What significant figures actually record
Significant figures are a shorthand for how well you know a number. A balance that reads 12.34 g is telling you the mass to the nearest hundredth of a gram; writing 12.340 g would claim a thousandth-of-a-gram balance you do not have. The digits that carry information about the measurement are the significant ones, and every other digit in the written number is a placeholder that only fixes the decimal point.
This matters because arithmetic cannot create precision. If you measure a rectangle as 12.3 cm by 4.5 cm, your calculator will happily report an area of 55.35 cm², but you never knew either side to a hundredth of a centimetre, so the last two digits are fiction. The honest answer is 55 cm². Reporting 55.35 is not a rounding preference; it is a false claim about your instrument.
Four counting rules cover every case you will meet. Every non-zero digit is significant. Zeros between non-zero digits are significant — 4506 has four. Leading zeros are never significant, because they only place the decimal point: 0.00250 has three, not six. Trailing zeros are significant when a decimal point is written — 0.00250 and 100.0 both end in a meaningful zero — and are conventionally read as placeholders when it is not, which is why 250 is normally taken as two figures.
How rounding to n significant figures is computed
Rounding to a number of significant figures is rounding to a decimal place whose position you have to work out first. That is the whole difference between this and rounding to two decimal places, where the position is handed to you.
Start with the decimal exponent, e = ⌊log₁₀|x|⌋. This is the power of ten of the leading digit: for 0.0234567 it is −2, because the number sits between 10⁻² and 10⁻¹. For 4506 it is 3. Keeping n figures means the last digit you keep sits at the 10e − n + 1 place, so that value is the size of your rounding unit. Divide by it, round to a whole number, multiply back.
Take 0.0234567 to three figures. e = −2, so the unit is 10−2 − 3 + 1 = 10⁻⁴ = 0.0001. Divide: 0.0234567 ÷ 0.0001 = 234.567. Round: 235. Multiply back: 0.0235. The same three lines work for 4506 to two figures — unit 103 − 2 + 1 = 100, 4506 ÷ 100 = 45.06, round to 45, multiply back to 4500.
One wrinkle: rounding can push a number over a power of ten. 9.97 to two figures gives 10., where the leading digit has moved a place. The calculator recomputes the exponent after rounding, which is why the scientific-notation output always shows exactly n digits in the mantissa even when the plain form looks like it gained one. If you want that notation on its own, the scientific notation calculator handles conversions in both directions.
Worked example: 0.0234567 × 1.20 to the correct precision
Suppose you measure a concentration of 0.0234567 mol L⁻¹ and multiply it by a dilution factor of 1.20 recorded on a volumetric flask. Work it through by hand.
- Count the figures in each value. 0.0234567: the two leading zeros are placeholders, leaving 2, 3, 4, 5, 6, 7 — six significant figures. 1.20: the 1 and 2 count, and the trailing zero counts because a decimal point is written — three significant figures.
- Round the first value to three figures on its own. e = ⌊log₁₀ 0.0234567⌋ = −2. Unit = 10−2 − 3 + 1 = 0.0001. 0.0234567 ÷ 0.0001 = 234.567 → 235 → 0.0235, or 2.35 × 10⁻².
- Check the rounding error. 0.0235 − 0.0234567 = 0.0000433, which is 0.185% of the original — small, but not zero, which is why you never round twice in the middle of a chain.
- Do the multiplication with full precision. 0.0234567 × 1.20 = 0.02814804. Keep every digit at this stage.
- Apply the retention rule. Multiplication keeps the smaller significant-figure count: min(6, 3) = 3.
- Round the product to three figures. e = ⌊log₁₀ 0.02814804⌋ = −2, unit = 0.0001, 0.02814804 ÷ 0.0001 = 281.4804 → 281 → 0.0281.
Notice what would have happened if you had rounded first: 0.0235 × 1.20 = 0.0282, which differs in the last digit from the correct 0.0281. Rounding intermediate values is the single most common way to lose a mark in a chemistry paper.
Which retention rule applies, and why they differ
Multiplication and division keep significant figures; addition and subtraction keep decimal places. These are not two versions of the same rule, and swapping them gives wrong answers in both directions.
The reason is what each operation does to relative and absolute uncertainty. In a product, relative uncertainties add, and relative uncertainty is what a significant-figure count expresses: a value known to three figures is known to roughly one part in a thousand whatever its magnitude. In a sum, absolute uncertainties add, and absolute uncertainty is what a decimal place expresses: a value good to the tenth is good to ±0.05 whether it reads 0.3 or 3000.3.
So 12.11 + 0.3 = 12.41 becomes 12.4. The 0.3 is only known to the tenths place, so the hundredths digit of the sum is unknowable, even though 12.11 has four significant figures and the answer has three. Subtraction can go the other way and destroy figures wholesale: 12.11 − 12.09 = 0.02, where two four-figure values produce a one-figure result. That collapse — subtractive cancellation — is a real numerical hazard, not a bookkeeping curiosity, and it is why analytical methods avoid taking small differences of large numbers.
Two categories of number are exempt. Exact counts — 12 flasks, 3 trials — have infinite significant figures and never limit a result. Defined conversion factors such as 1 in = 2.54 cm exactly, or 1 min = 60 s, are exact by definition and likewise do not limit anything. Only measured quantities constrain the answer.
Significant-figure counts for numbers that trip people up
| As written | Significant figures | Rule that decides it |
|---|---|---|
| 4506 | 4 | Interior zeros are always significant |
| 0.00250 | 3 | Leading zeros are placeholders; the trailing zero after a decimal point counts |
| 250 | 2 | Trailing zeros with no decimal point are conventionally placeholders |
| 250. | 3 | An explicit trailing decimal point makes the zeros significant |
| 2.50 × 10² | 3 | Scientific notation states the count with no ambiguity |
| 100.0 | 4 | All zeros follow a written decimal point |
| 1.0080 | 5 | Interior zeros count; the trailing zero counts after a decimal point |
| 0.000001 | 1 | Every zero is a leading placeholder |
| 6.022 × 10²³ | 4 | Only the mantissa digits count; the exponent is not a measurement |
| 12 flasks | infinite | An exact count never limits a result |
The 250 row is the only ambiguous case, and it is a convention rather than a law. If precision matters, write 2.5 × 10² or 2.50 × 10² and the question disappears.
The standard behind the rounding rule
ASTM E29, Standard Practice for Using Significant Digits in Test Data to Determine Conformance with Specifications, is the document most laboratories cite when a rounded result decides whether a material passes or fails. It specifies rounding to the nearest value and, on an exact tie, rounding so the last retained digit is even — the same rule IEEE 754 makes the default for binary floating-point arithmetic. NIST Special Publication 811 gives the parallel guidance for expressing SI measurement results.
The half-to-even rule exists because always rounding a tie upward biases a long series of results upward. Half-to-even splits ties between up and down and leaves the mean undisturbed. If you need a different tie rule for accounting or coursework, the rounding calculator lets you pick between half-up, half-even, ceiling, floor and truncation and shows what each one gives.
Mistakes that turn a right answer into a wrong one
- Rounding partway through a chain. Carry full precision to the end and round once. Rounding at each step accumulates error, as the worked example above shows in its final digit.
- Applying the multiplication rule to a sum. 12.11 + 0.3 is 12.4, not 12.1 — sums keep decimal places, not figures.
- Counting leading zeros. 0.00250 has three significant figures, not six. Leading zeros do nothing but position the decimal point.
- Treating exact counts as measurements. Dividing by 3 trials does not cut you to one significant figure; exact integers never limit a result.
- Writing trailing zeros you cannot support. Reporting 12.340 g from a centigram balance overstates the instrument by a factor of ten.
- Assuming a spreadsheet respects any of this. A cell displaying 55.35 has stored 55.35 in full; display formatting is not rounding, and the stored value is what feeds the next formula.
- Forgetting that subtraction can destroy figures. The difference of two close four-figure values may carry only one, and no rule restores what the subtraction removed.
Where significant figures stop being the right tool
Significant figures are a rough, cheap proxy for uncertainty, and they are the right tool for coursework, field notes and quick sanity checks. They are not the right tool for a formal measurement result. A count of figures cannot express that a value is 12.34 ± 0.02 rather than ± 0.005, cannot combine correlated uncertainties, and cannot carry a coverage factor. When any of that matters, propagate the uncertainties explicitly and report a standard uncertainty alongside the value, as NIST Special Publication 811 and the ISO Guide to the Expression of Uncertainty in Measurement describe.
Two practical habits sit between the two approaches. First, carry one or two guard digits through intermediate arithmetic and round only the reported result; every numerical library does this internally. Second, when you are comparing a measurement against a reference value, quote the discrepancy as a percentage rather than counting digits — the percent error calculator does that directly, and a percentage is far more informative than the statement that two values agree to three figures.
Finally, remember that significant figures describe a written decimal, and some quantities are better handled as exact rationals. A repeating decimal such as 0.8333… has no finite significant-figure representation at all; converted with the decimal to fraction calculator it is exactly 5/6, and staying in fractions until the final step avoids the question entirely. The same applies to irrational results from the square root calculator, where the digits continue indefinitely and only your inputs decide how many you may report.
Key terms
- Significant figure
- A digit in a written value that carries information about the measurement, as opposed to a placeholder that only positions the decimal point.
- Leading zero
- A zero before the first non-zero digit. Never significant — 0.0025 and 2.5 × 10⁻³ are the same two-figure value.
- Trailing zero
- A zero after the last non-zero digit. Significant when a decimal point is written, conventionally a placeholder when it is not.
- Guard digit
- An extra digit carried through intermediate arithmetic and discarded at the end, so that rounding error does not accumulate through a chain of steps.
- Half-to-even rounding
- The tie rule in ASTM E29 and IEEE 754: when the discarded part is exactly one half, round so the last retained digit is even. It removes the upward bias of always rounding ties up.
- Subtractive cancellation
- The loss of significant figures when two nearly equal values are subtracted, leaving a result with far fewer meaningful digits than either input.