7.4 Reading Strategy: Skim, Locate, Infer

7.4 Reading Strategy: Skim, Locate, Infer

Every Part 3 question can be answered with the same three-step drill. The goal is to fit each question within 45 seconds, leaving a 15-second buffer for hard ones within the 60-second per-question budget (Part 3 is 30 minutes for ~30 questions).

Step 1: Skim (10 seconds)

Capture three things from the document's skeleton:

• 件名 / 見出し — topic
• First sentence of each paragraph — structure
• Last sentence of the document — author's stance

You should walk away from the skim knowing: (a) what the document is about, (b) what type it is (email, notice, minutes, report, article), and (c) the author's stance (request, refusal, neutral, opinion).

Train with a timer: skim a 400-character email in 10 seconds. If you spend more than 15 seconds on the skim, you are reading the body, not skimming.

Step 2: Locate (15 seconds)

After the skim, read the question. Identify the keyword in the question (date, name, action verb, concept) and scan the document for that keyword. Read only the sentence(s) containing the keyword — not the surrounding paragraphs.

Train to locate a keyword in 5 seconds. If you spend more than 10 seconds, you are reading instead of scanning.

Step 3: Infer (10 seconds)

If the answer requires inference (not direct fact), look for these signals:

The most common inference question asks for the 筆者の主張, which is usually found after a しかし or in the final paragraph marked by 筆者は〜と考える.

Time Budget

Part 3 is 30 minutes for approximately 30 questions. That gives you 60 seconds per question, but you should aim for 45 seconds per question on easy/medium ones to bank 15 seconds for the hard ones. Within each 45-second slice:

• Skim: 10 seconds
• Locate: 15 seconds
• Infer (when needed): 10 seconds
• Verify and lock: 10 seconds

Trap Patterns

• Transition-but-ignore trap: An option quotes text before しかし when the conclusion is after しかし.
• Example-not-main-point trap: An option picks an example when the question asks for the main point.
• First-paragraph-not-conclusion trap: An option picks the first paragraph's topic when the conclusion is in the last paragraph.
• Opinion-when-fact-asked trap: An option picks an opinion when the question asks for a fact.

Combined Worked Example

You receive a 4-paragraph business email from a supplier. Question: 何が添付されていますか.

1. Skim (10 sec) — 件名 says 「お見積りとご案内」. Tone is report/notice hybrid.
2. Locate (15 sec) — keyword is 添付. Scan the document; the third paragraph contains 「添付ファイルにてお送りします」. Read only that paragraph.
3. Infer (10 sec) — the question is a fact question, not inference; lock the answer directly from the located sentence.

Total: ~35 seconds. The 25-second buffer rolls into the next hard question.

Reading-Pace Calibration

BJT Part 3 is a reading-pace test as much as a comprehension test. Train with a timer in three increments:

If you consistently exceed these targets, you are reading instead of skimming or scanning. The fix is mechanical: force your eyes to skip greeting and signature blocks, jump from 件名 directly to the first sentence of 本文, and only circle back when the question demands a detail you did not capture.

Common Inference Failure Modes

Three failure modes account for most inference errors in Part 3:

• Premature locking: Locking an option that matches a keyword without reading the conclusion marker that follows. Always read one extra sentence past the keyword.
• Over-reading a quote: Treating a quoted 社長 comment as the article's conclusion. Verify whether the article endorses or merely reports the comment.
• Confusing 例示 with 主張: Locking an example sentence (例えば, 〜として) when the question asks for the main point. Examples support the main point but are not the main point.

Vocabulary for the Workflow

件名, 見出し, リード文, 宛先, 添付, 同封, 别添, ご案内, お詫び, ご報告, お願い, 決定, 提案, 保留, 審議中, 見込み, 予想, 実績, 予定, したがって, そのため, ゆえに, しかし, だが, つまり, すなわち, べきだ, と考える, に違いない, に過ぎない, こそ.