In a series of articles and comments, Kenneth Bollen and his collaborators have incrementally refined an account of structural equation models that (a) model a latent variable as the effect of several observed variables and (b) carry an interpretation of the observed variables as, in some sense, measures of the latent variable that they cause. (For simplicity I follow this use of the term measurement with reservations described by Markus & Borsboom, 2013, Section 1.1.1.) Bainter and Bollen (this issue) contributed further to this line of research by providing a simulation study to expand and illustrate the argument offered by Bollen (2007) incorporating the distinction between causal indicators and composite indicators, offered by Bollen (2011) and Bollen and Bauldry (2011). Bainter and Bollen made a welcome contribution by addressing concerns about interpretational confounding and instability (Edwards, 2011; Edwards & Bagozzi, 2000; Howell, Breivik & Wilcox, 2007). Nonetheless, I will advance the following claims: 1. Bainter and Bollen underestimated the criticism that they intended to address. 2. The central notion of "conceptual unity" remains insufficiently clear. 3. Measurement requires a perspective broader than one limited to structural equation model specification.