The Ricci case, belatedly

I don't have a lot to say about the legal aspects of the Ricci case...I'll leave those to my sister. But there is one thing about the case that interests me.

The Ricci case revolves around a written test. The test was given to firefighters in New Haven; those who scored high enough would be promoted. However, the test results were scrapped when no black firefighters reached the promotion threshold; the city was afraid of being sued under "disparate impact laws." Disparate impact laws declare that tests that inherently favor one racial group over another are a form of racial preference, and therefore illegal.

So here's my question: How, theoretically, could a test inherently favor white people over black people?

One possible answer is "language". Many black Americans grow up speaking only in AAVE ("ebonics"), a dialect of English that is somewhat different from Standard English, and which, crucially, has no standard written format. Administering any written test, therefore, might be biased against African Americans in the aggregate. (This would also support my contention that most racial problems are reall language problems.)

But how could it be possible that some written tests are racially biased and some aren't? Are we to assume that blacks and whites think in fundamentally different ways? It seems to me that if we accept the idea of racially biased tests (as opposed to language-biased tests), we must accept the idea of group cognitive differences between different races. And that, for me at least, is a difficult idea to swallow.

Am I missing something? If we rule out fundamental black-white mental differences, is there some potential non-linguistic source of racial bias on written tests?