Command, Stem, Data: Reading the Exam Paper Before Answering It

In many UK science departments, past papers are used as rehearsal space for recall: a question is projected, pupils write, marks are counted. Yet before a pupil can recall anything, they must read. They must locate the command word, hold the stem in mind, interpret a table or graph, notice a constraint and decide which part of a multi-part question they are answering. The exam paper is not a transparent window onto the specification. It is a text type with its own conventions, and those conventions carry marks.

The paper as a genre

A GCSE science question is rarely a single sentence with a single instruction. It may include a rubric, a stimulus paragraph, a data table, a graph, a photograph, a command word, a mark tariff and several linked parts. The reading path is not linear. A pupil who reads from top to bottom in a straight line may answer the first part with information that the third part needed them to hold back. A pupil who ignores the mark tariff may write a paragraph where one mark was available, or a phrase where three were.

The reading demand is discipline-specific. It is not generic comprehension. It requires knowing that in a data-response question, the phrase using the data is an instruction to cite values, not an invitation to repeat taught content. It requires recognising that a question beginning with the stem in Figure 2 expects the figure to be read before the answer is planned. It requires noticing that a question says other than temperature, and that this exclusion changes the answer.

Command words are instructions

Suggest and explain are not interchangeable. Suggest usually invites pupils to apply familiar science to an unfamiliar context; the mark scheme often accepts a plausible mechanism rather than one exact phrase. Explain usually demands a causal chain: because, therefore, so. Describe asks for a pattern or feature. Compare requires both objects and a basis for comparison. Evaluate requires criteria and a judgement. These are tendencies rather than universal laws. Awarding organisations use command words differently across specifications, tiers and subjects, and the mark scheme is the authority.

A practical starting point is a command-word sort. Give pupils a set of past-paper questions and ask them to group the questions by what the command word is asking them to do, not by topic. Then ask them to rewrite each command word as a question to themselves: what does suggest require me to do here? What would count as an explanation rather than a description? The aim is not to memorise a glossary. The aim is to make the instruction visible before the science content dominates attention.

Stem, data and the multi-part trap

In a data-response question, the stem may contain context that is not needed. It may also contain the only clue about what the data mean. Pupils need a routine: identify what is being changed, what is being measured and what is being held constant. Ask what the table shows and what it does not show. Ask which row or column is relevant. Ask whether the units are the same throughout. Ask whether the graph shows a trend, an anomaly or a comparison.

Multi-part questions add another layer. Pupils should read all parts before writing. Later parts may depend on an earlier answer, or may contain a scaffold that makes the earlier part easier. A constraint such as using data from Table 1 or other than temperature may appear in part (b) but change how part (a) should be read. A pupil who answers part (c) before reading part (b) may miss the structure that the examiner built into the question.

A ten-minute reading routine

Past papers can become short reading lessons without becoming another full mock. The routine below works with a single data-response question and its mark scheme.

1. Select one question with a stem, a data source and at least two parts. Display it without answer lines.
2. Read the question aloud while pupils annotate. Model your own reading: I am reading the command word first. It says suggest, so I need to apply what I know to this unfamiliar context. I am underlining using data from Table 1 because my answer must include a value. I am reading part (b) before I write part (a) because the later part may tell me what the examiner expects.
3. Ask pupils to paraphrase the question in the margin in no more than ten words. Compare paraphrases in pairs.
4. Reveal the mark scheme. Ask pupils to match each mark to a reading move: a command word, a data value, a unit, a causal link or a constraint. Do not ask them to write the full answer.
5. Finish by asking what a pupil would need to notice first if they met this question in an exam.

Once a week for half a term is enough to make the routine familiar. The point is not to replace teaching of the science. The point is to make the reading of the paper an object of instruction.

Track misreading, not only misunderstanding

When marking, code errors rather than only counting them. A simple audit might use C for command word, S for stem constraint, D for data extraction, U for units, M for multi-part dependency and K for science knowledge. After three papers, patterns appear. If many pupils lose marks to C and D, a reading lesson is more efficient than reteaching the content. If the pattern is K, the department should reteach the science. A pupil who writes a perfect explanation of photosynthesis to a question asking for the effect of light intensity has a reading problem before a biology problem.

The language matters. Avoid labelling pupils as poor readers. Label the reading demand. The demand can be taught, and it can be assessed separately from recall.

The exam paper is a text. In 2026, as departments review schemes of work and assessment routines, reading the question deserves a place alongside revising the content. It is not a prelude to science. It is part of scientific literacy.