Against Richness of the Base: Evidence from Nganasan

Abstract

Since Optimality Theory is a highly output-oriented grammatical theory, the strongest hypothesis is that all systematic, language-particular patterns are the result of output constraints, and that there is no other place from which such patterns can derive. In particular, input is not a level of derivation that can be constrained. This principle is known as Richness of the Base hypothesis, and it states that there are no constraints on the input structure of words, and that all linguistic constraints are statements on the surface structure only. In other words, Richness of the Base attributes all systematic phonological patterns to constraint rankings, not to difference in inputs. In this paper, I consider some consonant gradation facts from a Uralic Samoyedic language Nganasan, and argue that (at least the strict interpretation of) the Richness of the Base hypothesis runs into problems when we deal with full range of relevant data from this language, namely isolated words, compounds, and borrowings.