In a credence goods game with an expert and a consumer, we study experimentally the impact of two devices that are predicted to induce consumer-friendly behavior if the expert has a propensity to feel guilty when he believes that he violates the consumer's payoff expectations: (i) an opportunity for the expert to make a non-binding promise; and (ii) an opportunity for the consumer to burn money. In belief-based guilt aversion theory the first opportunity shapes an expert's behavior if an appropriate promise is made and if it is expected to be believed by the consumer; by contrast, the second opportunity might change behavior even though this option is never used along the predicted path. Experimental results confirm the behavioral relevance of (i) but fail to confirm (ii). (C) 2013 The Authors. Published by Elsevier Inc. All rights reserved.