HSC Science Extension common mistakes
HSC Science Extension · Understand the traps and how to avoid them.
Describing the scientific method as a single fixed step-by-step recipe (question, hypothesis, experiment, conclusion) and applying it to every scenario, rather than distinguishing hypothesis-driven, observational, and exploratory approaches.
Textbook summaries and junior-science habits compress a messy, varied process into one flowchart, and students rarely see counterexamples such as descriptive taxonomy, modelling, or serendipitous discovery.
What to do instead
Practise classifying historical and contemporary case studies by research type (manipulative experiment, natural/observational study, modelling, meta-analysis) and write one sentence per case explaining which features of the method differ and why the question demanded it.
Treating a hypothesis as a prediction or a guess, and writing hypotheses that are not testable or falsifiable, e.g. 'plants will grow better' with no measurable variable or direction.
Everyday usage equates hypothesis with guess, and students are not trained to check that a proposed explanation generates an observable outcome that could contradict it.
What to do instead
Apply a two-part test to every hypothesis written: name the independent and dependent variables with units, and state explicitly what observation would prove it wrong. Rewrite any hypothesis that cannot fail that test.
Confusing correlation with causation or ignoring confounding variables when interpreting a reported association, and asserting a causal mechanism the evidence cannot support.
Causal language is common in media coverage and students absorb it, while the logic of controlled comparison and the role of alternative explanations is abstract and easily skipped under time pressure.
What to do instead
For each association encountered, list at least two plausible confounding variables and one study design (randomisation, matching, natural experiment) that would strengthen causal inference, then write a conclusion using calibrated wording such as 'is associated with' rather than 'causes'.
Writing a research proposal whose aim is too broad or vague to be feasible, e.g. 'investigate the effects of climate change on biodiversity' with no defined population, system, timeframe, or measurable outcome.
Students choose topics they find interesting rather than scoping them to what can be measured with available time, equipment, and access, and they conflate ambition with quality.
What to do instead
Draft the aim, then force it through a feasibility check: state the exact sample or system, the measurable dependent variable with units, the number of repeated measurements, and the resources required. Narrow the aim until every element is specified in one sentence.
Presenting a proposal with no articulated rationale or literature context, so the reader cannot see what gap the study fills or why the question matters.
Students treat the introduction as background trivia about the topic rather than an argument that builds to the research question, and they have limited experience locating and synthesising sources.
What to do instead
Structure the rationale as a chain: what is known, what remains unresolved or contradictory, and therefore what this study will determine. Cite at least three sources in that chain and end the section with a sentence beginning 'This study will therefore...'.
Omitting or mis-specifying risk assessment and ethical considerations, including no mention of consent, animal ethics approval, data privacy, or management of hazardous materials.
Ethics is often memorised as a list of approvals rather than reasoned about as part of design, and students assume school-level projects are exempt from scrutiny.
What to do instead
Include an ethics subsection that names each stakeholder or subject group, the specific risk (physical, psychological, privacy, environmental), its likelihood and severity, and the mitigation. State which approvals are needed and who grants them.
Failing to distinguish primary from secondary data, or treating a secondary dataset as if the student designed its collection, without acknowledging limitations of provenance, sampling, or measurement.
Secondary sources are convenient and appear authoritative, and students do not routinely interrogate who collected the data, when, for what purpose, and with what method.
What to do instead
For every dataset used, write a provenance note covering collector, purpose, sampling method, measurement instruments, and known limitations. In the discussion, state explicitly how those limitations constrain the conclusions that can be drawn.
Describing results with no measure of variability or uncertainty, reporting only means and treating small differences between groups as real.
Summary statistics are easier to compute and report than spread or inferential tests, and students lack confidence judging whether a difference exceeds natural variation.
What to do instead
Report an appropriate measure of spread (standard deviation, range, interquartile range) alongside every central value, state sample size, and include error bars. Before claiming a difference, compare it with the variability and, where appropriate, report a confidence interval or test statistic.
Drawing conclusions that overreach the data, generalising beyond the sampled population or conditions, or stating that results 'prove' the hypothesis.
Students want a decisive answer and mirror the confident language of popular science, and they have not practised calibrating claims to the strength of evidence.
What to do instead
Write conclusions using a claim-strength ladder: state the finding for the sampled conditions, then the inference, then the limit of generalisation. Replace 'proves' with 'supports' or 'is consistent with', and add one sentence naming what further evidence would be needed.
Producing a research report with misaligned structure and referencing, such as results interpreted in the results section, methods written as a narrative diary, or citations absent from the reference list and vice versa.
Students draft linearly and paste sections together without a final structural pass, and citation management is left to the end when errors are hard to trace.
What to do instead
Build the report to a fixed scaffold (introduction, methods, results, discussion, conclusion, references) and use a checklist pass: results contain only observations and analyses, interpretation appears only in the discussion, and every in-text citation is matched one-to-one against the reference list before submission.