Every measurement in IB Physics carries a built-in admission: the true value is never exactly known. Uncertainty is the interval within which that true value is expected to lie—a product of instrument resolution, method, and environment, and entirely distinct from a mistake. The purpose of uncertainty analysis is to quantify how much confidence a reported result deserves, then say so plainly.
The pattern that costs IB Physics candidates the most marks is treating uncertainty as a collection of disconnected formulas—something to apply at isolated moments rather than a continuous analytical thread running from data collection through to evaluative conclusion. Teacher-practitioner analyses of examiner reports have repeatedly flagged this fragmentation: students who cannot show integrated reasoning about uncertainty across an investigation lose marks not because their arithmetic is wrong, but because their analysis never connects. Treating uncertainty as a continuous reasoning discipline—one that runs through every stage, from instrument reading to final evaluative sentence—is what gives an analysis the internal consistency that higher-band mark schemes are specifically looking for.
Three Expression Forms — Absolute, Fractional, and Percentage, in the Order You Use Them
Absolute uncertainty should be assigned the moment a measurement is taken. For a single instrument reading, use the instrument’s resolution as the starting point: half the smallest division for an analog scale, or ±1 in the last displayed digit for a digital display. When repeated readings are available, compare their spread to the instrument-based value and switch to half-range uncertainty if the spread is larger. With a handheld stopwatch, the reaction-time component typically exceeds the display resolution, making the timing method itself the dominant source of uncertainty. Whatever basis applies, keep it consistent across the full dataset. Once the absolute value is set, round the uncertainty first—one significant figure as the default, or two when the leading digit is 1 or 2 and the extra precision changes how results compare. Round the measured value to the same decimal place, and present the result as (value ± uncertainty) unit.
Fractional uncertainty is the ratio of absolute uncertainty to measured value—dimensionless, and therefore directly comparable across quantities regardless of scale. A ±0.5 cm uncertainty on a 2 cm measurement gives a fractional uncertainty of 0.25; the same absolute uncertainty on a 50 cm measurement gives 0.01. Those two numbers represent very different confidence levels. Only the fractional form makes that difference immediately visible.
Percentage uncertainty—fractional uncertainty multiplied by 100—is the form most frequently requested in IB mark schemes and most directly tested in Paper 1B data questions. A ruler reading of (12.0 ± 0.5) cm carries a percentage uncertainty of 4.2%; a stopwatch interval of (5.0 ± 0.2) s carries 4.0%; a thermometer reading of (20.0 ± 0.5) °C carries 2.5%. The largest value identifies the dominant contributor to total experimental uncertainty. Stated to around 2 significant figures, percentage uncertainty points toward that dominant source rather than serving as a decorative number appended to a calculation.
Propagation — Matching the Rule to the Operation
Two rules cover every IB propagation calculation, and the operation—not the quantity type—determines which applies. When measured quantities are added or subtracted, their absolute uncertainties add. When they are multiplied or divided, their percentage uncertainties add. The logic is direct: addition and subtraction change absolute magnitudes, so absolute contributions accumulate; multiplication and division involve relative scaling, so it’s the relative contribution—captured by percentage uncertainty—that matters.
For a displacement calculated as the difference between x₁ = (45.0 ± 0.5) cm and x₂ = (12.0 ± 0.5) cm, the result is (33.0 ± 1.0) cm—absolute uncertainties add. For speed from d = (2.40 ± 0.02) m and t = (1.50 ± 0.05) s, the percentage uncertainties are 0.83% and 3.3%, respectively; they add to give 4.1%, so speed = (1.60 ± 0.07) m/s. Swap the rules—add percentage uncertainties through the subtraction, or add absolute values through the division—and both results are wrong on every application. Propagation errors sit among the most consistently penalized in IB mark schemes precisely because a misapplied rule produces a wrong result each time it is used, and that mistake recurs across every calculation built on that quantity. The fix is understanding what each rule preserves—and recognizing that a propagated uncertainty is inseparable from the value it qualifies; wherever that value appears next, its uncertainty must appear with it.
Graphical Uncertainty — From Error Bars to Gradient Uncertainty
Once a quantity carries a propagated uncertainty, that uncertainty determines the size of its error bar on a graph. If the plotted quantity is derived—speed, density, or any combination of raw measurements—the error bar must reflect the fully propagated combined uncertainty, not the instrument resolution of a single input variable. Plotting a derived quantity with bars drawn from raw resolution alone misrepresents the actual confidence the data support, and any analytical claim built on that graph inherits the same weakness.
Gradient uncertainty follows from the maximum and minimum gradient method: draw the steepest and shallowest lines that can still pass through the full set of error bars, then treat the range between those two gradients as the gradient uncertainty. This value then propagates into any physical constant or derived quantity the gradient represents. An Internal Assessment (IA) that quotes a gradient-derived result without its associated uncertainty range has an incomplete pipeline—the conclusion rests on a number with no stated confidence, and mark schemes penalize that gap directly.
IA Criteria and Paper 1B — When Calculations Become Judgments
Uncertainty plays a distinct role at each phase of an IA: Design requires identifying and justifying the primary uncertainty sources before data collection begins, Analysis demands consistent propagation through every calculated quantity and graphical gradient, and Evaluation asks whether the total uncertainty range supports the conclusion or overlaps with a competing explanation enough to limit the claim. What separates mid-band from upper-band performance is the interpretive step layered on top of this workflow. Mark schemes explicitly reward candidates who identify whether random or systematic uncertainty dominates and explain what that means for the result—recognizing, for example, that all data points sitting consistently above a theoretical line indicate a systematic instrument offset affecting the gradient or intercept, not random scatter. Vague references to ‘human error’ don’t earn credit; examiner commentary has flagged this repeatedly, and the candidates who lose marks here often had the right arithmetic.
- Find the dominant contributor — list the largest percentage uncertainties feeding your final result, and note any consistent offset pattern such as all data points shifted above or below a model line.
- Classify what dominates — decide whether random uncertainty (scatter and repeatability) or systematic uncertainty (consistent bias or offset) is the larger factor, then state it explicitly and explain why.
- Run the consistency check — if your result’s uncertainty interval overlaps the accepted or theoretical value, the data are consistent within uncertainty; if it does not, they are inconsistent beyond uncertainty, and connect this finding to the dominant source identified in step 2.
- Set conclusion strength — if the effect you are claiming is similar to or smaller than the combined uncertainty, limit the claim explicitly (trend suggested, not confirmed); if the effect clearly exceeds the combined uncertainty, state that the conclusion is supported by the data.
- Write the mark-earning evaluation sentence — “Because [dominant uncertainty source/type] dominates, the result is [consistent/inconsistent] with [model/accepted value] within uncertainty, so the conclusion is [supported/limited]; the most effective improvement would be [specific method change targeting that dominant source].”
Axis scales, error bar sizes, and table formatting are the language Paper 1B uses to encode uncertainty. Reading those features to reason about reliability or evaluate a proposed relationship is the same evaluative judgment the IA demands—just applied to an experimental context the candidate has never seen before.
The Six-Step Reporting Sequence
Start by assigning absolute uncertainty from instrument resolution for a single reading, switching to half-range when repeated readings produce a wider spread. From there, convert to fractional and then percentage uncertainty—the forms that make comparisons possible and propagation tractable. Apply the operation-matched propagation rule, carry the result into error bars on a graph, and extract gradient uncertainty using the maximum and minimum gradient method. The final step is reporting: (value ± uncertainty) unit, followed by one or two sentences on what the uncertainty range means for the conclusion.
The sequence applies at any scale, from a full IA to a single Paper 1B item. What it can’t survive is fragmentation—uncertainty treated as something to record at data collection and revisit nowhere else. Thread it through, and the analysis doesn’t just report a number; it makes a claim the examiner can actually evaluate.