A rounded number is evidence, not a promise
When a school plan needs to prove that a measurement meets a rule, arithmetic alone is not enough. We must also ask what the original number actually tells us. This G3 Mathematics guide builds carefully from decimal place value to justified threshold decisions. It follows Vol 0342’s reading of conditions and prepares a new Science application without repeating Vol 0340 on integer capacity. All practical scenarios here are invented teaching examples.
The number on the page is not always the exact number
A measurement of 12.4 metres may look decisive. Yet if a student rounded to the nearest tenth of a metre, the value that produced that report might have been anywhere from 12.35 metres up to, but not including, 12.45 metres under ordinary school rounding. The printed number is a useful summary, not perfect knowledge of the original measurement. This distinction becomes important when a route, container, exhibition stand or instrument reading must satisfy a limit. The difficult part is often not arithmetic. It is knowing what arithmetic is entitled to claim from the information supplied.
The exact skill this chapter teaches
This chapter is about how the precision of a number affects a decision. Earlier Mathematics chapters have covered measured travel, rates and bounds, as well as inequalities for whole-number capacity. Here we connect those foundations by asking a different question: if a measurement was rounded, can a student still prove a threshold was met? We will read measurements, infer legitimate intervals, calculate with appropriate caution, compare upper and lower limits, and write conclusions that do not promise more than the data. This is an examination-reasoning workshop grounded in rounding, measurement and inequalities, not an assertion that a separate advanced interval-arithmetic topic appears as a named examination chapter.
The 2027 examination context
The Singapore-Cambridge Secondary Education Certificate begins in 2027. For school candidates in G3 Mathematics, the official syllabus code is K310. The published specification lists two written papers, each with a two-hour-and-fifteen-minute duration, 90 marks and equal weighting of 50 percent. Candidates may use approved calculators in both papers, and they are expected to show essential working. Questions include direct skills and applications, with a real-world context included in the second paper. The mathematical tools in this guide are rounding, estimation, inequalities, numerical reasoning and sensible interpretation. Always follow the specific precision instructions attached to the real examination question.
What does reported precision mean?
An exact count of 18 chairs is normally an integer fact about 18 chairs. A length written as 18 metres to the nearest metre is different: it represents an interval of possible underlying lengths. The words “nearest,” “correct to,” “approximately” and “measured” should make you check the unit and the reporting convention. A written money amount of $18.00 might be an exact price charged in cents, not a rounded scientific reading. Do not automatically attach uncertainty to every decimal. The source tells you whether the number is exact, rounded, estimated or a nominal value. Identify that source status before deciding what interval, if any, is justified.
Start from the size of a rounding step
Rounding to one decimal place in metres uses increments of 0.1 metre. The maximum distance from the reported value, under ordinary rounding to nearest, is half of that step: 0.05 metre. Rounding to two decimal places uses steps of 0.01 and a half-step of 0.005. Rounding to the nearest whole centimetre uses a half-centimetre on each side of the displayed centimetre count. This half-step method is simple, but the units must be attached. It is easy to compute 0.05 and forget whether it represents centimetres, metres or seconds. Write the reporting unit first, then the step, then half the step.
Build a rounding interval carefully
Suppose a recorded board length is 2.6 metres to the nearest 0.1 metre. Then 2.55 ≤ true length < 2.65 metres under the usual half-up rounding convention. The lower endpoint is included; the upper endpoint is excluded because 2.65 rounds to 2.7. The original measurement is not necessarily at either endpoint. We are describing all compatible measurements, not claiming that the measuring device actually recorded an endpoint. A bounded interval is an account of what could be true given the rounding rule. It is not automatically a statement of the instrument’s physical accuracy or the reliability of the person measuring.
Do not confuse decimal places with significant figures
The number 4.30 has two decimal places but three significant figures. The number 4300 may have ambiguous reported precision unless the question specifies whether it is nearest ten, nearest hundred or something else. Scientific notation can make the intended significant figures clearer: 4.30 × 10³ communicates three significant figures. If 0.037 is given to two significant figures, its last significant digit represents a thousandth, and its rounding half-step is 0.0005. Students often start counting zeros from the decimal point and obtain the wrong interval. Ask what place value the final retained significant digit occupies. That is the rounding step you need.
Measured versus exact quantities
A class has exactly 30 registered students. A walk is reported as 30 metres to the nearest metre. A room has a maximum permitted attendance of 30 people. These three uses of 30 must not be treated as interchangeable. The first is a count, the second a rounded measurement, and the third an integer rule. An attendance limit cannot be adjusted by plus or minus 0.5 merely because the numeral is printed without decimals. Similarly, an exact fee of $30 is a money term, not a tape-measure result. Clarifying the type of quantity prevents the common mistake of inventing a bound for something that is not a rounded observation.
Why the upper bound is usually open
A value given as 7.2 to the nearest tenth corresponds to values at least 7.15 but less than 7.25 with conventional school rounding. The notation “less than” at the upper bound is not decoration. If a threshold is exactly 7.25, the reported 7.2 does not include 7.25. At the same time, the number may be arbitrarily close to 7.25 from below. Some practical decisions use strict limits, others use “at most” or “at least.” Before comparing values, translate the wording into the proper inequality. This avoids counting an endpoint where it is unavailable.
Three possible threshold outcomes
For a measurement interval, a pass-or-fail decision has three possible evidence states. The whole interval may lie safely within the permitted side of the threshold. The whole interval may lie outside it. Or the interval may cross the threshold, so the reported rounded value cannot settle the decision. For example, if a width is given as 2.4 metres to the nearest tenth, its interval crosses a minimum requirement of 2.42 metres. Some compatible true widths satisfy the requirement and others do not. The responsible answer is not “passes” or “fails.” It is “cannot determine from this rounded reading alone.”
Keep strict and inclusive limits separate
A requirement that a quantity be no more than 5.0 differs from a requirement that it be less than 5.0. The first permits equality; the second does not. If you have a rounded measurement, the boundary issue may decide whether a guarantee can be made. You need not invent a complicated notation for every question. Write the interval, copy the actual requirement and compare them directly. If the entire possible range is below a strict ceiling, compliance is established from the reported precision. If any possible value can reach or exceed the ceiling, the number is insufficient to guarantee compliance.
Do not round before a comparison
Consider a computed value of 14.96 and a required minimum of 15.0. Rounding 14.96 to the nearest whole number gives 15, but that displayed result should not be used to certify an exact minimum of 15.0. The original value remains 14.96, which is below the requirement. Carry sufficient digits through the calculation; apply the comparison to the unrounded working value; only then express the final answer to the requested precision. Rounding early can change a decision at a threshold, even if the numerical difference looks small. This is one reason showing working matters.
Estimated answers are not guarantees
A quick estimate can tell you whether a proposed solution is plausible. It rarely proves compliance when two values are very close. Suppose you estimate a cost at about $100 and the budget is $100. That is not proof that the final invoice will be within budget. A more useful examination answer might say that the estimate is too close to the ceiling to establish affordability and that the exact itemised cost should be checked. Estimation is a tool for testing scale and spotting absurd results. It is not an excuse to substitute a rounded guess for a decision that needs an exact calculation.
The difference between possible and certain
If a plank is 4.8 metres to the nearest tenth, a true length of 4.76 metres is compatible with the report, but 4.86 metres is not. The statement “The plank may be longer than 4.80 metres” is true. The statement “The plank is definitely longer than 4.80 metres” is not justified. In exam writing, words such as certainly, possibly, cannot, and not enough information have mathematical weight. Choose the weakest claim that fully answers the question without being evasive. A rigorous answer gives the interval or the decisive inequality so the reader can check why that degree of certainty is appropriate.
Two measurements can produce a range of totals
Imagine two boards, each reported to the nearest tenth: 2.4 metres and 3.1 metres. Their compatible intervals are [2.35, 2.45) and [3.05, 3.15). Adding lower endpoints gives 5.40 metres; adding upper bounds gives a total strictly below 5.60 metres. A simple reported sum of 5.5 metres is useful as a central estimate, but it conceals the range that the original rounding permits. If the combined length must be at most 5.5 metres, the rounded measurements alone do not guarantee that requirement. We do not know the exact pair of underlying lengths.
Products require more caution than sums
For positive lengths, the smallest possible rectangular area comes from the lower compatible lengths; the greatest upper limit comes from their upper limits. If sides are reported as 2.4 metres and 3.1 metres, each nearest 0.1 metre, the area is at least 2.35 × 3.05 = 7.1675 square metres and less than 2.45 × 3.15 = 7.7175 square metres. Multiplying displayed numbers gives 7.44 square metres, which is neither an exact area nor a reliable guarantee of a tight physical fit. When a question provides rounded dimensions, distinguish the arithmetic using given numerals from an analysis of all plausible underlying dimensions.
Units can change the boundary reading
A width of 125 centimetres to the nearest centimetre corresponds to lengths from 124.5 centimetres up to less than 125.5 centimetres. Converted into metres, that is [1.245, 1.255), not [1.25 − 0.5, 1.25 + 0.5) metres. The second expression mistakenly treats half a centimetre as half a metre. Convert either the interval endpoints or the rounding step with the same unit conversion. When converting square or cubic units, remember that the conversion factor is squared or cubed. If the question compares measurements from different sources, standardise units before comparing limits.
A useful visual without drawing
Imagine a number line with a permitted region marked on one side of a threshold. Place the lower and upper compatible measurements as an interval on the same line. If the interval sits wholly inside the permitted region, the guarantee is supported. If it sits wholly outside, it is ruled out. If it overlaps both, you need better precision. You can do this mentally or with a quick written inequality. The number-line model is often clearer than a long paragraph full of approximate numbers. It also forces you to identify whether a boundary point belongs to the permitted region.
Recorded precision is not instrument accuracy
A student may report a ruler reading to the nearest millimetre. That does not prove the ruler is perfectly calibrated or that there was no parallax error. The interval inferred from rounding describes a reporting convention, not every possible measurement uncertainty. The reverse mistake also occurs: a device with a fine display does not necessarily produce highly accurate measurements. In Science practical work, systematic errors and random scatter have separate explanations. Keep the mathematical question about rounding separate from the experimental question about trustworthiness. If the source provides an instrument tolerance, treat it explicitly rather than hiding it inside a decimal-place rule.
Use an error-checking example
A student writes “6.37 rounded to one decimal place is 6.3 because the first decimal digit is 3.” That is wrong. For one decimal place, inspect the next digit: 7 in the hundredths place makes 6.3 round up to 6.4. Another student writes “6.35 to one decimal place is definitely 6.3.” Under the standard half-up school convention, 6.35 rounds to 6.4. Always identify the retained position and the immediate next digit. For negative quantities or software-specific tie-breaking conventions, follow the stated rule rather than importing assumptions. The ordinary school rule works for the positive physical examples in this guide.
Why a count limit needs integers
Suppose a safety plan permits at most 37 people in a room. You cannot admit 37.4 people. The count is a whole-number decision and the limit is exact. If another part of the task uses a rounded distance for a queuing route, it belongs to a different layer of reasoning. Do not mix the two layers: one concerns whole-number attendance, the other compatible physical measurements. The earlier capacity-inequality guide handles the integer decision. Here the added skill is deciding whether the numerical evidence used to plan the situation is precise enough to support the claim made about it.
How to handle answer-rounding instructions
The K310 paper tells candidates to follow stated instructions and use appropriate numerical precision. A question may demand a specified decimal place, significant figure, or exact form. These are instructions about the answer you must present; they do not necessarily declare the given data to have that measurement uncertainty. In ordinary problems, a supplied figure may be taken as the exact numerical input for the mathematical model unless the problem identifies it as rounded or measured. Do not automatically turn every stated side length into an interval when the task simply says “calculate the area.” Introduce interval analysis when the wording or decision genuinely warrants it.
The real-world problem-solving habit
A useful response has four moves. State the given quantity and its reported precision. Identify the valid range if rounding is explicitly part of the task. Compare that range, or the appropriately precise calculated value, against the original rule. Finish with a conclusion that distinguishes guaranteed, impossible and unresolved. This process can fit in a few clear lines. Extra formulas are unnecessary unless they reveal an important relationship. A student who writes “not enough information” without showing why may be correct but incomplete; a student who shows a crossing interval demonstrates the reason the verdict is unresolved.
A common trap in percentage reporting
Suppose a survey reports 62 percent to the nearest whole percent. If the original rate was calculated and then rounded conventionally, the compatible rate lies from 61.5 percent up to less than 62.5 percent. This does not mean the respondent count itself has a half-person error. It also does not validate the sampling method. The numerator, denominator, data collection and reporting precision are distinct issues. A claim such as “more than 62 percent support the plan” cannot be proved from a rounded result of 62 percent, because compatible values lie both above and below 62. Keep a percentage as a ratio with a defined population before evaluating any headline.
Separate mathematical precision from decision policy
A school might choose to require a clear safety margin beyond a mathematical threshold. That is a policy, not a result derived from the rounded number. If a question gives no safety margin, do not invent one and silently change the criterion. If the question does specify a margin, include it in the actual inequality before calculating. A mathematically valid answer can still be practically unwise when the model omits important factors. The examination task is to reason from the stated assumptions while recognising that the model has limits. Name those limits without pretending that you have authority to approve a real installation.
When the report itself changes later
Imagine a field team first publishes a length as 18 metres to the nearest metre and then checks it with a better method, reporting 18.24 metres to the nearest centimetre. Do not treat those as two contradictory observations merely because the displayed numbers differ. The first permits a broad range; the second gives a narrower compatible range. Compare the intervals, dates and method notes. A higher-precision report may be more informative, but it is not automatically more accurate if the instrument or method is unreliable. The best conclusion is based on what each source establishes, not on the number of digits.
Repair before repeating the same error
When a question is missed, name the first wrong move. Did you identify the wrong rounding place? Forget the exclusive upper endpoint? Convert half a centimetre to half a metre? Compare a rounded final answer instead of the unrounded result? Treat a counted person as a measured fraction? Or assume a display’s precision guaranteed its accuracy? Correct the earliest faulty step, then solve the case again without looking at the answer. That is more useful than repeatedly doing full papers while making the same hidden assumption. Keep a small error log with one example and one corrected sentence for each pattern.
A careful handover from primary mathematics
In primary school, pupils learn decimal place value, measurement units, estimates and checking whether an answer makes sense. G3 students retain those foundations and use them to defend more complicated claims. If a student cannot reliably identify a tenth or hundredth, the right repair is to revisit place value, not to memorise interval notation. The PSLE learning guide provides the earlier route; the Mathematics owner covers broader secondary skills. The present chapter focuses on the one point where basic rounding becomes advanced judgment: a small numeric uncertainty may determine what can fairly be concluded.
The exit standard
At the end of this guide, you should be able to say why a rounded value is not automatically an exact value, construct the compatible interval when the rounding rule is known, compare it with a threshold, and avoid overclaiming certainty. You should also know when not to apply such an interval because a quantity is an exact count, a specified price or an ordinary mathematical input. Most importantly, you should be able to explain an unresolved result calmly. Good mathematical reasoning does not force every situation into yes or no. Sometimes the strongest result is a precise description of what remains unknown.
Worked problem clinic: eighteen different precision traps
Cover each explanation first. Write the interval or exact calculation, identify the rule and give the strongest conclusion supported by the evidence. The explanations are not shortcuts; they model how to justify a decision and how to recognise what is not yet known.
Case 1: A doorway measurement
Question. A doorway is reported as 0.9 m wide to the nearest 0.1 m. Can the report alone establish that its true width is at least 0.88 m?
Worked answer. No. The compatible interval is [0.85,0.95) m. It includes values below 0.88 and values above. The displayed 0.9 lies above the threshold, but that display is not the original exact measurement. A more precise observation is required to prove compliance.
Case 2: A strictly smaller limit
Question. A model diameter is 9.6 cm to the nearest 0.1 cm. A display cover accepts diameters strictly smaller than 9.7 cm. Is it certain to fit on that single condition?
Worked answer. Yes, under the stated rounding model, because the actual diameter is at least 9.55 cm and less than 9.65 cm. Every compatible value is below 9.7 cm. This settles only the specified diameter condition, not every physical factor such as irregular shape.
Case 3: A value crossing its own display
Question. A board is 3.0 m to the nearest 0.1 m. Must its actual length be at least 3.0 m?
Worked answer. No. Its compatible interval begins at 2.95 m and ends below 3.05 m. Some possibilities are less than 3.0 m. This is a common error: students confuse the centre of a reporting interval with a minimum.
Case 4: One decimal versus two
Question. The measurement 4.80 m is correct to two decimal places. Give the compatible range.
Worked answer. The range is 4.795 ≤ x < 4.805 metres. The last displayed digit is in the hundredths place, so the rounding step is 0.01 metre and its half-step is 0.005 metre. A range of 4.75 to 4.85 would be appropriate to one decimal place, not two.
Case 5: A sum with a ceiling
Question. Two cables are listed as 2.4 m and 3.1 m, both nearest 0.1 m. Does a 5.5 m maximum allow their combined true lengths in every compatible case?
Worked answer. No. Their total ranges from 5.40 m to less than 5.60 m. Some compatible totals exceed 5.5 m. The displayed sum 5.5 m is not evidence of a universal guarantee. Carry the two compatible intervals through addition before deciding.
Case 6: Rectangle area
Question. A plot measures 2.4 m by 3.1 m to the nearest tenth of a metre. What area range is supported by the reports?
Worked answer. Assuming positive side lengths and no extra measurement complications, the area is at least 2.35 × 3.05 = 7.1675 m² and less than 2.45 × 3.15 = 7.7175 m². The value 7.44 m² obtained from the displayed dimensions is one modelled estimate.
Case 7: Exact seats
Question. A room has 28 fixed seats. A student proposes that the actual number might lie between 27.5 and 28.5 because 28 is a whole number.
Worked answer. That is a category error. The source states a count, not a measurement rounded to the nearest whole seat. There are exactly 28 seats under the problem’s assumptions. Only the actual seating arrangement or another provided condition could change that count.
Case 8: A measured length in centimetres
Question. A track distance is 125 cm correct to the nearest centimetre. State its interval in metres.
Worked answer. The compatible distance is 124.5 cm ≤ x < 125.5 cm, or 1.245 m ≤ x < 1.255 m. The half-centimetre becomes 0.005 metres, not 0.5 metres. Converting units before comparing is the safe habit.
Case 9: Late rounding changes a decision
Question. A numerical calculation produces 14.96 units against a minimum of 15.00 units. Is a final report of 15 units enough to show compliance?
Worked answer. No. Rounding 14.96 to 15 for presentation does not change its value relative to the minimum. Compare the unrounded calculation first: 14.96 < 15.00. Only then present the answer to any required precision.
Case 10: Percentages and certainty
Question. A poll result is presented as 62% to the nearest whole percent. Does this establish that more than 62% approved?
Worked answer. No. Compatible true percentages range from 61.5% to less than 62.5%, including some below and above 62%. Moreover, percentage rounding alone says nothing about whether the sample represents the wider population. Keep numerical precision and survey quality separate.
Case 11: Significant figures
Question. A positive mass is reported as 0.037 kg to two significant figures. Determine the interval, assuming ordinary rounding.
Worked answer. The final significant digit is in the thousandths place, so the compatible interval is 0.0365 kg ≤ m < 0.0375 kg. The leading zeros are placeholders rather than counted significant figures. This differs from rounding to two decimal places.
Case 12: Nearest ten
Question. A crowd is estimated at 430 people to the nearest ten. What compatible integer counts can give that report under the ordinary convention?
Worked answer. The numerical interval is [425,435), but because a crowd is a count, the compatible integer counts are 425 through 434 inclusive. This combines a rounded estimate with the discreteness of people correctly.
Case 13: An exclusive endpoint
Question. A positive measurement reports 5.0 cm to the nearest tenth. Can the true value be exactly 5.05 cm under the conventional half-up rule?
Worked answer. No. A true value of exactly 5.05 cm rounds to 5.1 cm. The values that produce a display of 5.0 cm include 4.95 cm but are strictly below 5.05 cm. The exclusion of the upper endpoint can matter in threshold questions.
Case 14: Precise charge versus approximate estimate
Question. An invoice explicitly states a total charge of $18.40. May a student assume the amount could be $18.395 because it has two decimal places?
Worked answer. No. In an invoice, $18.40 is the stated amount payable in cents unless the task indicates an approximate conversion or measurement. Do not treat ordinary exact monetary charges as physical measurements merely because they use a decimal point.
Case 15: A ratio based on a real count
Question. A study records 12 qualifying observations out of 40, and a student says the fraction should be treated as a rounded measurement interval because 0.30 has two decimal places.
Worked answer. The underlying ratio is exactly 12/40 = 3/10 = 0.30 for the recorded observations. Rounding the display does not make the recorded fraction uncertain. Sampling bias, missing observations and generalisability are other questions requiring separate evidence.
Case 16: Two reported lengths, comparison
Question. Route A is reported as 12.4 m and Route B as 12.5 m, both to the nearest tenth. Can you assert that A is strictly shorter than B?
Worked answer. A ranges from 12.35 to below 12.45 m; B ranges from 12.45 to below 12.55 m. With the ordinary half-up convention, every compatible A value is below every compatible B value, although the gap can be arbitrarily small. This conclusion depends on the specified rounding rule and comparable units.
Case 17: Threshold at the reporting edge
Question. A volume is reported as 5.0 litres to the nearest tenth. Can the report guarantee a minimum of 4.95 litres?
Worked answer. Yes under the usual rounding model: the compatible interval begins at 4.95 litres, and the requirement says at least 4.95. If the requirement instead demanded strictly more than 4.95, the inclusive lower endpoint would prevent a universal guarantee.
Case 18: A realistic limit of the method
Question. A digital sensor displays 12.40, but the manufacturer says its readings may have a systematic bias of plus or minus 0.20. Is the two-decimal display alone reliable to within 0.005?
Worked answer. No. The display’s two decimal places describe resolution, not guaranteed instrument accuracy. The stated systematic bias is much larger. When practical tolerances are supplied, evaluate them in their own right; do not pretend that number formatting has removed measurement error.
Continue from the right foundation
Check the official SEAB 2027 G3 syllabus directory and the K310 Mathematics syllabus for current subject code, examination requirements and reporting conventions. These examples are independently written practice, not official SEAB specimen questions. Compare the earlier measurement-and-rates guide and EMS public-instructions audit. Use the G3 learner-guide hub and Mathematics owner for adjacent skills, with the PSLE route for place-value and measurement repair.
