The actual rule for adding numbers with significant figures
Most people mess this up because they apply the multiplication rule by habit. When you add numbers, the rule flips. You don't count total significant figures. You count decimal places. The answer gets rounded to match the least precise operand's decimal place, not the fewest total digits. Here's what that looks like in practice. Take 12.11 plus 0.3 plus 4.221. Your calculator spits out 16.631. But you can't report that. Look at the decimal places: 12.11 has two, 0.3 has one, 4.221 has three. The limiting term is 0.3 with just one decimal place. So your final answer rounds to 16.6. That single decimal place controls everything. I learned this the hard way back when I was processing spectrophotometry data in grad school. We were measuring absorbance values across a concentration series, and one of the standards had been recorded on a balance that only read to one decimal place in grams. The instrument gave readings to four decimal places in molarity, but the mass measurement was the weak link. I initially reported results to four significant figures across the board because I was focused on precision, not precision limits. My advisor flagged it immediately. The mass measurement's single decimal place dragged the entire calculation down. We had to redo the error propagation analysis because propagating uncertainty through addition follows the same principle — the least precise measurement dominates.
Here's the part nobody teaches properly. When numbers are written without explicit decimal points, ambiguity creeps in. 1500 could have two, three, or four significant figures depending on context. If you're adding 1500 plus 23.5, and 1500 has only two sig figs, that number is uncertain in the hundreds place. Your answer would round to the hundreds place, giving you 1500 again. The smaller number effectively disappears. This happens all the time in lab work where someone writes down a calibration factor as a round number and then adds measured values to it. You end up with garbage precision because the round number swallowed everything. Another thing that trips people up: intermediate rounding. When doing multi-step calculations, keep extra digits through every step and only round at the very end. If you round after each addition, you accumulate rounding error that can shift your final result by one or more digits. I've seen students lose points on reports for doing this repeatedly across five or six addition steps. The error compounds, and by the end the reported answer is wrong even though each individual step was done correctly. The workaround I use now is straightforward. I write every number with its uncertainty explicitly — like 12.11(1) or 0.3(1) — so there's no guesswork about which digit is the last significant one. Then I do the arithmetic with full calculator precision and apply the rounding only to the final result. It takes maybe ten seconds longer per problem but eliminates the ambiguity entirely.
There's also a scenario where the rule breaks down in a subtle way. If you're adding numbers that span very different magnitudes — say 0.001 plus 999 — the result is still governed by decimal places, not by the apparent precision of the larger number. 999 has no decimal places shown, so technically it's uncertain in the ones place. Your answer becomes 1000, with the trailing zeros being placeholders, not significant figures. It feels wrong because you're discarding the precision from the small number, but that's how the rule works. The large number's uncertainty in the ones place is what matters. For subtraction specifically, watch out for significant figure loss. When you subtract two close numbers, you can dramatically reduce your significant figures. 10.05 minus 9.98 gives 0.07 — that's one significant figure instead of the four each operand had. This is called catastrophic cancellation and it's the reason why high-precision work avoids subtracting nearly equal quantities whenever possible.
Get the Full Details
Quick reference
Align numbers by their decimal points. Identify which operand has the fewest decimal places. Perform the addition normally. Round the result to match that fewest decimal place count. Carry unrounded values through any subsequent operations. Round once at the end.
