Statistics Degrees of Freedom?

If the variances are umed equal then you have the right formula for df. If they aren't umed to be equal, you can have a more exact test by using a nasty formula to estimate d.f. that uses both sample variances. "but it says my answer is wrong... " - if the 'it" is some computer program it is just using a more general formula than most people would use (as it happens if the population variances are different this adjustment to df isn't exactly correct anyway).