Questions: Corpus Linguistics - Methodology

Question 1

A corpus is best described as:

• A. Any collection of written texts assembled for research
• B. A large, carefully designed, annotated collection of natural language texts with systematic sampling, quality control, and documented provenance
• C. A database of dictionary definitions
• D. A machine-readable version of a classic literary work

Answer: (B)

A true corpus involves systematic design: representative sampling, standardized annotation, documented source metadata, and size sufficient for statistical analysis. A random collection of texts is not a linguistic corpus. Careful design enables reliable, generalizeable findings.

Question 2

Why is corpus annotation (tagging for parts of speech, parsing for syntax, marking for semantic roles) important in corpus linguistics?

• A. To correct errors in the original texts
• B. To enable systematic searching and quantitative analysis of linguistic structures that are not visible in raw text
• C. Because raw text is uninterpretable
• D. To replace linguistic theory with empirical counting

Answer: (B)

Annotation enables systematic investigation. Raw text is interpretable but annotation allows automated searching for specific structures (e.g., 'all transitive verbs' or 'all relative clauses'), enabling large-scale pattern detection impossible through manual analysis. Annotation supports quantitative methodology.

Question 3

Corpus evidence that word X appears 1000 times more frequently than word Y definitively proves that speakers find word X more useful or psychologically salient.

• T. True
• F. False

Answer: False

Frequency is one clue but not definitive. Genre effects, register, writing conventions, and register-specific norms affect frequency. A word frequent in academic writing may be rare in conversation. Careful corpus analysis considers these confounds. Raw frequency alone cannot establish psychological properties.

Question 4

The use of corpora to study language has mainly confirmed theoretical predictions from introspective linguistics without substantially changing linguistic theory.

• T. True
• F. False

Answer: False

Corpora have revealed substantial gaps in introspective judgments. Collocation patterns, frequency distributions, and register variation were unpredicted by many theories. Corpora have reshaped understanding of grammar, showing many supposed absolute rules are actually tendencies. They've been transformative, not merely confirmatory.

Question 5

Explain why sampling strategy is crucial in corpus design and how poor sampling can lead to misleading conclusions about language.

Model answer: Corpora are meant to represent language use. If sampling is biased toward a particular register, genre, or population, findings may not generalize. A corpus of academic writing alone tells you about academic discourse, not language broadly. Systematic, representative sampling ensures findings are generalizable. Poor sampling creates systematic biases that appear as properties of language when they're actually properties of the sample.

Corpus evidence is only as good as sampling. If you want to study English broadly, random sampling of published English across genres and time periods is necessary. Convenience sampling (e.g., just what's available digitally) introduces systematic bias. Careful attention to sampling prevents misleading conclusions.