A difficult first part does not automatically block every later part of a PSLE question. Some later tasks need a number you have already found. Others return to the original passage, diagram or data. Still others supply a new value that can be used directly. The important skill is to identify the actual dependency rather than assume that the whole question is one unbreakable chain.
This guide teaches you to check which later parts really depend on your earlier answer. You will learn to continue with independent work, preserve a useful method when one quantity is unresolved and update the right later answers when an earlier result changes. The aim is not to hide a mistake or manufacture a number. It is to understand how the reasoning is connected.
Use the PSLE Learning Guide for the wider English, Mathematics and Science route. All examples in this workshop are original practice tasks, not official examination questions or promises about marking. Read the whole local task, follow any instruction about the required method and distinguish supplied facts from your own provisional working.
What a dependency means
Part B depends on Part A when completing B requires information established by A. In Mathematics, B may need the length calculated in A. In Science, an explanation may need the measured comparison identified earlier. In English, a question may explicitly ask you to justify the interpretation you have just given. In each case, the earlier output is an input to the later job.
Sharing a topic is not enough to create a dependency. Two questions about the same story may ask for different passage details. Two questions about the same rectangle may use different information. A later Science question about the controlled variables can sometimes be answered from the setup description even while you are unsure about the numerical result.
Use a simple test: cover your answer to the earlier part, but keep the question’s original information visible. Can you still explain how to answer the later part? If yes, the earlier answer may not be necessary. If no, identify the exact missing input. “I need the width” is more useful than “I cannot do any of this question”.
Distinguish given information from derived information
Given information is supplied by the question: a total, a labelled side, a sentence in the passage, an experimental condition or a table entry. Derived information is something you calculate or infer from those facts. Keep that distinction clear, especially when you are uncertain about your working.
Suppose a question states a total of 96 books and you calculate 36 fiction books. The 96 is given; the 36 is derived. If you later discover that you calculated the fiction count incorrectly, the printed total does not change. Only results that used the wrong fiction count need to be reconsidered.
This distinction is equally important in comprehension. Your first inference about a character is not a new fact added to the passage. A later answer should return to the text rather than treat your own earlier sentence as authoritative evidence. Otherwise one weak interpretation can create a chain of unsupported answers.
Worked example 1: a book problem with a separate branch
A library receives 96 books. Three eighths are fiction; the rest are non-fiction. Part A asks for the number of fiction books. Part B asks how many bundles of five can be made from all the non-fiction books. Part C asks whether four shelves, each holding at most 27 books, can hold the entire delivery.
For A, one eighth of 96 is 12, so three eighths is 36 fiction books. For B, subtract 36 from 96 to obtain 60 non-fiction books, then divide by five to obtain 12 bundles. Along this route, B uses the result from A. If you had written 30 fiction books by mistake, carrying that number into B would distort the later calculation.
Part C is different. Four shelves hold 4 times 27, or 108 books. The delivery contains 96 books, so it fits, with capacity for 12 more books. C uses the original total and the shelf capacity. It does not need the fiction count or the bundle count. Uncertainty in A need not stop you from completing C.
There is also an alternative route to B in this practice example. Non-fiction books are five eighths of the total, so five eighths of 96 gives 60 directly. Unless a question specifically requires a particular route, you can use a valid method that reaches the required quantity. A convenient dependency in your chosen solution is not always a necessary dependency in the mathematics.
After solving, describe the structure in plain language. A and B are connected through the fiction count on one route; C is a separate capacity check. That description helps you decide where to continue if one calculation becomes uncertain and which work must change if you later correct it.
A printed instruction can require a particular connection
Phrases such as “using your answer to Part A” or an explicit instruction to use a particular method tell you how the task should be approached. Read them carefully. Do not ignore a required connection just because you know a different route. The purpose of the dependency check is to understand the instruction, not to work around it.
Other wording may simply introduce a related task without demanding that your previous answer be used. The difference must come from the actual question, not a universal rule that every Part B uses Part A. Before deciding that you are blocked, identify whether the link is stated, logically necessary or merely a route you happened to choose.
During practice, your tutor can write two versions of the same problem: one requiring an earlier result and another allowing any valid method. Compare your approaches. This exercise builds flexibility while also teaching respect for task constraints. Efficient work is useful only when it still does the job the question asks for.
Worked example 2: repair only the descendants of a wrong length
A rectangle has perimeter 40 cm and width 6 cm. Part A asks for its length. Part B asks for its area. Part C asks the cost of placing a border around its perimeter at 15 cents per centimetre. Part D asks for the area after the width is increased to 7 cm while the length stays unchanged.
The correct length is 40 divided by 2, then minus 6, which gives 14 cm. The area in B is 14 times 6, or 84 square centimetres. The border in C costs 40 times 15 cents, or 600 cents, which is $6. The new area in D is 14 times 7, or 98 square centimetres.
Now imagine a learner initially writes 16 cm for A. That incorrect length produces 96 square centimetres in B and 112 square centimetres in D. When A is corrected to 14 cm, B and D must be checked because they used the wrong length. C need not change if it correctly used the printed perimeter of 40 cm.
Do not erase the whole page automatically. Trace where the wrong number travelled. A sentence or calculation using 16 cm is affected; a correct calculation using only the given perimeter and border rate is not. This is a selective repair, based on the actual reasoning rather than on the position of the answer on the page.
A forward check closes the repair. Two lengths of 14 cm and two widths of 6 cm give the stated 40 cm perimeter. Then the areas can be recomputed from the corrected dimensions. Changing only the final answer in B while leaving a working line that still uses 16 cm would leave the explanation inconsistent.
Worked example 3: one Science comparison, several different jobs
Two identical cups each contain the same volume of water at 60 degrees Celsius. P has an insulating sleeve and Q has no sleeve. They remain in the same cooler surroundings for ten minutes. P finishes at 48 degrees Celsius and Q at 40. Part A asks which cup has the smaller temperature decrease. Part B asks you to explain that difference. Part C asks for one stated condition kept the same.
A is answered by matching labels and calculating change: P decreases by 12 degrees Celsius and Q by 20, so P has the smaller decrease. B must explain the observed direction. The sleeve is linked to reduced heat transfer to the cooler surroundings, consistent with P’s smaller decrease in this example. B depends on understanding the comparison, whether you read it directly from the data or through A.
C asks a different job. Equal initial water temperature, equal starting volume or equal observation duration are stated conditions. You can identify one of them from the setup description even if you have not yet settled A. The unchanged condition does not become uncertain simply because your arithmetic is unfinished.
If you initially swapped P and Q, correcting A should make you inspect B’s direction. An explanation that says the sleeve made P cool more would contradict the corrected data. Your answer to C may still be correct. This is another case where a single error has specific consequences rather than invalidating every later sentence.
Use Vol 0049 on matching setups to their results when the dependency begins with a label-to-data error. Matching the evidence is the first repair; explaining the right difference comes afterward.
Worked example 4: later information must not rewrite an earlier evidence limit
An investigation initially gives one observation: a sample changes colour. Part A asks what can be concluded from that observation alone. Part B then supplies an additional test result and asks which of two explanations it supports. Part C asks for a follow-up investigation to distinguish the explanations further.
Part B can legitimately use the new result because it is supplied for that task. Part A is limited by the phrase “from that observation alone”. If the single observation cannot distinguish the explanations, your answer to A should not quietly import the extra test from B. Reading later parts is useful for understanding structure, but it does not erase evidence restrictions.
The existing guide on keeping later information from rewriting earlier Science reasoning develops this issue. Here, notice that dependencies can change because the question adds information. A later part may become answerable without making an earlier limited claim stronger.
In review, label which facts were available to each sub-question. This is similar to tracking what a character knew at a particular time in English. The right reasoning uses the evidence permitted for the current task. More information is useful only where the question allows it to be used.
Worked example 5: your first comprehension inference is not a passage fact
Read this original mini-passage: “Rafi reached the meeting point at nine. The others had not arrived. He checked the clock twice and started to type a message. Before he sent it, a new message appeared: the meeting had been moved to half past nine. He put the phone away and sat down on the bench.”
Part A asks why Rafi checked the clock twice. Part B asks what information the new message gives. Part C asks how his behaviour changed after reading it. These questions share a passage, but they do not all require the same answer. A requires a supported interpretation of the repeated clock-checking; B asks for directly stated information; C compares behaviour before and after the message.
A reasonable answer to A is that Rafi was checking the time because the others had not arrived when he expected them. Avoid claiming that they deliberately abandoned him; the passage does not establish that. B can be answered directly: the meeting was moved to 9:30. C can describe that he stopped preparing the message, put his phone away and sat down after learning the new time.
If you are uncertain about the best wording for A, B remains accessible from the text. You do not need to invent a reason for the delay to report the new time. Likewise, C should be built from the actual actions, not from an emotional label you guessed in A.
This prevents a common chain error: first writing an unsupported motive, then using that motive as the evidence for several later answers. Your response is a claim to be checked against the passage. It does not become part of the source merely because you wrote it earlier.
Worked example 6: a later restriction uses the cases, not just their count
Choose two different activity sessions lasting 30, 45, 60 or 75 minutes. Their combined activity time may not exceed 105 minutes, and order does not matter. Part A asks for all qualifying pairs. Part B adds a 15-minute break while retaining the 105-minute overall limit and asks how many pairs remain possible.
A’s valid pairs are 30 with 45, 30 with 60, 30 with 75, and 45 with 60. There are four. In B, the activity time must be no more than 90 minutes because the break uses 15 minutes. Only 30 with 45 and 30 with 60 remain, giving two pairs.
Notice what B actually needs. It does not need only the number four. It needs the identities and durations of the candidate pairs, or another complete method for finding those that satisfy the tighter limit. A learner who preserves only the count may have to reconstruct the cases before proceeding.
This is a different kind of dependency from carrying one numerical answer forward. The useful earlier output is a structured set of possibilities. Vol 0048 on complete lists without repeated cases explains how to make that set reliable.
If A is incomplete, blindly filtering it can miss valid solutions in B. Check that the earlier list covered every permitted branch. Alternatively, solve B with a fresh complete search using the new 90-minute limit. The earlier work is useful only to the extent that it contains the information the later job requires and was established correctly.
Use a small dependency note, not a complicated diagram
During practice, write a few words beside each part: “needs length from A”, “uses original total”, “uses new data” or “returns to paragraph two”. These notes make the structure visible without turning the question into a separate organisational project. A simple arrow can show that one result feeds another.
The note should identify information, not just letters. “B uses A” is less helpful than “B uses the remaining amount found in A”. Naming the quantity allows you to search for an alternative valid route or recognise when the question provides the same value elsewhere.
Once the habit is stable, reduce the notes. In the examination, you may need only a label next to an intermediate value or a short reminder beside a part you plan to revisit. The purpose is to protect thinking and continuity, not to fill the margin with arrows that take longer to interpret than the question itself.
What to do when an earlier value is unresolved
First, read the later part and identify the missing input precisely. If it requires the area and you have not found the length, ask whether the original diagram supplies another relationship that can establish the area. Do not assume that your first attempted route is the only route, but follow any method instruction the question explicitly gives.
Second, complete genuinely independent parts. A question about a stated control variable, a directly supplied time or a separate capacity check may not use the uncertain value. Moving to it is not abandoning the whole question. It is recognising that a page can contain several reachable jobs.
Third, preserve useful reasoning for a dependent part without presenting an invented value as fact. You may be able to write a valid relationship, such as “remaining books = total books minus fiction books”, and leave space to complete it after the missing quantity is settled. Keep the unknown clearly labelled. Do not choose a convenient number merely to make the working look finished.
If you use a tentative result that came from a genuine method, recognise that its consequences remain tentative until checked. This is a working-state decision, not a guarantee of marks for later answers. The safest aim is still to correct the underlying result and make all dependent working consistent with it.
Do not rely on a blanket rule about follow-through marks
Assessment rules depend on the question and subject. This guide does not promise that a wrong earlier answer will receive credit when reused correctly later. It also does not suggest that an explanatory paragraph earns marks simply because it is long or uses scientific terms.
The 2026 PSLE Mathematics syllabus describes marking according to item type and asks for clear mathematical working where required. Use the current official document and the actual question instructions rather than an assumed universal rule about error carry-forward.
The practical advice here is narrower: write what you genuinely know, show a valid method when required, label unfinished quantities honestly and keep working on accessible parts. That makes the reasoning understandable. Whether a particular response earns credit is determined by the assessment, not by a slogan about showing working.
When an earlier answer changes, trace the consequences
Do not update later answers by changing every visible number that happens to match the old one. Identify which values were derived from the corrected result. A 12 appearing as a given measurement is not automatically affected because an unrelated earlier calculation changed from 12 to 14.
Return to the rectangle example. Changing the derived length from 16 to 14 affects the area calculations that used it. It does not change the printed width of 6 cm, the printed perimeter of 40 cm or the border price per centimetre. Repairing by number matching rather than reasoning can create fresh errors in correct parts.
A useful check is to read the affected chain aloud during practice: corrected length gives corrected area; unchanged perimeter gives unchanged border cost. This makes the scope of the repair explicit. In Science, do the same with claims: corrected setup identity gives corrected comparison and mechanism direction, while the stated control variable remains the same.
In English, revise later explanations that depend on a changed inference, but recheck them against the passage independently. Do not assume that replacing one character trait requires replacing every observation. Facts about what the character said and did may stay correct even while your interpretation improves.
Worked example 7: a new value can restart a later branch
An original practice task asks you to calculate the mass of a sample in Part A. Part B then says, “For this part, use a sample mass of 250 g,” and asks for the total mass of six such samples. That supplied instruction creates a clear input for B: six times 250 g gives 1500 g, or 1.5 kg.
Do not replace 250 g with your unfinished answer from A. The task has specified the value to use for B. If A remains unresolved, B can still be completed from its own instruction. The same principle applies when a later part introduces a different object, a new experimental condition or a new passage detail.
However, read the wording carefully. “Using the mass you found in Part A” is not the same as “use a mass of 250 g”. One creates an explicit dependency; the other supplies an input. A vague memory that later parts often use earlier answers should not override the actual sentence.
For a second practice version, make the supplied mass 200 g and keep six samples. The answer becomes 1200 g. Only the branch governed by the new input changes. This small exercise shows how dependency tracking follows information flow rather than the habit of carrying every earlier number forward.
Worked example 8: a follow-up conversation changes the job
The same idea can help in English oral practice. A partner first asks whether you would join a class activity. You answer yes and give a reason. The partner then asks what difficulty an organiser might face. The topic is shared, but the second question does not depend on repeating your personal choice.
Answer the new job: perhaps too few helpers arrive, materials are missing or instructions are unclear. Develop one relevant difficulty using the practice situation. Do not assume that every follow-up is a request to defend the first answer in more words. Listen for whether the question asks for a reason, an example, a limitation or a different viewpoint.
If you realise that your earlier answer was off task, make a small repair rather than carrying the mismatch into the follow-up. Vol 0047 on repairing an oral answer provides original exchanges for this situation. Spoken continuity is useful, but it must not become automatic repetition of an earlier response.
Practice set: identify the blocked and open parts
Question one: a rectangular garden has perimeter 36 m and width 5 m. A asks for length; B for area; C for the cost of fencing the stated perimeter at $4 per metre. A learner is unsure about the length. Which later part can be completed without that length? Solve all parts when ready.
Question two: a class has 80 cards. One quarter are blue, and the remaining cards are red. A asks for the blue count. B asks for the red count. C asks whether five boxes holding 18 cards each can hold all the cards. State which results use the blue count and which use the original total directly.
Question three: in a Science practice task, A asks which setup had the greatest change in temperature. B asks for a controlled condition stated in the method. C asks for an explanation of the greatest change. If the results table is temporarily covered, which task might still be answerable from the method? Explain why.
Question four: a comprehension task asks for a character’s likely feeling, the location of an event explicitly stated in the passage and a reason for the feeling. Which task does not require the feeling inference? Why should the feeling answer not be used as a substitute for passage evidence?
Question five: Part A asks you to find a length. Part B explicitly says to use a length of 12 cm and a width of 4 cm to find an area. Your answer to A is 10 cm. Which length belongs in B? Explain how the wording decides the input rather than the order of the parts.
Question six: a list in A contains three valid cases but is not complete. B asks how many cases remain after an extra condition is added. Can you obtain a reliable answer merely by crossing out invalid entries from those three? State what must be established first or what alternative route is available.
Practice answers with the dependency made explicit
For question one, the length is 36 divided by 2 minus 5, giving 13 m. The area is 65 square metres. The fencing costs 36 times $4, or $144. C uses the given perimeter and rate, so it remains accessible without the calculated length. B needs the length on the usual area route.
For question two, there are 20 blue cards and 60 red cards. B can use 80 minus the blue count, or calculate three quarters of 80 directly. The boxes hold 90 cards, enough for all 80, with capacity for ten more. C does not need the colour split. The chosen route can create a dependency that an alternative valid route avoids.
For question three, the control-variable task may remain answerable because its evidence is in the method. The greatest-change comparison needs the relevant readings, and the explanation needs the correct observed direction. Do not let the unavailable table make a clearly stated setup condition disappear.
For question four, the location question returns directly to the passage. The feeling and its reason are connected, but both must be supported by textual evidence. A feeling you wrote down does not become a fact that can prove itself. For question five, B explicitly supplies 12 cm, so the area is 48 square centimetres; use that instruction for B.
For question six, filtering an incomplete list may miss valid survivors. Establish a complete list for A or conduct a fresh complete search under B’s new conditions. Counting only the surviving entries you happened to find is not enough to prove the final count.
A later answer is not automatically an independent check
Suppose you calculate a rectangle’s length as 16 cm, multiply by its 6 cm width to obtain an area of 96 square centimetres, then divide 96 by 6 and recover 16. The calculations agree, but the agreement does not confirm that the original length is correct. The area came from that same length. The check has travelled around a loop built from one assumption.
Return instead to the independent given condition: the perimeter is 40 cm. A rectangle measuring 16 cm by 6 cm has perimeter 44 cm, so the proposed length fails that condition. The correct length is 14 cm. This check can disagree with the original result because it uses the question’s supplied perimeter rather than a later value generated from the disputed length.
Dependency tracking therefore improves verification as well as continuation. Before using B to check A, ask whether B was itself calculated from A. If it was, the reverse calculation may catch an arithmetic slip but will not necessarily expose a wrong starting assumption. Choose a given condition, a genuinely different representation or another independent piece of evidence when that is the risk you need to test.
The same warning applies to English. Writing a character trait in A, then explaining it in B by repeating A in different words, does not create new support. Return to the passage’s actions or dialogue. In Science, a predicted outcome derived from your own mechanism is not a new observation confirming that mechanism. Keep the direction of support visible.
Practise the decision before increasing the time pressure
Begin with a short multi-part task and ask the learner to identify the input for each part without solving it. This separates dependency recognition from subject calculation. An adult can then check whether the learner has correctly identified supplied facts, derived quantities and independent evidence.
In the next round, introduce one genuine unresolved result. Ask which later parts can still be attempted and why. Do not reward random movement around the page. The learner should name the evidence or relationship that makes a later part accessible. That explanation is the skill being trained.
On another day, change the subject and remove the dependency prompts. A Mathematics number chain can become a Science evidence chain or an English inference-and-support task. Look for the same decisions: what does this part need, where can that information be found and does the earlier uncertainty actually affect it?
Finally, use a timed mixed section. Afterward, review one place where the learner continued appropriately and one place where a dependency was missed. The goal is not to complete every possible annotation. It is to prevent one local difficulty from becoming an unnecessary block on unrelated work.
Guidance for parents and tutors
Ask “What does this part need?” before saying “You must finish A first.” Sometimes A is genuinely required. Sometimes the learner can use original data, a new given value or another valid relationship. Make the decision from the task rather than from a fixed teaching order.
When an earlier answer changes, ask the learner to find the affected work. Do not simply replace all later answers for them. Tracing the consequences helps the learner understand why some results must be updated and why others should remain untouched. This is part of mathematical and scientific reasoning, not just neat correction.
Keep feedback proportionate. A wrong first calculation may cause several incorrect final numbers, but that does not necessarily mean the learner lacks every concept used later. Conversely, a consistent chain using a wrong assumption is not fully correct just because each later arithmetic step follows neatly. Diagnose the original error and the dependency handling separately.
Frequently asked questions
Must I always answer the parts in printed order?
Follow the examination instructions and any required method. Within an ordinary multi-part task, a later independent part may be accessible while an earlier part is unresolved. Read it before assuming it is blocked, and make your answers clearly identifiable.
How can I tell whether a dependency is real?
Name the information needed by the later part. If it must come from the earlier result, the dependency is real. If the same information is supplied directly or can be established through another permitted route, the earlier answer may not be necessary.
Should I invent a number so I can show the next method?
No. Write a relationship or label the unknown where that is useful, but do not present an invented number as given or established. A genuine provisional result should be checked, and this guide does not promise credit for carrying an error forward.
What if I correct A but have little time left?
Check the later parts that actually used A’s old result. Prioritise the connected chain rather than rewriting independent correct work. Update the relevant working and final values consistently; changing only one visible number can leave contradictions.
Can a later question help me understand an earlier one?
It can reveal useful structure or a possible check, but respect any restriction on the evidence available to each part. Later supplied information must not be used to strengthen an earlier conclusion when that earlier task explicitly limits its source.
How does this relate to checking every part?
Checking coverage asks whether every requested job has an answer. Dependency checking asks what each answer relies on and what must change if an input is corrected. Both matter. A fully filled page can still contain a chain built on one unsupported assumption.
Continue with the connected practice routes
Use the English Learning Guide, Primary 6 Mathematics Learning Hub and PSLE Science Learning Guide for deeper subject support. The PSLE Learning Guide keeps the wider examination route together.
The practical rule is to trace information, not just question letters. Find what the later part needs, use the correct source, continue where the work is genuinely independent and repair every result that depended on a corrected input. One unresolved answer should block only the work that truly needs it.
