Arithmetic operations and uncertainty propagation

A lab table, a metre scale, a stopwatch — and every reading you take is a little bit wrong. Not carelessly wrong, just physically wrong: no ruler is infinitely fine, no eye is infinitely steady, no stopwatch click is infinitely timed. So when you write down a length as $16.2\ \text{cm}$ you are really saying "somewhere close to $16.2$ , plus or minus a bit I can't shrink away." The honest question is not "did I get the exact answer" — nobody ever does — but "how far can my final answer be trusted, given how far each measurement can be trusted?" That question has a precise arithmetic answer, and it is one of the most quietly powerful ideas in the whole Units and Measurements chapter. Add two uncertain lengths and their uncertainties add. Multiply an uncertain mass by an uncertain volume and their fractional uncertainties add. Raise a measurement to a power inside a formula — length cubed, for a density calculation — and that fractional uncertainty gets multiplied by the power before it adds in. None of this is guesswork; it is a small, learnable set of rules that CBSE tests every single year, in exactly the combinations this article works through. This article builds directly on significant figures and rounding, the article before it: there you learned how many digits in a single number are meaningful, here you learn what happens to that meaning when numbers are combined by arithmetic. We keep every worked number from the earlier version of this article intact — the $16.2\ \text{cm} + 3.8\ \text{cm}$ sum, the $16.2 \times 3.8$ product, and the $4\%$ density-error calculation — and build outward from them: more worked drills, a full derivation of the propagation rule, a figure, quiz questions, and the exam-card recipe you actually pull out the night before the test.

Key takeaways

Every measured number carries an uncertainty, an absolute error $\Delta a$ , and reporting a result means reporting both the value and how far it can be trusted.
Addition and subtraction: absolute errors add. For $Z = A \pm B$ , the uncertainty is $\Delta Z = \Delta A + \Delta B$ — always added, never subtracted, because the worst case is when both errors push the same way.
Multiplication and division: relative (fractional) errors add. For $Z = A^{p}B^{q}/C^{r}$ , $\dfrac{\Delta Z}{Z} = p\dfrac{\Delta A}{A} + q\dfrac{\Delta B}{B} + r\dfrac{\Delta C}{C}$ .
Powers matter most. A quantity raised to a higher power — like $L^3$ in a density formula — contributes that many times its own fractional error to the total.
Significant-figure rules mirror the error rules. Addition/subtraction keeps the fewest decimal places; multiplication/division keeps the fewest significant figures — this is exactly why the error rules work the way they do.
The final answer can never be more precise than its shakiest ingredient. Reporting extra digits beyond what the propagated error justifies is a fabricated precision, not real information.
Key idea: absolute errors add under addition/subtraction while relative errors add (each weighted by its power) under multiplication/division, so $\Delta Z/Z = p\,\Delta A/A + q\,\Delta B/B + r\,\Delta C/C$ for $Z = A^{p}B^{q}/C^{r}$ tells you exactly how much trust survives in a calculated result.

The core insight

When measured quantities are combined arithmetically, absolute errors add under addition/subtraction while fractional errors add (weighted by their powers) under multiplication/division and exponentiation, so the propagation rule $\Delta Z/Z = p\,\Delta A/A + q\,\Delta B/B + r\,\Delta C/C$ turns individual measurement uncertainties into the trustworthy precision of a final calculated answer.

Where it fits in the syllabus

This is the third sub-topic of CBSE Class 11 Chapter 1, Units and Measurements, sitting right after significant figures and rounding and right before dimensions of physical quantities. It is squarely inside the prescribed 2025-26 scope: absolute, relative and percentage error; the addition rule $\Delta Z = \Delta A + \Delta B$ ; and the general power-law propagation rule $\Delta Z/Z = p\,\Delta A/A + q\,\Delta B/B + r\,\Delta C/C$ for $Z = A^p B^q / C^r$ . It is core exam material, tied closely to laboratory practicals — every experiment where you compute a density, a spring constant, or an acceleration from several separately-measured quantities uses exactly this arithmetic. Formal statistical treatment of error (standard deviation, Gaussian distributions, least-count theory beyond the basic idea) is explicitly out of scope at this level and is left for later years; this article stays entirely within what CBSE actually asks.

The physical picture

Every measurement is a comparison against an imperfect instrument, read by an imperfect observer, so the number you write down is never exactly the true value — it differs from it by some error. If several readings $a_1, a_2, \ldots, a_n$ of the same quantity are taken, their mean $a_{\text{mean}}$ is the best available estimate of the true value, and the absolute error in any one reading is how far that reading sits from the mean.

\Delta a_i = |a_{\text{mean}} - a_i|

where: $a_i$ is the $i$ -th individual reading, $a_{\text{mean}}$ is the arithmetic mean of all the readings, and $\Delta a_i$ is the absolute error of that reading — always taken as a positive magnitude, since we care about the size of the discrepancy, not its direction. In practice a single representative absolute error $\Delta a$ (often the largest of the $\Delta a_i$ , or a stated instrument tolerance) is carried forward for the whole measurement. Once you have $\Delta a$ for a measured value $a$ , two more numbers follow immediately: the relative (fractional) error, which compares the absolute error to the size of the quantity itself, and the percentage error, which is the same comparison expressed out of a hundred.

\frac{\Delta a}{a} \qquad \text{and} \qquad \frac{\Delta a}{a}\times 100\%

where: $a$ is the mean (best-estimate) value of the measurement and $\Delta a$ is its absolute error. Relative error is dimensionless — a $0.5\ \text{g}$ error on a $50.0\ \text{g}$ mass is a relative error of $0.01$ , exactly the same "shakiness" as a $0.05\ \text{cm}$ error on a $5.00\ \text{cm}$ length, even though the raw absolute numbers look completely different. This is precisely why relative error, not absolute error, is the right currency for combining measurements through multiplication and division: it is scale-independent, so it can be added across quantities with different units.

Figure

Two error bars side by side, one for a length measurement and one for a mass measurement, each shown as a central dot with a shaded band around it representing the absolute error, with a note that the same relative error can look like very different absolute error bars

Derivation ABuilding the power-law error propagation rule

The goal is to show, using only ordinary algebra and the definition of relative error, why a derived quantity $Z$ built from measurements $A$ , $B$ , $C$ through a formula $Z = A^{p}B^{q}/C^{r}$ ends up with fractional error $\Delta Z/Z = p\,\Delta A/A + q\,\Delta B/B + r\,\Delta C/C$ — that is, why the errors add as fractions, each scaled by the power of the quantity it belongs to, regardless of whether that quantity sits in the numerator or the denominator. This matters because it is the single formula that turns a table of individually-measured lengths, masses and times into a trustworthy uncertainty on any calculated physical quantity — density, speed, spring constant, anything built by multiplication, division and powers of measured numbers.

Step 1. Start from the maximum possible fractional error in $Z$ . Write the measured values with their uncertainties as $A \pm \Delta A$ , $B \pm \Delta B$ , $C \pm \Delta C$ , and consider the worst case where $A$ and $B$ are as large as possible while $C$ is as small as possible (since $C$ sits in the denominator, a smaller $C$ makes $Z$ larger). The largest possible value of $Z$ is then built from $(A+\Delta A)^p(B+\Delta B)^q/(C-\Delta C)^r$ .

Step 2. Expand each factor using the small-error approximation. Because $\Delta A/A$ , $\Delta B/B$ , $\Delta C/C$ are all small fractions (typically a few per cent), each power can be pulled out and approximated: $(A+\Delta A)^p = A^p\left(1+\dfrac{\Delta A}{A}\right)^p \approx A^p\left(1+p\dfrac{\Delta A}{A}\right)$ , keeping only the first-order term and dropping the tiny squared and higher terms. The same step applies to $(B+\Delta B)^q$ and to $(C-\Delta C)^{-r} \approx C^{-r}\left(1+r\dfrac{\Delta C}{C}\right)$ — note the sign flips to a plus here because $C$ is in the denominator and we are looking for the largest $Z$ .

Step 3. Multiply the three approximations together. $Z + \Delta Z \approx A^pB^qC^{-r}\left(1+p\dfrac{\Delta A}{A}\right)\left(1+q\dfrac{\Delta B}{B}\right)\left(1+r\dfrac{\Delta C}{C}\right)$ . Multiplying out the brackets and again dropping the products of two small fractions (they are second-order and negligible compared to the individual terms), only the sum of the three first-order terms survives: $Z+\Delta Z \approx Z\left(1+p\dfrac{\Delta A}{A}+q\dfrac{\Delta B}{B}+r\dfrac{\Delta C}{C}\right)$ , using $Z = A^pB^qC^{-r}$ .

Step 4. Subtract $Z$ from both sides and divide by $Z$ . $\Delta Z \approx Z\left(p\dfrac{\Delta A}{A}+q\dfrac{\Delta B}{B}+r\dfrac{\Delta C}{C}\right)$ , so dividing through by $Z$ gives the result: $\dfrac{\Delta Z}{Z} = p\dfrac{\Delta A}{A}+q\dfrac{\Delta B}{B}+r\dfrac{\Delta C}{C}$ . Every term is added with a positive sign — there was never a subtraction anywhere in the derivation, because at every step we deliberately chose the combination that made $Z$ largest, which is exactly the worst-case (maximum) error a measurement could produce. This is why the rule always adds fractional errors, whether the quantity sits in a numerator or a denominator.

Worked examples

Board level - sum and product of two measured lengths

Two lengths are measured as $l_1 = 16.2\ \text{cm}$ and $l_2 = 3.8\ \text{cm}$ . Find their sum and the product $l_1 \times l_2$ , each rounded to the correct number of significant figures or decimal places.

Step 1. Add the two lengths directly: $l_1 + l_2 = 16.2 + 3.8 = 20.0\ \text{cm}$ .

Step 2. Apply the addition rule for significant figures: the result keeps as many decimal places as the least-precise term. Both $l_1$ and $l_2$ have 1 decimal place, so the sum is kept to 1 decimal place, which it already satisfies. Answer: $l_1 + l_2 = 20.0\ \text{cm}$ .

Step 3. Multiply the two lengths directly: $l_1 \times l_2 = 16.2 \times 3.8 = 61.56\ \text{cm}^2$ .

Step 4. Apply the multiplication rule for significant figures: the result keeps as many significant figures as the term with the fewest. $l_1$ has 3 significant figures, $l_2$ has only 2, so the product must be rounded to 2 significant figures: $61.56 \to 62$ . Answer: $l_1 \times l_2 = 62\ \text{cm}^2$ .

Competitive level - percentage error in a calculated density

The density of a metal cube is calculated using $\rho = m/L^3$ , where the mass is $m = (50.0 \pm 0.5)\ \text{g}$ and the side length is $L = (5.00 \pm 0.05)\ \text{cm}$ . Find the percentage error in the calculated density, and then find the density itself with its uncertainty correctly rounded.

Step 1. Write $\rho$ in power-law form: $\rho = m \cdot L^{-3}$ , so $p = 1$ for the mass and the power on $L$ is $3$ . The propagation rule gives $\dfrac{\Delta\rho}{\rho} = \dfrac{\Delta m}{m} + 3\dfrac{\Delta L}{L}$ .

Step 2. Compute each relative error: $\dfrac{\Delta m}{m} = \dfrac{0.5}{50.0} = 0.01$ and $\dfrac{\Delta L}{L} = \dfrac{0.05}{5.00} = 0.01$ .

Step 3. Combine, weighting the length term by its power of 3: $\dfrac{\Delta\rho}{\rho} = 0.01 + 3(0.01) = 0.04$ , so the percentage error is $\dfrac{\Delta\rho}{\rho}\times 100\% = 4\%$ . Answer: the density carries a maximum percentage error of 4%, and the length term (weighted by the power 3) contributes three times as much as the mass term.

Step 4. Find the central value of the density itself: $\rho = m/L^3 = 50.0 / (5.00)^3 = 50.0/125.0 = 0.400\ \text{g/cm}^3$ . The absolute error is $\Delta\rho = 0.04 \times 0.400 = 0.016\ \text{g/cm}^3 \approx 0.02\ \text{g/cm}^3$ . Answer: $\rho = (0.40 \pm 0.02)\ \text{g/cm}^3$ , and note the last digit of $\rho$ is rounded to match the precision the error actually justifies — writing more decimal places than this would be a false claim of precision. Timing note: under two minutes. The chain is identify the power-law form of the formula, list each relative error, weight by its power, add, then convert the percentage error back into an absolute error on the final rounded value if the question asks for it. The one trap to watch: the power of $L$ in the formula ( $L^3$ , so weight 3) is not the same as the number of significant figures in $L$ (3 s.f.) — these are two different "3"s that happen to coincide here and are easy to conflate under time pressure.

Common traps

Subtracting errors instead of adding them when a quantity sits in a denominator (like $L^{-3}$ above) — fractional errors from division always add, never subtract, because the propagation rule is built from the worst-case (maximum-error) combination, not an average or a cancellation.
Forgetting to weight by the power. A quantity raised to a higher power in the formula contributes that many times its own relative error to the total — missing this weighting (treating $L^3$ like plain $L$ ) is the single most common mark-loser in density and volume-based error questions.
Mixing up the two significant-figure rules. Addition/subtraction uses the "fewest decimal places" rule; multiplication/division uses the "fewest significant figures" rule. Applying the wrong rule to the wrong operation is an easy, frequent slip.
Reporting excess digits in the final answer. If the percentage error works out to $4\%$ , quoting a density to four decimal places is meaningless — the propagated error itself tells you how many digits of the final answer can actually be trusted.
Forgetting to convert the answer back from percentage error to absolute error (or vice versa) when the question specifically asks for the value in the form $(\text{value} \pm \text{uncertainty})$ .

You should be able to…

State and apply the addition rule $\Delta Z = \Delta A + \Delta B$ for a sum or difference of measured quantities.
State and apply the power-law propagation rule $\Delta Z/Z = p\,\Delta A/A + q\,\Delta B/B + r\,\Delta C/C$ for $Z = A^{p}B^{q}/C^{r}$ , including recognising the power of each quantity correctly.
Explain, in words, why the rule always adds fractional errors even when a quantity appears in the denominator.
Correctly choose between the "decimal places" and "significant figures" rounding rules depending on whether the operation is addition/subtraction or multiplication/division.
Convert between absolute error, relative error and percentage error for a single measured quantity.
Identify which term in a propagated-error calculation dominates the total error, and explain why powers amplify uncertainty.

Ask the tutor: Want to practice propagating error through a formula with three or four variables at once, like the ones that show up in practicals — ask the AI tutor to generate a step-by-step drill for you.

Why do errors add even when I'm dividing, not adding?

Because the rule for multiplication and division is about the fractional errors, not the raw arithmetic operation. When you divide $A$ by $B$ , the worst case is $A$ coming out too large at the same time $B$ comes out too small — both mistakes push the ratio $A/B$ in the same direction, making it too large. So even though the operation is division, the two fractional errors still stack up in the same direction and must be added, never subtracted. The addition/subtraction rule (absolute errors add) and the multiplication/division rule (relative errors add) are two separate rules for two separate situations — the word "add" appears in both, but it is adding different things.

Why does a cubed term hurt the error so much more than a linear term?

Because the propagation rule multiplies each quantity's relative error by its power in the formula before adding. If $L$ appears as $L^3$ , a $1\%$ error in $L$ becomes a $3\%$ contribution to the error in the final result — the power effectively amplifies the uncertainty three times over. This is exactly why, in a real lab, you are told to measure the side of a cube far more carefully than its mass if you are computing density: the cubed quantity dominates the final uncertainty, so tightening that one measurement improves your final answer the most.

How is percentage error different from percentage difference between my result and the textbook answer?

They measure completely different things. Percentage error, as used in this article, is $\Delta a/a \times 100\%$ — it tells you how far a single measurement (or a quantity propagated from several measurements) is spread around its own best-estimate mean, based purely on your instrument's precision and your reading habits. Percentage difference from a "known" or textbook value is a separate comparison entirely, checking your experimental result against an accepted standard, and it can be large even if your propagated error is small (pointing to a systematic mistake) or small even if your propagated error is large (a lucky coincidence). CBSE practicals usually ask for the propagated percentage error, not a comparison to a textbook value, unless the question explicitly says so.

Do I need to know standard deviation or Gaussian error analysis for this chapter?

No — that level of statistical treatment is explicitly beyond the CBSE Class 11 prescription for this chapter. What you do need is the simpler picture used throughout this article: take the mean of several readings as your best estimate, find each reading's absolute error as its distance from that mean, and then use the addition rule or the power-law rule to combine errors when readings are combined by arithmetic. Standard deviation, Gaussian (normal) distributions and formal least-count theory belong to more advanced treatments you may meet later, not to this chapter.