While many methods exist to measure the policy positions of political parties in multi-party settings, reliable methods to measure the policy positions of individual legislators from text remain relatively undeveloped. This paper addresses several outstanding methodological issues for text scaling models that distinguish them from voting models and more generally those based on item response theory assumptions. These are: the nature of the text scaling models and their assumptions about how political words are generated, how to effectively represent uncertainty in position estimates, and how to add extra information to influence text-based position estimates.
