This paper presents a series of reference game experiments (Frank and Goodman, 2012) and fits the results to a number of Bayesian computational models in order to explore the role of linguistic and perceptual bias in iterative pragmatic reasoning. We first discuss the modeling choices made by Franke and Jaeger (2016) and others who have used similar frameworks to model reference game tasks. We introduce a space of different plausible Bayesian models based on this work, and compare models’ fit to new experimental data to replicate the basic findings of Franke and Jaeger regarding the strong role for perceptual salience (e.g., the primacy of color over shape as a differentiating property for possible referents) and linguistic category (e.g., a preference for nouns over adjectives) in pragmatic reference resolution. We then uncover an additional possible effect of what we call labeling, whereby a hearer may simply ignore non-salient, non-differentiating semantic properties.