Scoring inference items on an informational Gagauz passage
A mock Bacalaureat paper in Gagauz carries a 180-word informational passage about the falling number of weekly Gagauz lessons in village schools. Item 1 asks candidates to locate the author's stated reason for the fall. Item 2 asks what the phrase 'ana dili' means in sentence 3, where the author contrasts it with the language children use in the schoolyard. Item 3 asks what the author implies about parents' attitudes. Item 4 asks for the function of the final sentence. The scheme gives 2 points per item: 2 = complete and text-based, 1 = partial or partly unsupported, 0 = wrong or off-text. Candidate A answers item 1 correctly in one line, glosses item 2 as 'the language of the mother', answers item 3 by copying sentence 5 verbatim, and writes 'it summarises' for item 4. Candidate B gives two contradictory reasons for item 1, glosses item 2 as 'mother's tongue', paraphrases the implication in item 3 but adds a claim the passage never makes, and identifies the concluding recommendation in item 4. Both scripts total 5 of 8.
1. Score both scripts item by item and total them. Show the reasoning behind each of the eight decisions.
Read the worked answer
Item 1 (locating the stated reason). Candidate A gives the reason the passage actually states: 2. Candidate B offers two reasons that cancel each other, so the examiner cannot tell whether the right one was located: 1 (partial).
Item 2 ('ana dili'). Both candidates translate word by word. 'The language of the mother' and 'mother's tongue' show the phrase was decomposed, not understood: it means the language a child acquires at home, which is precisely what the contrast with schoolyard language is doing. Neither reaches the 2-point descriptor 'explains the meaning in context'; both get 1.
Item 3 (implied attitude). A copies sentence 5. Copying proves location, not inference: 1. B performs the right operation (paraphrase of an implication) but adds an unsupported claim, which is exactly the 1-point descriptor 'partly unsupported': 1.
Item 4 (function of the last sentence). 'It summarises' does not distinguish a summary from a recommendation: 1. B names the function: 2.
Totals: A = 5, B = 5. The tie is real and should not be forced apart.
2. Both scripts total 5/8. Decide which performance is stronger and state what the examiner should record.
Read the worked answer
The totals hide opposite profiles. A is accurate but literal: it locates and copies, and its only failure is vagueness. B attempts inference twice, succeeds once, and over-reads once. Since the construct being measured is comprehension, including inference, B's script shows the higher-order operation, while A's shows only retrieval.
The decisive question is which defect matters more for an informational text. B's invented claim is a comprehension failure, not a stylistic slip: a reader who adds propositions to a factual passage will misreport it. A's errors are under-specification, which is the milder fault.
Recommended record: enter sub-scores (literal items 1-2; inferential items 3-4) alongside the total, and note in the moderation log that B's item 3 answer was 'inference with added unsupported content'. Rationale: the single total rewards copying with 2 and inference with 1 in places, which inverts the construct. A third anchor script should settle borderline cases rather than an examiner's impression.
3. Rewrite item 2 so that a word-for-word gloss cannot earn full marks, and state what the new item tests.
Read the worked answer
New wording: 'In sentence 3 the author writes ana dili rather than gagauz dili. What does this choice add to the meaning of the sentence?'
Why this works: the old item asked for a definition, which a bilingual dictionary answer can satisfy. The new item asks about a choice, so the candidate must compare two expressions and relate the contrast to the surrounding sentence about the schoolyard. A gloss alone does not answer it.
Descriptors: 2 = explains that ana dili names the home language acquired from the family, and notes that the author is contrasting home acquisition with school use; 1 = identifies the meaning 'native language' but does not connect it to the contrast; 0 = literal translation with no explanation.
What it tests: lexical inference in context plus awareness that word choice carries stance. A second acceptable variant asks which words in sentence 3 signal that the phrase is being used figuratively, which shifts the item towards textual evidence rather than meaning.