Series route: This guide belongs to the complete PSLE Learner’s Guide volume directory. The original PSLE Learning Guide supplies the wider English, Mathematics and Science route, while PSLE English provide verified subject-specific continuation.
Find your route through this guide
This volume solves one precise learner problem: the learner chooses what and which by habit, without checking whether the question opens an unrestricted field or asks for selection from a known limited set. The aim is not to memorise a slogan. It is to understand the mechanism well enough to recognise the idea when the wording, numbers, diagram or context changes. A Primary 6 learner should be able to explain the decision, perform it under time pressure, and detect a tempting answer that only looks familiar.
- Understand the mechanism before memorising a form.
- Study the worked examples and compare near-misses.
- Repair common misconceptions with targeted tests.
- Complete graded independent practice before reading the answers.
- Use delayed transfer and examination control to make the skill durable.
- Use the parent and tutor notes only after the learner has attempted the work.
The problem beneath the mistake
“What book are you reading?” asks openly for the book’s identity. “Which book are you reading—the blue one or the green one?” places the answer inside a small named set.
The distinction is not simply that which is more formal. It is a choice-structure distinction. What opens the field; which points into a bounded field.
A set need not be written after the question word. If three folders are visible on a desk, “Which folder is yours?” is natural because the shared scene supplies the choices.
A category can still be broad even when a noun follows. “What language do you speak at home?” does not normally present a fixed menu. The learner supplies a category member from general knowledge.
Which can ask about more than two alternatives. A list of ten activities is still bounded. The number of options matters less than whether the set is identifiable.
What can introduce exclamations and noun clauses as well as questions: “What a surprise!” and “I know what she chose.” Those structures require separate grammar checks, although the open-information idea can still help in noun clauses.
Which can act as a relative word: “The path, which was narrow, led uphill.” Here it connects a relative clause rather than requesting a selection. Punctuation and sentence role matter.
Questions in comprehension may use what even when the passage contains the answer. “What caused the delay?” asks for information, not for selection from a displayed menu.
A multiple-choice layout does not automatically force which inside the question stem. The original wording may ask an open information question that is later answered through provided options.
When both forms sound possible, meaning changes subtly. “What subjects interest you?” invites any subjects. “Which subjects interest you?” implies a known curriculum, list or previously mentioned set.
The answer form helps diagnose the question. If the natural answer begins “one of these…” or points to a visible option, which is likely. If it supplies a name, type or description without a defined menu, what is likely.
The reliable routine is identify the requested information, locate any bounded set, test whether pointing or selecting would answer, then check whether the word is actually a question word or a relative connector.
Core decision: Use what when the speaker asks openly for information or category; use which when the answer is chosen from a particular limited set that is stated, visible or already shared.
A useful rule has three parts: a trigger, an operation and a check. The trigger tells you when the rule applies. The operation tells you what to do. The check tells you whether the result still matches the meaning of the question. Learners who remember only the operation often apply it in the wrong place; learners who remember only an example often freeze when the surface details change.
Mechanism: why the rule works
1. Name the requested information
State whether the question asks for an object, category, action, reason or description.
Diagnostic question: What must the answer supply? Reliable response: Write a neutral answer frame.
2. Search for the choice set
Look in the sentence, previous lines, diagram and shared scene for named or visible alternatives.
Diagnostic question: Are the possible answers presented as a set? Reliable response: List the actual candidates if they exist.
3. Test open generation
Imagine answering without seeing a menu.
Diagnostic question: Could many unlisted answers be equally possible? Reliable response: If yes, prefer what.
4. Test pointing or selection
Imagine pointing to one member of the known set.
Diagnostic question: Would that complete the task? Reliable response: If yes, prefer which.
5. Check grammatical role
Decide whether the word asks a question, introduces a noun clause or begins a relative clause.
Diagnostic question: Is a selection even being requested? Reliable response: Do not apply the question rule to every which.
6. Preserve intended scope
When rewriting, keep the answer field equally open or equally bounded.
Diagnostic question: Did the revision add or remove a choice set? Reliable response: State the implied options explicitly if needed.
Worked examples: make the decision visible
Do not race through these examples. Before reading each explanation, name the evidence that matters and the information that does not. Then compare the correct route with the tempting route. That comparison is where the transferable learning sits.
Worked example 1: Open identity
Task. Choose: What/Which is your favourite story?
Reasoning. No limited list has been supplied, so the listener generates a title from an open field. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. What is your favourite story?
Independent check. A later answer can be any story, not one of stated options. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 2: Visible selection
Task. Three pencils lie on the table. Ask the owner to identify hers.
Reasoning. The shared scene supplies a bounded set of three. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. Which pencil is yours?
Independent check. The answer can point to one of the visible pencils. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 3: Named alternatives
Task. Choose: What/Which route will you take, A or B?
Reasoning. A and B form an explicit set. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. Which route will you take, A or B?
Independent check. The answer must select from the two routes. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 4: Open cause
Task. Choose: What/Which caused the glass to crack?
Reasoning. The question asks for causal information without presenting candidates. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. What caused the glass to crack?
Independent check. Do not assume a list merely because the passage contains clues. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 5: Known syllabus list
Task. A form lists Art, Music and Drama. Ask about the pupil’s selection.
Reasoning. The printed form creates the bounded field. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. Which activity did the pupil select?
Independent check. The answer names one listed activity. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 6: Relative which
Task. Interpret: The ruler, which was bent, gave an unreliable reading.
Reasoning. Which introduces extra information about the ruler; it does not ask a question. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. The relative clause explains the ruler’s condition.
Independent check. The commas help mark the clause. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 7: Noun-clause what
Task. Interpret: Explain what the character realised.
Reasoning. What introduces the content to be explained. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. State the realisation using passage evidence.
Independent check. No direct question mark is required. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Worked example 8: Meaning contrast
Task. Compare: What colour do you like? Which colour do you like?
Reasoning. The first is generally open; the second implies a known palette or set in context. The controlling idea is whether the listener must generate an answer from an open field or select one or more members from an identified set. This prevents a learner from choosing an answer merely because one word, number or observation resembles a remembered example.
Answer. Both can be grammatical, but their implied answer fields differ.
Independent check. Supply a context that makes each natural. A good check approaches the conclusion from a different direction. Repeating the same steps more slowly may repeat the same hidden assumption.
Deepening lab: compare cases, conditions and limits
A method becomes independent when you can say where it applies, where it stops and which change would alter the conclusion. The cases below deliberately vary one structural feature at a time. For each case, pause before the analysis and predict whether the core rule remains available. Then name the exact condition that controls the decision. This is harder than repeating a worked solution, but it is also closer to what an unfamiliar examination item demands.
Deepening case 1: Open occupation
Case. What job interests you?
Analysis. No menu is supplied. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Give one possible answer. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 2: Interview shortlist
Case. Which job from the shortlist interests you?
Analysis. The shortlist bounds the field. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Name the limiting phrase. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 3: Visible cups
Case. Several cups are on a tray.
Analysis. Shared visibility can license which. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Point to the chosen cup. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 4: Unknown cause
Case. The cause has not been offered as alternatives.
Analysis. What invites an explanation. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Write a cause question. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 5: Known causes
Case. A table lists heat, force and chemical action.
Analysis. Which asks the learner to select from the list. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Cite the selected row. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 6: Many options
Case. Twenty books appear in a catalogue.
Analysis. The set is large but still bounded. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Use which naturally. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 7: Broad category
Case. What sport do you play?
Analysis. The answer field is the broad category of sports. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Do not invent a fixed list. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 8: School subjects
Case. Which subject do you prefer?
Analysis. A school timetable may supply the understood set. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. State the context. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 9: Multiple choice
Case. The page provides four options after an open stem.
Analysis. Option layout and stem grammar are separate. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Keep the original question meaning. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 10: Which as relative
Case. The experiment, which lasted an hour, was repeated.
Analysis. No selection occurs. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Identify the antecedent. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 11: What as exclamation
Case. What a careful answer!
Analysis. The form expresses evaluation, not a request. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Classify the sentence type. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 12: What as noun clause
Case. I remember what she said.
Analysis. The clause names remembered content. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Convert to a direct question. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 13: Which one
Case. Which one is correct?
Analysis. One points to the bounded candidates. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Name the candidate set. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 14: What kind
Case. What kind of material is this?
Analysis. The question asks for classification from a broad field. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Use evidence in the answer. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 15: Which kind
Case. Which kind shown in the key matches the sample?
Analysis. The key supplies bounded categories. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Follow the mapping. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 16: Both possible
Case. What/which route is safest?
Analysis. Context determines whether routes are open or already mapped. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Write two contexts. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 17: Rewriting risk
Case. Changing what to which can imply a list that was not present.
Analysis. Meaning changes even if grammar remains possible. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Restore scope. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 18: Answer diagnostic
Case. “The second diagram” points into a set.
Analysis. A selection answer suggests which. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Check the diagram labels. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 19: Open answer diagnostic
Case. “A sudden loss of support” supplies information.
Analysis. An explanatory answer suggests what. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. State the mechanism. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Deepening case 20: Final audit
Case. Information target, set, generation, selection, grammatical role and scope complete the check.
Analysis. Each blocks a different error. Notice that the answer depends on whether the listener must generate an answer from an open field or select one or more members from an identified set; changing a decorative detail would not be enough to change the reasoning.
Challenge. Audit ten questions. State the conclusion in one precise sentence, then add one sentence that names the evidence or relationship that makes it valid.
Finish the lab with a blank-page recall. Write the rule, its conditions, one confirming example, one near-miss and one independent check from memory. Compare that page with the article only after finishing. If the rule is correct but the condition is missing, the learning is not yet safe: an overgeneralised rule can produce confident errors. If the condition is correct but no example can be generated, practise construction rather than rereading.
Common misconceptions and targeted repairs
Misconception 1: Which is simply formal what
What the learner may think. The words differ only in register.
Why it fails. They normally shape the answer field differently. The error feels efficient because it removes a decision, but it also removes the evidence boundary that makes the answer valid.
Targeted repair. Identify open versus bounded choice. Use this short test immediately: Where is the set?
Misconception 2: Only two choices use which
What the learner may think. Which applies only to pairs.
Why it fails. Any identifiable limited set can support which. The error feels efficient because it removes a decision, but it also removes the evidence boundary that makes the answer valid.
Targeted repair. Count boundaries, not merely options. Use this short test immediately: Is the list finite and shared?
Misconception 3: A noun after the word settles it
What the learner may think. What language and which language follow the same automatic rule.
Why it fails. The surrounding choice structure controls meaning. The error feels efficient because it removes a decision, but it also removes the evidence boundary that makes the answer valid.
Targeted repair. Inspect context and implied options. Use this short test immediately: Open field or known list?
Misconception 4: Multiple choice requires which
What the learner may think. Every MCQ stem must use which.
Why it fails. The stem can ask an open information question while options provide answer selection later. The error feels efficient because it removes a decision, but it also removes the evidence boundary that makes the answer valid.
Targeted repair. Separate grammar from format. Use this short test immediately: What does the stem request?
Misconception 5: Every which asks a question
What the learner may think. Which always selects.
Why it fails. It can introduce a relative clause. The error feels efficient because it removes a decision, but it also removes the evidence boundary that makes the answer valid.
Targeted repair. Find the antecedent and punctuation. Use this short test immediately: Is there a question mark?
Misconception 6: Both are always interchangeable
What the learner may think. If both sound grammatical, meaning is unchanged.
Why it fails. One can add or remove an implied set. The error feels efficient because it removes a decision, but it also removes the evidence boundary that makes the answer valid.
Targeted repair. Paraphrase the answer field. Use this short test immediately: Could an unlisted answer fit?
Graded independent practice
Attempt each item without looking ahead. Write one line that identifies the controlling evidence before writing the final answer. Items 1–3 check recognition, 4–7 require explanation, and the later items change the surface form so that memory alone is insufficient.
Practice 1: Open field
Question. Choose: What/Which instrument can produce this sound? No list is given.
Explained answer. What.
Why this answer earns trust. The learner must identify an instrument from an open field. Transfer variation: Add a list and revise.
Practice 2: Bounded field
Question. Choose: What/Which of these instruments uses a vibrating string?
Explained answer. Which.
Why this answer earns trust. Of these explicitly limits the set. Transfer variation: Name the selected item.
Practice 3: Visible context
Question. Three doors are labelled A, B and C. Ask about the exit.
Explained answer. Which door is the exit?
Why this answer earns trust. The visible labels create a finite choice set. Transfer variation: Explain why what would sound more open.
Practice 4: Cause question
Question. Complete: ___ made the plant bend?
Explained answer. What made the plant bend?
Why this answer earns trust. The question requests causal information without a menu. Transfer variation: Supply a scientific answer only with evidence.
Practice 5: Relative clause
Question. Does which request a choice in “The lamp, which was hot, was switched off”?
Explained answer. No; it introduces a relative clause about the lamp.
Why this answer earns trust. The antecedent and commas show the role. Transfer variation: Remove the extra clause.
Practice 6: Noun clause
Question. Explain the role of what in “I know what happened.”
Explained answer. It introduces the content that is known.
Why this answer earns trust. It is not a direct interrogative in that sentence. Transfer variation: Turn it into a direct question.
Practice 7: Scope change
Question. How does “What method worked?” differ from “Which method worked?”
Explained answer. What is open unless context narrows it; which implies selection from known methods.
Why this answer earns trust. The answer fields differ. Transfer variation: Supply a methods table.
Practice 8: Large set
Question. Can which refer to one item in a list of fifty?
Explained answer. Yes, if the list is the shared bounded set.
Why this answer earns trust. Limited does not mean only two or three. Transfer variation: Point to the relevant entry.
Practice 9: Rule recall
Question. What is the fastest distinction?
Explained answer. What opens the answer field; which selects from a known limited set.
Why this answer earns trust. The set is the deciding evidence. Transfer variation: Recall after two days.
Practice 10: Transfer
Question. Edit fifteen mixed what and which questions.
Explained answer. Every choice identifies the requested information and the source of any bounded set.
Why this answer earns trust. Explanation prevents register guessing. Transfer variation: Include two relative clauses.
Delayed transfer: prove that the learning lasts
Transfer test 1: Answer-field sort
Sort twenty prompts into open generation, bounded selection and non-question clauses. Do this after a delay, without rereading the worked examples. The purpose is to test retrieval and decision-making, not immediate imitation.
Success criterion. Every classification cites the choice-set evidence. If the criterion is not met, return to the smallest missing decision rather than repeating the entire chapter.
Transfer test 2: Delayed context writing
After one week, write paired contexts where what and which are both grammatical but imply different scopes. Do this after a delay, without rereading the worked examples. The purpose is to test retrieval and decision-making, not immediate imitation.
Success criterion. Each pair makes the boundary visible. If the criterion is not met, return to the smallest missing decision rather than repeating the entire chapter.
Transfer test 3: Worksheet audit
Check questions, captions and relative clauses in a mixed worksheet. Do this after a delay, without rereading the worked examples. The purpose is to test retrieval and decision-making, not immediate imitation.
Success criterion. No which is treated as a question unless its grammatical role and set support that reading. If the criterion is not met, return to the smallest missing decision rather than repeating the entire chapter.
Examination control
1. Read the job before the story. Underline the command and the noun it governs. “Explain the change,” “state the word,” “find the fraction” and “identify the evidence” are different jobs even when the surrounding context is similar.
2. Separate given information from remembered information. Use only the facts supplied for the current task plus secure subject knowledge. Do not let a previous question, a familiar worksheet or an expected pattern silently become new evidence.
3. Write enough working to expose the decision. A marker or teacher should be able to see the relationship that justifies the answer. Extra sentences are not automatically stronger; the strongest working is concise, relevant and checkable.
4. Reserve a different-route check. For language, read the sentence for meaning and grammar. For mathematics, estimate or reverse the operation. For science, check the claim against the stated observation and a conservation or fair-test principle.
5. Stop when the requested job is complete. Unrequested speculation can contradict a correct answer. Add detail only when it explains the mechanism or satisfies another explicit part of the question.
Under time pressure, use a compact four-part margin routine: trigger → evidence → operation → check. Trigger names the feature that activates the method. Evidence identifies the exact words, values or observations allowed. Operation performs the language, mathematical or scientific step. Check asks whether the final statement answers the requested job, preserves units or referents, and avoids a stronger claim than the evidence supports.
A compact error log
After practice, do not record only “careless mistake.” Name the broken decision. Useful labels include trigger missed, evidence imported, relationship reversed, referent lost, unit dropped, operation correct but job wrong, and check repeated the same route. Add one prevention cue beside the label. A short accurate log is more useful than a long emotional account because it tells the learner what to do on the next item.
Review the log after two days and again after a week. Reattempt only one or two representative items first. If the method transfers, move to a mixed set. If it fails, repair the exact link between trigger and operation. This spacing matters because a solution can feel obvious while it is still visible and disappear when the page is closed.
Notes for parents and tutors
Ask “Where are the choices?” before correcting the question word. If the learner can point to a list, display or shared set, which becomes meaningful.
Use real menus, timetables and trays of objects for bounded choices, then contrast them with open questions about causes, names and categories.
Accept either form only when the learner can describe the implied context. The goal is deliberate scope, not a rigid one-word substitution rule.
The adult’s most useful question is not “Why were you careless?” but “Which decision did the evidence require?” Ask the learner to point to the controlling word, value, relationship or observation. If the learner can explain that link, reduce support. If the learner cannot, model one example and then change the surface details for the next attempt.
Frequently asked questions
When should I use what?
Use it for an open request for information or category when no limited set controls the answer.
When should I use which?
Use it to select from a particular stated, visible or shared set.
Does which mean there are only two choices?
No. The set can contain many items as long as it is bounded and identifiable.
Can both words be grammatical?
Yes, but they may imply different answer fields.
Does every which ask a question?
No. Which can introduce a relative clause.
What is the quickest check?
Ask whether an unlisted answer could fit or whether the answer must come from a known set.
Next routes
Return to the complete volume directory to locate the series, use the original PSLE Learning Guide for the foundation route, and continue through PSLE English for verified subject guidance. The next useful task is a mixed set in which this idea is sometimes relevant and sometimes irrelevant; that is the point at which a learner must choose, not merely copy.