Wait, What? The most important thing you write in a laboratory may be the thing that went wrong.

A smudged endpoint, a bubble trapped in a burette tip, a thermometer that had not equilibrated, a sample that was visibly different from the others — these details can matter more than a beautifully copied method. A laboratory record is not a polished story written after the experiment. It is a contemporaneous record of what was actually done, seen, measured and changed.

That habit matters because practical science is not only about producing a result. It is about preserving enough evidence for someone — including your future self — to judge what the result deserves to mean.

Observation is not inference

Observation: “The blue solution became colourless after 18.6 cm³ was added.” Inference: “The reacting species had been consumed at the endpoint.”

Observation: “A white solid formed.” Inference: “An insoluble product was produced.”

The inference may be scientifically reasonable, but it is still an interpretation. Strong laboratory records make this boundary visible. That protects students from one of the most common reasoning mistakes in Science: writing a model as though it were directly seen.

What should a useful laboratory record contain?

• The question or purpose of the investigation.
• The variables and important conditions.
• What was actually done, including meaningful deviations from the intended method.
• Raw measurements with units and appropriate precision.
• Qualitative observations that may help explain the result.
• Calculations, graphs or transformations derived from the raw data.
• Problems, anomalies and procedural events noticed during the work.
• A conclusion supported by the evidence.
• An evaluation tied to specific mechanisms of error and realistic improvements.

The exact format varies by subject and laboratory, but the principle is stable: preserve the evidence trail.

Do not rewrite history

Suppose the planned method says “heat for five minutes,” but the burner went out after three minutes and the sample cooled before reheating. A weak record silently writes the intended method. A useful record notes what actually occurred.

This is not about confessing mistakes for punishment. It is about scientific traceability. If the result later appears strange, the recorded event may explain why. Science becomes less trustworthy when the final account is cleaner than reality.

Evaluation: stop writing “human error”

“Human error” is usually too vague to be useful. Which action? By what mechanism? In what direction might it change the result? How would the proposed improvement reduce it?

Weak: “Human error when timing.” Better: “Reaction time introduces variability when starting and stopping the stopwatch. Timing a longer interval or using an automated sensor would reduce the fractional contribution of the start-stop delay.”

Weak: “Heat was lost.” Better: “Energy transferred from the reaction mixture to the cup and surrounding air, so the measured temperature rise was smaller than it would be under better thermal isolation. Using a lid and better insulation would reduce this transfer.”

The Royal Society of Chemistry’s evaluation guidance similarly pushes students beyond generic labels toward control variables, significant sources of error, anomalies and realistic improvements. Read the RSC evaluating-experiments resource.

An improvement must attack the error mechanism

If parallax is the problem, “repeat more times” does not necessarily solve it. The improvement is to change the viewing geometry or use an instrument that avoids that reading problem. If heat loss is the problem, “use a more precise thermometer” does not stop energy leaving the system. If the sample is unrepresentative, a finer balance does not repair the sampling design.

This gives you a powerful evaluation test: Does my proposed improvement actually interrupt the mechanism that produced the limitation?

Errors can affect results differently

Some limitations increase scatter. Some shift readings systematically. Some reduce the measured effect. Some inflate it. Some make the interpretation ambiguous without producing a predictable direction.

A high-standard evaluation tries to predict the effect where scientifically justified. If a calorimetry experiment loses heat to the surroundings, the measured temperature change will usually be smaller in magnitude than the idealised change associated with the process. If a burette contains an air bubble in its tip initially and the bubble fills during delivery, the burette reading may indicate more liquid delivered than actually reached the flask.

Negative and null results belong in science

If the expected colour change does not occur, or if two conditions show no clear difference, do not manufacture a trend. A null result can mean the hypothesis is unsupported under those conditions; the effect may be smaller than the experiment can resolve; the chosen range may be poor; or the model may need revision.

The record should allow these alternatives to be considered. “Nothing happened” is often scientifically richer when expanded into exactly what was observed and what the method was capable of detecting.

Evidence boundaries

Suppose an experiment shows that seedlings exposed to one light condition grew taller over seven days than seedlings in another condition. The evidence supports a statement about those seedlings, conditions, duration and measurements. It does not automatically establish a universal law for all plant species, all light spectra, all temperatures and all stages of growth.

Good conclusions respect the boundary between what this experiment showed and what the wider scientific literature establishes.

Secondary → JC progression

Secondary: record raw observations clearly; distinguish qualitative and quantitative data; identify specific errors; state realistic improvements; connect conclusion to results.

JC: separate observation, transformation and inference; discuss assumptions and model limitations; distinguish random and systematic contributions; evaluate whether an improvement changes uncertainty materially; recognise sampling and calibration issues; state the domain over which a conclusion is supported.

Checkpoint: evaluate the evaluation

A student measures the enthalpy change of combustion of a fuel by heating 100 g of water with a small spirit burner. Her evaluation says:

“There was human error. I should repeat the experiment more times and use a thermometer with more decimal places.”

• Why is “human error” weak?
• Name a likely major physical limitation.
• Predict how that limitation affects the measured temperature rise.
• Suggest an improvement that targets it directly.
• Would extra thermometer digits necessarily solve the dominant problem?

Answer key and WHY reasoning

“Human error” does not identify a mechanism. A likely major limitation is energy transfer to the surrounding air and apparatus instead of only to the water. This generally reduces the water’s measured temperature rise relative to an idealised system. Better insulation, a lid and a geometry that improves energy transfer to the water can reduce the loss. A thermometer with finer displayed resolution may improve temperature reading resolution, but it does not prevent heat from escaping and therefore may leave the dominant limitation almost unchanged.

A practical-study method students can actually use

After every school practical, take five minutes and write a “second-pass record” with four questions: What did I actually observe? What did I infer? What was the largest limitation and why? What single change would most improve the evidential quality?

Over time, compare these notes across experiments. You will start seeing recurring families of problems: endpoint judgement, thermal loss, reaction time, parallax, uncontrolled variables, insufficient range, poor sampling, instrument resolution and invalid assumptions. Practical science then becomes learnable as a system rather than a collection of unrelated experiments.

Singapore practical-assessment connection

The Singapore-Cambridge O-Level Chemistry practical framework assesses planning; manipulation, measurement and observation; presentation of data and observations; and analysis, conclusions and evaluation. It explicitly expects candidates to identify significant sources of error, explain how they affect results and explain how they can be reduced. Read the current SEAB syllabus document.

Authoritative next steps

• SEAB 2026 O-Level syllabus directory
• SEAB 2026 A-Level syllabus directory
• Royal Society of Chemistry: evaluating experiments
• NIST measurement uncertainty guidance

Teaching Guide

For parents and teachers: when marking practical reflection, reject vague labels gently but consistently. Ask students to complete this causal chain: specific event or limitation → physical/chemical/biological mechanism → expected effect on data → targeted improvement. That one structure raises the quality of evaluation dramatically and transfers across Physics, Chemistry and Biology.