Abstract

On the hierarchical picture of models, theoretical models are constructed on the basis of theory and assessed by comparison to distinct models constructed from empirical data. Using the determination of the structure of the folded polypeptide chain as a case study, I instead argue that information from theory and data alike can be interpreted as constraints in the construction of models of information. On this view, more reliable information ought to be prioritized, sometimes forcing reinterpretations of less reliable information; information from theory and data is thus interdependent. I show how the reliability of information can be assessed, arguing that the evidence for a planar peptide bond was stronger and more secure than the evidence for a repeating subunit every 5.1 Å. I further show how models are assessed alongside interpretations of information in a coherentist manner: a better model accommodates more information, particularly reliable information; a model’s inability to accommodate some information necessitates reinterpreting that information.