A plausible answer is not automatically a supported inference. Primary 3 students are beginning to move beyond literal comprehension, but this creates a new risk: treating anything that “could be true” as something the passage actually supports.
This preserved 2019 Punggol URL is rebuilt around one reader job: calibrating inference confidence—possible vs supported. It is not a duplicate P3 tuition landing page. For current programme information, families can use Primary 3 English Tuition in Small Groups at Punggol or the broader Primary English Tuition in Punggol page.
Historic tutor, phone and centre claims from the 2019 version are not carried forward. eduKate Singapore is an independent tuition provider and is not affiliated with MOE, SEAB or any school.
Three Confidence Levels
| Level | Meaning |
|---|---|
| Literal | Directly stated |
| Supported inference | Not stated directly but justified by clues |
| Possible only | Plausible but insufficiently grounded |
Students should learn to say which level their answer belongs to.
The Clue Requirement
An inference needs at least one relevant clue, and stronger inferences often use more than one.
- action;
- dialogue;
- reaction;
- description;
- sequence;
- contrast.
If the student cannot point to a clue, confidence should drop.
Possible Does Not Mean Wrong
A possible interpretation may be reasonable in real life. The issue is whether the passage gives enough evidence.
This distinction matters because comprehension rewards evidence-controlled meaning, not imaginative expansion.
Use the Confidence Ladder
- What is directly stated?
- What clue suggests more?
- How strong is the clue?
- Could another interpretation fit?
- What is the strongest conclusion the evidence justifies?
Overconfidence Example
If a character checks the clock twice, “she was waiting for something” may be supported. “She was terrified she would miss an important flight” is possible but too specific unless the passage provides more clues.
Underconfidence Example
Some students refuse to infer unless the exact answer appears in the text. They need the opposite repair: learning that several clues can support a reasonable conclusion even when the wording is not copied directly.
Inference Is a Bounded Claim
A useful pattern is:
Clue(s) → interpretation → boundary.
The boundary asks: what does the passage not prove?
Calibrate Emotion Words
- worried vs terrified;
- annoyed vs furious;
- uncertain vs afraid;
- pleased vs overjoyed.
Vocabulary precision helps inference precision because stronger words require stronger evidence.
Calibrate Motive
A character may act because of a goal, belief or feeling. Students should avoid inventing motives simply because they fit the story.
- What happened before the action?
- What did the character say?
- What changed after new information?
A Confidence Matrix
| Answer pattern | Interpretation | Repair |
|---|---|---|
| No clue | Guess | Find evidence first |
| One weak clue | Possible | Lower claim strength or find more support |
| Several converging clues | Supported | Build concise inference |
| Exact statement in text | Literal | Do not call it inference unnecessarily |
How a 3-Pax Class Calibrates Confidence
In a maximum three-student class, learners can rate different answers as literal, supported or merely possible, then defend the rating with evidence. This makes confidence visible rather than hidden.
Current MOE English Context
The MOE Primary English Language Syllabus 2020 develops students’ ability to construct meaning from texts and apply language strategies across contexts. Calibrated inference is one part of disciplined reading comprehension.
Signs Inference Confidence Is Improving
- Students cite evidence before committing to a claim.
- Overly strong emotion words decrease.
- Literal and inferred information are separated.
- Alternative interpretations are considered.
- Answers become more concise and better supported.
Possible vs Supported: Almost-Code Summary
CLAIM:
literal?
inferred?
IF inferred:
find_clues()
compare_alternatives()
choose_strongest_supported_claim()
set_boundary()
IF no_support:
downgrade_to_possible()
OUTPUT:
calibrated_inference
less_guessing
stronger_evidence_control