Analyzing Maximum Entropy typologies through implicational universals with negated consequents

In recent work, Anttila and Magri have extended the Greenbergian implicational universals to probabilistic phonology and have argued that these universals are unsuitable to analyze ME typologies. In this paper, we introduce a new class of universals that are weaker because they capture what phonological grammars cannot do, rather than what they must do. We then develop the formal theory of these new, weaker, negative universals and show how they can be easily computed for ME typologies. Finally, we present preliminary evidence that these new universals might provide a better tool to analyze ME typologies.