COLLEGE PHYSICS • PROBLEM-SOLVING & QUANTITATIVE TOOLS

Experimental Uncertainty & Error Propagation

Every measurement carries doubt—learning to quantify and propagate that doubt is what separates data from knowledge.

Historical Context & Motivation

No physical measurement is ever exact. From the earliest astronomical observations of Tycho Brahe to the nanoscale metrology of modern semiconductor fabrication, experimenters have wrestled with the same fundamental question: how much can we trust our numbers? The formal study of experimental uncertainty grew out of centuries of effort to reconcile conflicting observations in astronomy, geodesy, and physics. Understanding how uncertainties arise—and how they combine when measurements feed into calculations—is not a pedantic bookkeeping exercise; it is the backbone of scientific credibility. Without a rigorous framework for error propagation, we cannot meaningfully compare theory with experiment, replicate results, or design instruments that push the frontiers of precision.

1632
Galileo & Repeated Measurement
Galileo Galilei systematically repeated inclined-plane experiments, noting discrepancies between trials and implicitly acknowledging that no single measurement tells the full story.
1809
Gauss & the Normal Distribution
Carl Friedrich Gauss published Theoria Motus, deriving the Gaussian (normal) distribution and the method of least squares to optimally combine astronomical observations burdened by error.
1893
Karl Pearson & Formal Statistics
Karl Pearson introduced the standard deviation as a named concept and developed rigorous statistical tools that would become the language of experimental uncertainty in physics and beyond.
1993
GUM — The Modern Standard
The Joint Committee for Guides in Metrology published the Guide to the Expression of Uncertainty in Measurement (GUM), establishing the internationally accepted framework—Type A and Type B evaluations—used in laboratories worldwide today.

The central question that this lesson addresses is deceptively simple: if the inputs to a calculation each carry some uncertainty, what is the uncertainty of the output? Answering that question requires us first to classify the kinds of error that afflict measurements, then to develop a mathematical machinery—rooted in calculus—for tracking how uncertainties flow through formulas. By the end, you will be equipped to report results that honestly reflect the limits of your apparatus and methodology.

Core Principles & Definitions

Before diving into formulas, it is essential to build a precise vocabulary. Physicists distinguish between the true value of a measurand (the quantity being measured), the measured value obtained in practice, and the uncertainty that quantifies how far the measured value might plausibly lie from the true value. Crucially, uncertainty is not a mistake; it is an intrinsic feature of every measurement process.

1

Accuracy vs. Precision

Accuracy measures how close a result is to the true value, while precision measures reproducibility. A set of tightly clustered results may be precise yet inaccurate if a systematic bias is present.
2

Systematic Error (Bias)

Systematic errors shift all measurements in the same direction—e.g., a miscalibrated scale or a persistent parallax. They do not decrease with repeated trials and must be identified and corrected.
3

Random Error (Statistical)

Random errors cause readings to scatter unpredictably about the mean. They arise from uncontrollable fluctuations (vibrations, electronic noise) and can be reduced by averaging over many trials.
4

Absolute vs. Relative Uncertainty

Absolute uncertainty (δx) carries the same units as the measurement. Relative (fractional) uncertainty (δx/x) is dimensionless and makes it easy to compare the 'quality' of different measurements.
5

Significant Figures Convention

The number of significant figures in a reported result should match the precision of the measurement. Uncertainties themselves are conventionally reported to one or two significant figures, and the result is rounded to match.
KEY TAKEAWAY
Think of uncertainty like the margin of error in a GPS reading. Your phone might show you at latitude 40.7128°, but it also says '±5 m.' That '±5 m' is the uncertainty—it doesn't mean the GPS is broken; it means the device is honestly telling you the limits of its knowledge. In physics, every reported value should come with its own '±' to be scientifically meaningful.

Visualizing Accuracy, Precision & Distributions

The classic target analogy for accuracy and precision is well-known, but a more quantitatively useful visualization is the Gaussian distribution of repeated measurements. When random errors dominate, repeated measurements of the same quantity cluster around the mean in a bell curve whose width is characterized by the standard deviation σ. The following diagram shows how the distribution relates to the concepts of accuracy, precision, and the standard deviation of the mean.

A Gaussian bell curve centered on the sample mean x̄. The violet bracket spans ±1σ (68.3 % of data), and the pink bracket spans ±2σ (95.4 %). The orange arrow shows the systematic bias—the offset between the mean and the true value.

In this diagram, the width of the distribution reflects precision: a narrow bell curve means tightly clustered readings (high precision), while a broad curve indicates large scatter. The horizontal offset between the dashed cyan line (x̄) and the dashed amber line (true value) represents systematic bias—the measure of inaccuracy. Importantly, taking more data points narrows the standard deviation of the mean = σ/√N), but it cannot reduce systematic bias—only careful calibration or redesign can do that.

Mathematical Framework of Error Propagation

The mathematical heart of this topic lies in answering the question: given a derived quantity q = f(x, y, z, …) where each input variable carries an uncertainty, what is δq? The general answer comes from a first-order Taylor expansion of f about the measured values, assuming that the uncertainties are small relative to the values themselves. We begin with the standard deviation of a single set of measurements, build up to the standard deviation of the mean, and then derive the general propagation formula.

STANDARD DEVIATION
σ = √[ (1/(N−1)) Σᵢ (xᵢ − x̄)² ]
Here N is the number of measurements, xᵢ is the i-th reading, and x̄ is the sample mean. The factor (N−1) is Bessel's correction, accounting for the fact that the sample mean is estimated from the data itself.
STANDARD DEVIATION OF THE MEAN
σ_x̄ = σ / √N
This tells us that the uncertainty in the average decreases as 1/√N. Quadrupling the number of trials halves the uncertainty in the mean.
GENERAL PROPAGATION FORMULA
δq = √[ (∂f/∂x)² δx² + (∂f/∂y)² δy² + (∂f/∂z)² δz² + … ]
Each term is the partial derivative of f with respect to a measured variable, evaluated at the best estimates, multiplied by the uncertainty in that variable. This assumes the errors in x, y, z, … are independent (uncorrelated). If correlations exist, covariance terms must be added.

Common Special Cases

The general formula simplifies to memorable rules for the two most common function forms. For sums and differences (q = x ± y), the absolute uncertainties add in quadrature: δq = √(δx² + δy²). For products and quotients (q = xy or q = x/y), the relative (fractional) uncertainties add in quadrature: δq/|q| = √( (δx/x)² + (δy/y)² ). For a quantity raised to a power (q = xⁿ), the relative uncertainty scales linearly with the exponent: δq/|q| = |n| × (δx/x). These rules are not separate axioms—they are all derivable from the general propagation formula via straightforward partial differentiation.

SUM / DIFFERENCE RULE
q = x ± y → δq = √( δx² + δy² )
Absolute uncertainties combine in quadrature. Note that subtraction does not cancel uncertainties—it adds them just like addition does.
PRODUCT / QUOTIENT RULE
q = x × y or q = x / y → δq/|q| = √( (δx/x)² + (δy/y)² )
Relative (fractional) uncertainties combine in quadrature. This means the least precise input dominates the output uncertainty.

Detailed Breakdown of Propagation Rules

To solidify your understanding, the diagram below maps out the decision tree you should follow when propagating uncertainties through any formula. The key insight is to decompose complex expressions into a chain of elementary operations (addition, multiplication, powers, functions) and apply the appropriate rule at each stage, combining intermediate uncertainties in quadrature.

Decision flowchart: start by identifying the operation type, apply the corresponding rule (absolute uncertainties for sums, fractional for products, derivatives for general functions), then combine in quadrature if multiple variables are involved and report the final result with its uncertainty.
Summary of propagation rules for elementary operations
OperationFormula for qPropagation RuleWhat Adds in Quadrature
Additionq = x + yδq = √(δx² + δy²)Absolute uncertainties
Subtractionq = x − yδq = √(δx² + δy²)Absolute uncertainties
Multiplicationq = x × yδq/|q| = √((δx/x)² + (δy/y)²)Relative uncertainties
Divisionq = x / yδq/|q| = √((δx/x)² + (δy/y)²)Relative uncertainties
Powerq = xⁿδq/|q| = |n| × (δx/x)Scaled relative uncertainty
General functionq = f(x)δq = |df/dx| × δxDerivative × absolute unc.
Common Pitfall
When you subtract two quantities of nearly equal magnitude, the result can be very small while the absolute uncertainty remains large, producing an enormous relative uncertainty. This is sometimes called catastrophic cancellation. For example, if x = 10.0 ± 0.5 and y = 9.8 ± 0.5, then q = x − y = 0.2 with δq = √(0.25 + 0.25) = 0.71, giving a relative uncertainty of 350 %. Always watch for subtractions of similar-sized quantities.

Worked Example: Density from Mass and Volume

Suppose you measure the mass and dimensions of a small aluminum cylinder to determine its density. You find: mass m = 45.2 ± 0.3 g, diameter d = 2.10 ± 0.02 cm, and height h = 3.90 ± 0.03 cm. What is the density and its propagated uncertainty?

Finding ρ ± δρ for an Aluminum Cylinder
1
Step 1 — Write the formulaDensity is ρ = m / V, where the volume of a cylinder is V = π(d/2)²h = πd²h/4. So ρ = 4m / (πd²h).
2
Step 2 — Calculate the volumeV = π × (2.10)² × 3.90 / 4 = π × 4.41 × 3.90 / 4 = π × 17.199 / 4 = π × 4.2998 ≈ 13.51 cm³.
V ≈ 13.51 cm³
3
Step 3 — Calculate the densityρ = m / V = 45.2 / 13.51 ≈ 3.346 g/cm³. (We'll hold extra digits until the final rounding.)
ρ ≈ 3.346 g/cm³
4
Step 4 — Identify the propagation structureIn ρ = 4m / (πd²h), the constant 4/π has no uncertainty. The formula is a product/quotient of m, d², and h. Using the quotient/product rule, the relative uncertainty in ρ is: δρ/ρ = √( (δm/m)² + (2 × δd/d)² + (δh/h)² ). Note the factor of 2 in front of δd/d because d is raised to the power 2.
5
Step 5 — Compute each fractional uncertaintyδm/m = 0.3/45.2 = 0.00664. δd/d = 0.02/2.10 = 0.00952, and 2 × δd/d = 0.01905. δh/h = 0.03/3.90 = 0.00769.
Fractional uncertainties: 0.664 %, 1.905 %, 0.769 %
6
Step 6 — Combine in quadratureδρ/ρ = √(0.00664² + 0.01905² + 0.00769²) = √(4.41×10⁻⁵ + 3.63×10⁻⁴ + 5.91×10⁻⁵) = √(4.66×10⁻⁴) = 0.0216, or about 2.2 %. The diameter contributes the most because of the exponent.
δρ/ρ ≈ 2.2 %
7
Step 7 — Compute absolute uncertainty and reportδρ = 0.0216 × 3.346 = 0.072 g/cm³. Rounding the uncertainty to one significant figure gives δρ ≈ 0.07 g/cm³, and we round ρ to match: ρ = 3.35 ± 0.07 g/cm³.
ρ = 3.35 ± 0.07 g/cm³
💡 Sanity Check
The accepted density of aluminum is about 2.70 g/cm³. Our result of 3.35 ± 0.07 g/cm³ is outside the accepted range—suggesting either a systematic error (perhaps the cylinder isn't pure aluminum) or a measurement mistake. This is exactly the kind of critical evaluation that uncertainty analysis enables: without δρ, you couldn't know whether the discrepancy is significant.

Strengths, Limitations & Common Misconceptions

The quadrature-based propagation framework is remarkably powerful, but it rests on assumptions that you should be aware of. Knowing when the standard approach works well—and when it breaks down—is a mark of experimental sophistication.

Strengths and limitations of standard error propagation
StrengthsLimitations
Simple, closed-form formulas for common operations—easy to apply by hand.Based on a first-order Taylor expansion; breaks down when uncertainties are large relative to the measured values.
Adding in quadrature correctly accounts for partial cancellation of independent random errors.Assumes inputs are independent (uncorrelated). If two variables share a common source of error, covariance terms are needed.
Provides a quantitative basis for comparing experimental results with theory.Does not handle systematic errors—those must be estimated separately and reported alongside the statistical uncertainty.
Works seamlessly for multi-step calculations by chaining intermediate propagations.Assumes underlying distributions are approximately Gaussian; for highly skewed distributions, Monte Carlo methods may be more appropriate.
KEY TAKEAWAY
Think of error propagation like an engineering tolerance analysis for a machine assembly. Each part has a tolerance (uncertainty), and you need to know whether the final assembly will fit within spec. The propagation formula is the mathematical tool that tells you whether the accumulated tolerances will blow up your final result—or stay within acceptable bounds. Just as engineers distinguish between tolerance stackup (random) and a machining offset (systematic), physicists separate random from systematic uncertainty.

Connection to Advanced Uncertainty Analysis

The propagation rules you have learned constitute a first-order, analytic approach that works beautifully for most introductory physics scenarios. However, modern metrology and experimental research often face situations where these rules are insufficient. Three advanced frameworks extend the basic approach: Monte Carlo simulation, Bayesian inference, and the full GUM framework with correlation. Understanding where the basic approach fits in the landscape of uncertainty analysis will prepare you for upper-division lab courses and research.

Standard propagation vs. advanced methods
FeatureStandard PropagationMonte Carlo / Bayesian
Mathematical basisFirst-order Taylor expansion of f(x, y, …)Repeated random sampling (MC) or Bayes' theorem (Bayesian)
Handles nonlinear functionsOnly approximately (linear regime)Exactly—samples the full distribution of outputs
Handles correlationsRequires explicit covariance termsNaturally included through joint sampling or prior specification
Computational costNegligible (pen and paper)Moderate to high (thousands to millions of iterations)
Typical use caseIntro lab courses, quick estimatesResearch-grade metrology, complex models, non-Gaussian distributions

In a Monte Carlo uncertainty analysis, you generate thousands (or millions) of synthetic data sets by randomly sampling each input variable from its probability distribution, compute the output for each set, and then examine the resulting distribution of outputs. The standard deviation of that output distribution gives δq without any Taylor-expansion approximation. Bayesian methods go further by incorporating prior knowledge about parameters and producing full posterior probability distributions. These approaches are increasingly accessible through computational tools like Python's uncertainties package or dedicated Monte Carlo libraries. As you advance in your physics career, these techniques will become indispensable, but the analytic propagation rules remain the essential foundation upon which they are built.

Practice Problems

PROBLEM 1CONCEPTUAL
A student measures the length of a table five times and obtains values of 1.52 m, 1.53 m, 1.51 m, 1.52 m, and 1.52 m. She then discovers that her meter stick was manufactured 2 mm too short. Classify the scatter in her readings and the meter-stick defect as either random or systematic error, and explain whether taking more readings would help with each.
PROBLEM 2BASIC CALCULATION
You measure two forces: F₁ = 12.4 ± 0.3 N and F₂ = 8.7 ± 0.2 N. Find the net force F = F₁ − F₂ and its uncertainty δF.
PROBLEM 3INTERMEDIATE
The period of a simple pendulum is measured as T = 1.42 ± 0.03 s, and its length as L = 0.50 ± 0.01 m. Using g = 4π²L/T², propagate the uncertainties to find g ± δg.
PROBLEM 4APPLIED
In an electrical experiment, you measure voltage V = 5.20 ± 0.05 V and current I = 0.245 ± 0.008 A. You also compute the power dissipated as P = V × I and the resistance as R = V / I. Find P ± δP and R ± δR. Which derived quantity has the larger relative uncertainty, and why?
PROBLEM 5CRITICAL THINKING
Derive the propagation formula for the function q = A × e^(−Bx), where A ± δA and x ± δx are measured quantities and B is an exact constant. Then discuss: under what conditions does the uncertainty in x dominate the result, and how does this relate to the concept of sensitivity coefficients?

Lesson Summary

Every physical measurement carries uncertainty—an honest quantification of how much the measured value might differ from the true value. Uncertainties arise from random errors (statistical scatter, reducible by averaging) and systematic errors (persistent biases, correctable only by calibration or redesign). The standard deviation σ characterizes the spread of repeated measurements, while the standard deviation of the mean σ/√N quantifies the uncertainty of the average and decreases with more data.

When measured quantities feed into formulas, uncertainties propagate according to the general propagation formula δq = √[Σ(∂f/∂xᵢ)² δxᵢ²], which simplifies to memorable rules: for sums and differences, absolute uncertainties add in quadrature; for products, quotients, and powers, relative uncertainties add in quadrature (with exponents as multipliers). The variable with the largest sensitivity coefficient dominates the output uncertainty, guiding experimenters on where to invest effort in improving precision. For highly nonlinear or correlated situations, advanced techniques such as Monte Carlo simulation and Bayesian inference extend the framework beyond the first-order approximation.

Varsity Tutors • College Physics • Experimental Uncertainty & Error Propagation