Correct in the Margin, Blank in English

The data point away from grammar.

• Literal 78% vs inference 19% is a 59-point gap. Grammatical weakness is diffuse: it depresses everything. A pupil who cannot handle a sentence cannot retrieve a stated fact either, yet literal retrieval is the class's strongest band. Grammar is therefore not the binding constraint.
• Vocabulary-in-context at 44% sits between the two and behaves like a barrier rather than a skill: pupils who miss 'cracked', 'leak' or 'bend' cannot build the inference the writer intends, so part of the 19% is a vocabulary failure in disguise.
• The margin notes are decisive. A pupil who writes the right reason in Urdu and leaves the English blank has made the inference. The comprehension is intact; the production channel is blocked.

Diagnosis: the bottleneck is academic vocabulary plus the bridge from thought to English output - not the tense system. The 41 non-standard forms ('The water is finish') are the visible surface of that same production pressure: under time and risk, pupils fall back on L1 syntax.

A grammar revision week targets the one component for which there is least evidence. It is also the lowest-value intervention available: it would raise the 78% further and leave the 19% where it is.

Design rule: every slot must move a pupil from reason held in Urdu to reason written in English, and must recycle the same three question types the exam uses.

Mon - the bridge, using a 60-word text. Pupils answer one inference question by writing the reason in Urdu if they wish, then translate it into one English sentence using a frame: 'The writer mentions ___ because ___.' Collect three frames, praise the reasoning, correct only the frame.

Tue - vocabulary in context, six words. Teach the context strategy explicitly ('cover the word, read the sentence, guess the job the word does') on six words from the pump passage. Pupils write a guess, then check. No dictionary during the guess.

Wed - frame bank and substitution. Build a class bank of six frames on the board (because / so that / in order to / this shows / the writer wants us to see / although). Pairs substitute new content into two frames. This converts a one-off sentence into a reusable tool.

Thu - paired reconstruction. Pupil A gives the reason aloud in Urdu, Pupil B renders it into English, then swap roles on a new item. You circulate and note the two most common breakdowns to teach on Friday. This is the highest-value slot: it rehearses the exact transition that failed in the exam.

Fri - timed rehearsal, exam format. Eight minutes, a 120-word passage, one literal, one vocabulary, one inference. Same layout as the examination paper so the format is not an extra load.

Measure the bottleneck, not the average. Re-run the same three item types on a parallel 180-word passage after two weeks and report the three percentages side by side:

• Inference: 19% is the target. Anything above roughly 35% in two weeks is real movement; a jump to 70% would suggest the post-test was easier, so keep the two passages matched in length and question type.
• The Urdu-only response count: this is the cleanest indicator of the bridge working. Record how many inference items are answered with a margin note and a blank line. If that number falls, the intervention is doing its job even if marks rise slowly.
• Vocabulary in context: 44% should move before inference moves much; if it does not, the vocabulary slot needs strengthening.

What must not get worse: literal retrieval at 78% and the general accuracy of short answers. A common failure mode of inference coaching is that pupils start writing long speculative answers and stop answering the actual question. Keep literal items in the post-test as a guard.

Report to the head in one table, three rows, two columns, with the sample size (55) stated. Then the argument about grammar is settled by data rather than by impression - and if the head still wants the grammar week, ask for it in addition to, not instead of, this work.