Dear Andrew,

I think the above contrast vector you give may be not suitable to give a correct F-test on whether any two of the four conditions are different. In your above example, I think the contrast vector [0 1 1 1] in your above example tests the null hypothesis: (B - mean) + (C - mean) + (D - mean) = 0; This null hypothesis has never been equal to (B - mean) = (C - mean) = (D - mean) = 0, which I think is the correct form of a F-test.

If I am somewhere wrong, please let me know.

Best.