In this commentary, Betsy McCoach and David Kenny state they are in general agreement with Bainter and Bollen (this issue) that causal indicators are not inherently unstable. Herein, they outline several similarities and differences between latent variables with reflective and causal indicators. In their examination of the two models, they find merit in both. Ultimately, they believe that the choice between RI and CI models should be primarily based on conceptual considerations. However, they also note that the causal indicator approach appears to function best when the paths from ?[subscript 1] to dependent variables are moderate to large and that when these paths are too small, the causal indicator approach may result in negative error variances of the disturbance ?[subscript 1].