The Copying Habit: When Students Answer Every Question as a Retrieval Task
At a senior high school in Tamale, a Form One class of 40 has two 50-minute English lessons a week. On a mock paper built around a 700-word passage about plastic pollution, the department's item analysis shows 78% correct on literal retrieval, 61% on vocabulary in context, 34% on inference and 22% on a question asking students to judge the writer's attitude with supporting evidence. Scripts show that on inference items 26 of the 40 students copied a sentence from the passage verbatim; nine wrote personal opinions such as 'I think plastic is bad'; only five quoted a phrase and then explained it. In class, students read silently, answer the numbered questions, and the teacher reviews the answers on the board. The head of department, wanting better results, has ordered two extra periods a week of past-paper practice from now until the examination.
1. What does the score pattern reveal about these students' reading, and why is two extra periods of past-paper practice unlikely to fix it?
Read the worked answer
What the pattern shows. The scores form a ladder: 78% literal, 61% vocabulary, 34% inference, 22% evaluation. This is not a general 'weak comprehension' problem. It is a specific gap between locating information and building meaning from it.
The script evidence locates the gap precisely. Twenty-six of forty students answered inference questions by copying a sentence: they have learned that a reading question is answered by finding matching words. Only five quoted a phrase and explained it, which is the actual move required. Nine substituted personal opinion for evidence - a different failure, showing they do not yet accept that an answer must be authorised by the text. The 22% on attitude is the far end of the same ladder: attitude is rarely stated outright, it is carried by word choice, and these readers are not yet reading words as choices.
Why more past papers will not work. Past-paper practice is assessment, not instruction. It offers forty more opportunities to copy a sentence, and it rewards the retrieval habit that is causing the problem, because literal scores will creep up and create the appearance of progress. The students do not lack exposure to questions; they lack a model of the thinking a question demands.
Dealing with the head of department. Do not simply refuse. Negotiate with evidence: keep one period for timed practice, convert the other to strategy teaching, and offer to bring the same item analysis back in six weeks so the comparison is fair.
2. Plan a 50-minute lesson on the same passage that teaches inference rather than testing it. Specify the sequence, the questions you would ask, and what students are doing at each stage.
Read the worked answer
Principle. The lesson must make the invisible reasoning visible before asking anyone to produce it. One passage, few questions, heavy modelling.
0-5 minutes: activate. Ask: 'A friend says "I'm fine" in a flat voice. How do you know whether to believe her?' Collect two answers. Name the target: today we read for what is suggested, not only what is stated.
5-15 minutes: model with two sentences. Put two sentences from paragraph two on the board. Think aloud: 'The writer calls the ocean a plastic soup. Soup is thick and everything is mixed in. So the writer is not just saying there is plastic - she is saying it is everywhere and cannot be separated out.' Then point to the reasoning moves: notice the word, unpack it, link to the claim, check against another line.
15-25 minutes: paired work with question sorting. Give four questions. Students label each one: answer is right there in one sentence, or answer needs two parts of the text joined. This teaches them to distrust the copy-the-sentence reflex.
25-40 minutes: written answer with a warrant. Students answer: 'What does the phrase plastic soup suggest that the word pollution would not? Use evidence from the passage.' They must write claim, quotation, and one sentence of explanation.
40-50 minutes: exemplar. Read one strong and one copying response anonymously. Ask the class which one answers the question and why. Close with the rule: a quotation alone is not an answer.
3. How would you assess inference and evaluation fairly in this class, given that only 22% scored on the attitude question and general vocabulary is weak?
Read the worked answer
Separate the constructs. A reading test should not silently become a vocabulary test or a writing test. Score comprehension and vocabulary on separate sections so that a student with limited word knowledge is not recorded as unable to infer.
Use short constructed responses with a three-part rubric. For each item, award up to three marks: one for a defensible claim, one for relevant textual evidence with a line or paragraph reference, one for explanation linking evidence to claim. A copied sentence earns the evidence mark but not the explanation mark - which immediately distinguishes the five students who already reason from the twenty-six who retrieve. Publish the rubric before the task.
Do not rely on one item. A single attitude question yields an unreliable judgement, especially with 40 students. Use four to six items of the same type across two short texts, and treat the pattern as the finding.
Offer an oral route. Where writing is the barrier, a two-minute conference question - 'show me the words that tell you how the writer feels, and tell me why' - reveals an inference that a written script may hide. Record the outcome on the same rubric.
Do not penalise spelling or grammar here. Deducting marks for surface error adds construct-irrelevant variance and will depress exactly the students whose reasoning is developing.
Measure growth with parallel items, not the same past paper. Pre- and post-tests should use fresh short texts with matched item types, or improvement will be confounded with familiarity.