This dissertation is about the learnability of different generative, Separationist approaches to nominal morphosyntax. The core of my investigation is number, gender, and declension class, as manifested across nouns, adnominals (adjectives, numerals, demonstratives, and quantifiers), and articles. An extreme position would require that all of these categories be based on fixed and innate morphological, syntactic, and semantic representations (Strong Universalism). But this absolutist proposal is untenable for modeling declension-class based patterns of inflectional allomorphy, which I argue should be encoded using representations that are induced from language-particular input. Both of the positions that I model and compare seriously in this dissertation start from this position. Where the two differ is on the status of grammatical gender representations: Is grammatical gender "also" encoded via language-particular, induced features (Distributional Inductivism)?, or are all syntactic features -- including gender -- necessarily grounded in semantics (Grounded Syntactic Features)? I first articulate a detailed proposal for what an Induced Distributional Features Theory looks like architecturally, and how a learner in that style might induce gender and declension class features based solely on formal cues. This is important groundwork for the second goal of this dissertation: to bring out learning consequences as a new playing field for theory comparison between architecturally different theories of the grammar. After all, those theoretical choices are not just constraints for the linguist working in a particular framework as they seek an analysis of some phenomenon in some language or other; these are proposed fundamental constraints on the individual hypothesis spaces of every learner of every language. This dissertation therefore concretizes these theoretical views to propose algorithmically precise learning mechanisms, which I then computationally implement as part of testable learning models. Their comparable performance serves as a proof of concept that grammatical gender (and declension class) systems can be learned without relying on innately provided semantic representations. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://bibliotheek.ehb.be:2222/en-US/products/dissertations/individuals.shtml.]