Abstract
Given the broad range of frameworks for syntactic theorizing within generative linguistics the question sometimes arises as to whether philosophical, or best theory, criteria could help us choose between those frameworks. For example, some frameworks are claimed to be more conceptually simple, others are claimed to be more formally rigorous, and some are claimed to be better grounded empirically or conceptually. Such cross-theoretical considerations need to be handled with care. Consider the question of whether some theories handle more data or better data or are more “driven by data” and are thus more empirically sound. This quickly leads us to the question of what data is important and what relation it bears to theory choice. I argue that there is no atheoretical notion of best or most important data - whether data is interesting or important depends upon the theoretical questions that we are investigating and what the ultimate goal of our work is. Furthermore, I argue that data only really exists in the context of some theoretical framework, and thus claims of data-driven theorizing and atheoretical or framework-free data gathering needs to be viewed with some skepticism. In addition, I consider the claims made about simplicity and formal rigor, and then offer a positive proposal for cross-theoretical evaluation: the best theory is the one that enables the theorist to tackle problems of interest with the most efficiency, and I argue further that the choice of best theory is a decision that is best left up to the individual theorist. Cross-theoretical superiority claims involving simplicity and formal rigor are vacuous unless understood as claims about ease of use, or alternatively as tools to facilitate reduction to a more basic science or scientific unification generally.
