How to write the IB Physics IA

The Physics IA is the one piece of coursework your Physics grade is marked on internally — worth 20% of your final grade at both SL and HL. Most students lose marks not because they can't do physics, but because they pick a question with no underlying model, or they never turn their data into a linearised graph whose gradient means something. The single move that separates a top-band Physics IA from an average one is this: choose a relationship you can write as an equation, rearrange it into a straight line, and read a physical quantity straight off the gradient with its uncertainty. This guide walks you through the whole thing: what the IA is, how it's marked, exactly how to write each part, and what separates a top-band investigation from an average one.

Before you set up any apparatus, it helps to understand why the IA is built the way it is. The four criteria reward the full arc of a scientific investigation — designing it, analysing what came out, drawing a defensible conclusion, and judging honestly how far you can trust it. Because they are equally weighted, an elegant experiment with a careless evaluation scores no better than a modest experiment evaluated rigorously. The students who do best treat all four criteria as deliverables from the very first planning session, rather than rushing the design and hoping the analysis rescues them. Plan backwards from the linearised graph and the gradient you intend to extract, and you will almost always design a stronger investigation than someone who plans forwards from "an interesting idea".

The IB Physics IA at a glance

Under the current syllabus (first assessment 2025) the Physics IA is called the Scientific Investigation: a single, focused, individual investigation reported concisely — typically 6–12 pages within the 3,000-word guideline. It is marked out of 24 across four equally weighted criteria, and the same four criteria are used for Biology, Chemistry and Environmental Systems & Societies. There is no separate mark for "personal engagement" any more — your engagement should show through the relevance and ownership of your investigation. What distinguishes Physics is the central role of the governing equation: the examiner expects your question to rest on a physical law, your graph to test that law in linearised form, and your gradient to deliver a measurable physical quantity you can compare to an accepted value.

How the Physics IA is marked: the four criteria

Every mark comes from one of these four criteria, each worth 6. Write your IA criterion by criterion and check what each rewards:

How to write a Physics IA, step by step

1. Choose a research question with a governing equation. Pick a relationship you can write as an equation, with one variable you can change over a range and one you can measure. Phrase it: "How does [independent variable, range + units] affect [dependent variable, how measured]?"
2. Linearise the governing equation. Rearrange it into y = mx + c so the model predicts a straight line whose gradient yields a physical quantity — for a simple pendulum, T = 2π√(L/g) becomes T² = (4π²/g)L.
3. Identify and control your variables. Name your control variables, tie them to the assumptions of the model (for the pendulum, the small-angle approximation and negligible air resistance), and explain how each is held constant.
4. List materials and equipment with uncertainties. Record every instrument with its measurement uncertainty so each raw value carries an absolute uncertainty from the start.
5. Write a reproducible method. Develop it through trials and write it so another student could repeat the experiment exactly.
6. Collect enough raw data. At least five values of the independent variable across a wide range, each repeated several times, recorded with units and uncertainties.
7. Propagate uncertainty through every calculation. Add absolute uncertainties when adding or subtracting, add percentage uncertainties when multiplying or dividing, and multiply the percentage by n when raising to the power n (so squaring a period doubles its percentage uncertainty).
8. Plot a linearised graph with error bars and gradients. Add error bars from the propagated uncertainties, a best-fit line, and maximum and minimum gradients so the gradient is quoted as m ± Δm.
9. Extract the physical quantity and compare it. Calculate the quantity from the gradient (for the pendulum, g from 4π²/gradient) and compare it with the accepted value within experimental uncertainty.
10. Evaluate honestly. Weigh limitations by impact, name the dominant percentage uncertainty, and propose specific improvements and an extension.

Physics IA structure: what goes in each section

There is no single mandated layout, but the clearest structure that maps onto the criteria is:

• Research question & background — the question, the governing equation, and why it is worth investigating.
• The model & linearisation — the equation rearranged into y = mx + c, with what the gradient will represent.
• Variables — independent (range + units), dependent (how measured), and controlled (tied to the model's assumptions).
• Materials & equipment — every instrument listed with its measurement uncertainty.
• Method — a reproducible procedure, developed through trials, with safety where relevant.
• Raw data — a clear table with units and uncertainties, plus qualitative observations.
• Data processing — sample calculations with uncertainty propagation, including the rule for powers.
• Linearised graph — processed data with error bars, best-fit line and maximum/minimum gradients.
• Conclusion — the quantity extracted from the gradient, compared to the accepted value within uncertainty.
• Evaluation — limitations by impact, the dominant percentage uncertainty, improvements, extension.
• References — a consistent citation style throughout.

How to choose a research question that scores

More Physics IAs are decided at the question stage than at any other point, so it is worth slowing down here. A scoring question has four properties, and you should be able to point to each one before you commit. First, it rests on a governing equation — a physical law you can write down and use to predict a relationship, rather than a vague "does it change?" enquiry with nothing to test. Second, it has a clearly named independent variable with a range and units you can manipulate, and a measurable dependent variable. Third, and most importantly, the relationship should linearise: you should be able to rearrange the equation into the form y = mx + c so that a straight-line graph's gradient yields a physical quantity such as g, a resistivity, a wave speed or a spring constant. Fourth, it must be feasible with school apparatus and within your time budget, with uncertainties small enough that the trend is not lost in the scatter.

A reliable way to pressure-test a question is to write down the governing equation first, then ask "what would I plot to get a straight line, and what would the gradient mean?" If you can answer that in one sentence, the question is probably strong. If you cannot — if the only graph you can imagine is a curve you would describe but never analyse — then either choose a different relationship or work out the linearisation before you commit. The classic example is the simple pendulum: T = 2π√(L/g) looks like a curve when you plot T against L, but squaring it to T² = (4π²/g)L turns it into a line whose gradient hands you g. Favour questions where a single straight-line graph delivers your result, avoid those that depend on one dramatic reading rather than a trend across a wide range, and steer clear of relationships whose dominant uncertainty would swamp the effect you are trying to measure.

What a strong vs weak Physics IA looks like

The fastest way to lift your marks is to see the difference. Here is the same work done two ways.

The research question

Linearisation and the graph

The evaluation

Common mistakes that cost marks

• No governing equation. A question with no underlying physical model gives you nothing to predict or test.
• A curve left un-linearised. Plotting raw quantities that produce a curve wastes the gradient — linearise the equation first.
• No uncertainty propagation. Raw and processed data without absolute and percentage uncertainty caps Data analysis.
• No error bars or gradients. The max and min gradients are where the uncertainty in your result comes from — leaving them off loses easy marks.
• No comparison to an accepted value. A gradient that is never measured against the literature can't reach the top band in Conclusion.
• "Human error." Examiners read this as "I didn't analyse my method." Always name the dominant percentage uncertainty.
• Going over 3,000 words. Examiners stop reading at the limit — be concise.

Physics IA — frequently asked questions

Managing the 3,000-word limit

The word count is a guideline, not a wall, but examiners are explicit that they stop reading at the limit — so anything past 3,000 words simply will not be marked. The skill is spending words where they earn marks. As a rough budget, give your introduction, governing equation and research question around 350–450 words, your variables and method around 600–700, your uncertainty propagation and graphical analysis the largest share at roughly 700–900, your conclusion around 350–450, and your evaluation around 450–600. Raw-data tables, sample calculations, graphs and references generally sit outside the word count, so push detail into well-labelled tables and figures rather than narrating every number in prose. Cut anything that reads like a textbook derivation the examiner already knows; the marks are for your design, your analysis and your judgement, not for re-deriving standard physics. If you find yourself over the limit, the fastest savings almost always come from trimming background theory and tightening the method, not from cutting analysis.

One more discipline pays off here: write in the third person and the past tense throughout, because that is how a scientific report reads and it keeps the prose lean. Replace "I think g came out a bit high probably because of timing" with "The measured g of 9.82 m s⁻² agreed with the accepted 9.81 m s⁻² within the propagated uncertainty, with reaction-time error the dominant contribution." The second version is shorter, more confident and more obviously a piece of analysis — exactly what the conclusion and evaluation criteria reward.

Guidance written by experienced IB examiners and aligned to the current Physics guide (first assessment 2025). Not affiliated with or endorsed by the International Baccalaureate Organization.