The existing literature acknowledges that a mismatch between the experimenter s and the subjects models of an experimental task can adversely affect the interpretation of data from laboratory experiments. We primarily focus on experiments designed to test a hypothesis by comparing behavior across two or more games and highlight the drawbacks of using a between-subjects design or a within subjects design. An alternative design for laboratory experiments is proposed which may alleviate this concern especially in studies of social preferences. We argue that the proposed design does not introduce any conceptual problems in addition to those that are already present in the between and within subjects designs and may alleviate some of the problems inherent to these designs in certain cases. We also use the proposed design to answer some questions that have attracted continued attention in the literature on social preferences in general and reciprocity in particular.