A significant figure is a digit the measuring instrument could actually resolve. The counting rules say which digits those are, and the two arithmetic rules keep an answer from claiming more resolution than the numbers that went into it. They are different rules because addition and multiplication damage a measurement in different ways.
That last sentence is the part almost no explanation gives, and it is why the rules feel arbitrary and get forgotten between units. Learn the reason once and the two rules stop being a pair of things to memorize.
What is a significant figure actually claiming?
That the instrument could tell that digit apart from its neighbors. Writing 12.4 g says the balance resolved tenths of a gram and could not resolve hundredths. Writing 12.40 g says it resolved hundredths, and is therefore a different and stronger claim about the equipment, not a cosmetic difference.
OpenStax puts the boundary of that claim at the last digit: a measured value carries digits that are certain plus one that is estimated, and that final estimated digit is why 12.4 and 12.40 are not the same statement.
So the whole system is one idea with three corollaries. Do not report a digit the measurement could not support, do not throw away a digit it could, and do not let arithmetic manufacture digits that were never there.
How do you count significant figures?
Start at the first nonzero digit on the left and count it and everything to its right. That single sentence, which is how OpenStax states it, covers every case people usually learn as four separate rules.
- All nonzero digits count. 254 g has three.
- Zeros between nonzero digits count. OpenStax calls these captive zeros and says they result from measurement and are therefore always significant. 1.008 has four.
- Leading zeros never count. OpenStax is flat about it: leading zeros are never significant. 0.00250 L has three, because counting starts at the 2.
- Trailing zeros after a decimal point count. That is the same 0.00250, whose final zero is there only because someone measured it. 100.0 mL has four.
- Trailing zeros before the decimal point are ambiguous. OpenStax says the number of significant figures is uncertain in a number that ends with a zero to the left of the decimal point, and resolves it with scientific notation: 1,300 g could be written 1.3 x 10^3 for two figures, 1.30 x 10^3 for three, or 1.300 x 10^3 for four.
That ambiguity is a real one, not a trick question. If a problem hands you 1,300 g with no further information, you are being asked to assume something, and scientific notation is how a source avoids making you guess.
Which numbers have no significant figures at all?
Exact ones, and they are more common in chemistry than in any other first-year subject. OpenStax notes that counting is the only type of measurement free from uncertainty, and that defined quantities are also exact.
Three places this matters:
- Coefficients in a balanced equation. The 3 in 3 CO2 is a count of molecules, not a measurement, so a stoichiometry chain never loses figures at the mole ratio.
- Defined conversions. There are exactly 1000 meters in a kilometer because the prefix is a definition. OpenStax makes the same point about exact conversion factors in its treatment of unit conversion. Chaining those factors is dimensional analysis, where the units decide which one comes next.
- Counted objects. Four flasks is exactly four.
The practical consequence is that an exact number never limits the answer. Look at the measured quantities only.
Why is the addition rule different from the multiplication rule?
Because the two operations carry different kinds of uncertainty forward. LibreTexts states the principle directly: when we add or subtract measurements we propagate their absolute uncertainties, and when we multiply or divide measurements we propagate their relative uncertainties.
Absolute uncertainty lives in a decimal place. Relative uncertainty lives in a proportion. That is the whole difference, and one pair of numbers shows it.
Take 12.11 and 0.3. The second value is stated to the tenths place, so it stands for something between 0.25 and 0.35. Treat 12.11 as exactly 12.11 for a moment, so that only the second number’s uncertainty is in play.
Added. The sum is 12.41, and the true value lies between 12.36 and 12.46. The hundredths digit could be anything, so it goes. The answer is 12.4, which is OpenStax’s addition rule: round to the same number of decimal places as the number with the fewest decimal places. Notice that the answer has three significant figures even though 0.3 has one. Significant figures never entered the reasoning.
Multiplied. The product is 3.633, but the true value lies between 12.11 x 0.25 = 3.0275 and 12.11 x 0.35 = 4.2385. Only the leading digit survives that spread, and even it is shaky. The answer is 4, which is OpenStax’s multiplication rule: round to the same number of digits as the number with the fewest significant figures.
Same two numbers. One answer keeps three figures, the other keeps one. Nothing about that is arbitrary once you see that 0.3 is a small absolute uncertainty and a very large relative one.
Where do you round in a multistep problem?
At the end, once. LibreTexts says it plainly: when carrying out calculations that involve multiple steps, you should avoid doing any rounding until you obtain the final result.
This is the rule most often broken in practice, and the damage is quiet. A percent yield computed from a theoretical yield of 50.148 g gives 89.7%. Rounding that intermediate to 50.1 g first gives 89.8%. Neither digit is wrong by much, and both cannot be right, and an answer key computed from the unrounded chain will disagree with yours for a reason that has nothing to do with chemistry.
Carry the extra digits through every line and round when you write the final answer. On paper that means keeping one or two guard digits. On a calculator it means not retyping the display.
One thing to decide before you round: what happens on an exact 5. OpenStax uses the round-half-to-even convention, where a dropped 5 with nothing after it rounds toward whichever retained digit is even, so 92.85 becomes 92.8. Many courses teach round-half-up instead, which would give 92.9. Both are conventions rather than facts, they differ only in the last digit, and the one that matters is the one your course uses. Say which one you are using if there is any doubt.
How many significant figures does a pH have?
Only the digits after the decimal point, which catches nearly everyone the first time. OpenStax gives the rule for logarithms in general: when taking the log of a value, keep as many decimal places in the result as there are significant figures in the value.
Work it through. A solution with [H3O+] = 2.5 x 10^-3 M has a concentration known to two significant figures, so:
pH = -log(2.5 x 10^-3) = 2.6020599…, reported as 2.60
Two decimal places, because the concentration had two significant figures. The 2 before the decimal point is doing the job of the exponent, which is positional information rather than measured information, so it does not count.
Going back the other way, a pH of 3.60 has two decimal places, so:
[H3O+] = 10^(-3.60) = 2.5118864… x 10^-4, reported as 2.5 x 10^-4 M
Two significant figures. The pH relation page carries the conditions this arithmetic sits inside, including that pH + pOH = 14.00 holds at 25 degrees Celsius and not at other temperatures, which OpenStax states explicitly because the autoionization constant is temperature dependent.
How many significant figures does a molar mass have?
As many as the atomic weights you used, and those come with published uncertainties. The IUPAC abridged standard atomic weights give carbon as 12.011 plus or minus 0.002 and sulfur as 32.06 plus or minus 0.02, so a sulfur-containing molar mass is genuinely less precisely known than a carbon-and-hydrogen one.
In practice the sample mass is almost always the weaker number, so the molar mass rarely limits the answer. Use the atomic weights at the precision the table gives them, do not round the molar mass before dividing, and let the mass you weighed set the figures in the final answer. Looking up an atomic weight rather than recalling it is the habit that makes this true, and the formula sheet by topic links every relationship on this site to a page that names its source values.
The three errors worth watching for
Reporting the calculator. A display of 74.85032995 is not a measurement. If the data carried three figures, the answer carries three.
Rounding at every line. Covered above, and worth stating twice because it is the one that survives into second-year courses.
Applying the multiplication rule to a sum. 104.6 g + 2.5 g + 17.17 g is 124.27 g before rounding and 124.3 g after, because 104.6 g and 2.5 g are each known only to the tenths place, and tenths is the fewest decimal places in the set. Counting significant figures instead would see two of them in 2.5 g and report 120 g, throwing away two digits the balance actually measured.
When an answer disagrees with the key and the chemistry looked right, the last digit is the first place to check. Working the same problem through Flasky shows how many figures each line kept and where the single rounding happened, which separates a routing error from a rounding habit in about ten seconds. Those two need different fixes, and telling them apart is most of the value.
The formula pages state the conditions on every relationship used above, which is the other half of getting an answer that holds up: the right number of digits on a number that was computed under the wrong assumption is still the wrong answer.
Sources
- 1.5 Measurement Uncertainty, Accuracy, and Precision, OpenStax Chemistry 2e
- 1.6 Mathematical Treatment of Measurement Results, OpenStax Chemistry 2e
- 14.2 pH and pOH, OpenStax Chemistry 2e
- 4.6 Significant Figures and Rounding, Chem1 (Lower), LibreTexts Chemistry
- 4.3 Propagation of Uncertainty, Analytical Chemistry 2.1 (Harvey), LibreTexts Chemistry
- Abridged Standard Atomic Weights, IUPAC Commission on Isotopic Abundances and Atomic Weights