ABSTRACT
How should one decide between two accounts of a psychological phenomenon such as the word superiority effect? This question is at the heart of the scientific enterprise, regardless of whether verbal theories are being tested in experimental settings or computational models are being evaluated in simulations. Psychologists have proposed selection criteria that reveal a range of dimensions on which model evaluation is thought to be important, dimensions that make the task of model selection formidable (e.g., Jacobs & Grainger, 1994, for a review). Falsifiability or testability (Popper, 1959) is a necessary precondition for testing a theory, yet determining whether there are potential outcomes incompatible with the theory is not always straightforward. Simplicity, the ability to capture a phenomenon in the least complicated way (e.g., with the fewest parameters) should be favored (Jeffreys, 1957). A theoretical explanation must be reasonable and consistent with established findings (explanatory adequacy) and must provide a good description of the data (descriptive adequacy). In addition, a theory must generalize correctly to novel situations.
