The scientists studied MHC data from 90 married couples, and compared them with 152 randomly-generated control couples. They counted the number of MHC dissimilarities among those who were real couples, and compared them with those in the randomly-generated 'virtual couples'. "If MHC genes did not influence mate selection", says Professor Bicalho, "we would have expected to see similar results from both sets of couples. But we found that the real partners had significantly more MHC dissimilarities than we could have expected to find simply by chance." Within MHC-dissimilar couples the partners will be genetically different, and such a pattern of mate choice decreases the danger of endogamy (mating among relatives) and increases the genetic variability of offspring. Genetic variability is known to be an advantage for offspring, and the MHC effect could be an evolutionary strategy underlying incest avoidance in humans and also improving the efficiency of the immune system, the scientists say.
These are intriguing results. But there is also some data for selective positive assortative mating; e.g., females being attracted to the smell of those who have their father's MHC alleles! So though the balance of research for humans does seem to suggest some aspect of "opposites attracting," it isn't quite a slam dunk. Also, contrary to what the authors state assortative mating is still alive and kicking. The study uses a Brazilian sample, which is likely as close as one can get to panmixia in a racial mixed society, but in the United States well over 95% of whites still marry other whites. That being said, I do know that outmarriage rates are highest among American Indians, who have relatively homogeneous MHC profiles. Unfortunately their relatively small numbers, and already high white admixture, serve as barriers toward drawing too much from this data.