In Northern Iroquoian languages, a nominalizer (nlzr) is typically required to transform a verb into a noun, either for noun incorporation or to create a full DP. In some cases, the nominalizer is required only for noun incorporation and not for the formation of a DP. Interestingly, the converse is never found. That is, there are no lexical roots that require the nominalizer for the formation of a DP, but not for noun incorporation. With this asymmetry in mind, we examine the categorial properties of roots in Northern Iroquoian. We discuss three common theories of the categorization of roots: (i) the traditional theory, in which all roots are specified as nouns or verbs (or adjectives for languages that have this category), (ii) the Bare Root Hypothesis, in which all roots are acategorial, and (iii), the Roots as Nouns Hypothesis, in which all roots are nouns. We show that the Northern Iroquoian facts are not amenable to any of these theories. We propose instead that some roots in Northern Iroquoian are categorially specified (some as nouns, some as verbs), while others are truly bare.