This paper evaluates syntactic models of Jespersen’s Cycle (Jespersen, 1917). Two types of model are examined: Neg-criterion based approaches, and morphosyntactic feature-based approaches. These two types of analysis characterise the parameters involved in Jespersen’s Cycle in different ways. Different morphosyntactic options are in competition within the two syntactic models. So the two analyses make different predictions concerning the patterns of variation and change we ought to find in historical data. Detailed quantitative study of diachronic data from Middle English tests these predictions, and provides an empirical basis to evaluate different structural analyses. The Middle English data provide evidence that two forms of ne should be distinguished within Jespersen’s Cycle. I argue that a morphosyntactic feature-based analysis makes the required distinction, and propose an account of Jespersen’s Cycle based on morphosyntactic features.