SRDP Language in Use Task Generator (B1 / B1+)

Question 3 — Topic: "Any preferences for the topic?"

• A specific topic (e.g., "healthy eating", "social media", "travel")
• From the SRDP Themenpool (I'll choose)
• Tied to a specific Way2Go unit
• Surprise me

Question 4 — Grammar/Vocabulary Focus: "Should I target specific grammar or vocabulary areas?"

• General mix of grammar and vocabulary (Recommended)
• Grammar-heavy (tenses, conditionals, passive, relative clauses, etc.)
• Vocabulary-heavy (collocations, phrasal verbs, word families)
• Specific focus (e.g., "present perfect vs. past simple", "phrasal verbs")

If the user has already specified parameters in their initial message, skip those questions.

Overview: What You Generate

Each Language in Use task consists of:

1. A text with gaps — an authentic-sounding English text with strategically placed gaps
2. Items — the gaps themselves, each testing a specific grammar or vocabulary point
3. An answer key — with correct answers, acceptable alternatives, and rationale

The Five SRDP Language in Use Task Formats

Step-by-Step Workflow

Step 1: Read the Appropriate Reference File

Before writing anything, read the reference file for the chosen format. This is non-negotiable.

Also read:

• references/b1-vs-b1plus-liu.md to calibrate difficulty
• references/item-validity-firewall.md to internalize ALL known item validity traps before constructing any items. This prevents generating items with defensible alternative answers, broken gaps, or pattern-guessable transformations.

Step 2: Write the Text

THE FUNDAMENTAL DIFFERENCE FROM RC TEXTS

LiU texts serve a different purpose than RC texts. In RC, the text is the content being tested. In LiU, the text is the VEHICLE for testing grammar and vocabulary. This means:

• The text must be natural enough that gaps don't feel forced
• The text must be information-rich enough to provide context for gap-solving
• But the text is NOT the star — the GAPS are the star
• You are designing gaps first, then wrapping a natural text around them

The construction order matters:

1. Choose 10–14 grammar/vocabulary targets appropriate to the level
2. Draft a text on the topic that naturally contains those structures
3. Place gaps at the target structures
4. Verify the text reads naturally with gaps filled in
5. Verify each gap has exactly one defensible correct answer

TEXT AUTHENTICITY FOR LiU

LiU texts face the same "AI-sounding" risk as RC texts. Apply these principles:

Text Personality: LiU texts need personality too. They should feel like real blog posts, magazine articles, emails, or narratives — not texts constructed solely to test grammar.

Text Types That Work Best for LiU:

• Personal blogs and narratives (natural tense variety, phrasal verbs, collocations)
• Advice/how-to articles (imperatives, conditionals, modal verbs)
• Magazine feature articles (varied structures, semi-formal register)
• Formal and informal emails (register contrast, fixed phrases)
• Biographical accounts (past tenses, time expressions, reported speech)
• News-style reports (passive voice, reported speech, formal vocabulary)

Text Types That DON'T Work Well:

• Highly technical texts (specialized vocabulary overwhelms general language testing)
• Poetry or very literary texts (ambiguity makes gaps too open-ended)
• Dialogue-only texts (lacks narrative context for prediction)

Authenticity Markers for LiU Texts:

• Named people, places, or institutions (not "a man" but "James, a 24-year-old from Bristol")
• Specific details: numbers, dates, costs, names
• A clear genre identity (the text should clearly BE a blog post, article, email, etc.)
• Natural paragraph transitions (not "Moreover," "Furthermore," "Additionally,")
• At least one moment of personality: humor, opinion, surprise, self-deprecation
• Register consistency (except where intentional shift is the point)

AI-Tells to Eliminate (same as RC):

• "X is becoming increasingly popular" → Open with a specific example
• "In today's world / modern society" → Delete. Start with content.
• Every paragraph starting with a topic sentence → Let some start with examples or continuations
• Overly balanced both-sides arguments → Give the writer a stance

TEXT LENGTH AND GAP DENSITY

B1+ texts are approximately 20% longer than B1 texts at the same density.

TEXT-LEVEL CALIBRATION (B1 vs B1+)

Read references/b1-vs-b1plus-liu.md for the full calibration guide. Key differences:

Step 3: Construct the Items (Gaps)

Read the format-specific reference file and follow its construction principles precisely.

Universal Item Requirements

• Item 0 is always a completed example with the answer filled in
• Gap placement follows text order — item 1's gap appears before item 2's gap in the text
• Even distribution across the text — gaps should be spread across all paragraphs, not clustered in one section
• Mix of grammar and vocabulary items — aim for approximately 40–60% grammar, 40–60% vocabulary (the exact ratio depends on the format — see reference files)
• Each gap must have exactly ONE defensible correct answer (for MC: one clearly best option; for OC: one word, though spelling variants may be accepted; for WF: one correct transformation)
• No two consecutive gaps should test the same linguistic feature — if gap 3 tests a preposition, gap 4 should test something different (tense, collocation, etc.)

The Grammar/Vocabulary Mix — What to Test

B1 Grammar Targets (select 5–7 per task):

• Past simple vs. present perfect (the central B1 distinction)
• Present perfect simple vs. continuous
• Irregular past tense forms
• Passive voice (present and past)
• Relative clauses (who/which/that/where/whose)
• First and second conditionals
• Modal verbs (should, must, may, might, could for possibility/obligation)
• Gerunds vs. infinitives (enjoy -ing, want to, stop -ing)
• Reported speech (basic — said that, told me to)
• Articles (a/an/the — functional, not tricky)
• Prepositions of time, place, and direction
• Connectors: although, because, however, while, so

B1 Vocabulary Targets (select 4–6 per task):

• Collocations: make friends, do homework, take a break, catch a bus, have a choice
• Phrasal verbs: find out, turn up, break down, come across, look after, cut out, end up
• Fixed phrases: for instance, on the other hand, as well as, in addition to
• Adjective + preposition: afraid of, interested in, good at, aware of, proud of
• Verb + preposition: depend on, look for, care about, belong to, apply for
• Word families (for WF): strong→strength, use→useful, educate→education

B1+ Additional Grammar Targets:

• Past perfect (had + past participle)
• Third conditional (if + had + pp → would have + pp)
• Complex passive (present perfect passive, modal passive)
• Fronted adverbials ("Despite the evidence,", "Having spent three years,")
• Concession structures (although, despite, in spite of, while)
• Hedging (tend to, appear to, seem to, it is argued that)
• Inversion after negative adverbials (rare, but appears in formal texts)

B1+ Additional Vocabulary Targets:

• Semi-formal collocations: conduct research, raise awareness, draw conclusions
• Academic vocabulary: implications, consequences, evidence, perspective
• Hedging expressions: to a certain extent, arguably, in many cases
• More sophisticated phrasal verbs: come up with, make up for, put up with

Step 4: Create the Answer Key

For every task, provide:

1. Correct answers for all items
2. For MC Cloze: Complete distractor rationale — why each wrong option is wrong AND why it might tempt a student. Label each distractor with its type (see references/mc-cloze.md).
3. For Open Cloze: The correct word plus any acceptable alternatives (e.g., "also" or "too" if both work). Explicitly state what does NOT count.
4. For Word Formation: The correct transformed word, the transformation process (e.g., "strong → strength: adjective → noun via suffix change"), and common student errors.
5. For Banked Cloze: Why each distractor option doesn't fit any remaining gap.
6. What each gap tests: State the linguistic feature (e.g., "Tests: present perfect vs. past simple", "Tests: collocation 'aware of'", "Tests: phrasal verb 'find out'")

Step 5: Mandatory Verification Procedure

THE TWO-PASS RULE: Write the task as a DRAFT, verify it, then output the corrected FINAL VERSION if any errors were found.

Verification 1: Gap Distribution Map

Number the paragraphs. For each gap, identify:

Gap 0 (example): Paragraph X — Tests: [linguistic feature]
Gap 1: Paragraph X — Tests: [linguistic feature]
Gap 2: Paragraph X — Tests: [linguistic feature]
...

Distribution across paragraphs: P1=N, P2=N, P3=N...
Any paragraph with 0 gaps? → Add one
Any paragraph with >3 gaps? → Redistribute
Consecutive gaps testing same feature? → Swap or replace

Verification 2: Grammar/Vocabulary Balance

Grammar items: [list gap numbers] = N total
Vocabulary items: [list gap numbers] = N total
Ratio: X% grammar / Y% vocabulary
Target: 40-60% / 40-60%

If heavily skewed, replace items to rebalance.

Verification 3: Answer Validity — THE DEVIL'S ADVOCATE TEST

MANDATORY: Read references/item-validity-firewall.md before running this verification. It catalogs every known LiU trap type with real examples from failed test generation.

THE DEVIL'S ADVOCATE TEST — run for EVERY gap regardless of format:

For each gap, argue that a DIFFERENT answer (not the intended one) is correct. If you can make a reasonable case → the gap is invalid → REDESIGN IT.

For MC Cloze — the SUBSTITUTION TEST + DEVIL'S ADVOCATE:

For each gap, insert EACH option into the full sentence and read it aloud:

Gap X:
  "[sentence with option A]" — grammatical? ✓/✗  natural-sounding? ✓/✗
  "[sentence with option B]" — grammatical? ✓/✗  natural-sounding? ✓/✗
  "[sentence with option C]" — grammatical? ✓/✗  natural-sounding? ✓/✗
  "[sentence with option D]" — grammatical? ✓/✗  natural-sounding? ✓/✗
→ Exactly 1 option works? YES/NO → if NO, REDESIGN
→ Native speaker test: Would a native speaker naturally use any distractor here? YES/NO
  → if YES for any distractor, it's a valid answer → REDESIGN

Additional MC sub-checks:

1. Transitivity check: Does the sentence have a wh-word or object that requires a transitive verb? Check that the "correct" answer has the right transitivity.
2. Rationale verification: For every distractor rationale that makes a grammar claim (e.g., "X requires Y"), verify the claim is actually true. Wrong rationale = broken item.
3. Level-appropriateness: At B1, if distinguishing correct from distractor requires native-speaker intuition → accept multiple answers OR redesign.
4. Answer distribution: Count correct answer letters across ALL gaps:

   A=_  B=_  C=_  D=_
   Each letter at least 2? YES/NO. Any letter >3? YES/NO.
   → if skewed, swap answer positions (not content)
   

For Open Cloze — the AMBIGUITY TEST + ALTERNATIVES: For each gap, try at least 3 different words:

Gap X: intended answer = "___"
  Alternative 1: "___" — fits? YES/NO
  Alternative 2: "___" — fits? YES/NO
→ If any alternative fits: restructure sentence to force one word OR list all in answer key
→ Content word check: Is the target a function word? If it's a content word → wrong gap type

For Word Formation — the FORM TEST + BROKEN GAP CHECK:

Gap X: base word = "___"  target word = "___"
  Step 1: Does the sentence make sense WITH the target word? YES/NO
    → if NO → GAP IS BROKEN → redesign sentence around a word that works
  Step 2: List ALL derivations of the base word:
    [derivation 1], [derivation 2], [derivation 3]...
  Step 3: Try each derivation in the sentence. Only one must fit.
    → if 2+ fit → add context to force one
  Step 4: Does the target word sound natural in the sentence? (Read aloud)
    → if forced/awkward → redesign sentence

Additional WF sub-checks:

1. Consecutive-type check: List the transformation type for each gap in sequence. No two consecutive gaps may use the same type. Use at least 5 different types across 10 gaps.

   Gap 1: [type] → Gap 2: [type] → ... consecutive same? YES/NO
   

2. Spelling check: Verify correct spelling of every target word.

For Banked Cloze — the UNIQUENESS TEST: Each option in the bank must fit exactly ONE gap. Verify by trying every unused option in every remaining gap.

Option X: fits gap [N] ✓. Also fits gap [M]? YES/NO → if YES, redesign one gap sentence

For Error Correction — the AMBIGUITY TEST: Each error line: verify only ONE word is wrong and only ONE correction exists. Each CORRECT line: try to argue it contains an error. If you can → redesign.

Verification 4: Authenticity Check

Re-read the complete text with all gaps filled in. Does it sound natural?

• Any awkward sentences created by gap placement? → Rewrite
• Does the text read like a real blog/article/email? → If not, revise
• Any AI-tells? (generic openings, "Moreover/Furthermore" paragraph starters) → Fix
• Named people/places present? → If not, add
• Register consistent throughout? → Verify

Verification 5: Level Calibration

For B1:

• Could a 15-year-old Austrian student with B1 English complete 60–70% of this?
• Are all grammar structures within the B1 syllabus?
• Are vocabulary items high-frequency and communicatively useful?

For B1+:

• Is this noticeably harder than B1? (more complex structures, more subtle distractors)
• Are there 2–3 items that require understanding implied context?
• Does the text include hedging, multiple perspectives, or evaluative language?

Output the Final Version

After all verifications, output the complete corrected task:

## FINAL VERSION
[Complete task with all corrections applied]

SRDP Themenpool (Topic Areas for LiU)

B1 Topics (5. Klasse / Year 5)

• Sports, fitness, and leisure activities
• School life, education, and study habits
• Food, nutrition, and healthy eating
• Communication, social media, and relationships
• Rules, laws, and regulations (including strange laws, parenting)
• Travel, transport, and tourism
• Animals, nature, and environment (introductory level)
• Jobs, careers, and work experience

B1+ Topics (6. Klasse / Year 6)

• Arts, culture, and creative industries
• Traditions, trends, and lifestyle changes
• Environment, conservation, and sustainability
• The English-speaking world and intercultural issues
• Media literacy, journalism, and digital culture
• Science, technology, and innovation
• Ethics, social justice, and contemporary debates
• Health, psychology, and wellbeing

Output Format

Present the task in this structure:

## [Topic] | [Level] | [Format]

### Text

[Title]
[Subtitle/byline if appropriate]

[Text with numbered gaps]

[For MC: options table after text or inline]
[For BC: word bank listed before or after text]
[For WF: base words in brackets at each gap]

### Answer Key

[Complete answer key with rationale for every gap]

### Verification Log

[All 5 verifications with results]

Core Principles

The gap is the assessment. Every gap must test a specific, identifiable linguistic feature. If you can't name what a gap tests in 5 words or fewer ("present perfect vs. past simple", "collocation: aware of", "phrasal verb: find out"), the gap is unfocused and must be redesigned.

Distractors are diagnostic. In MC format, each distractor should represent a specific type of student error. If a distractor is wrong for no identifiable reason — if a student would never plausibly pick it — it wastes an option slot. Every distractor must have a "temptation mechanism" that exploits a real learner confusion.

Context enables, doesn't give away. The surrounding text should provide enough context for a student with the target grammar/vocabulary knowledge to answer correctly, but NOT so much that any student can guess correctly without that knowledge. If the answer is obvious from context alone (regardless of grammar knowledge), the gap tests reading comprehension, not language use.

One answer, no ambiguity. Each gap must have exactly one defensible correct answer. For MC, exactly one option should be clearly best. For OC, one word should be clearly expected (though minor variants may be accepted). Ambiguous gaps are the single most common failure mode in LiU task construction.

Natural text, strategic gaps. The text must read naturally with gaps filled in. If the text feels like it was written around the gaps (rather than gaps being placed in a natural text), it has failed. The best LiU texts work as readable, engaging texts on their own — the gaps are invisible when filled.

Level is in the grammar, not just the topic. B1 and B1+ differ primarily in the grammar structures tested and the sophistication of distractors — not just in topic choice. A text about social media could be B1 or B1+; what determines the level is whether the gaps test basic collocations or complex passive constructions.