Home › IA guide
SL and HL · 2025 syllabus
The IA, step by step
Your internal assessment is a scientific investigation: you ask your own question, collect real data and analyse it. This page explains what examiners look for, shows a full worked example of the analysis, and ends with a free self-check you can use before you hand in.
Go to the self-checkLab or simulation?The IA on one pageThe four criteriaRead an annotated exemplar
The IA on one page
The whole IA as a mind map: the four criteria, what each one needs, and the mistake I see most often in each. Click any box to jump to that part of this guide.
- Focused question: variables, range, units and the system
- Short theory you will really use: an equation to test
- Pilot run first: choose the range, 5+ values, repeats
- Equipment table: uncertainty and calibration of each instrument
- Control variables: why it matters, how kept, how checked
- Safety, ethical and environmental points
⚠ From my marking: No calibration, no instrument uncertainties, shallow controls
- Raw data tables: quantity, unit and ± in every heading
- Consistent significant figures in every column
- Standard deviation when you have 5 or more repeats
- Carry uncertainties through every calculation
- Graph: error bars, best fit, steepest and least steep lines
- Final result ± uncertainty, rounded correctly
⚠ From my marking: No standard deviation, mixed significant figures
- Answer the question with numbers ± uncertainty
- Describe the relationship your data show
- Scientific context: published value, theory or paper
- Cite the source so a reader can find it
- Agree? Is the accepted value inside your range?
- Explain any odd point with physics
⚠ From my marking: No scientific context to compare with
- Specific weaknesses: random or systematic?
- Uncertainty budget: which source matters most
- Put a number on the size of each weakness
- A realistic improvement for each weakness
- Say how much it would help (e.g. 7 % → 3 %)
- Limits of the conclusion: range, system, assumptions
⚠ From my marking: General sentences with no numbers
The rules in one minute
The requirements and the criteria are the same for both levels. HL students do not need a harder topic.
You must collect and analyse quantitative data. Add qualitative observations when they help.
A lab experiment, fieldwork, a spreadsheet model, data from a database, or a simulation. You can combine them. See the lab and simulation guides.
Group work is allowed, but each student has their own research question, their own raw data and writes their own report.
Your teacher reads one draft and gives advice, but does not edit it. The next version you hand in is your final version.
Title, your candidate code, the codes of any group members, and the word count. No cover page or contents page is needed.
Not in the word count: graphs and diagrams, data tables, equations and calculations, citations, the bibliography and headings. So use tables and graphs freely, and keep the writing short and clear.
Six steps from idea to final report
- Pick a topic you care aboutChoose something you can really measure with the equipment, time and data you have. A simple idea done carefully scores better than an exciting idea that does not work.
- Write a focused research question, then try itDo a short trial run (a pilot). It tells you a sensible range, how many readings you need, and what goes wrong. Change the plan if you need to.
- Plan the methodChoose the range and number of values of your independent variable (aim for at least 5, spread widely), the repeats (at least 3 for each value), the instruments, the control variables and how you will keep them constant, and any safety, ethical or environmental points.
- Collect the dataRecord the raw data straight away, with units and uncertainties, and note anything unusual you see. Keep a photo of the set-up.
- AnalyseProcess the data, carry the uncertainties through, draw a graph with error bars, find the gradient and its uncertainty. The worked example below shows how.
- Write, check, hand inWrite the conclusion and evaluation, then go through the self-check on this page. Hand in one draft, act on your teacher's advice, and submit the final version.
The four criteria, in plain English
Each criterion is worth 6 marks. Your teacher reads your report and picks the description that fits it best. The top band does not mean perfect: it means the work does what is asked clearly and well.
1. Research design 6 marks
Is your question focused, and could someone else repeat your investigation from your report?
What a top report does
- Names the independent and dependent variables, with units and the range, and the system being studied.
- Gives short background theory that is used later, usually leading to an equation and a prediction.
- Explains the choices: why this range, this number of values, these repeats, these instruments.
- Lists the control variables and says how each one is kept constant and why it matters.
- Mentions safety, ethical and environmental points that apply.
- Describes the method with enough detail to repeat it (sizes, models, settings), without obvious steps.
- Gives the uncertainty of every instrument and how each one was calibrated or zero-checked (example).
Mistakes that cost marks
- A broad question such as “How does length affect a wire?”
- Pages of general theory that never connect to the analysis.
- Control variables listed with no reason and no method: “temperature was kept constant” (see a deep version).
- Too few values (fewer than 5) or a very narrow range.
- A vague method: “measure the length”, “use a ruler”.
- No instrument uncertainties and no calibration or zero check.
- No mention of safety when there is a hazard (heat, mains, lasers, heavy masses).
🔒 Criteria 2, 3 and 4 are part of the full course.
Full course resources are available to enrolled students through an access code.
Data analysis, Conclusion and Evaluation: what a top report does and the mistakes that cost marks.
Enrolled student? Enter your access code.
You only need to do this once on each device.
Research design: what students forget
In research design, students forget to say how they calibrated their instruments, and they do not give the uncertainty of each piece of equipment. Without these, the method cannot be repeated properly and the uncertainties later in the report have no starting point.
The control variables are also very shallow. Students write “temperature was kept constant”, but they do not explain why it would change the result or how exactly they kept it constant.
— Ahmed Gomaa
Use an equipment table, not a shopping list
For every measuring instrument, give what it measures, its range, its uncertainty and how you checked or calibrated it. Here is the table for the wire experiment in the worked example:
| Instrument | Measures | Range used | Uncertainty | Calibration or check |
|---|---|---|---|---|
| Metre rule (mm scale) | length L of wire | 0.200–1.000 m | ± 0.001 m (± 0.5 mm at each end) | Checked that the zero is at the very end of the rule; read at eye level. |
| Micrometer screw gauge | diameter d of wire | 0.40 mm | ± 0.01 mm | Closed the jaws and read the zero error (+0.01 mm); subtracted it from every reading. |
| Digital multimeter (20 Ω range) | resistance R | 0.8–4.0 Ω | ± 0.01 Ω | Touched the two leads together and recorded the lead resistance (0.05 Ω) before starting. |
| Thermometer | room temperature (control) | about 22 °C | ± 0.5 °C | Checked in melting ice: read 0.0 °C. |
Ways to calibrate or check an instrument
Read the instrument when the true value is zero: close a micrometer or vernier, tare a balance, zero a force sensor or a light sensor in the dark. Correct every reading by the zero error.
Measure something you already know: a thermometer in melting ice (0 °C), a balance with a standard mass, a meter with a standard resistor, a sensor against a reference.
For a sensor or a home-made instrument, record two known values and draw a calibration line, then use it to convert every reading.
One sentence for each instrument is enough: what you checked, what you found, and how you corrected for it.
Control variables: always say why and how
For each control variable, a top report answers three questions: why would it change the dependent variable (use physics), how exactly did you keep it constant, and how did you check that it stayed constant?
| Control variable | Why it matters | How it was controlled | How it was checked |
|---|---|---|---|
| Temperature of the wire | Resistivity rises with temperature, so R would rise. | Small meter current; readings taken quickly. | Room temperature recorded at start and end. |
| Material and diameter | R = ρL/A: a different ρ or A changes the gradient. | All lengths cut from the same reel of wire. | Diameter measured at one point only: a weakness, discussed in the evaluation. |
| Contact between clips and wire | A loose contact adds extra resistance to the reading. | Same clips, cleaned, pressed on the same way each time. | Repeat readings agreed within 0.05 Ω. |
| Straightness of the wire | A kink makes the measured length shorter than the real length. | Wire taped straight along the metre rule. | Checked by eye before each reading. |
Uncertainties: the other big place marks are lost
In data analysis, most marks are lost on uncertainties and errors. Students mix up the types of error, forget to carry uncertainties through their calculations, and most of them never use the standard deviation, even when they have enough repeats to do it.
They are also not consistent with significant figures: 1.5 in one row and 2.04 in the next, or a resistance written as 121.95122 Ω from readings with three figures.
— Ahmed Gomaa
Errors and uncertainties are not the same thing
Readings scatter above and below the true value, for example reaction time with a stopwatch. Repeats and averaging reduce its effect. It shows up as spread in your repeats.
Every reading is shifted the same way, for example a meter that does not read zero. Repeats do not help. It often shows up as an unexpected intercept on your graph.
The range within which you think the true value lies, written as ± after the value. It is a number you calculate, not a mistake.
Step 1 · Uncertainty of one reading
- Digital instrument: ± 1 in the last digit shown (a meter reading 2.38 Ω gives ± 0.01 Ω).
- Analogue scale: ± half of the smallest division (a ruler marked in mm gives ± 0.5 mm), unless reading it is harder than that.
- Timing by hand: your reaction time matters more than the stopwatch display. Time many oscillations or a long interval, and use repeats.
Step 2 · Uncertainty from repeats: use the standard deviation
When you have 5 or more repeats, use the standard deviation (σ) as the uncertainty in the mean. It uses every reading, not just the two extremes. With only 3 repeats, half the range is acceptable, but plan for 5.
Example: the time for 10 oscillations of a pendulum, measured 5 times.
| Trial | t / s ± 0.01 | t − mean / s | (t − mean)² / s² |
|---|---|---|---|
| 1 | 15.42 | −0.004 | 0.00002 |
| 2 | 15.38 | −0.044 | 0.00194 |
| 3 | 15.51 | +0.086 | 0.00740 |
| 4 | 15.45 | +0.026 | 0.00068 |
| 5 | 15.36 | −0.064 | 0.00410 |
| Sum | 0.01414 | ||
σ = √( Σ(t − mean)² ÷ (n − 1) ) = √(0.01414 ÷ 4) = 0.059 s
t = (15.42 ± 0.06) s → period T = t ÷ 10 = (1.542 ± 0.006) s
In a spreadsheet: =AVERAGE(B2:B6) gives the mean and =STDEV.S(B2:B6) gives σ. Show one calculation by hand in your report, then let the spreadsheet do the rest. Note that the spread here (σ = 0.06 s) is much bigger than the stopwatch resolution (0.01 s), so the spread is the real uncertainty.
Step 3 · Carry the uncertainty through every calculation
| If you… | then… | Example |
|---|---|---|
| add or subtract | add the absolute uncertainties | (20.0 ± 0.5) °C − (15.0 ± 0.5) °C = (5.0 ± 1.0) °C |
| multiply or divide | add the percentage uncertainties | v = d ÷ t: 1 % in d and 3 % in t gives 4 % in v |
| raise to a power n | multiply the percentage uncertainty by |n| | A = πd²/4: 2.5 % in d gives 5 % in A |
| use a graph | draw error bars; use the steepest and least steep lines for the gradient | see the worked example |
Step 4 · Write the final answer correctly
- Round the uncertainty to 1 significant figure (2 if the first digit is 1).
- Give the value to the same decimal place as the uncertainty: (1.542 ± 0.006) s, not (1.5424 ± 0.0059) s.
- The % difference from a published value is not your uncertainty. Compare the two in your conclusion.
Step 5 · Be consistent with significant figures
Examiners check this in every table, calculation and final answer. The rules are simple; the mark is lost when they are applied in some rows and not in others.
| Inconsistent | Consistent | ||||
|---|---|---|---|---|---|
| V / V | I / mA | R / Ω | V / V ± 0.01 | I / mA ± 0.1 | R / Ω |
| 1.5 | 12.3 | 121.95122 | 1.50 | 12.3 | 122 |
| 2.04 | 16.8 | 121.4 | 2.04 | 16.8 | 121 |
| 3 | 24 | 125 | 3.00 | 24.0 | 125 |
- Raw data: the instrument decides the decimal places, and every row of a column has the same number. A meter reading to 0.01 V gives 1.50 and 3.00, not 1.5 and 3.
- Processed data: give the same number of significant figures as the least precise measurement used. Here I has 3 significant figures, so every R has 3.
- Working: keep all the calculator digits in the middle of a calculation and round only the value you write down at the end.
- Final answer: round the uncertainty first, then give the value to the same decimal place (Step 4).
- Check before you hand in: scan each column from top to bottom. Every number in it should look the same.
Scientific context: where most marks are lost
The conclusion is where my students lose marks most often. The reason is nearly always the same: they forget the scientific context. They give their result, but they never find a published value, a theory or a research paper to compare it with.
— Ahmed Gomaa
What counts as scientific context?
A constant or property you measured, such as g, a resistivity, a specific heat capacity or a spring constant, taken from a data handbook or a reliable website.
An equation from the course that predicts your relationship, for example T = 2π√(L/g) predicts that T is proportional to the square root of L.
A study that measured the same effect. You compare the trend or the size of the effect with yours.
The manufacturer's values for a component you used, such as a thermistor, an LED, a solar cell or a spring.
Where to look
- Your class notes and the data booklet for the equations and the theory behind your question.
- Data handbooks and the websites of national measurement laboratories and universities for published values.
- Google Scholar (scholar.google.com) for research papers. Search with the names of your two variables and the system, for example “squash ball temperature coefficient of restitution”. Reading the abstract and the results graph is often enough.
- The supplier's website for data sheets of the components you used.
Start looking before you collect data. If you know what the published result is, you can choose a range and method that let you compare properly.
How to compare: three cases
| Your result is… | Compare it like this | Example |
|---|---|---|
| a value (a constant or property) | Check whether the published value lies inside your uncertainty range. If it does not, compare the % difference with your % uncertainty and suggest a systematic cause. | ρ = (4.9 ± 0.3) × 10⁻⁷ Ω m; published 4.9 × 10⁻⁷ Ω m lies inside the range, so they agree. |
| a relationship | Compare the shape, gradient or power with the one the theory predicts. | Theory: T ∝ L0.5. Your log–log gradient is 0.48 ± 0.03, so 0.5 is inside the range. |
| a trend with no simple equation | Compare the direction and size of the trend with a research paper, and explain the physics behind it. | A paper reports that bounce height rises with ball temperature; your data show the same trend, levelling off above 60 °C. |
Cite it so it can be traced
Give the author or organisation, the title, the year, and the page or link. A reader must be able to find the exact value you used. A value with no source does not count as scientific context.
“The resistivity of the wire was (4.9 ± 0.3) × 10⁻⁷ Ω m. The published value for constantan at 20 °C is 4.9 × 10⁻⁷ Ω m [2], which lies inside our uncertainty range, so the result agrees with accepted data.”
🔒 The rest of this guide is part of the full course.
Full course resources are available to enrolled students through an access code.
You have seen the first 8 of 16 sections. The other 8 are part of the full course.
Enrolled student? Enter your access code.
You only need to do this once on each device.
Free to try, no code needed: A.1 Kinematics · A.2 Forces · B.1 Thermal energy transfers · C.1 Simple harmonic motion · D.1 Gravitational fields · E.1 Structure of the atom · IA guide