IB Chemistry IA / Scientific Investigation: From Question to Evaluation

The work commonly searched for as the IB Chemistry IA is officially called the Scientific Investigation in the current course. It is not the story of a perfect experiment. It is a data-justified chemical answer to an individual research question. Much of the report’s quality is determined before writing begins: the question, measurement range, instrument resolution, repeats and pilot limit what can later be concluded.

This guide supports planning and self-review; it does not produce assessed work for a student. The research decisions, analysis and argument must reflect the student’s own work; any permitted quotations or externally generated material must be clearly credited.

The current task at a glance

The 3,000 words are a maximum, not a target. The IB Chemistry guide excludes items including data tables, equations, calculations, citations and bibliography from the word count, but this does not justify overcrowded figures or hiding prose in tables. Follow the school’s submission instructions too.

The four criteria are marked separately, but they are not four isolated essays. One evidence chain connects them: the question determines the method, the method determines data quality, the data determine analysis, and the analysis limits the conclusion and evaluation.

The four assessment criteria

Each criterion has a maximum of 6 marks, making 24 in total. This is an explanatory summary, not a replacement for the official descriptors supplied by your school.

• Research design: connect a focused, contextualised question to justified variables, measurement choices and a reproducible method.
• Data analysis: present relevant raw and processed data clearly; justify processing and account for uncertainty in the answer being developed.
• Conclusion: answer the question consistently with the analysis and compare the finding with relevant, referenced scientific context.
• Evaluation: explain the relative effects of specific weaknesses or limitations, then justify improvements that address those causes.

Build one evidence chain

The investigation is one evidence chain

1. QuestionMeasurable and chemically meaningful
2. MethodControls the relevant variables
3. RangeResolves the expected change
4. Data qualityEnough precision and repetition
5. AnalysisAnswers the question
6. ConclusionClaim justified by evidence
7. EvaluationSpecific limits and improvements

Each link carries a different risk:

• Question: too broad, purely descriptive or not measurable.
• Method: the dependent variable is not measured directly or reproducibly enough.
• Range and resolution: the expected change is smaller than uncertainty, or the reaction is too fast.
• Data quality: too few repeats, inconsistent controls or missing raw readings.
• Analysis: an attractive graph appears, but its quantity does not answer the research question.
• Conclusion: the claim is stronger than the evidence.
• Evaluation: generic “human error” appears without a specific effect or improvement.

Editing prose late cannot recover data that were never measured. The most valuable checks therefore occur before full data collection and immediately after it.

Filter the research question

Name the chemical system, the quantity varied, the quantity measured and the investigated range. An elaborate instrument does not make a question strong by itself. The research-question and pilot guide owns the eight detailed filters and the worked progression from broad interest to measurable question.

Use the pilot to change a decision

A pilot tests whether the method can produce evidence of the required quality. If a reaction is too fast for reliable timing, change the method or permitted range and test the revision. Preserve the reason for that decision. The pilot article contains the detailed thiosulfate example and decision sheet. Because that example produces SO₂, it requires institutional laboratory supervision, prior risk assessment, suitable ventilation and authorised equipment.

Move from raw data to a chemical claim

Data analysis is not a competition to perform more calculations. It is an auditable route from measurement to the answer.

1. Raw data: every direct reading, unit, instrument and relevant qualitative observation. Do not record only averages.

2. Instrument and resolution: state the display increment and how measurement uncertainty is treated. Many decimal places on a digital display do not guarantee accuracy.

3. Repeats: repetition reveals spread and helps identify an anomalous result. Exclude a value only for a documented scientific reason.

4. Processed quantity: mean, rate, concentration, energy, equilibrium constant or another value belongs only when it advances the research question.

5. Graph or model: axes, units, uncertainty and fit should be clear. Do not force a linear model merely because a trendline is convenient.

6. Chemical claim: state what the trend shows, which values support it, which chemical model interprets it and where the evidence stops.

An auditing reader should be able to trace one graph point back to its original readings and understand why it is relevant to the research question.

Treat uncertainty as information

Uncertainty is not a compulsory “errors” paragraph. It shows how strong a comparison is and how precise a conclusion can be.

With random spread, repeated readings differ. Timing, reading noise or inconsistent sample handling may contribute. Repetition and averaging can reduce the influence of random variation, but cannot repair insufficient instrument resolution.

A systematic bias can shift results in a similar direction: an incorrectly calibrated pH meter, continuing heat loss or a stock solution of incorrect concentration. More repeats do not move the result closer to the true value; calibration, control measurement or method change is needed.

The range is the largest repeat minus the smallest; half-range is half that difference. Neither is standard deviation, standard error of the mean or a confidence interval. A display increment of 0.01 cm³ does not by itself establish a total uncertainty of ±0.01 cm³.

For example, repeated readings of 10.02, 10.03 and 10.08 cm³ have a range of 0.06 cm³. That calculation does not establish instrument accuracy or statistical significance between conditions. Identify which uncertainty you estimate, state the assumptions, and represent that quantity on the graph. Averaging suitable independent repeats can reduce the random uncertainty of a mean; it does not correct a calibration bias or a consistently late endpoint.

Write a defensible conclusion

A strong conclusion has four parts:

1. Claim: a direct answer, limited to the investigated range.
2. Data: a specific trend, value, gradient, model parameter or uncertainty.
3. Chemistry: the relevant theoretical relationship and comparison with accepted scientific context.
4. Limit: what the data do not show and the conditions attached to the claim.

Weak: “Reaction rate increased with temperature, which agrees with theory.”

More defensible, if these are the observed results: “Across the investigated 25–45 °C range, the reciprocal-time rate proxy increased consistently; the difference between the 25 and 45 °C means exceeded the spread of repeats. This is consistent with a larger fraction of particles possessing energy at least equal to the activation energy at higher temperature. The conclusion is nevertheless limited to reciprocal time measured using this approximate optical endpoint.”

Do not claim causation from a correlational method. Do not generalise beyond the measured range. Uncertainty does not automatically weaken the conclusion; it defines how large a claim is justified.

Evaluate cause, effect, improvement and trade-off

Evaluation: four linked parts

1Cause

What specifically created the limitation?

2Affected data

Which measurements or derived values changed?

3Direction

Was the result raised, lowered or made less precise?

4Improvement + trade-off

What would reduce the effect, and what would it cost?

“Human error” is not an evaluation because it identifies no mechanism. “Use more precise equipment” is also insufficient unless the affected data and expected improvement are clear.

Worked evaluation: two timing delays with opposite effects. Let T be the time from actual mixing to the defined optical endpoint, δstart the delay before starting the clock, and δend the delay in noticing that endpoint. Then t = T − δstart + δend. For an illustrative T = 10.0 s:

A late start shortens measured time and raises 1/t; a late endpoint lengthens time and lowers 1/t. With both delays, the net direction depends on their difference. The same 0.6 s late start changes a true 60.0 s interval to 59.4 s, raising the reciprocal by about 1.0% rather than 6.4%. Shorter trials therefore suffer the larger relative effect. Do not automatically claim a flatter temperature trend without identifying which delay occurred.

Here 1/t is a rate proxy, not automatically an initial reaction rate: it has unit s⁻¹ and averages progress towards an optical threshold. Comparisons need the same threshold, geometry, concentrations, volume and other relevant controls, with a defensible link between the threshold and comparable chemical progress. A scattering endpoint can itself depend on mixing and particle formation.

A synchronised reagent-addition/start trigger plus logged light transmission at a fixed threshold can reduce manual timing delays. It still requires calibration and checks for trigger latency, mixing, detector response, temperature and the optical meaning of the threshold. A light sensor alone does not correct an unsynchronised start.

For every important limitation ask:

• which data were affected;
• in which direction and with what likely magnitude;
• whether it changed validity or mainly precision;
• whether the proposed improvement actually treats that cause;
• what new limitation or resource demand the improvement creates.

Two or three detailed, evidence-based evaluations are more useful than a list of ten generic errors.

Protect authenticity and academic integrity

The IB guide is explicit: the submitted Scientific Investigation must be the student’s own work. Support is allowed, but decisions and writing cannot be outsourced.

Potentially appropriate, subject to school rules:

• teacher explanation of criteria and scientific concepts;
• questions that help the student notice a gap in the method;
• feedback on a draft within IB and school limits;
• technical support for safe equipment use;
• teaching source-search and referencing methods;
• a language or structure checklist that does not rewrite the student’s argument.

Not appropriate or a serious risk:

• inventing a research decision, data or result;
• analysing the student’s data for them;
• submitting paragraphs written by someone else as the student’s own;
• taking material without citation;
• submitting AI-generated prose as the student’s own;
• rewriting the report until it no longer reflects the student’s thinking.

The IB's statement on AI and assessment does not impose a blanket ban on AI tools. It requires transparency: AI-generated text, images or graphs included in work are not the student's own and must be credited in the text and referenced. A citation does not turn outsourced thinking into independently assessed work. Follow the current IB and school rules, document permitted assistance, and verify every source. ChemistryTutor teaches concepts and methods; it does not generate a personalised submission, analyse your assessed data for you or invent results.

What can parents do?

A parent can support the process without writing Chemistry or prose:

• make milestones and school deadlines visible;
• ask whether the pilot happened before full data collection;
• check that safe, authorised equipment access exists;
• remind the student to preserve raw data, versions and sources;
• ask about decisions: “Which pilot result made you change the method?”

Do not rewrite the conclusion, recalculate and select the “better-looking” data, or search for a finished IA to copy. Student ownership and thinking are part of the assessment.

A twelve-week workflow with two decision gates

This is an illustrative schedule spread across 12 calendar weeks, not an official IB timetable or prescribed teaching-hour allocation. School deadlines and supervisory instructions take priority.

A four-phase, twelve-week workflow

Weeks 1–2Question

Background, variables, feasibility and safety.

Weeks 3–4Pilot gate

Test the method; approve or redesign before full collection.

Weeks 5–7Evidence gate

Collect, process and audit sufficient data.

Weeks 8–12Argument

Conclude, evaluate, reference and complete the integrity check.

Phase 1 — question and feasibility, weeks 1–2

Background theory, variables, safety, equipment and an initial method outline. The output is a pilotable question, not a polished introduction.

Phase 2 — pilot gate, weeks 3–4

Test extreme and middle points, timing, resolution and controls. Week 4 gate: approve, modify or redesign the method. Full collection begins only after a documented decision.

Phase 3 — data and evidence gate, weeks 5–7

Collect repeats, preserve raw data, process a sample calculation and test whether the graph is meaningful. Week 7 gate: are quantity, quality and traceability sufficient to answer the question? If not, collect what is missing now—not at the end of the programme.

Phase 4 — argument and integrity, weeks 8–12

Build the conclusion from data and accepted context, complete the evaluation, then use the final two weeks to check the full evidence chain, references, word limit, figures, authenticity and any AI declaration required by the school.

Every week should end with a tangible output: decision sheet, pilot note, raw-data backup, analysis audit or revised section. “Worked on it” is not a milestone.

Final submission check

Question and design

• The question is precise, contextualised, measurable and chemically relevant.
• Independent and dependent variables, range, intervals, repeats and controls are justified.
• The method is reproducible without unnecessary recipe prose.
• Safety, ethical and environmental decisions are specific.

Data and analysis

• Every graph point traces back to raw readings.
• Units, significant figures, instrument resolution and uncertainty are consistent.
• Calculations and graphs answer the research question; no analysis is decorative.
• Any excluded value has a documented scientific reason.

Conclusion and evaluation

• The conclusion directly answers the question using specific data and uncertainty.
• Comparison with accepted chemical context is referenced.
• Each important limitation links cause, affected data, direction, improvement and trade-off.
• The claim does not exceed the measured range or evidential strength of the method.

Submission and integrity

• The report is no more than 3,000 words and states the word count.
• Every source is cited appropriately and the bibliography is complete.
• Any AI declaration required by the school is complete.
• Versions, raw data and calculations are traceable.
• Every figure is readable, labelled and contributes directly to answering the research question.
• The scientific decisions, analysis and argument are the student’s own; permitted borrowed material is clearly identified and referenced.

A strong Scientific Investigation does not pretend the experiment was flawless. It shows that the student recognised the limits of the evidence, interpreted the data using appropriate Chemistry and made exactly the claim that the investigation can defend.

Online work can develop question design, processing, uncertainty and argument. Actual apparatus handling, risk assessment and measurement belong in supervised laboratory work. These original synthetic examples must not be submitted as collected experimental data.