📄 Claudia Lange & Sven Leuckert, Corpus Linguistics for World Englishes, A Guide for Research

Summary

This is a review by Élise Guillemin-Bonniec (Lexis, 2022) of Lange and Leuckert's Routledge guide. The book connects two fields, the study of English varieties worldwide and corpus methods. The authors argue that computers and the internet spread English as a lingua franca and also made large corpora possible.

The guide has eight chapters in three parts:

• the history, theory and terminology of English varieties;
• corpus linguistics, its aims and its tools;
• how corpora shed light on varieties, including learner English.

Each chapter offers keywords, definitions, exercises with worked answers and further reading, so students, lecturers and newcomers to research can all use it.

The authors take a clear position. They reject any ranking of varieties and treat them as different rather than deviant. The reviewer finds the book clear and teachable but disputes some of its claims:

• the colonial explanation it gives for all variation;
• its treatment of learner English as one more variety;
• a paradox in its phrasal-verb example.

The review matters to anyone who builds or uses corpora of non-standard English.

Key points

• The book has 217 pages and eight chapters in three parts. It is non-linear, and each chapter opens with keywords.
• It covers Kachru's three concentric circles (inner, outer, expanding) and his split between genetic and functional nativeness. It notes the model's weakness: one country can contain several kinds of English users, as post-Apartheid South Africa shows.
• It presents Schneider's 2003 Dynamic Model, in which postcolonial varieties move from exonormative to endonormative norms. The final stage is differentiation from the parent variety.
• The learner corpora discussed are ICLE, ICELv2, LINDSEI and ICNALE. They carry error annotation and learner variables such as L1 and years of English study. Pronunciation is left out because the book relies on written corpora.
• In the phrasal-verb case study, learners and speakers of New Englishes use far fewer phrasal verbs than native speakers. The closing caveats warn that a bigger corpus is not automatically more representative and that annotation quality differs between corpora.

Interesting ideas

• I1. Describing a variety as different rather than wrong keeps systematic regional patterns from being counted as noise.
• I2. Learner corpora record speaker and task variables (L1, years of study, task type) next to the text, which lets results be split by population.
• I3. The warning that size does not buy representativeness, and that annotation quality varies by corpus, argues for judging each source on its own merits.
• I4. Halliday's lexicogrammar, as the book uses it, treats lexicon and grammar as one continuum rather than separate layers.
• I5. The reviewer notes that comparing phrasal-verb frequencies against native usage quietly brings back a norm. A frequency gap alone should therefore not be read as a deficit.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (tree): Treebank and dialect-conversion tooling could tag regional constructions with a variety label instead of a repair rule, so the parser keeps them as valid input.
• I2 (learn): Evaluation sets could carry similar source metadata, so parser or reader accuracy can be broken down by text origin rather than reported as one number.
• I3 (learn): Silver data from different sources could be weighted by a measured per-source agreement rate instead of being pooled by volume.
• I4 (frames): Verb classes and argument frames could be bound to whole constructions such as phrasal verbs, not to bare lemmas, so particle choice drives role assignment.
• I5 (frames): A phrasal verb and its single-word equivalent ("go in" / "enter") could map to the same frame, so meaning extraction does not depend on which variety the writer prefers.