STATS MAKE ME CRY

Not exactly a regression problem, but...

Suppose there are two ability tests, on which scores are normally distributed and correlated .7 with one another. Suppose that there's a criterion which requires scores at some level, say 2 standard deviations above the mean on one of these tests. We know that if only one test is allowed, then the probability of a person scoring above 2 SDs may be simply calculated from tables of the normal distribution or the function available in spreadsheets.

However, suppose the problem is to determine the percentage who would achieve the criterion on one or both tests when allowed to take both.

My solution is to use the bivariate normal, and to calculate the probability using Mathematica or some other program which does bivariate normal calculations. I have answers for various versions and contingencies of this problem, but have run into someone who argues very vehemently that the bivariate normal does not model this this problem, and is not appropriate. I don't buy it, but wonder whether he has something and has just not argued the position properly.

I can find nothing in any references to indicate that this is an inappropriate usage.