UC Berkeley

This dissertation uses quantitative and geographic analysis techniques to examine the historical and geographical processes that have shaped California's linguistic diversity. Many questions in California historical linguistics have received diminishing attention in recent years, remaining unanswered despite their continued relevance. The studies included in this dissertation reinvigorate some of these lines of inquiry by introducing new analytical techniques that make effective use of computational advances and existing linguistic data. These studies represent three different scales of historical change -- and associated geographic patterns -- and demonstrate the broad applicability of new statistical and geographic methodologies in several areas of historical linguistics.

The first of these studies (Chapter 2) focuses on the dialect scale, examining the network of internal diversity within the Eastern Miwok languages of the Central Sierra Nevada foothills. This study uses dialectometric measures of linguistic differentiation and geographic distance models to characterize the dialect geography of this language family and examine how human-environment relations have influenced its development. This study finds three primary linguistic divisions in the Eastern Miwok dialect network, corresponding to Plains Miwok, Southern Sierra Miwok and Northern/Central Sierra Miwok, as well as a number of smaller patterns of regional variation. It also identifies elevation, vegetation, and surface water as influences on the dialect network in the region and establishes the utility of cost distance modeling for studying historical linguistic contact networks in situations where our historical knowledge is limited.

The second study (Chapter 3) evaluates the hypothesis that the small families and isolate languages of California form a few, deep genealogical ``stocks''. While attempts to validate two of these -- Hokan and Penutian -- have not met with widespread approval, the classifications themselves have been adopted widely. This study examines the statistical evidence for such deep, stock-level relationships among California languages by implementing a metric of recurrent sound correspondence and a Monte Carlo-style test for significance. The multilateral comparison involved in the clustering component of this method makes it particularly sensitive to the types of large clusters that might represent "Hokan" and "Penutian" groups. However, this test finds no evidence for such groupings and casts further doubt on the genealogical status of these categories.

The scale of the final study (Chapter 4) is broader both temporally and geographically. Chapter 4 examines the idea that Northern California functions as a linguistic area. Uncertainty regarding the genealogical and contact-related influences on individual languages in the region and links between Northern California and other linguistic areas make it difficult to evaluate existing proposals about the region's areal status based only on the regional similarities such studies offer as evidence. This chapter uses measures of spatial autocorrelation to determine whether the spatial patterns exhibited by individual features and cumulative patterns in the region as a whole are likely to reflect a history of geographic trait diffusion. While there is good evidence for areal feature spread in Northern California, and particularly in the Northwestern California and Clear Lake areas, many of the features that occur in Northern California extend up the Pacific coast and suggest that Northern California may be better characterized as a peripheral part of the better-supported Northwest Coast linguistic area.