A student reads:
Mei Ling placed the unopened envelope beneath a pile of books and began wiping a table that was already clean.
The question asks:
What can you infer about Mei Ling?
One answer says:
She hates cleaning.
Possible? Technically. Supported? Not really.
A stronger answer says:
She appears to be avoiding the envelope because she may be anxious about what it contains.
The second answer connects two details: the unopened envelope and the unnecessary task.
Inference is evidence-based interpretation, not imaginative guessing.
What inference actually is
An inference is a conclusion not stated directly but supported by information in the text, sometimes together with reasonable background knowledge.
A useful model is:
textual evidence + connection → justified conclusion
Not:
interesting idea → answer
Why inference feels difficult
Literal questions are reassuring. The answer is visible.
Inference questions ask the student to cross a small gap.
Harun checked the departure board again, although his train number had not changed.
The passage does not say he is anxious.
Repeated checking is evidence that the departure matters strongly to him.
The student must move from behaviour to possible state.
Stay close to the evidence
Weak answers fail in two directions.
Too close: they copy the passage and add no interpretation.
Too far: they invent a story the passage cannot support.
The strong answer sits between those extremes.
Think of inference as a bridge
One side is the text.
The other side is the interpretation.
Evidence forms the bridge.
If the evidence is missing, the answer collapses.
Inference is probabilistic
A person who checks a phone repeatedly might be anxious, impatient, expectant or bored.
The surrounding text narrows the possibilities.
Good readers ask:
Which interpretation best explains the available evidence?
Evidence has different strengths
“Of course I’m pleased for you,” Mara said, closing the drawer harder than necessary.
The spoken words are positive. The physical action may contradict them.
That contrast supports an inference such as jealousy or irritation more strongly than the spoken words alone.
Students need to judge not only whether evidence exists, but how strongly it supports the claim.
Common inference types
Emotion or attitude
Gestures, silence, repetition, hesitation and tone may suggest an internal state.
Motivation
Earlier events, goals and fears may explain behaviour.
Relationship
Formality, interruption, avoidance and trust may reveal social distance or power.
Situation
Objects, setting and sequence may reveal what has happened without stating it.
Background knowledge helps — but cannot replace the text
If the lights go out and everyone lifts a phone, world knowledge helps us infer that phone screens or torches provide light.
But if a passage merely mentions a hospital, the reader should not invent an emergency.
The text remains the anchor.
A three-step method
- Find the clue.
- Say what the clue normally suggests.
- Check competing explanations.
Then choose the most defensible interpretation.
Use “because” as a training device
I think Daniel is reluctant to enter because he pauses outside the room after hearing his parents arguing.
The word forces the evidence into view.
Inference is different from writer’s effect
The wind clawed at the shutters.
Inference:
the weather is violent or threatening.
Writer’s effect:
“clawed” gives the wind animal-like aggression, making the storm feel hostile.
Same evidence. Different question job.
Inference is different from prediction
Inference explains what the current evidence suggests.
Prediction asks what may happen next.
They are related but not identical.
Why students over-infer
They answer from personal experience, notice one clue but ignore others, or try to make the answer dramatic.
Language such as:
suggests
appears
likely
may indicate
can preserve appropriate uncertainty.
Why students under-infer
They copy because they fear being wrong.
Training should teach them to move exactly one justified step beyond the text.
A distance-from-text check
Zone 1 — Copy: no interpretation.
Zone 2 — Inference: evidence-bound added meaning.
Zone 3 — Speculation: invented information.
Hilary Mantel as a high-level lens
One broad lesson from Hilary Mantel is that small observable details can carry larger meaning: who waits, who looks away, who is admitted, what is left unsaid.
The transferable lesson is close observation without reckless invention.
Inference connects English to real-world reasoning
People infer constantly from incomplete evidence.
New evidence arrives. Good reasoning remains willing to revise.
This pattern appears in Science, History, relationships, news reading and AI evaluation.
AI makes evidence discipline more important
An AI system can produce a confident interpretation.
The student should still ask:
Which exact evidence supports that claim?
If the answer cannot point back to the text, the interpretation may be plausible but ungrounded.
Connections eduKateAI can learn
textual clue → candidate interpretation → evidence check → competing explanation → best-supported inference
new evidence → revise interpretation
reading inference ↔ scientific hypothesis testing ↔ historical source interpretation
The public principle is that inference is constrained interpretation. It adds meaning, but the added meaning remains accountable to evidence.
Related eduKateSG resources
- Why Good English Is Really the Skill of Making Better Decisions
- What Good Writers Notice Before They Write
Research basis
This article draws on current Cambridge First Language English guidance, secondary comprehension-strategy research and recent middle-school inference intervention research.