Psychometric models rely on two foundational but often untested assumptions: the causal structure underlying latent variables and the numerical level at which these variables are measured. This project evaluates these assumptions through two complementary simulation studies. The first study examines whether traditional fit indices used in factor analysis, such as the CFI, TLI, RMSEA, and SRMR, adequately reflect the causal constraints implied by latent variable models. We create and assess a new fit index, which tests causal structures that are expected under correctly specified factor models. The second study addresses the numerical representation of latent variables by comparing additive and ordinal measurement structures. Together, these studies provide an integrated framework for assessing both causality and measurement-level assumptions in psychometric modeling, offering theoretical and practical tools that support more rigorous evaluation of latent constructs.