1. Let x and y be random variables with finite expectations. We shall say that x dominates y if e I{(x) } > E {+(y) } whenever q5 is a continuous convex function on the real line R1. (The expectations ?{+(x) } and ?{+(y) } are always well defined if + oo is admitted as a value.) Assume now that xi and x2 are independent and dominate respectively the independent random variables yi and Y2. Let q5 be a continuous convex function on R2 and denote by Fi and Gi the distribution functions of xi and yi. We have