discriminate (usually through the use of dolls or puppets) with respect to ethnic identification by the age of 3-4 years, findings regarding ethnic preference are more complex. Researcher W.C. Banks has implied that research designs that compare ethnic preferences across groups cannot lead to valid inferences. Alternatively, it is proposed here that such research can be useful if two commonly used measurement approaches, one based on a comparison of preference choices and the other on chance frequencies, are considered as evaluating two different phenomena. When 21 studies are analyzed, different results are obtained by the use of different statistical approaches, one comparing percentages of own group choices. Most notably, it can be seen that whites consistently make more favorable own group responses than do blacks. This finding raises questions about the validity of white "control" groups for research on ethnic preferences. (Author/GC)