Advanced English II (BFE19013)

Corpus Critical Discourse Analysis

Instructor: Miao Zhang 張淼

Using corpus to do CDA

The playbook we are replicating

  1. Context first — what do we already know about the topic?
  2. CL map — collocation → concordance
  3. Downsample — representative texts → a small sample
  4. CDA close read — the five DHA strategies
  5. Iterate — each pass challenges the other

(O’Halloran’s procedure in miniature.)

The synergy

→ what did the numbers reveal that close reading would have missed?

← what did the close reading reveal that the frequencies hid?

Series A

State of the Union: how presidents talk about “we/us” and “they/them”

State of the Union addresses (SOTU) are a rich, longitudinal corpus for studying presidential discourse.

Corpus note: SOTU

  • SOTU: one speech per file, 228 speeches
  • The filename is your metadata: 1947_Truman.txt = year + president
  • Change over time shows up as the year in the concordance’s File column

Topic 1: Constructing the Other

Question: How have SOTU addresses nominated and predicated out-groups — Nazis -> communists/Soviets → terrorists → China, illegal immigrants?

Topic 1: method

  1. Concordance of the target terms — lines arrive in file order, so the years march forward on their own. The Concordance Plot shows it even faster: hit density per speech, 1790 → today
  2. Downsample to 2–3 speeches where the term clusters → the five CDA strategies

Topic 2: “We the people”

Question: Who is included in we / us / our — and who is excluded?

Method:

  1. Concordance of we / us / our / they / them
  2. Quantify inclusive vs. adversarial uses
  3. Close-read where “we” does argumentative work — promising, blaming, uniting

Topic 3: Topoi and metaphors of crisis

Question: Which topoi (danger, threat, burden, humanitarianism, history/legacy) recur — and through which metaphors (war, journey, disease, disaster)?

Topic 3: method

  1. Key words: one president vs. the rest of SOTU (reference = the whole folder)
  2. Group the key words into semantic fields (war? economy? health?)
  3. Concordance → metaphor identification

Topic 3: anchor & payoff

Anchor: Baker et al.’s key words + semantic fields — their water metaphors (pour, flood, stream) dehumanized RASIM.

Your version: which metaphor family does a crisis speech ride on?

Topic 4: The “style of power”

Question: Which lexical patterns persist across all presidents, regardless of topic?

Method — replicate c-collocates manually:

  1. Pick 3–4 presidents spanning the timeline; collocates of America / American for each
  2. Keep only those present in ≥ X of Y presidents (set the threshold; justify it!)
  3. What survives the filter = the enduring style of power

Topic 4: setting the threshold

Why a threshold at all? Baker et al. kept collocates appearing in at least 7 of 10 yearly sub-corpora to filter out “seasonal” hits.

Your version: 3 of 4 presidents. What’s your rule — and why?

Topic 4: anchor & payoff

Payoff: Fairclough, quoted in Baker et al. §5 —

“A single text on its own is quite insignificant: the effects of media power are cumulative, working through the repetition of particular ways of handling causality and agency, particular ways of positioning the reader, and so forth.”

Our takeaway: the frame outlives every speech.

Break

Series B

State of the Union, continued — the DHA strategies we haven’t spent yet

Topic 5: Reporting other voices

Question: When presidents quote critics, “some say,” opponents — whose viewpoint, and how fairly?

Method:

  1. Concordance: say / said / critics / some argue / my opponents
  2. Classify each hit: ally quoted / opponent quoted / strawman?
  3. Close-read: does the quote do the attacking?

Topic 5: anchor & payoff

Anchor: DHA strategy 4 — perspectivation / discourse representation (Baker et al. 2008, Table 1, p. 280).

Payoff: reporting a critic’s words lets a president demolish a position while staying above the fight.

Topic 6: Nominalization and agency

Question: Where do speeches hide actors (“mistakes were made,” “the deficit grew”) vs. assign them?

Method:

  1. Concordance of nominalizations + agentless passives (decision, growth, relief; was done, were lost)
  2. Classify each hit: agent present / absent / recoverable
  3. Discuss the legitimation effects: who is responsible — anyone?

Topic 6: anchor & payoff

Payoff: the grammar of evasion, spotted at scale.

Baker et al.’s water metaphors work the same way — RASIM constructed as an “out-of-control, agentless… natural disaster.”

Topic 7: SOTU vs. academic writing

Question: When presidents and academics write about similar ground — economy, education, health, security — how does the discourse differ?

Corpus note: BAWE

  • BAWE = British Academic Written English: 2,761 university texts, ~6.5M words, across disciplines
  • Local copy: Corpora/BAWE_txt/ (plain text, tags stripped) — load it as the reference corpus
  • Licensed for research/teaching — do not redistribute the files

Topic 7: method

  1. Keyness: target = SOTU (or one era folder), reference = BAWE → the presidential signature
  2. Flip on negative keywords → the academic signature
  3. Group both lists into semantic fields (reuse Topic 3’s method)
  4. Pick 2–3 words the two corpora share on a topic → concordance both: what changes?

Topic 7: anchor & payoff

Anchor: the same Baker et al. keyness logic — but the lens now separates registers, not eras.

Payoff: the contrast is the finding. Where SOTU says we / will / must and argues by metaphor, BAWE hedges, nominalizes, and argues with however / therefore / evidence.

Scaffolding

Two sessions, one report

Session A — the CL pass (50 min)

  1. Claim your topic (groups of 3–4: facilitator / recorder / presenter); RQ on the board — 5 min
  2. Build your map: keyness / collocates / concordance — 20 min
  3. Build your category table — 15 min
  4. Downsample: pick 2–3 speeches; one sentence on why these — 10 min

✅ Checkpoint: show the instructor your category table.

Break

Session B — the CDA pass (50 min)

  1. Close-read your sample with the five DHA strategies (handout: Baker et al. 2008, Table 1) — 25 min
  2. Compare passes: where do numbers and close reading agree? Where do they fight? — 10 min
  3. Draft your one-page synergy report — 10 min
  4. Rehearse a 2-minute version → your Sessions 10–11 presentation — 5 min

The deliverable: 1-page synergy report

Blocks 1–3:

  1. Research question + corpus slice (data, president/terms searched, size)
  2. The quantitative map — category table + one pattern
  3. The close reading — one passage, analysed with the five DHA strategies

The deliverable: block 4 — the synergy

One finding in each direction:

→ what did the numbers reveal that close reading would have missed?

← what did the close reading reveal that the frequencies hid?

Assessment

Where this practice leads

Course component Weight How this series feeds it
Participation 15% Session A/B checkpoints + synergy report
Group presentations (Sessions 10–11) 30% The 2-minute rehearsal becomes your full presentation
Essay 40% Your report = a ready-made pilot study: question, method, finding

Your presentation paper does the same loop

Hardaker & McGlashan (2016), “‘Real men don’t hate women’: Twitter rape threats and group identity” — your Sessions 10–11 reading.

They triangulate frequency, collocation, and keyness to find the discourse, then close-read to see how identity is constructed.

Your synergy report is the same design on a different corpus.

Wrap-up: what you practiced

  • The Baker et al. loop runs on any corpus: speeches, tweets, news.
  • Nomination, predication, topoi, perspectivation, intensification — the same five strategies see all of them.
  • The reference corpus is a lens: Topic 7 split registers — SOTU against BAWE showed what “presidential” language is.
  • CDA framing: social practice, not score-keeping — critique the discourse, not the speaker.

References (1/2)

  • Baker, P., Gabrielatos, C., KhosraviNik, M., Krzyżanowski, M., McEnery, T. & Wodak, R. (2008). A useful methodological synergy? Combining critical discourse analysis and corpus linguistics to examine discourses of refugees and asylum seekers in the UK press. Discourse & Society 19(3): 273–306.

References (2/2)

  • O’Halloran, K. (2010). How to use corpus linguistics in the study of media discourse. In A. O’Keeffe & M. McCarthy (eds.), The Routledge Handbook of Corpus Linguistics (pp. 563–577). Routledge.
  • Hardaker, C. & McGlashan, M. (2016). ‘Real men don’t hate women’: Twitter rape threats and group identity. Journal of Pragmatics 91: 80–93.
  • Nesi, H. & Gardner, S. (2012). Genres across the Disciplines: Student Writing in Higher Education. Cambridge University Press. (BAWE corpus.)

Thank you

Save your tables, screenshots, and corpus slices — bring them to Sessions 10–11.

Advanced English II