Refinements in Tabular Models of Variation in Regional American English

Gordon R. WoodCOLING 1973

TL;DR

This paper introduces a refined tabular model for analyzing regional vocabulary variation in American English, moving beyond simple word counts to a more dynamic, systemic view of word choice. By employing computer-assisted analysis of a large dataset, the authors demonstrate how to represent the relationships between synonyms and their regional preferences, offering a more nuanced understanding of dialectal differences.

What Problem Does This Paper Address?

Traditional studies of regional American English, like those by Kurath and Atwood, relied on tabular presentations of word usage. While these methods effectively cataloged the presence and frequency of specific words in different localities, they offered a static snapshot. The challenge was to develop a model that could capture the relationships between word choices within a dialect and represent the systemic nature of vocabulary variation, rather than just isolated word occurrences. The authors aim to create a more dynamic and implicationally richer model for understanding how regional vocabulary functions.

Methods

The core methodology involves computer-assisted analysis of a large dataset of regional vocabulary choices. The process begins with a printed questionnaire asking participants to select synonyms for specific concepts. This data is then coded and transferred to punch cards for computer processing.

Key methodological refinements include:

  • Computer Sorting: Initial sorting by demographic factors (age, education, sex) and then by geographical location (county-by-county).
  • Tabular Representation: Moving from qualitative descriptions (e.g., "common," "rare") to quantitative counts of synonym choices.
  • Systemic Modeling: The crucial refinement is the development of a "closed model" where the entire vocabulary is treated as a system. This allows for the analysis of the relative importance and preference of each synonym within a set, not just its absolute frequency.
  • Percentile Calculations: To facilitate comparison and highlight relative importance, the data is transformed into percentile calculations, showing the proportion of choices for each synonym within a given region or sub-region.
  • Sub-Pattern Analysis: The data is further broken down into four sub-patterns to examine relationships between adjacent states, within states, between north/south and east/west halves of states, and within local areas.

Datasets

The primary dataset used is derived from the work of the proposed Linguistic Atlas of the United States and Canada, specifically focusing on the southern United States. This includes data from:

  • Tennessee
  • Florida
  • Alabama
  • Mississippi
  • Louisiana
  • Arkansas
  • Oklahoma
  • Georgia (included as an eastern anchor)

Texas was excluded due to the near completion of Atwood's survey at the time. The dataset consists of responses from approximately 1,000 individuals, with each response coded for biographical data and synonym choices from a printed questionnaire.

Key Insights and Ideas

  • Vocabulary as a System: The most significant insight is the conceptualization of regional vocabulary not as a collection of independent words, but as a closed system where the choice of one synonym affects the perceived importance or preference of others.
  • Dynamic Representation: The refined model allows for a more dynamic representation of dialectal variation, showing how synonyms compete and coexist within a region.
  • Relative Importance Matters: Simply knowing a word is present is less informative than understanding its relative preference compared to other synonyms for the same concept. This is crucial for identifying subtle regional distinctions.
  • General vs. Regional Words: The analysis confirms that generally preferred words often outweigh more regionally specific ones, but it also reveals that even general words exhibit variations in regional preference.
  • Refining Dialect Boundaries: The computer-assisted analysis and mapping of these refined data allowed for a redefinition of technical terms like "Midland" and "Southern" dialect as they apply to the studied region.

Key Results

The paper presents a methodology for generating detailed tabular data that displays the presence and relative preference of synonyms across different regions. An example is provided for the concept "a woman who helps at childbirth," showing the percentage of preference for "granny woman" and "midwife" in different halves of Tennessee and Georgia.

  • The tabular extracts demonstrate how the model can highlight regional preferences, such as east Tennessee showing a stronger preference for "granny woman" and "midwife" compared to west Tennessee.
  • The analysis confirms that general words are often preferred over regional ones, but also reveals that general words themselves have varying regional preferences.
  • The computer printouts of percentile calculations served as the basis for detailed maps that aided in redefining dialectal boundaries.

Limitations and Future Work

  • Limitations:
    • The authors acknowledge that a search at a different time, with a different questionnaire, or among different people would shift the entries, though it would not alter the fundamental systemic design.
  • The original intention to sort by age, education, and sex was partially superseded by the need to create a manageable publication format.
  • The sheer volume of raw data from 1,000 people and 1,000 words made full publication of the detailed county-by-county listings impossible.
  • Future Work:
    • The paper implies that further application of this refined model to other regions and concepts could lead to a more comprehensive understanding of American English dialectology.
  • The authors suggest that dialectology must consider the characteristic of regional preference variation in general words, alongside the presence of distinctively local words.

Summary updated

BibTeX
@inproceedings{wood1973refinements,
  author = {Gordon R. Wood},
  title = {{Refinements in Tabular Models of Variation in Regional American English}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1973},
}
Figure