Nancy Dorian's foundational work on the loss of Gaelic in the East Sutherland communities continues to provide important insights into the nature of the process of language change in situations of obsolescence. In this article I look at a subset of Dorian's data from the perspective of current syntactic theory, and argue that the connected loss of such apparently different constructions as objects of non-finite verbs, inalienable possessive structures, and a range of passives, and the concomitant restructuring of the grammar, all follow from the interaction between a reduction in agreement features on a functional head and the broad syntactic ecology of the language. This approach makes sense of why these apparently disparate constructions all undergo the particular kinds of change that are seen, changes which are mysterious from the perspective that an obsolescing language should alter to become more like the dominant language (in this case English) which is replacing it.

This article outlines the differences in goals, methods and results that variationist researchers may encounter when exploring and/or documenting a threatened language variety, and underscores special considerations and aspects of the research program that linguists must work to control for when working with endangered varieties of Western languages. In particular, it examines questions and strategies for dealing with sparse data for longitudinal studies; fewer speakers for stratified samples; the inverse relation between linguistic fluency and age; social network constraints in small speech communities; literacy-centric exercises in oral language communities; and larger project protocols designed for stable linguistic communities. Throughout the paper, the collection and analysis of Louisiana French liaison data from 1939, 1977, and 2010 provide an application of the proposed methods.