I study a repeated buyer-seller relationship for the exchange of a given good. Asymmetric information over the buyer's reservation price, which is subject to random shocks, may lead the seller to use a rigid pricing policy despite the possibility of making higher profits through price discrimination across the different satates of the buyer's reservation price. The existence of a flexible price subgame perfect equilibrium is shown for the buyers sufficiently locked-in. When the seller faces a population of buyers whose degree of involvment in the relatioship is unknown, the flexible price equilibrium is not necessarily optimal. Thus tipically the seller will prefer to use the rigid price strategy. A learning process allowing the seller to screen the population of buyers is derived abd the existence of a switching point between the two regimes (i.e. price rigidity and price felxibility) is shown.