In the focus article of this issue, von Davier, Naemi, and Roberts essentially coupled: (1) a short methodological review of structural similarities of latent variable models with discrete and continuous latent variables; and (2) 2 short empirical case studies that show how these models can be applied to real, rather than simulated, large-scale data sets. In this commentary, the author shares a few conflicting impressions. These impressions concern the alignment between what he felt von Davier, Naemi, and Roberts set out to do rhetorically, on the one hand, and how they went about accomplishing their goal, on the other hand. In short, this author argues that both aspects of their work are valuable in isolation but that a stronger evidentiary basis, grounded in richer validation arguments, needs to be brought to bear to flesh out the nuances of their targeted rhetoric. Statistically, the author was hoping that their argument would be based, in part, on a more comprehensive evaluation of model-data fit that would go beyond the relative level--even though that is clearly useful--and would also focus on absolute/global as well as local levels (e.g., item level, item pair level, person level). Moreover, he was anticipating a more in-depth use of covariates along with a comparison of their empirical findings with those of previous researchers at a finer level of detail. Finally, the author was also looking forward to a rigorous evaluation of competing data sets from one instrument, or potentially multiple instruments, that target the same corners of a nomological network so that they could have contrasted narratives about respondents arising from different latent variable models and data sets. Instead, von Davier, Naemi, and Roberts focused almost exclusively on information-based indices for relative model-data fit assessment, classification error indices, and a few basic correlational results for 2 studies with rather distinct constructs.