Exploration: Entering the World of Secondary Science · Lesson 6 of 8
Prediction Evidence and Scientific Testing
“Predictions face evidence like contestants at an audition: convincing ideas stay, and weak ones go back for revision.”
• Distinguish a reasoned scientific prediction from a vague guess. • Turn an informal claim into a measurable, testable statement. • Identify changed, measured and controlled variables in a simple investigation. • Evaluate a claim by asking about mechanism, comparison and repeatable evidence. • Explain how mismatched predictions lead to better models or measurements.
Dark Clouds at Lunchtime
Varsha looks outside and says, ‘It will rain this afternoon because the clouds are dark.’ Meghna could simply agree, disagree or wait. Instead, she asks: how humid is the air, is pressure falling, what direction is the wind moving, and what happened on earlier days with similar conditions? The dark clouds begin a useful observation, but a scientific prediction needs measurable evidence and a clear time or outcome.
A good prediction is not guaranteed to be correct. Its strength is that we can tell what result would support it and what result would challenge it. A statement designed so that every possible outcome counts as success is not a strong scientific test.
A specific expected outcome derived from observations, evidence, a model or an explanation, stated so that later observations can support or challenge it.
A tentative, testable explanation for an observation or pattern. A hypothesis should lead to predictions that can be checked with evidence.
| Statement | Problem | Improved scientific form |
|---|---|---|
| This fertiliser is best. | ‘Best’ is undefined; no comparison or measure. | Bean plants receiving 2 g of fertiliser weekly will gain more mean height in 21 days than identical plants receiving none. |
| The lake is polluted. | No indicator or threshold is named. | Water downstream will contain less dissolved oxygen than water upstream when measured at the same time. |
| The phone battery is bad. | No use condition or outcome is specified. | With screen brightness fixed at 50%, this phone will lose more charge in one hour of video playback than the comparison phone. |
| Dark clouds mean rain. | Vague condition and no time window. | If humidity exceeds the chosen threshold and pressure continues falling, rain will begin before 5 pm. |
Problem
A packet claims that fertiliser X makes bean plants grow faster. Design a fair, school-level test.
- 1.Define the claim. Let ‘grow faster’ mean greater increase in plant height over 21 days.
- 2.Form a prediction: plants given the stated amount of fertiliser X will show a greater mean height increase than plants given no fertiliser.
- 3.Choose groups with several similar seedlings, not just one plant in each group, because individual plants vary.
- 4.Change the fertiliser condition. This is the changed or independent variable.
- 5.Measure height increase at fixed intervals. This is the responding or dependent variable.
- 6.Keep other major conditions similar: species, starting size, pot size, soil, water, light and measurement time.
- 7.Record all results, calculate the group mean and compare—not just the tallest plant.
- 8.If the fertilised group does not grow more, check dose, measurement and assumptions. Do not rewrite the prediction after seeing the result.
Changed variable: the main condition deliberately altered. Measured variable: the response recorded as evidence. Controlled variables: other important conditions kept as similar as practical so the comparison is fair.
Problem
A weather model predicts rain five days ahead, but the observed weather is dry. Why does one mismatch not prove that weather science is useless?
- 1.Recognise that weather depends on many interacting quantities, including temperature, pressure, humidity and wind.
- 2.Measurements have limited resolution and are taken at particular places and times.
- 3.Small differences in starting conditions can grow as the model calculates farther into the future.
- 4.Compare forecast confidence: a near-term forecast may be more reliable than a detailed forecast far ahead.
- 5.After a mismatch, scientists compare predicted and observed data to improve measurements, model equations and assumptions.
- 6.Evaluate many forecasts over time rather than judging the entire method from one event.
Checking Viral Claims
Suppose a post claims that food becomes harmful during an eclipse. Repeating the claim does not make it evidence, and mocking the claim does not test it. Scientific thinking asks what ‘harmful’ means, what physical, chemical or biological mechanism is proposed, what measurable change should occur, and what fair comparison would isolate the effect of an eclipse from ordinary time, temperature and storage conditions.
Problem
Outline how the claim could be investigated without assuming it true or false in advance.
- 1.Define a measurable outcome, such as microbial count, pH, temperature or another agreed indicator of spoilage.
- 2.Prepare equal portions of the same food at the same time using clean containers.
- 3.Keep storage duration, temperature and handling as similar as possible. Include an appropriate comparison sample.
- 4.Record environmental measurements during the eclipse and during the comparison period rather than treating ‘eclipse’ as an unexplained force.
- 5.Measure outcomes using the same method without changing the standard after seeing results.
- 6.Repeat and look for a consistent difference larger than ordinary variation.
- 7.If no physical, chemical or biological mechanism or repeatable difference appears, the evidence does not support the claim. This conclusion remains open to better evidence.
Searching only for examples that agree with a belief is confirmation bias. Decide the prediction and measurement method first, record all outcomes, and actively ask what evidence would show the idea is mistaken.
Before testing, complete this sentence: ‘My claim would be challenged if…’ If no possible observation could challenge it, the claim has not yet been made scientifically testable.
No. It is scientific when it follows from a stated idea and can be checked. An incorrect prediction can reveal a weak assumption or incomplete model.
Quiz
Which prediction is most testable?
In a fertiliser experiment, what is the measured variable if growth means height gained?
What should scientists do when prediction and observation disagree?
Which description best matches Scientific prediction?
Which description best matches Hypothesis?
Practice Problems
- Rewrite ‘This phone has a good battery’ as a testable prediction.
- In the bean-fertiliser investigation, name the changed variable, measured variable and two controlled variables.
- Why is ‘The clouds look dark’ not enough evidence for a strong rain prediction?
- A study tests only one fertilised plant and one unfertilised plant. Identify two weaknesses.
- Use the five-question claim checker to evaluate one viral health or household claim without deciding the conclusion in advance.
Key Takeaways
• A scientific prediction states a checkable expected outcome. • A hypothesis is a testable explanation that generates predictions. • A fair comparison changes the relevant condition while controlling major alternatives. • Variables must be identified before collecting data. • Claims should be evaluated through mechanism, measurement, comparison and repetition. • Mismatches drive revision of models, assumptions or measurements. • Evidence must be recorded honestly, including results that challenge the preferred idea.
Estimation and the Connected World of Science Evidence is not always exact. Next, we learn how rough but reasoned estimates check scale—and why real problems require several branches of science together.