On the one hand, inflectional morphology can encode semantic features, such as numerosity or sex; on the other, its has a functional role. The agreement of morphological features disambiguates the relations between constituents in sentence parsing, and reduces processing effort by favoring word predictions (Dye et al., 2017; Wicha et al., 2004). Ideally, those processes should be favored by consistency between form and features. Instead, the inflectional features, as Gender and Number, often show allomorphy and syncretism and only seldom surface in consistent morphophonological values (such as -s for plural nouns in English; from now on referred to as morphemes). We will discuss how the distributions of inflectional features and morphemes can respond to general coding and processing principles. We will also suggest that the encoding of referential information can still partially affect the distributions of inflectional features, even in a fundamentally functional category like inflection. Finally, we will point to some questions that arise from the observed data.