Exam blueprint
Structural constants confirmed across 2015 and 2018 papers
I
Vocabulary · 20 pts
II
Grammar · 20 pts
III
Error ID · 20 pts
IV
Cloze · 20 pts
V
Reading · 20 pts
Section anatomy fixed format · confirmed both years
Section I — 10 single-word vocabulary items in sentence context. 2 pts each = 20 pts. All four options are grammatically valid; discrimination is purely collocational.
Section II — 10 grammatical phrase/structure items. 2 pts each = 20 pts. Tests prepositions, conjunctions, relative clauses, conditionals, causative verbs, and comparatives.
Section III — 10 paragraphs, each with 4 underlined segments (A–D). One segment is ungrammatical. 2 pts each = 20 pts. Errors are syntactic or morphological — never lexical.
Section IV — One continuous academic passage, 10 numbered blanks, closed word/phrase list. 2 pts each = 20 pts. Blanks target discourse connectors, idiomatic verb phrases, and aspectual adverbs.
Section V — Two reading passages (Part I + Part II), 5 comprehension questions each. 2 pts each = 20 pts. Question sequence: factual detail (Q1–2), inference (Q3–4), global TRUE statement (Q5).
Core skill per section non-interchangeable
Each section tests a distinct cognitive operation. Applying the wrong approach to the wrong section is the most common structural error.
| Section | Core skill | What does NOT help |
|---|---|---|
| I | Collocational precision — which word bonds with which in natural English | Synonym knowledge, grammar rules |
| II | Structural precision — which grammatical form fits a specific syntactic slot | Vocabulary meaning, general reading |
| III | Error detection — identifying one syntactic/morphological flaw in a paragraph | Checking for wrong word meaning |
| IV | Discourse coherence — filling blanks to maintain logical flow across a passage | Item-by-item vocabulary guessing |
| V | Inferential reading — distinguishing supported claims from plausible-sounding ones | Speed-scanning for matching words |
Section I · vocabulary precision
10 items · 2 pts each · collocational knowledge is the only discriminator
Question formula
The stem provides a complete sentence with a syntactic slot. All four options are grammatically valid — the stem never filters by grammar. The only discriminator is the collocational bond between the correct word and its anchor in the sentence.
Four confirmed item subtypes
Type A · near-synonym filter
All 4 options share a broad semantic field. Correct answer has the tightest collocational bond to the anchor noun or verb. E.g. "symptoms / diagnoses / medicines / patients" — "symptoms of hay fever are a runny nose…" anchors to "symptoms."
Type B · register filter
Options span formal/informal or domain-specific registers. Correct answer fits the sentence's register and domain. E.g. "tangible" (formal adjective collocating with abstract nouns like "tension") vs concrete alternatives.
Type C · intentionality filter
Adverb options differ only in whether the agent intends the action. Context always specifies. "Keenly aware" = deliberate attention. "Inadvertently aware" = accidental noticing. The sentence's narrative voice signals the correct one.
Type D · fixed phrase
Correct answer is part of a fixed collocation or phrasal verb. Distractors break the idiom. E.g. "figure out how to," "in advance," "advent of genetic engineering," "take issue with."
Operational framework — 4 steps
1
Isolate the collocational anchor before reading the options — the noun, verb, or adjective in the sentence that the blank must bond with. Write your own best answer first.
2
Test intentionality for adverb blanks — does the agent know/intend the action? The surrounding context (narrative tense, subject's role) always signals this unambiguously.
3
Check fixed-phrase status — if the blank is flanked by a specific preposition or particle, a fixed phrase is required. Eliminate options that break the idiom's structure, not just its meaning.
4
Never select by elimination alone — if three options are clearly wrong, still verify the remaining one genuinely fits the specific collocational context before marking it.
Section II · grammatical structures
10 items · 2 pts each · structural precision in context
Recurring grammar targets confirmed 2015 + 2018
Both papers draw from the same finite pool of structures. The exam does not test obscure grammar — it tests high-frequency structures at their precision-demanding edge cases.
Relative pronouns in restrictive clauses — "whoever comes," "where salt is deposited," "those of France and Austria." Error type: wrong pronoun for its grammatical role (subject / object / possessive / locative).
Conditional inversions without "if" — "Had it not been for your advice," "Should you be unable to attend." The exam tests recognition of inversion as a conditional signal. Distractors offer "Could" or "Will" which cannot invert into conditionals the same way.
Causative constructions — "have it cashed," "had his watch repaired." Tests active vs passive causative form: have + object + past participle = passive causative. Distractors corrupt the participle form or introduce an infinitive.
Expect / want + infinitive complement — "I expect you to remember to finish." Distractors swap gerund for infinitive, or conflate the two senses of "remember" (remember + gerund = past memory; remember + infinitive = future obligation).
Comparative reference: "those / that of" — "less densely equipped than those of France and Austria." Tests whether the comparison referent is singular (that) or plural (those) and requires a stand-in pronoun to avoid repeating the head noun.
Passive + preposition collocate — "is referred to as a founder." Distractors drop the required preposition, use the wrong auxiliary, or produce an active form where a passive is needed.
Discourse questions and exclamatives — "How come you are leaving so early?" vs "Why" / "Why not" / "How about." Tests fixed pragmatic phrases that cannot be substituted by near-equivalents in informal registers.
Operational framework — 4 steps
1
Identify the grammatical category of the blank — conjunction, pronoun, preposition, verb form, or phrase structure. This constrains the answer before reading any option.
2
For pronoun blanks: determine the grammatical role in the clause — subject → who/whoever; object → whom/that; possessive → whose; locative → where.
3
For verb-form blanks: check whether the main verb's complement takes infinitive or gerund, and whether the structure is causative-active or causative-passive (have + obj + past participle).
4
For conjunction blanks: map the logical relation (concession, condition, sequence, contrast, reason) and eliminate options that produce the wrong semantic relationship between the two clauses.
Section III · grammatical error identification
10 paragraphs · 1 error each · the 2018 answer key reveals the correction, exposing the exact error taxonomy
Complete error taxonomy derived from 2018 corrections
The 2018 answer key provides both the error location (A/B/C/D) and the correction, making the error type unambiguous. Seven types cover 100% of observed errors across both papers.
| Error type | Example (2018) | Rule |
|---|---|---|
| Wrong article | "the walk" → "a walk" | Generic, non-specific actions take the indefinite article. "The" presupposes a specific referent already established. |
| Verb tense | "win" → "won" | Narrative past-tense paragraphs require consistent simple past. Present tense verbs in past context are errors. |
| Stative vs dynamic verb | "be died" → "be dead" / "have died" | "Die" is dynamic (action); "be dead" is stative (state). "Be died" combines a stative auxiliary with a dynamic participle incorrectly. |
| Noun number | "robot" → "robots"; "a faces" → "a face" | Countable nouns used generically require plural with no article. Singular with article requires morphological agreement. |
| Wrong preposition in idiom | "in the front lines" → "on the front lines"; "over performances" → "into performances"; "to" → "than" in comparative | Fixed prepositional idioms do not allow substitution. Comparatives require "than," not "to." |
| Passive morphology | "is borrow" → "is borrowed" | Passive construction requires the past participle form (-ed / irregular). Present form after "is/are/was/were" is always wrong. |
| Verb complement | "struggle to engagement" → "struggle to engage" | After "struggle to," a bare infinitive is required. A noun or gerund following "to" in this pattern is ungrammatical. |
Operational framework — 4-pass method
1
Pass 1 · verb scan — examine every underlined verb for tense consistency with the paragraph's time frame, correct active/passive form, and whether the participle is correctly formed.
2
Pass 2 · noun scan — examine every underlined noun for article correctness (a/the/zero) and number (singular/plural). Generic countable nouns take plural with no article; specific singular nouns need "the" or "a."
3
Pass 3 · preposition scan — examine every underlined preposition against its governing verb, noun, or fixed phrase. Specifically check "in/on" for positional idioms and "than/to" for comparative structures.
4
Pass 4 · complement scan — examine every verb-complement relationship: to + infinitive vs -ing; stative vs dynamic verb choices; adjective vs participle after copular verbs.
Hard constraint: The error will never be a wrong word meaning (lexical error). If your candidate error is "this word means something slightly different," keep scanning. The error is always syntactic or morphological.
The complexity-distraction trap
Options B, C, and D often contain syntactically complex structures — subordinate clauses, inversions, appositives — that look suspicious due to their density. This draws attention away from the simpler, more invisible error, which frequently sits in option A or B in a function word (article, preposition, auxiliary). The brain autocorrects familiar phrases at speed. Slow down on prepositions and articles specifically.
Section IV · cloze passage
10 blanks · closed word list · discourse coherence is the primary skill
What the cloze actually tests
Unlike Sections I and II where each item is self-contained, Section IV blanks are interdependent — a wrong choice at blank 2 may not be detectable until blank 6. Both papers use academic or intellectual texts (2015: history of linguistics; 2018: English grammar rules and adjective ordering). The blanks target discourse connectors, idiomatic verb phrases, and aspectual/temporal adverbs — not vocabulary meaning.
Confirmed blank types
Logical connector — concession, addition, contrast, condition. E.g. "even though," "because," "however." Wrong answers choose a connector with plausible broad meaning but the wrong logical relation to surrounding sentences.
Idiomatic verb phrase — "points one out," "lay out the rule," "take issue with," "come as a surprise." Distractors are near-synonyms that break the specific idiom. The collocational bond governs, not general meaning.
Temporal/aspectual adverb — determines the precise time relationship or logical condition between two clauses. Wrong answers change the implied time frame or conditionality of the claim.
Pronoun / quantifier — "nobody can say why" (2018 Q5-C). Tests whether the logical scope is universal, particular, or negative. Distractors offer "everybody," "somebody," or "anybody" — only one fits the logical content.
Operational framework — 4 steps
1
Read the full passage once without filling any blanks — establish the main argument and the author's logical position. Every passage has a clear thesis that governs which logical connectors are appropriate.
2
Identify the discourse function of each blank — is it connecting two clauses (connector), naming an action (verb phrase), specifying scope (quantifier), or indicating time/degree (adverb)?
3
Check right context as carefully as left context — many blanks are governed by what follows. E.g. "take ___ with" is governed by "with" appearing later; "despite / although" is governed by whether the following clause is a noun phrase or a full clause.
4
Re-read the completed passage — verify it flows as one coherent argument. A discourse mismatch (contradiction, illogical sequence, broken idiom) signals an error in a preceding blank.
Section V · reading comprehension
2 passages × 5 questions · predictable question sequence per passage
Question sequence formula consistent across both papers
Q1–2 · factual detail
Stem maps to a specific passage sentence. Correct answer paraphrases, never copies. Distractors use passage wording but distort the meaning, switch the referent, or describe a different detail from the same paragraph.
Q3–4 · inference
Answer is not stated directly. Correct answer follows from combining two or more passage sentences. Distractors are true in isolation but non-inferential, or are plausible-sounding but absent from the text entirely.
Q5 · global TRUE statement
Four statements, three of which contradict, go beyond, or reverse the passage's claims. Correct answer is explicitly supported by passage content. Distractors frequently use stronger language than the passage does.
Occasional visual Q
Appears in some years (2015 Q4 on tally marks). Requires precise mapping from a text description to a visual format. The passage always fully specifies the correct diagram — no inference needed, only careful literal reading.
Operational framework — 4 steps
1
Read the questions before the passage — all 5 questions. Identify which paragraph likely addresses each one. This directs your reading attention before you start.
2
For factual questions: locate the specific passage sentence, then test each option against it. Eliminate any option that adds information not present in the passage, regardless of how plausible it sounds.
3
For inference questions: identify the two or more sentences whose combination produces the answer. If an option is supported by only one sentence, it is almost certainly a distractor — inference requires synthesis.
4
For TRUE/FALSE global questions: treat each option as a testable claim. Find passage support for each one individually. Immediately eliminate any option that uses language stronger (more absolute) than the passage uses.
Distractor architecture
How wrong answers are engineered — every pattern confirmed across both papers
Section I + II distractors
The near-synonym trap. All options share a broad semantic field. Only collocational bond or fixed-phrase knowledge eliminates the distractors. E.g. "diagnoses / medicines / patients / symptoms" — all illness-related, but only "symptoms" collocates with "of hay fever are a runny nose, watery eyes, and a headache."
The plausible-but-defective grammar trap. In Section II, an option is semantically sensible but grammatically broken in the specific syntactic environment. E.g. "A · has refers" appears at speed to be a present perfect, but is a hybrid non-form. Reading for meaning will miss this — only reading for structure catches it.
The partial-phrase trap. An option contains part of a required fixed phrase but not all of it. E.g. "of France and Austria" looks like a complete comparative reference, but the correct answer requires "those of France and Austria" — the pronoun stand-in is mandatory to make the comparison grammatical.
The wrong-complement trap. For verbs that take either infinitive or gerund depending on meaning (remember, try, stop), the distractor exploits the less-familiar sense. E.g. "remember finishing" (past memory) vs "remember to finish" (future obligation) — the sentence context determines which sense is required.
Section III distractors
The complexity-distraction trap. Options B, C, and D frequently contain syntactically complex structures (inversions, subordinate clauses, appositives) that appear suspicious due to density. This draws attention away from the simpler error in option A or B, which sits in a function word the eye passes over.
The autocorrect trap. The brain autocorrects familiar phrases when reading for meaning. "In the front lines" reads fluently despite "in" being wrong (correct: "on the front lines"). "Go for the walk" feels natural despite the article being wrong (correct: "go for a walk"). Slow down specifically on articles and prepositions.
Section V distractors
The true-but-irrelevant trap. The option is factually supported by the passage, but it does not answer the specific question asked. This is the most common distractor type in Q1–2. Always verify the option answers the question, not just that it is passage-consistent.
The extreme-language trap. The distractor uses absolute language ("always," "never," "all," "completely," "every") while the passage uses qualified language ("most," "some," "often," "generally," "many"). The correct answer mirrors the passage's hedging. Any option stronger than the passage's claim is false.
The passage-word transplant trap. A distractor lifts exact words or phrases from the passage to appear authoritative, but applies them to the wrong referent, time frame, or logical relationship. Always verify the full propositional content of an option, not just whether its vocabulary appears in the passage.
The single-sentence inference trap. For inference questions (Q3–4), a distractor offers a claim that is supported by exactly one passage sentence, which appears to make it correct. The actual correct answer requires synthesising two or more sentences — if an option follows from one sentence alone, treat it with suspicion.
Decoder cheat sheet
Question phrasing → what the examiner actually wants
Section I + II decoders
When you see
4 options from the same semantic field
The examiner wants
The one that collocates with the anchor word — ignore meaning, focus on bond
When you see
Adverb options (earnestly / keenly / inadvertently / slowly)
The examiner wants
You to determine whether the agent intends or does not intend the action — read the subject's role in context
When you see
A blank before a preposition ("take ___ with," "in ___ of")
The examiner wants
The word that completes a fixed phrase — the preposition after the blank is your clue
When you see
Who / which / whose / where / whoever in a relative clause blank
The examiner wants
You to identify the grammatical role: subject=who, object=whom/that, possessive=whose, place=where, free relative=whoever
When you see
Conditional options: Could / May / Should / Will at start of sentence
The examiner wants
An inverted conditional — "Should you be unable to attend" = "If you should be unable to attend." Only "Should" inverts this way.
When you see
have / get + object + verb-form options
The examiner wants
Causative passive: have + object + past participle. "Have it cashed" not "have it cashing" or "have it cash."
Section III decoders
When you see
An underlined noun without an article
The examiner wants
You to check: is this noun generic and countable? If yes, it needs plural (no article). If it's a specific singular, it needs "the" or "a."
When you see
An underlined "in / on / at / to" next to a verb or fixed phrase
The examiner wants
You to verify it against the exact idiomatic form: "on the front lines" not "in the front lines"; "than" in comparatives not "to."
When you see
An underlined verb in a past narrative paragraph
The examiner wants
Consistent simple past. A present-tense verb in a past-tense paragraph is the error — check tense first, form second.
When you see
is/are/was/were + underlined verb
The examiner wants
The past participle (-ed or irregular). If the underlined word is a base form or present form, it's the error: "is borrow" → "is borrowed."
Section V decoders
When you see
"According to the article / passage / research…"
The examiner wants
Strict factual retrieval. Correct answer is directly supported. Eliminate anything that adds to or extends what the passage actually says.
When you see
"Which of the following statements is TRUE?"
The examiner wants
You to treat each option as a claim and find explicit passage support. Options with "always / never / all / only" are almost always wrong — the passage uses hedged language.
When you see
An option with exact words from the passage
The examiner wants
You to check the full propositional content, not just the vocabulary. Passage words transplanted into the wrong context or applied to the wrong referent = distractor.
When you see
An inference question where one option is directly stated in one sentence
The examiner wants
Synthesis from multiple sentences. A single-sentence support usually means the option is a distractor — true inference requires combining at least two passage ideas.
Targeted practice drills
Designed around the internal logic of each section — not generic comprehension practice
Drill 1 · collocation mapping for Section I
Take any Section I question. Before reading the options, write your own best answer in the blank. Then compare your answer to the four options. If it matches one option exactly, you identified the collocation correctly. If not, identify which collocational bond you missed and why. Do 10 items per session, and track which subtype (near-synonym / register / intentionality / fixed-phrase) causes the most errors.
Practice template: "The tension between those countries was almost ( )."
Write your word first → then check options A–D → identify why "tangible" bonds with "almost" + abstract noun "tension" while "discrepant / flippant / feasible" don't. The anchor is the abstract noun requiring a physical-metaphor adjective.
Drill 2 · error injection for Section III
Take a grammatically correct paragraph (any source). Deliberately introduce exactly one error from the confirmed taxonomy: wrong article, wrong verb tense, wrong preposition in an idiom, noun number, passive morphology error, or verb-complement mismatch. Exchange with a study partner to find and correct. The objective is internalising where in a sentence each error type tends to hide — not recognising the corrected form, but recognising the broken form at speed.
Focus areas for injection: articles before generic countable nouns; "on/in" for positional idioms; past tense verbs in narrative paragraphs; "-ed" participial form in passives; "to + bare infinitive" after complement-taking verbs.
Hard rule: do not inject vocabulary errors. The exam never does. Only syntactic and morphological errors are valid targets for this drill.
Drill 3 · discourse-connector mapping for Section IV
Take any academic passage (MEXT past papers are ideal — same register and discourse style). Remove all conjunctions, connectors, and linking adverbs, replacing them with blanks. Attempt to restore them without consulting the original. Compare your answers to the original. Mismatches reveal where your discourse-logic intuition breaks down.
Critical sub-drill — "even though" vs "because": take 5 pairs of clauses and insert both connectors. Identify which logical relationship the original encodes and whether the alternative connector produces coherent or incoherent meaning. This is the single most commonly tested contrast in Section IV.
Secondary focus: "nobody / somebody / everybody / anybody" in negative-scope contexts. "Though nobody can say why" vs "though anybody can say why" — the scope of the quantifier is determined by the implied logical claim, not by sentence surface form.
Drill 4 · option autopsy for Section V
For every reading question you answer, do not stop at selecting the correct answer. Write one sentence explaining why each wrong option fails. Categorise each failure using the four distractor types: (a) true but irrelevant, (b) extreme language / stronger than passage, (c) passage words applied to wrong referent, (d) unsupported or single-sentence inference. After 20 questions, tally which distractor type catches you most — that is your specific vulnerability and the target for your remaining preparation time.
Apply immediately to: 2015 V Part II Q3 (why people know more about Babylon than other civilisations). Correct answer = D.
Autopsy: A is true (Babylonian discoveries are used today) but describes a consequence of their knowledge, not the reason we know more about them. B has no direct passage support. C inverts the passage — the passage says they were organised record-keepers, not that they tried to dominate the world. D directly states the cause: they wrote on clay tablets that survived.
Drill 5 · conditional + causative production for Section II
Write 10 sentences from scratch using each of the following structures. Then corrupt one element per sentence (introduce a plausible error), and identify exactly what you changed and why it matters. This builds error-detection reflex for Section II at exam speed.
Target structures: (1) Inverted conditionals — "Had it not been for…," "Should you be unable to…," "Were it not for…" (2) Causative — "have/get + object + past participle": "have it repaired," "get the document signed," "had her car checked."
Corruption targets: replace "Should" with "Could" in an inversion; replace the past participle with an infinitive in a causative ("have it repair"); drop the object pronoun from a causative. These are the three most common distractor constructions the exam uses for these structures.
Drill 6 · quantifier scope mapping for Section IV + V
Collect 10 sentences using "nobody / anybody / somebody / everybody" and "none / any / some / all / most / many / few." For each, identify the logical scope the quantifier asserts (universal positive, universal negative, existential positive, existential negative). Then test each with a negation to verify the logical relationship. This directly addresses the "nobody can say why" type blank in Section IV and the "which statement is TRUE" type in Section V, where the examiner tests whether you can distinguish hedged from absolute claims.