Comparing Constraints on Contraction Using Bayesian Regression Modeling

Area of Science:

• Sociolinguistics
• Phonological Variation
• Grammatical Theory

Background:

• Contraction patterns in English verbs like 'is' and 'has' are complex.
• Existing sociolinguistic models often simplify multi-way variation.

Purpose of the Study:

• To identify factors influencing 'is'-contraction in American English.
• To compare these factors with those affecting 'has'-contraction.
• To propose a grammatical model for variable forms based on shared contraction patterns.

Main Methods:

• Bayesian Markov chain Monte Carlo (MCMC) modeling was employed.
• Regression modeling was used to analyze non-binary response variables.
• Quantitative analysis of sociolinguistic variation in verb contractions.

Main Results:

• Novel predictors for 'is'-contraction were identified.
• Many contraction patterns for 'is' were also observed for 'has'.
• Ternary variation in 'has'-contraction is not adequately explained by sequential binary choices.

Conclusions:

• The quantitative patterns of 'has'-contraction support a model with three distinct, co-varying grammatical forms.
• Shared contraction patterns between 'is' and 'has' offer insights into grammatical representation.
• Findings have implications for sociolinguistic methodologies and theories concerning ternary variables.