A Guide to Build (ING) GLMM Trees in Canadian Maritime English: Part 2, Linguistic Factors

This second paper in a two‐part methodological guide demonstrates how Generalised Linear Mixed Model (GLMM) tree analysis can be used to explore linguistic conditioning in sociolinguistic variation. Building on Part 1, which introduced the dataset and illustrated how GLMM trees reveal social patterning in (ING) variation, Part 2 focuses on the internal linguistic factors governing the alternation between [ɪn] and [ɪŋ] in Canadian Maritime English (CME). Using over 10,000 tokens from more than 300 speakers—the largest single‐dialect (ING) dataset to date—this study shows how GLMM trees accommodate overlapping or collinear predictors such as grammatical category, phonological context, and lexical frequency. The analysis confirms the dominant morphological constraint (verbs favouring [ɪn], non‐verbal forms favouring [ɪŋ]) and identifies additional phonological and frequency effects that emerge only in specific prosodic and lexical environments. These results demonstrate that GLMM trees offer a straightforward, statistically robust way to model complex sociolinguistic data while maintaining interpretability. The paper concludes by comparing hierarchical tree‐based and ‘flat’ regression models, validating the stability of the GLMM trees through cross‐validation, and highlighting how this approach can clarify the interplay of linguistic and social constraints in variationist research.