How to Write a Lab Report That Reads Like Science

IMRaD: the four-part skeleton

IMRaD stands for Introduction, Methods, Results, Discussion — the skeleton every experimental report wears, from first-semester chemistry to senior thesis work. The order mirrors how a replicator thinks: why the question was asked, how the test was run, what the instruments recorded, what the recordings mean. Each section answers exactly one of those questions, and material answering a different question belongs elsewhere. Most introductory courses ask for 1,000 to 1,500 words of body text plus title page and references — a shorter document than the essay you wrote to get in.

Write the sections in working order, not reading order. Methods first, while the bench details are still fresh; Results second, straight from your data tables; then the Discussion, then the Introduction, and the abstract dead last. An abstract written before the results exist predicts findings the report never delivers — as a tutor I watched that happen in roughly half the reports students brought me, and the abstract was usually the only section written the night before.

• Abstract: 100-150 words, one sentence each for purpose, method, headline result and implication — written last
• Introduction: about 250 words, narrowing from the field's general problem to your specific hypothesis
• Methods: about 350 words, written so a classmate who missed the lab could repeat the experiment
• Results: about 400 words, every figure and table called out in order, no causes assigned
• Discussion: about 400 words — interpretation, comparison with published values, and your dominant error source

What students get wrong about lab reports

The most common failure is writing an essay with a beaker in it. Students pad the Introduction with 500 words of textbook history before reaching the hypothesis, while Methods shrinks to 80 words that say 'we followed the procedure.' Graders read that ratio as a missing skill, because a lab report's scientific value lives in Methods and Results — the introduction is context, not a chapter. The passive-voice myth makes it worse: students warp every sentence to avoid 'we,' although APA 7 and most current journals accept first person without complaint.

The second failure is copying the lab handout's procedure word for word. That is patchwriting — a plagiarism risk your instructor's similarity checker will catch — and it dodges the actual assignment, which is translating what you did into replicable prose. Third comes interpretation smuggled into Results: 'this proves osmosis occurred' is a Discussion sentence wearing the wrong label. In my experience grading alongside TAs, those two errors cost more points than a wrong hypothesis ever does, because a wrong result analyzed honestly is still good science.

Methods: the replication recipe

Run the replication test: hand your Methods section to a classmate who missed the lab and see whether they could repeat the experiment without asking you anything. Passing requires specifics — 'three 250 mL beakers,' not 'some beakers'; 'five trials (N = 5),' not 'several runs'; the balance model, the solution concentrations, the temperature, the software and its version. Everything sits in past tense: 'we heated the mixture to 78 °C,' not 'heat the mixture.' Present tense turns your report back into a handout.

Document deviations, and document them without apology. 'Trial 3 was repeated because the stopcock leaked' reads as competence, not failure — real lab work always contains retries, and instructors quietly reward the students who say so. Leave out the false starts that changed nothing, but never silently drop a trial that failed: a note in Methods explaining the exclusion protects you in the Discussion. One habit worth stealing from published papers: name the lab manual as your base procedure and describe only what you changed. That single move cuts Methods by a third and raises its precision.

Results: data, not interpretation

Results answers one question: what did you measure. Report central values with their uncertainty, sample sizes, trends and any statistics the course expects — means, standard deviations, a t-test if the rubric names one. Call out every figure and table in the text, in numerical order, and give the numbers rather than the adjectives: 'the solution reached 78.4 °C after six minutes' beats 'it got hot fast' every single time. A Results section in the 1,500-word report carries about 400 words and usually two or three figures or tables.

Negative results belong here too, not in a footnote or a trash bin. If trial four sat 9% above the other five, it goes in the table, in the mean calculation (or in a documented exclusion), and eventually in the Discussion. Keep causal vocabulary out of this section entirely — 'because,' 'proves,' 'explains' all belong one section later. Raw data tables, if your instructor wants them, ride in an appendix behind the References. The next section gives you the sentence-level test for drawing that boundary.

Show, don't interpret: the sentence-level test

Here is the test I give every tutee: if a reader could argue with your sentence while staring straight at your data, that sentence is interpretation and it belongs in the Discussion. Compare two versions. Results voice: 'Group B averaged 12.3 seconds, 18% slower than Group A.' Discussion voice: 'Group B's slower times probably reflect fatigue, since error rates climbed after minute eight.' The first sentence reports; the second explains. Same experiment, same numbers — different sections of the report.

The hard line exists because interpretation needs evidence, and the evidence — your analysis, your comparison with published values — has not been presented yet. Some TAs literally circle every 'because' in a Results section, and the circles mark sentences to move, not sentences to defend. Moving them is cheap: the Discussion needs material anyway, and a Results section stripped of argument reads faster and scores cleaner. My honest opinion is that this one boundary separates A reports from B- reports more reliably than any other habit in this guide.

Uncertainty & error analysis

Every measurement carries uncertainty, so report values with it: 9.62 ± 0.05 m/s², not a naked 9.62. When your course compares against a standard, compute percent error = |experimental - accepted| ÷ accepted × 100. Worked example: you measure gravitational acceleration as 9.62 m/s² against the accepted 9.81 m/s². The math runs |9.62 - 9.81| ÷ 9.81 × 100, which lands at about 1.9% — report it to two significant figures, no more.

Then sort your errors into the two families that matter. Random error scatters both directions — reaction-time jitter across trials is the classic case — and it shrinks when you average more trials. Systematic error pushes everything one direction: a balance reading 0.2 g high biases every mass in the data set, and no amount of averaging fixes it. Name the dominant source and its direction in one plain sentence, such as: 'The largest source of systematic error was balance calibration, which biased all mass readings upward by roughly 0.2 g.' That sentence earns more credit than a bulleted list of every error that theoretically exists. My opinion, and I will defend it: error analysis graded well is error analysis that makes a claim.

Figures and captions

Two placement rules cover most of the formatting points: figure captions go below the figure, and table captions go above the table. Both run as complete sentences with units and conditions, written so a reader skimming only the visuals still understands the experiment. A caption that says 'Figure 1' alone is a caption that forfeits points. The split is standard across APA, MLA and most journal styles, so it is one of the cheapest wins available on the rubric. Check the figures once more after final formatting, too: captions drift when images reflow, and a caption stranded on the wrong page helps nobody.

• Number figures and tables in separate series (Figure 1, Figure 2, Table 1), ordered by first mention
• Call out every figure and table in the text; an uncalled-out visual loses points on most rubrics
• Label both axes with quantity and unit — Time (min), Absorbance (AU) — and never leave an axis unitless by accident
• Keep captions below figures and above tables even when the layout looks odd; the rule does not bend
• State the sample size in the caption: 'Mean mass across five trials (N = 5).'
• Skip 3D pie charts and decorative gradients; they distort proportions and read as filler

Discussion that faces anomalies

Open the Discussion by answering your hypothesis in plain words — supported, partially supported, or not supported — before any hedging. Then put your numbers next to published ones: your 9.62 m/s² against the handbook's 9.81 is a 1.9% gap, and the Discussion is where you explain it. Anomalies get faced, not buried: 'Trial 4 ran 9% high; the stopcock leaked visibly, and keeping it in the mean shifts the result by 2%.' That sentence shows judgment, which is the thing actually being graded.

Close with limitations framed as design facts rather than confessions — sample size, instrument resolution, one uncontrolled variable — and end with one concrete next experiment, such as repeating the run with a sealed system and N = 10. Before submitting, run the draft through EssayAtlas's free Readability Checker to catch the 40-word sentences that creep into Discussion sections, and hold each section to its word budget with the Word Counter on this page. Both tools are free, and both catch what a tired brain at 11 p.m. will not.