Skip to notes
K223 / K225 / 2027
Data and Experimental Skills

Lesson 3 of 7

Evaluate evidence

Explain limitations and target improvements.

SEC G2 Science K223 / K225 (2027)

  • I can evaluate methods and suggest improvements.Syllabus

    Syllabus K223 / K225, C4. Evaluate methods and suggest improvements

Evaluate evidence and explain the direction of an error

An improvement is useful when it targets a specific limitation.

Different limitations need different responses
IssueEffect on evidenceUseful response
Random variationRepeated results scatter unpredictablyRepeat independent measurements; compare their spread and use a justified mean
Systematic biasResults are displaced consistently, for example by an incorrect calibrationCheck against a suitable standard or correct the apparatus/method; repetition alone does not remove the bias
Uncontrolled variableTwo possible causes change togetherHold that variable constant or design an additional comparison
Limited range or sampleA relationship may be supported only locallyCollect appropriate additional values; restrict the conclusion to the evidence

Accuracy concerns closeness to a valid reference value; precision concerns the closeness of repeated measurements. A tightly grouped set can still be inaccurate. An anomalous point is a reason to investigate the method and repeat a measurement, not permission to remove inconvenient evidence silently. Record an exclusion and its reason.

Write evaluation as a causal chain: limitation -> affected measurement -> effect on the calculated result -> targeted improvement. "Human error" and "use better equipment" do not identify what happened. For example, gas escaping before collection lowers measured gas volume; assembling a closed collection system before starting the reaction addresses that loss.

Timing by hand is one such limitation. You start and stop a stopwatch yourself, so your reaction time adds an error to every timing. When a reaction is very fast, as at a high temperature, that error is a large part of the time measured.

A result must be repeatable before anyone trusts it. In March 1989, Martin Fleischmann and Stanley Pons of the University of Utah announced "cold fusion" in a test tube. Physicists in several countries tried to repeat their results and failed.

Check your understanding

A thermometer reads 2 degrees too high at every temperature. Will averaging five readings remove that error?

  1. No. Averaging works with an even number of readings, so take six readings instead.
  2. No. Averaging reduces random scatter, but a fixed offset stays in every reading.
  3. Yes. Five readings are enough for their errors to cancel each other out.
Show the answer

No. Averaging reduces random scatter, but a fixed offset stays in every reading.

Every reading is 2 degrees high, so their average is too. Calibrate the thermometer against a reference. A temperature change measured with the same thermometer can cancel a constant offset.

Check your understanding

A pipette should deliver 25.00 cm3. Five measured deliveries are 24.60, 24.62, 24.61, 24.60 and 24.62 cm3. How would you describe these readings?

  1. Both accurate and precise, because the five readings agree closely
  2. Precise but not accurate: they agree closely but are all too low
  3. Accurate but not precise: their average is close enough to 25.00 cm3
Show the answer

Precise but not accurate: they agree closely but are all too low

Precision is how close repeated readings are to each other; these agree within 0.02 cm3. Accuracy is closeness to the true value, and all are about 0.4 cm3 low.

Watch out for this

A student says: "The reaction cannot be studied at 90 degrees C, because it would be over in less than two seconds." What is wrong with this?

Show the answer

That gives the time but not the problem it causes. Explain the effect on the measurement. The time is too short to measure accurately with a stopwatch. Your reaction time in starting and stopping it is a large part of so short a time. Examiners did not credit answers that only said the reaction would take less than one or two seconds.