Calling models with causal indicators "measurement models" implies more than they can deliver

We are delighted to see Bainter and Bollen’s excellent paper as a focus article in Measurement. In our view, psychological researchers who use SEM rely too reflexively on reflective measurement, without sufficiently considering whether their indicators are likely to be caused by the latent construct. When causality flows from indicators to the construct, fitting a reflective model will result in model misfit, misspecification, incorrect parameter estimates, and fruitless discussion about the underlying factor structure of a questionnaire. This is a topic that requires greater awareness.
Bainter and Bollen defend the position that causal measurement is just as valid as reflective measurement and that it leads to no great interpretational problems. While we wholeheartedly agree that not all indicators should be treated as reflective, we find their position potentially misleading. The relation between causal indicators and the latent variable they indicate is insufficiently restrictive to allow causal indicators to uniquely pick out a latent variable.
In fact, the identification of a latent variable as the target of a measurement instrument requires effect indicators. However, calling the causal indicator side of a model a "measurement model" implies that the causal model is in fact sufficient to identify a latent variable, obscuring the need for additional effect indicators. In our view, the debate over interpretational confounding is just one consequence of the ensuing confusion, and much would be gained by simply abandoning the terminology of measurement in discussing models with causal indicators.