A series of recent papers has demonstrated "mirror extreme" variation in measures of brain size, between individuals with deletions compared to duplications, for copy number variation loci. For most loci, ASD-associated CNVs involve larger brain sizes, and schizophrenia-associated CNVs involve smaller brain sizes, the same patterns also found for idiopathic cases of ASD and schizophrenia. However, some CNVs have been linked with both disorders. Taken together, these findings appear paradoxical and require explanation. One possibility is that the paradoxical CNV data represents, to some degree, false positive diagnoses. A second possibility is that the paradox is based on false premises. I review current evidence salient to the two hypotheses, and suggest future studies to test them in a definitive way.