The conclusion and the evaluation
What a conclusion must do
- The standard says the conclusion must relate to the aim. So it answers the aim, in the aim's own words, and it is the shortest section in the report.
- Put the aim's own wording in the first sentence. The aim was to find out how and why the channel of the Waipuna Stream changes between the foothills and the coast. It changes in four measured ways, and the main cause is increasing discharge.
- Answer every sub-question briefly, in the order you asked them. If one could not be answered, say so and say why — an honest gap beats a vague sentence pretending to fill it.
- State how confident you are. Strongly supported by four variables moving together and suggested by forty questionnaire responses are different levels of confidence, and saying so is judgement, not weakness.
- Keep it to one or two paragraphs. A conclusion is a verdict, not a summary of the report.
What a conclusion must not do
- Do not introduce new data. Anything that matters belongs in the findings.
- Do not just repeat the findings. A conclusion that lists all six results again has not concluded anything.
- Do not answer a different question from the aim. This happens when the interesting data turns out to be somewhere other than where the aim pointed. If your report ended up answering a better question, say that in the evaluation — do not quietly swap it into the conclusion.
- Do not overclaim. One catchment on one day says nothing about rivers in general, and a marker reads an overclaim as a failure of judgement.
Evaluating the process, stage by stage
- The standard asks for strengths and weaknesses of the research process, and how they affect the validity of the findings. Two halves, and the second is the one that is skipped.
- Evaluate the process, not the outcome. We got a clear trend is a result. The 2 km spacing made distance usable as an axis is a process strength.
- Work through the seven stages and give each a sentence or two: the aim, the plan, the collection, the recording, the presentation, the findings, and the analysis itself.
- Be specific and use your field log. The weather was bad is worthless; rain began between sites 4 and 5, so the lower two sites were measured at a higher flow than the upper four is evidence.
- Strengths must be argued too. Three velocity timings per site let me tell a real difference between sites from stopwatch error, because my repeats spread about 9 per cent while sites 1 and 6 differ by over 100 per cent.
- Quantify a weakness where you can. Forty responses out of an invented population of 900 is about 4 per cent, and twelve refusals in one street means one distance band is under-represented.
Linking a weakness to validity
- Validity is whether your data really measures what your aim asks about. A weakness only matters as far as it changes what your findings can claim.
- Use a three-part sentence: name the weakness, state the effect on the data, then state the effect on the finding.
- Because two tributaries join between sites 2 and 4, distance downstream and tributary input cannot be separated, so I can claim that the channel widens downstream but not that distance alone causes it.
- Some weaknesses do not threaten your finding, and saying so is a strength. A float measures surface velocity, so all my velocity figures are too high — but the error is in the same direction at every site, so the trend between sites is unaffected.
- Sort your weaknesses by how much they matter. Three that change what you can claim, discussed properly, beat eight listed.
- Do not confuse a weakness with a mistake. Sample size is a limit of a school field day; writing 1,400 cm in a column of metres is an error. Both belong in the evaluation, labelled correctly.
The Excellence sentence: how the process could be improved
- Excellence adds one requirement: discussing ways the research process could be improved. It is the most commonly missing sentence in the standard, and it is the cheapest grade in Level 2 Geography.
- An improvement must be specific, achievable and tied to a weakness you already named. Collect more data is not an improvement. Add a site immediately above and immediately below each tributary junction, so tributary input and distance downstream can be separated is.
- Rank them. Best repair a confound, because those limit what you may claim; next repair coverage, such as a second field day at a different flow; last repair method, such as a flow meter instead of a float.
- Say what each improvement would let you claim that you currently cannot. That sentence is the difference between listing improvements and discussing them.
Worked Example
Worked example
A student's evaluation of the invented Waipuna Stream study reads:
Strengths: we worked well as a team and collected lots of data. Weaknesses: not enough time, small sample size, human error. Next time we would do more.
Rewrite it so it would reach Excellence.
Answer:
Step 1 — replace the strengths with process strengths that are argued.
Worked well as a team is not a research strength. Replaced with: Systematic 2 km spacing across all six sites made distance a usable x-axis, so the trends I report belong to the stream rather than to where we chose to stop. And: three velocity timings per site gave a spread of about 9 per cent, so a difference of over 100 per cent between sites 1 and 6 is far outside my own measurement error and can be reported confidently.
Step 2 — turn each vague weakness into a named one with an effect on the data.
Small sample size becomes: forty questionnaire responses out of roughly 900 residents is about 4 per cent, and twelve refusals fell in one street, so the distance band nearest the channel is under-represented.
Human error becomes: the tape blew sideways at site 4 and the first width reading was repeated, and a float measures surface velocity, so every velocity figure is above the true channel mean.
Step 3 — add the second half the original omitted entirely: the effect on validity.
The float error is systematic and in the same direction at every site, so the downstream trend is unaffected and only the absolute values are too high. By contrast, the two tributaries joining between sites 2 and 4 mean distance downstream and tributary input cannot be separated, so I can claim the channel widens downstream but not that distance alone drives it. And the questionnaire's uneven response rate means the perception gradient I found may be understated near the channel.
Step 4 — replace we would do more with specific improvements tied to those weaknesses.
- Add a site immediately above and immediately below each tributary junction. This converts the confound into a measurable step change, so I could then claim how much of the widening is tributary input.
- Repeat the transect on a second day at higher flow. My one day was low flow, so this would show whether the pattern holds under the conditions that actually shape the channel.
- Use a flow meter rather than a float, which removes the surface-velocity bias and lets me report absolute velocities, not only the trend.
- Pilot the questionnaire and re-visit the refusing street at a different time of day, which would even out the response rate across the distance bands.