In this note we study the implications on a bivariate normal Roy model of two sets of monotonicity hypotheses proposed recently by Manski and Pepper (Econometrica, Vol. 64 (2000), pp. 997-1011). In that simple context, we show that these hypotheses imply strong restrictions on the correlations structure between the decision and the rewards. Copyright Blackwell Publishing Ltd and The Victoria University of Manchester, 2004. (This abstract was borrowed from another version of this item.)