On the use of hypothesis-driven reduced models in auditory neuroscience

There are a number of detailed models of auditory neurons that are able to reproduce a wide range of phenomena. However, using these models to test hypotheses can be challenging, as they have many parameters and complex interacting subsystems. This makes it difficult to investigate the function of a mechanism by varying just one parameter in isolation, or to assess the robustness of a model by systematically varying many parameters. In some cases, by limiting the scope of a model to testing a specific hypothesis using a particular set of stimuli, it is possible to create a reduced mathematical model with relatively few, independent parameters. This has considerable advantages with respect to the problems above. In particular, if a certain behavior is robust and does not depend on finely tuned parameters, then different implementations are more likely to produce the same results—a key property for reproducible research. In addition, the code for these models is typically simpler and therefore more readable...