Wissenschaftliche Arbeitsweisen / Scientific practices
This domain is assessed inside every other topic: it is the marks awarded for designing a controlled experiment, handling data honestly, and evaluating a model. Answers are judged on whether variables, controls and conclusions are stated precisely in English.
Experimental design
- State the independent variable, the dependent variable (with unit and how it is measured) and at least two controlled variables.
- Include a control group, random allocation or random sampling, replicates, and a sample size large enough for the conclusion drawn.
- Distinguish accuracy (closeness to the true value) from precision (reproducibility), and reliability (repeatability of the whole method) from validity (does the design measure what it claims?).
Data analysis
- Report the mean and a measure of spread (standard deviation, range, error bars). Overlapping error bars or a stated statistical test result (e.g. p < 0.05) is the standard for claiming a difference.
- Calculate rates as change in the dependent variable divided by time; read values off axes with the correct unit and sensible precision.
- Correlation is not causation: identify confounders and look for a plausible mechanism or a controlled intervention.
Models and evaluation
- Models (membrane diagrams, energy pyramids, cladograms) simplify; evaluate by naming what is omitted and what the model therefore cannot predict.
- Evaluate method limits: sampling bias, measurement error, uncontrolled variables, and whether the range of the independent variable covers the biologically relevant span.
Worked example
Two fertiliser treatments are compared: mean shoot length 12.4 cm (SD 1.8) versus 14.1 cm (SD 2.1), n = 10 each, error bars shown as ±1 SD and clearly overlapping. Reasoning: the difference in means is 1.7 cm, but the spread is larger than that difference and the error bars overlap, so the data do not support a claim of a real effect; increasing n or reducing variability would be needed. If the same data came from an uncontrolled field trial, confounding by light or soil would also weaken any causal claim.
Traps that cost marks
Writing 'proves' instead of 'supports'; describing a graph ('it goes up') without quoting values and identifying the trend; giving a conclusion without the mechanism; omitting units; proposing a method with no control group; answering 'human error' instead of a specific source of error.