Primary 5 Science Tuition | Bedok is a year-specific guide for families comparing Primary Science tuition Singapore, P5 Science tuition in Bedok, a Science tutor or Science tuition centre for upper-primary learners, and 3-pax small-group tuition that prepares students for the cumulative demands of Primary 5 and the PSLE runway. At P5, Science becomes less forgiving of isolated memorisation. Students need to connect concepts across topics, reason from experiments, interpret diagrams, tables and graphs, use scientific vocabulary precisely, discriminate among MCQ options and construct structured explanations that link evidence to mechanism.
Searches for P5 Science tuition Bedok, Primary 5 Science tutor Bedok, Primary Science tuition Singapore, MOE Primary Science syllabus, SEAB PSLE Science, process skills, scientific inquiry, MCQ, structured questions, open-ended reasoning, keywords, experiments, fair tests, diagrams, tables, graphs, data interpretation, application, answering techniques, exam preparation, PSLE readiness and 3-pax small-group Science tuition are all pointing towards the same core problem: can a learner recognise and use the relevant scientific relationship when the question changes form? Strong tuition should align with the current MOE Primary Science syllabus while building toward the assessment objectives and revised format described by SEAB for PSLE Science.
This Bedok article is a local discovery and routing page within eduKateSG’s established Science system. It does not create a competing subject hub and it does not claim that eduKateSG has a physical tuition centre in Bedok. The broad owners remain the Science Learning Hub, Primary Science Tuition Singapore and the Primary Science Tuition branch. Bedok already has broad legacy Science owners on eduKateSG, including Bedok Primary Science Tuition and Primary Science for Bedok Families; this page narrows the intent to Primary 5 rather than replacing those broader resources.
Primary 5 Is Where Science Becomes Cumulative
Primary 5 is a change in load, not merely a change in chapter number. Earlier concepts do not disappear when new ones are introduced. The learner is expected to carry prior knowledge forward, integrate it with newer systems and processes, and recognise which ideas matter inside increasingly unfamiliar questions. A student who survived lower-primary Science by remembering recent worksheets can suddenly appear to decline even when effort has not changed.
The reason is often structural. A child may know individual facts but not relationships. When a question combines a diagram, a data table and an experiment, the student must coordinate several operations: understand the setup, identify what changed, retrieve the relevant concept, compare evidence, infer a relationship and communicate it precisely. A failure at any step can make the whole answer look wrong.
Good P5 tuition therefore needs two clocks. One clock teaches the current school syllabus. The other keeps earlier knowledge retrievable. If every lesson follows only the newest chapter, revision debt accumulates silently. By P6, the student may be trying to rebuild years of Science under examination pressure. P5 is the year to prevent that debt from compounding.
The Current MOE Science Framework Is More Than Content Coverage
The MOE Primary Science syllabus frames learning through Core Ideas, Practices and Values, Ethics and Attitudes. The five themes—Diversity, Cycles, Systems, Energy and Interactions—provide recurring structures across topics. A tuition programme that merely races through chapter content can therefore miss the deeper purpose: students need to use scientific ideas, evidence and inquiry practices together.
At P5, the themes become especially useful because they let learners connect older and newer topics. Systems thinking can travel from plant and human structures into circuits and other functional arrangements. Interactions encourage students to ask which factor affects which outcome. Energy questions can be understood by tracing sources, transfers and observable changes. Cycles demand attention to sequence, repetition and conditions. Diversity demands classification by evidence rather than resemblance.
The Primary 5 Learning Target: Build Scientific Models That Travel
A scientific model is a compact explanation of how a system or process works. It may be expressed in words, arrows, labels, a simple diagram or a relationship between variables. The purpose of a model is not to look impressive. It gives the learner something stable to operate when surface details change.
For example, a student should not need the exact classroom apparatus to reason about a fair test. The objects can change, but the logic stays: identify the factor deliberately changed, identify the outcome measured, keep other relevant factors comparable, compare results and decide whether the evidence supports a conclusion. That model travels across contexts.
The same principle applies to systems. A learner who understands part → function → connection → consequence can use that structure across plants, human body systems and electrical systems. A learner who understands condition → process → effect → evidence can use it across many interaction and energy questions. P5 tuition should make these transferable structures explicit.
Knowledge With Understanding Is Not the Same as Recognition
SEAB’s current Science assessment objectives distinguish knowledge with understanding from application of knowledge and scientific inquiry. That distinction is useful well before P6. A student may recognise a correct sentence in notes yet fail to produce it independently. Another may produce a fact but fail to apply it to evidence in the question.
P5 tuition should therefore move students through a sequence: learn, retrieve, recognise, apply, explain, vary and revisit. The learner first builds an accurate concept. Next, the concept must be recalled without the page open. Then it must be selected in an unfamiliar situation. Finally, the student has to explain why it applies and return to it after a delay.
A Diagnostic Framework for P5 Science
A total mark is a summary; it is not a diagnosis. Useful analysis separates at least eight possible failure points. Concept: the scientific idea is inaccurate. Retrieval: the idea was learned but is unavailable. Selection: the learner cannot choose the relevant concept from a mixed context. Evidence: visual or numerical information is misread. Inquiry: variables, controls or method logic are misunderstood. Inference: the conclusion does not follow from the evidence. Communication: the answer is scientifically incomplete or imprecise. Execution: timing, attention or checking breaks down.
Two children with the same percentage may therefore need completely different tuition. One may have broad conceptual gaps; another may understand the Science but lose marks through graph reading and incomplete explanations. A programme that gives both students the same correction sheet may increase practice without increasing accuracy.
Bedok Search Intent and the Existing eduKate Science Ecosystem
Current search results around Bedok and eastern Singapore commonly surface tuition centres and tutors promising MOE-aligned Primary Science, P4–P6 progression, PSLE preparation, experiments, answering techniques, scientific keywords and small-group support. Those are legitimate comparison terms, but families should also ask a harder question: what exactly changes after a student gets a question wrong?
eduKate already has broad Bedok Science pages across its ecosystem. This year-specific P5 route therefore does not need to behave as another broad “Bedok Science tuition” owner. Its job is narrower: explain the learning demands of Primary 5, diagnose the transition into upper-primary Science and connect that intent back to the central Science architecture. This reduces unnecessary competition among eduKate’s own pages.
Adrian: P5 Transfer Requires More Than Another Similar Example
Adrian understands a concept when the new question resembles the worked example. When apparatus, labels or surface details change, he hesitates. His tutor does not solve this by showing him twenty more near-identical examples. Instead, the tutor places two different-looking questions side by side and asks Adrian to identify the relationship they share.
Next, Adrian explains which details are scientifically important and which are decorative. Then one condition is changed. He predicts what should happen and justifies the prediction. Later, the same concept reappears in a mixed paper. The purpose is to loosen the concept from the original page and make it portable.
Jo: Build Mechanism Chains Before Writing Full Answers
Jo often knows the right keyword but stops too early. If a question asks why an outcome occurs, she names the process but does not connect the process to the stated evidence. Her tutor teaches her to build a short mechanism chain before writing.
The chain might take the form condition → process → change → observable outcome. In another question it may be structure → function → effect. Jo does not copy a universal sentence template; she uses the chain to ensure that the causal middle is present. Her final answer often becomes shorter because the thinking has become clearer.
Ben: Experiments Are Arguments Built From Comparisons
Ben can identify a changed variable when the worksheet asks directly, but he struggles when the experimental design is embedded inside a long scenario. His tutor changes the question: “What claim is this experiment trying to test, and what comparison would make that claim believable?”
That wording helps Ben understand the purpose of control. A fair test is not a vocabulary exercise. It is an attempt to make one explanation more credible by preventing other relevant factors from changing at the same time. Once Ben sees experiments as arguments from evidence, variables become roles rather than labels.
Aisha: Mixed Retrieval Prevents Revision Debt
Aisha revises faithfully but tends to study one chapter at a time until it feels fluent. Months later, earlier topics fade. Her tutor introduces a cumulative retrieval system. Each lesson begins with a few questions from older themes, including one that requires explanation rather than recognition.
The purpose is not constant testing for its own sake. Delayed retrieval tells Aisha which knowledge remains available. It also normalises the idea that old Science is still live Science. By the end of P5, she should not need a complete rebuild before P6 revision begins.
Ryan: MCQ Review Should Reveal Why a Distractor Looked Plausible
Ryan marks MCQ questions as right or wrong and moves on. That wastes diagnostic information. A well-designed distractor often represents a common misconception, a reversed relationship, a missed condition or an attractive but irrelevant fact. His tutor asks him to explain why his wrong option seemed reasonable.
Ryan’s correction record therefore includes the decision he made, the clue he ignored and a rule for the next attempt. “I chose the biggest number” becomes “I must compare changes from the starting values, not only final values.” That replacement decision can transfer to another dataset.
Mira: Convert Visual Information Into a Working Representation
Mira can read lengthy passages but becomes uncertain when information is distributed across a diagram, labels and a table. She learns to compress the representation. She marks what changes, what is measured, the direction of movement, the relevant units and the comparison that matters.
This is not decorative annotation. It reduces the amount of information that must be held mentally. Once the structure is visible, Mira can apply the scientific model. P5 questions often reward the learner who can transform a complicated display into a simpler relationship.
Clara: Answer the Question Asked, Not the Topic Remembered
Clara knows a great deal and likes to show it. On structured questions, she sometimes writes everything she remembers about a topic. The answer may contain true Science but still fail to address the exact comparison or causal relationship required.
Her tutor makes her identify the command, evidence and required endpoint before writing. “Compare” needs the relevant relationship between cases. “Explain” needs the mechanism. “Predict” needs a stated outcome plus scientific justification. “Suggest” needs a plausible response grounded in the scenario. Knowledge becomes useful only when it is scoped correctly.
Ethan: Build a Start Routine for Unfamiliar Multi-Part Questions
Ethan is comfortable with single-step questions but freezes when a page contains a long setup and several parts. His tutor teaches an entry routine: identify the system, identify what changes, identify the evidence, identify what each sub-question asks, then map the parts onto known concepts.
The routine prevents him from treating the entire page as one giant problem. Each sub-question has a job. Sometimes an early part asks for observation while a later part asks for explanation. Sometimes the same setup supports both an MCQ-like decision and a constructed response. Breaking the task into roles lowers cognitive load.
Scientific Inquiry: Prediction Is Not Guessing
A scientific prediction should arise from a model. The learner identifies what will change, applies a known relationship and states the expected outcome. P5 tuition can train prediction by changing one condition after a standard problem is solved. What if the amount changes? What if the material changes? What if a component is removed? What if the starting condition is different?
These counterfactual questions reveal whether the student understands the mechanism. A memorised answer often collapses when one condition changes. A working model adapts.
Hypotheses: Connect a Factor to a Measurable Outcome
When students formulate a hypothesis, vague language creates vague investigations. A useful hypothesis identifies a relationship between a factor and an observable or measurable outcome. The exact wording should suit the level and context, but the underlying logic is consistent: if this condition changes, what outcome is expected, and why is that expectation scientifically reasonable?
The tutor should also discuss what evidence would fail to support the hypothesis. Scientific inquiry is not about making the experiment prove what the student already believes. It is about designing a comparison that can produce informative evidence.
Fair Tests: Ask Which Alternative Explanation Has Been Removed
P5 students should move beyond chanting “keep everything the same except one variable.” In real questions, not everything needs to be identical; what matters is whether relevant competing factors could explain the result. The student should be able to say why a particular control is necessary.
For example, if two setups receive different amounts of light and also begin with different quantities of material, the outcome cannot be attributed confidently to light alone. The second difference creates an alternative explanation. This reasoning prepares students for method-evaluation questions and for the application-and-inquiry demands described in the SEAB Science assessment objectives.
Reliability, Repeated Trials and Variation
Students often assume that one measurement is automatically trustworthy. P5 is a good stage to introduce the idea that biological and physical observations can vary, instruments have limits and repeated measurements can strengthen confidence. Repeated trials do not magically make a bad method fair, but they can show whether a result is consistent.
A tutor can ask students to compare two datasets: one with a single dramatic reading and another with repeated measurements that cluster. Which gives stronger support, and why? This develops evidence judgement without requiring advanced statistics.
Tables: Compare Like With Like
P5 table questions increasingly punish casual comparison. Students need to inspect headings, units, starting conditions and the variable being changed. Comparing final values may be inappropriate if starting values differ. Looking at one row may be insufficient if the question asks for a trend.
A reliable sequence is: read the structure, select the relevant data, describe the pattern, then explain only if required. This keeps evidence and interpretation separate long enough for each to be checked.
Graphs: Describe Before You Explain
Students often look at a graph and immediately tell a scientific story. The safer sequence is to establish what the graph actually shows. Read both axes and units. Note scale. Identify the interval or comparison. Describe the relationship. Then bring in the scientific concept if the command asks for explanation.
This discipline matters because expectations can distort reading. A learner may believe a trend should increase and therefore overlook a plateau or decrease. Data interpretation requires the evidence to lead.
Diagrams: Ask What Information Is Encoded Spatially
A diagram may encode distance, connection, direction, sequence, structure or relative position. P5 students should learn to ask what information the picture contains that the prose does not. Labelling everything is not always useful; selective marks should reveal the relationship needed for the question.
For systems, arrows can show flow or transfer. For forces or light, a sketch can externalise spatial reasoning. For life processes, a sequence can prevent steps from being reversed. Representation is part of thinking, not merely presentation.
Concept Pairs That Need Explicit Contrast
Many errors come from concepts that live close together. Heat and temperature are related but not interchangeable. Observation and inference are different. Mass and force are different quantities. A structure and its function are not the same statement. Evaporation and boiling should not be collapsed into one idea. Food for a plant and substances absorbed from the soil are not identical concepts.
Contrast practice is efficient. Put two related concepts side by side and ask: what do they share, how do they differ, what evidence distinguishes them, and what wrong sentence would reveal confusion? This makes the boundary between concepts explicit before a distractor exploits it.
Scientific Vocabulary: Build Meaning, Boundary and Use
A P5 vocabulary entry should do more than define a term. It should show the concept’s boundary. What does the word mean? What does it not mean? Which evidence would make it relevant? Which other term is commonly confused with it? What sentence would use it correctly in context?
This approach prevents “keyword hunting,” where a child scans a question for a familiar word and then writes a memorised statement. In strong Science, vocabulary follows reasoning. The student first identifies the relationship, then uses the term that names it precisely.
MCQ: Train Discrimination Before Speed
Under the revised Standard PSLE Science format examined from 2026, multiple-choice performance remains a major component of the examination. P5 tuition should not turn into constant timed PSLE papers, but it should train the reasoning that MCQs compress: identify the scientific principle, read every condition, reject distractors for a reason and avoid changing a correct answer without new evidence.
For selected questions, require the student to justify the correct option and reject the strongest distractor. Ask what misconception the distractor represents. Ask which word or condition changes the answer. These habits improve discrimination before speed is emphasised.
Structured Questions: Build the Causal Middle
The current PSLE Science framework also requires constructed responses in structured questions. P5 students should gradually learn to produce explanations rather than rely on recognition. A strong response often needs an explicit relationship between evidence and outcome.
One useful thinking sequence is evidence → concept → mechanism → requested conclusion. Not every answer needs four visible clauses. The scaffold ensures that the causal middle has not been skipped. The final wording should stay proportional to the question.
Open-Ended Reasoning Is Not a Model-Answer Copying Exercise
Model answers can illustrate completeness, but they should not become scripts detached from the science. A learner who memorises one sentence for one question may fail when the context changes. A learner who understands the mechanism can reconstruct an appropriate answer.
SEAB’s 2026 article on thoughtful PSLE assessment design is useful here because it explains how scientifically correct understanding may be demonstrated through more than one valid approach when the reasoning is sound. See What Thoughtful Assessment Design Looks Like in the PSLE. The teaching implication is clear: answer technique should support understanding, not replace it.
Answering Techniques Should Be Derived From Science
Technique can help students manage questions, but it should not become a substitute for understanding. A rigid phrase may work on one familiar question and fail when conditions change. Good technique is conditional: identify the command, locate the evidence, select the scientific relationship, construct the needed link, then check whether the endpoint has been answered.
The tutor should explain why a response works. Students learn the architecture of an answer rather than memorising cosmetic sentence shapes. This makes the technique transferable and reduces dependence on one teacher’s preferred wording.
Interleaving: Remove the Chapter Label
Blocked practice is useful when a concept is new because it reduces selection demands. But staying blocked for too long gives the learner an unrealistic clue. A worksheet titled “Forces” announces which concept family to use. An assessment does not.
P5 interleaving can begin with two or three topic families, then broaden. Students learn to diagnose the question before solving it. The difficulty often rises temporarily because the prompt has been removed. That temporary struggle is informative: it shows whether recognition has become independent.
Spaced Retrieval: Make Earlier Science Stay Available
Spacing should be built into the lesson calendar rather than left for year-end revision. A small retrieval set after several days, another after several weeks and a mixed checkpoint later can reveal which concepts are decaying. Concepts that are repeatedly forgotten deserve different treatment from concepts that remain stable.
The goal is not endless testing. It is maintenance. By P6, the learner should have a large body of Science that remains available without re-learning from zero.
Build a Living Knowledge Map
P5 students benefit from seeing how ideas connect. A living knowledge map can start with the five themes and add concepts, common evidence types, typical misconceptions and relationships as the year progresses. It should remain compact enough to use rather than becoming another giant set of notes.
The map helps when a mixed question arrives. Instead of searching memory for a page, the student asks which family of relationship is active. Over time, this builds faster concept selection and makes revision more coherent.
Error Correction Needs a Second Attempt
Reading a model answer after making a mistake can create an illusion of repair. The student recognises the correct reasoning while it is visible but may not be able to reconstruct it later. Every important correction should therefore include a second attempt on a changed question after the answer is removed.
The sequence can be short: diagnose the error, explain the replacement idea, solve one near-transfer item, then revisit the concept after a delay. A correction is complete only when the learner can perform without the model in front of them.
Three-Pax Small-Group Science: Personalisation Must Be Observable
A three-student tutorial should allow the tutor to see the reasoning of every learner. If one student needs concept reconstruction, another needs evidence-reading practice and a third needs answer compression, the lesson should make room for those differences. Small groups are useful when the feedback loop is short.
Adrian can explain why two different contexts share a concept. Jo can challenge a missing causal step. Ben can identify what makes an investigation fair. Aisha can retrieve an older concept. Ryan can explain why a distractor is tempting. Mira can translate a diagram into a relationship. Clara can edit an overlong answer. Ethan can demonstrate an entry routine for a complex question. The group becomes a reasoning environment, not three children silently completing identical pages.
A 90-Minute P5 Science Tutorial Architecture
A productive session might begin with ten minutes of cumulative retrieval. The tutor then spends twenty minutes repairing or teaching one model. Guided application follows, with questions chosen to expose common misconceptions. Students then complete an independent transfer set containing a changed context. The final segment reviews errors by mechanism and assigns a narrow retrieval target.
This is not a rigid timetable. Some misconceptions need longer reconstruction. Some groups need more data work. The architecture matters because each lesson should connect new learning to old knowledge, independent performance and future retrieval.
Homework Should Be Small Enough to Review Properly
Homework is useful when the tutor can explain why each item is there. One set may target retrieval; another may target graph interpretation, fair-test logic, scientific vocabulary, mixed concept selection or structured explanations. High volume without analysis can hide recurring errors because the learner simply moves from one page to the next.
A good homework loop includes review. Which items were wrong? Which errors share a cause? Which corrected concept needs a fresh question? What should be retrieved again next week? That loop turns homework into evidence about learning.
School Examination Review: Do Not Waste the Script
A returned school paper is a map of actual performance under constraints. Categorise the losses. Which marks came from missing concepts? Which came from reading the wrong variable? Which were due to incomplete mechanisms? Which MCQ distractors repeatedly worked? Which questions were left late? Which topics were old rather than recently taught?
Then choose one or two high-leverage repair targets. Reworking the whole paper from start to finish can feel thorough but may blur the patterns. The purpose of review is to change future decisions.
A Twelve-Week P5 Science Build
Weeks 1–2: Baseline and concept repair. Use mixed diagnostic questions and identify the first weak links. Weeks 3–4: Representation. Move between words, diagrams, tables and graphs. Weeks 5–6: Inquiry. Work on variables, fair tests, prediction and method evaluation.
Weeks 7–8: Explanation. Build mechanism chains and concise responses. Weeks 9–10: Interleaving. Mix old and new topics and remove chapter cues. Weeks 11–12: Delayed transfer. Revisit repaired concepts under new contexts and moderate time pressure. The result should be a cleaner learning system, not just a temporary score increase.
What P5 Students Should Be Able to Do by Year End
- Retrieve major earlier concepts without extensive re-teaching.
- Recognise which concept family applies in mixed-topic questions.
- Read diagrams, tables and graphs before forming a conclusion.
- Explain why an investigation is or is not a fair test.
- Distinguish observation, inference, prediction and explanation.
- Use scientific vocabulary inside correct relationships.
- Reject MCQ distractors for scientific reasons.
- Construct structured responses with evidence and mechanism.
- Review errors by cause instead of copying model answers.
- Carry a manageable cumulative revision system into Primary 6.
How Parents Can Support P5 Science
Parents do not need to supply model answers. Better questions are often simpler: “What evidence are you using?” “What changed?” “What did the experiment measure?” “What else could have caused this?” “Can you draw the mechanism?” “What would happen if one condition changed?” These prompts keep ownership with the child.
Parents can also look for patterns over time. If every Science session becomes a long rescue operation, the child may be carrying too much revision debt. If marks fluctuate sharply, ask whether performance depends on recently taught topics. If open-ended answers are weak despite strong oral explanations, target written output specifically.
Bedok Families: Convenience and Instructional Quality Need to Work Together
A weekly P5 class must fit school, travel, meals and rest. Proximity is therefore a real constraint. But travel convenience should not be the only criterion. Families should ask how quickly the tutor can identify a misconception, whether older topics are retrieved, whether students actually write and explain during class, and whether corrections are tested again rather than merely marked.
Current search language around Bedok Science tuition often stresses experienced tutors, MOE-aligned notes, PSLE preparation and small groups. Those features are meaningful only when they produce observable learning behaviours. The strongest evidence is not the brochure phrase but the student’s improved ability to reason independently on a changed problem.
What to Ask When Comparing P5 Science Tuition in Bedok
- How are P3–P4 concepts kept retrievable while P5 content is taught?
- How does the tutor diagnose misconceptions rather than only mark answers?
- How are experiments, variables, fair tests and method evaluation taught?
- How often do students work with unfamiliar diagrams, tables and graphs?
- How are MCQ distractors used diagnostically?
- How are structured explanations taught without over-reliance on memorised scripts?
- How is scientific vocabulary connected to mechanisms?
- How is mixed-topic practice introduced before P6?
- What does individual feedback look like in a 3-pax class?
- How does the programme distinguish P5 foundation-building from final PSLE drilling?
Frequently Asked Questions: Primary 5 Science Tuition | Bedok
Is P5 the right time to start PSLE Science preparation?
P5 is an excellent time to build PSLE capabilities without turning every lesson into full-paper drilling. Cumulative retrieval, mixed recognition, scientific inquiry, data interpretation and structured explanation can all be strengthened before final-year pressure increases.
Should P5 students start doing complete PSLE papers?
Occasional exposure can be informative, but large volumes of full papers may be inefficient if major concepts or processes are still unstable. Targeted sets often repair weaknesses faster. Full-paper practice becomes more useful as the student’s knowledge base and stamina mature.
Why does my child score well in topical worksheets but poorly in exams?
Topical worksheets provide a hidden clue about which concept to use. Exams mix topics and require concept selection. Interleaved practice and unfamiliar contexts help train recognition.
Are keywords enough for open-ended Science?
No. Scientific terms matter, but they must sit inside a correct relationship. A keyword without the causal or comparative link can leave the answer incomplete.
How should a P5 student revise old topics?
Use spaced retrieval rather than restarting every chapter from the beginning. Attempt a small mixed set from memory, identify what has decayed, review only the weak components and test them again later.
Does eduKateSG operate a physical Bedok Science centre?
This page does not make that claim. It is a local discovery and learning route for families searching from Bedok. Current teaching locations, formats and availability should be confirmed directly with eduKateSG.
The Bedok Primary 5 Science Route
Primary 5 should convert earlier knowledge into a connected, retrievable and transferable system. The learner needs to maintain old concepts, build new models, reason from experiments, interpret representations, reject plausible distractors, explain mechanisms and develop a correction process that changes future decisions. That is the bridge between knowing chapters and being ready for Primary 6.
Use the Science Learning Hub, Primary Science Tuition Singapore and the Primary Science Tuition branch for broader subject navigation. The Bedok sequence also includes Primary 4 Science Tuition | Bedok, Primary 6 Science Tuition | Bedok and PSLE Science Tuition | Bedok. For exact cohort rules, syllabus details and examination format, official MOE and SEAB documents remain the final reference.