The critical aim of this dissertation is to show the lack of explanatory value of typological generalizations in generative research paradigms, and the constructive aim is to propose an alternative conception of typology which gives a justifiable place to typological facts. My contention is that we cannot conclude that the human language faculty (HLF) lacks the means to generate a linguistic phenomenon from only the lack of such a phenomenon in the languages of the world. The temptation to do so arises from equivocation regarding the term Language as used within different generative paradigms: the classical generative paradigm, and the generative-parametric paradigm. The former characterizes Language, understood as HLF, the mental object which allows us to produce and understand languages. For the latter, however, Language also includes the distribution of linguistic structures in the world. HLF is a natural kind; the distribution of linguistic structures in the world is not. Equivocation of the term ‘Language’ occurs when one notion is exchanged for the other within an argument. The problem: only natural kinds support induction. The goal of characterizing HLF is discovering what is necessarily true of HLF. The distribution of linguistic phenomena in the world, although constrained by what HLF allows us to acquire, is also constrained by historical contingency. Generalizations based on these accidental factors are valueless in characterizing HLF: I show this in two case studies, which deal with syllable structure and verbal morpheme order. I argue that the study of the distribution of linguistic phenomena in the world is a historical science, which requires a different set of assumptions than an experimental science such as the classical generative paradigm. The alternative I offer is called ideal-typology. Ideal-typology replaces inductive inference based on natural kinds with pragmatic explanation based on ‘ideal-types’. Ideal-types are convenient fictions, purpose-built to manipulate our cognitive systems into understanding the diversity of historical-scientific data. I illustrate the practice of ideal-typology by showing how the diversity of Chinese tone systems can be measured and organized by the use of ideal-types. Beyond increasing understanding of the data themselves, ideal-typology yields hypotheses that experimental sciences can test.
More detail can easily be written here using Markdown and $\rm \LaTeX$ math code.