Results. We used structural equation modeling to test our hypotheses regarding the bonding process that results from voicing a recommendation. Structural equation modeling is well suited for testing complex causal processes with experimental data (for an application, see Hennig-Thurau et al. 2006; Wagner, Hennig-Thurau, and Rudolph 2009) and has several advantages over traditional analysis of variance approaches (Bagozzi and Yi 1989). In particular, it accounts for measurement error in the latent variables and thereby reduces the chance of type II errors. In addition, it "allows for a more complete modeling of theoretical relations, whereas traditional analyses are limited to associations among measures" (Bagozzi and Yi 1989, p. 282).