Internal Assessment · Biology · Chemistry · 10 min read
IB science IA criteria explained: what each of the four criteria rewards
What each of the four criteria in the Biology, Chemistry and Physics IA rewards, what typically holds reports in the middle band, and how to move up, with a word budget and a final checklist.
On this page
The IB science IA (the "scientific investigation" in Biology, Chemistry and Physics) is marked on four criteria worth 6 marks each: Research design, Data analysis, Conclusion and Evaluation, for 24 marks and 20% of your final grade at both SL and HL. In short: Research design rewards a focused question in a specific scientific context and a method whose choices are explained; Data analysis rewards clear and precise processing with uncertainties; Conclusion rewards an answer that follows from your own data and is compared with accepted science; Evaluation rewards specific weaknesses, how much each one mattered, and improvements tied to them.
Half the marks are for the last two criteria, which are often written in a rush at the end. This guide explains each criterion in our own words, using the current Biology and Chemistry guides (first assessment 2025), shows what typically keeps a report in the middle band, and gives a before-and-after fix for each. For a complete report annotated section by section, see the annotated science IA example.
How is the science IA marked?
| Criterion | Marks | The question it asks |
|---|---|---|
| Research design | 6 | Is the question focused and set in context, are the method choices explained, and could someone repeat it? |
| Data analysis | 6 | Are the data recorded, processed and presented clearly and precisely, with uncertainties, in a way that answers the question? |
| Conclusion | 6 | Does the answer follow from your analysis, and is it compared with accepted science using traceable sources? |
| Evaluation | 6 | Are specific weaknesses identified, is their relative impact explained, and are the improvements tied to them? |
A few rules about how marks are given, all from the subject guides:
- Bands of two. Each criterion has levels of 0, 1–2, 3–4 and 5–6. The teacher finds the level that describes the work best, then picks the upper mark if the work fits that level strongly and the lower if it only just reaches it.
- Best fit, not a checklist. A report can meet parts of a criterion at different levels; the mark reflects the overall balance. The top level does not require a flawless report.
- Criteria are independent. A strong method does not raise your evaluation mark, and a weak conclusion does not drag down your data analysis.
- Who marks it. Your teacher marks the report out of 24; the IB then moderates the marks externally, so the final mark can differ from your teacher's.
- One draft. Your teacher gives advice on one draft without editing it; the next version is final.
The levels climb through command terms. In Research design, the question moves from stated with no context, to outlined in a broad context, to described in a specific one, and the method choices from stated to described to explained. A conclusion moves from stated to described to justified; an evaluation from stating generic weaknesses to explaining the relative impact of specific ones. Mid-band reports usually stop one step short: they say what they did or found, not why.
What does Research design reward?
Three things: a research question placed in a specific context, method choices that are explained rather than listed, and a method detailed enough to be repeated. The guide's clarifications say that context means naming the variables (or the two correlated variables), briefly describing the system, and giving background theory that bears directly on the question. Method choices cover how you measure the variables, the range, interval and number of repeats, how you control other variables, and any safety, ethical or environmental issues.
What holds students in the middle band: a general textbook introduction ("enzymes are biological catalysts…"), a variables table with no reasons, and a method that leaves out a step someone else would need.
Loses marks"The temperatures used were 20, 30, 40, 50 and 60 °C. Each was repeated three times."
Earns marks"Five temperatures from 20 to 60 °C cover the rise in rate, the optimum reported for potato catalase [citation] and the denaturation phase. Five repeats per temperature were chosen after a pilot showed a spread of about 10% between repeats, too large for three runs to give a reliable mean."
The second version explains why that range and why that number of repeats, and ties both to evidence. For writing the question itself, see how to write an IA research question.
What does Data analysis reward?
Processing that is both clear and precise, uncertainties considered appropriately, and processing that is accurate and relevant to the question. In the guide's terms, clear means a reader can follow how you processed the data; precise means you follow conventions: units, consistent decimal places and significant figures, labelled tables and graphs.
The guide also separates two kinds of problem. A major omission or error stops a valid conclusion being drawn at all. A significant one still allows a conclusion but limits how far it can be trusted. The middle band is roughly "clear or precise, but not both", with significant gaps.
What holds students in the middle band: averages and a graph but no processing that actually answers the question; uncertainties in the table headers that are never used again; error bars with no explanation of what they show.
Loses marksA table of mean times at each temperature and a bar chart of the means.
Earns marksRates calculated as 1/t with propagated uncertainties, plotted against temperature as a scatter graph with error bars (stated as ±1 standard deviation), and Q10 calculated for each 10 °C interval below the optimum, with one fully worked example.
See uncertainties in the IA and graphs for the IA for the mechanics, and which statistical test to use if your design compares groups or looks for a correlation.
What does Conclusion reward?
A conclusion that is justified: relevant to the research question, fully consistent with your analysis, and supported by a relevant comparison with the accepted scientific context. Two clarifications matter. "Fully consistent" means you interpret your processed data including its uncertainties. And the scientific context (published values, textbooks, papers, course notes) must be cited in enough detail that a reader could find the source.
What holds students in the middle band: a conclusion that says the hypothesis was "supported" without quoting numbers; a claim stronger than the data allow (a trend called "proportional" when the error bars overlap); a vague nod to theory ("this agrees with collision theory") with no comparison of values.
Loses marks"The results show that as temperature increases the rate increases, which supports my hypothesis and collision theory."
Earns marks"Rate rose from 0.021 ± 0.002 s−1 at 20 °C to 0.083 ± 0.009 s−1 at 50 °C. The Arrhenius plot gives Ea = 54 ± 6 kJ mol−1. The published value I found for this reaction lies just outside this range [citation], so random error alone does not explain the difference; the temperature drop during each run, discussed in the evaluation, would bias Ea in this direction."
The numbers are illustrative. What earns the marks is the structure: the answer with values and uncertainties, a comparison with a cited value, and an honest statement of whether the difference falls within the uncertainty.
What does Evaluation reward?
Specific weaknesses or limitations of your method, an explanation of their relative impact on your results, and realistic improvements explained and linked to those weaknesses. The guide's clarifications give useful categories. Weaknesses can be about controlling variables, the precision of measurements or the variation in the data. Limitations can be about how far the conclusion reaches: the range you tested, the system you studied, the assumptions you made.
At the bottom of the scale are generic weaknesses, the kind that could be written about any experiment. In the middle, weaknesses are specific but simply described. At the top, you explain which weakness mattered most and why.
What holds students in the middle band: a list of five weaknesses of equal weight, none ranked; improvements such as "use better equipment"; nothing about how the weakness changed the actual numbers.
Loses marks"Human error in timing. Improvement: be more careful and do more repeats."
Earns marks"The biggest weakness was judging the end point by eye. At 60 °C, runs lasted about 10 s, so a 0.5 s reaction-time error is about 5% of each reading, compared with about 1% at 20 °C. This makes the high-temperature rates least reliable, and they set the slope of the Arrhenius plot. A light sensor logging absorbance would remove the judgement. The ±1 °C drift in the water bath had a smaller effect (about 3% on rate) and matters less."
A table with columns for weakness, effect on these results, size of the effect and improvement makes ranking easy to show; the annotated example uses one.
How do the four criteria depend on each other?
The criteria are marked independently, but your report is not written independently. Decisions in the design set the ceiling for everything after it.
- Design limits analysis. Three levels and one run each cannot support a gradient with an uncertainty or a statistical test, however well you write.
- Analysis limits the conclusion. If you never processed uncertainties, the conclusion cannot say whether a difference from the literature is real.
- Design feeds evaluation. A method that recorded the uncontrolled variables (room temperature, light, time between runs) gives you evidence for the evaluation; one that did not leaves you guessing.
So spend a lesson on a pilot before you collect the real data. It is the cheapest way to raise three criteria at once.
How should you spend 3,000 words?
The report is capped at 3,000 words. Charts, diagrams, data tables, equations, calculations, citations, the bibliography and headers do not count. The start of the report gives the title, your candidate code, the codes of any group members, and the word count; no cover page or contents page is needed. A split that matches the weight of the criteria:
| Section | Criterion | Suggested words |
|---|---|---|
| Research question, context and background theory | Research design | 400–500 |
| Variables, method choices with reasons, safety, ethics, environment, procedure | Research design | 450–550 |
| Text explaining processing, graphs and statistics | Data analysis | 300–450 |
| Conclusion with values, uncertainties and literature comparison | Conclusion | 500–650 |
| Evaluation: ranked weaknesses, impact, improvements | Evaluation | 600–750 |
These are our suggestions, not IB rules; even the upper ends total only 2,900 words. The shape to avoid is the reverse: 1,200 words of background and 300 words of evaluation.
What are the rules for group work?
Collaboration is optional. If you collaborate, the group is at most three students, and you may develop part of the method together. Each of you must still answer your own research question in one of three ways: a different independent variable, the same independent variable with a different dependent variable, or a different selection from a larger data set the group collected. No two students may present the same raw data, and every report, including the method, is written individually; group reports are not allowed. Even without formal collaboration, classmates can help you collect data. If your school builds a shared class database with standard protocols and you use it, your IA counts as a database investigation, and your method focuses on how you selected and filtered the data.
Final checklist before you hand in
- The research question names the independent and dependent variables, the range and the specific system.
- The background leads to a prediction; nothing in it is there just to fill space.
- Every method choice (range, interval, repeats, controls, equipment) has a reason.
- Safety, ethics and environment are addressed where relevant.
- Someone else could repeat the method from your description alone.
- Tables have units and uncertainties in headers, with consistent decimal places.
- Processing goes beyond averages and is shown with one worked example.
- Error bars are on the graphs, and the caption says what they represent.
- The conclusion quotes your values with uncertainties and answers the question directly.
- At least one cited, traceable source is compared numerically with your result.
- Weaknesses are specific, ranked by impact and linked to realistic improvements.
- The title, candidate code(s) and word count are at the top, and the count is at most 3,000.
For the patterns that most often pull marks down, with before-and-after fixes, read the most common science IA mistakes. And if you are still choosing what to investigate, start with our Biology IA ideas or Chemistry IA ideas.
Questions students ask
Is there still a personal engagement or communication criterion?
No. The science IA on the current syllabus (first assessed 2025) has four criteria worth 6 marks each: Research design, Data analysis, Conclusion and Evaluation. Personal engagement and communication were criteria in the old IA and no longer exist as separate criteria, although clear communication is now part of Data analysis.
Are the criteria different at SL and HL?
No. SL and HL students write the same kind of report against the same four criteria, out of 24, and the IA is worth 20% of the grade at both levels.
Do I need a hypothesis?
None of the criteria asks for one by name. What Research design rewards is a research question set in a specific context, including relevant background theory. A prediction reasoned from that theory is a natural way to show the context, and it gives your conclusion something to test.
Can my teacher read my draft?
Yes, once. Your teacher can give written or spoken advice on one draft but cannot edit it, and the next version you hand in is the final one. Make the draft complete so the feedback is worth having.
What happens if my report is over 3,000 words?
The guide sets 3,000 words as a maximum. Treat it as a hard ceiling and ask your teacher how over-length work is handled at your school; trimming is almost always possible without losing marks.
Try one: Biology and Chemistry IA
A student timed a colour change by eye at five temperatures, from 20 °C (about 50 s per run) to 60 °C (about 10 s per run). Which evaluation point would be credited at the highest level of the Evaluation criterion?
- There may have been human error in the timing, and more repeats would make the results more reliable.
- The water bath was only accurate to ±2 °C, which is a weakness. A better water bath could be used.
- The end point was judged by eye. A reaction-time error of about 0.5 s is about 1% of a 50 s run at 20 °C but about 5% of a 10 s run at 60 °C, so the high-temperature rates, which set the steepest part of the curve, are the least reliable. A light sensor logging absorbance would remove this judgement.
- The experiment was a success because the results followed the expected trend.
Show the answer and why
Answer: C
Option C names a specific weakness, quantifies how much it affects different parts of the data, explains its effect on the conclusion and ties a realistic improvement to it. Option A is generic: it could be written about any investigation. Option B is specific but does not explain the effect or why a better bath would matter. Option D is not an evaluation of the method at all.