Hypotheses, Investigations and Evaluating Evidence
A newspaper headline says a new tablet "cures colds" because 90% of people who took it got better within a week. Should you believe it? Most colds clear up in about a week anyway. In this page you will learn how biologists plan investigations, measure carefully and judge evidence, so that you can answer this question yourself.
Hypotheses and testable predictions
A hypothesis is a proposed explanation for an observation, which can be tested by investigation. It is not just a guess: it is based on what is already known.
A hypothesis is testable if you can collect data that would either support it or show it to be wrong. This means it must refer to things you can measure or observe, and it must be possible for the result to go against it.
-
Testable: "Increasing the temperature from 10 °C to 40 °C increases the rate of catalase activity."
-
Not testable: "Enzymes like warm places." ("Like" cannot be measured.)
From a hypothesis you make a prediction: what you expect to see if the hypothesis is correct. Evidence can support a hypothesis, but a hypothesis is never "proven" for certain, because a later investigation might give a different result.
Define the term hypothesis* and state what makes a hypothesis testable*.
Define the term hypothesis and state what makes a hypothesis testable.
Variables and fair tests
An investigation changes one factor and watches its effect. Every factor that can change is a variable.
| Variable | Meaning | Example: temperature and enzyme rate |
|---|---|---|
| Independent | The one you deliberately change | Temperature of the water bath |
| Dependent | What you measure as the result | Time taken for the substrate to be broken down, or volume of product per minute |
| Controlled | Everything else, kept the same | pH, enzyme concentration, substrate volume, reaction time |
A test is fair only if just the independent variable changes. If two things change at once, you cannot tell which caused the effect.
A control (control experiment or control group) is a set-up where the independent variable is absent or at its normal value. It gives a baseline for comparison. For example, boiled enzyme in place of active enzyme shows that the effect is due to the enzyme and not something else.
Distinguish between the independent variable*, the dependent variable* and a controlled variable* in an experiment, using the example of an investigation of the effect…
Distinguish between the independent variable, the dependent variable and a controlled variable in an experiment, using the example of an investigation of the effect of temperature on the rate of an enzyme reaction.
A student wants to test whether light intensity affects the rate of photosynthesis in pondweed. a) Write a testable hypothesis. b) State the independent and dependent…
A student wants to test whether light intensity affects the rate of photosynthesis in pondweed.
- a) Write a testable hypothesis.
- b) State the independent and dependent variables and how each would be measured.
- c) State three variables that should be controlled.
Data, accuracy, precision and error
Good data must be both close to the truth and consistent.
-
Accuracy: how close a measurement is to the true value.
-
Precision: how close repeated measurements are to each other.
-
Reliability: results that can be repeated and give similar values. Repeating an experiment and finding the mean reduces the effect of chance variation, so it improves reliability.
Mean and spread
The mean is the sum of the readings divided by the number of readings. A large spread in repeat readings means low precision.
Types of error
-
Random error: unpredictable differences between readings, making them scatter either side of the true value (for example, misjudging when a bubble has formed). Repeating and averaging reduces it.
-
Systematic error: the same fault pushes every reading the same way (for example, a thermometer that always reads 2 °C too high, or heat from a lamp warming the water). Repeating does not remove it, so it reduces accuracy. You must fix the method or calibrate the equipment.
An anomaly is a result that does not fit the pattern. Check whether there was a mistake, repeat it, and say so in your report. Do not simply delete it.
Explain the difference between accuracy* and precision* in measurements.
Explain the difference between accuracy and precision in measurements.
A student measured the volume of oxygen released by pondweed in 5 minutes, repeating the experiment five times: 4.2 mL, 4.4 mL, 4.3 mL, 4.1 mL and 4.5 mL. a) Calculate…
A student measured the volume of oxygen released by pondweed in 5 minutes, repeating the experiment five times: 4.2 mL, 4.4 mL, 4.3 mL, 4.1 mL and 4.5 mL.
- a) Calculate the mean volume.
- b) Explain why repeating the experiment improves reliability.
- c) Suggest one source of error in measuring the gas volume and how it could be reduced.
The results of an investigation of enzyme activity at different temperatures show a bell-shaped curve with an optimum at 37 °C, but there is a large spread in the…
The results of an investigation of enzyme activity at different temperatures show a bell-shaped curve with an optimum at 37 °C, but there is a large spread in the replicate readings at 60 °C.
- a) Suggest two reasons for the larger spread at 60 °C.
- b) Distinguish between a random error and a systematic error, with one example of each for this investigation.
- c) Evaluate whether a study published in a magazine with no information about the method or sample size should be given the same weight as a peer-reviewed article. Justify your answer.
Sample size and statistical analysis
Living things vary, so one plant or one patient proves little. A larger sample size and replicates reduce the effect of chance and make a conclusion more reliable. Samples should be chosen randomly and be representative of the population.
To decide whether a difference between two groups is real or just chance, scientists use a statistical test. For two means a t-test is common. It gives a p-value, the probability of getting a difference at least this large by chance if there is really no difference.
-
If $p < 0.05$, the difference is statistically significant: less than a 5% probability it is due to chance.
-
If $p \ge 0.05$, the data do not show a significant difference.
Significance says the difference is unlikely to be chance. It does not say it is large or important. Correlation does not prove causation either: two things changing together may both be caused by a third factor.
A researcher claims that a new fertiliser increases the yield of potatoes. She grows 10 plants with fertiliser in a sunny field and 10 without fertiliser in a shady…
A researcher claims that a new fertiliser increases the yield of potatoes. She grows 10 plants with fertiliser in a sunny field and 10 without fertiliser in a shady field, then reports a 30% greater yield for the fertilised plants.
- a) Identify two flaws in the design of this investigation.
- b) Redesign the investigation so that the conclusion would be valid, including sample size, controls and replication.
- c) Explain what statistical analysis might be used to decide whether the difference is significant, and what the result would mean.
Models, sources and bias
Models
A model is a simplified representation of something (a diagram, equation, physical object or computer simulation) used to explain it, predict what will happen, or study something too small, large or complex to see directly. Models are never perfect, but they can be modified when new evidence arrives.
Evaluating sources
Not all claims are equally reliable. Ask:
-
Who wrote it, and is it peer-reviewed (checked by independent experts before publication)?
-
Is the method, sample size and raw data given, so others could repeat it?
-
Was there a control group?
-
Who paid for it? Is there a conflict of interest?
-
Is it supported by other independent studies?
Bias is a systematic leaning towards a particular result, which makes evidence unfair. It can come from the people running the study (expecting a result), the participants (knowing they got the "real" treatment), the choice of sample, or the funder. It is reduced by random selection, a control group, and blind trials (participants do not know which treatment they get) or double-blind trials (neither participants nor researchers know).
A newspaper headline states: "New tablet cures colds: 90% of people who took it recovered within a week." a) Identify two pieces of information missing that you would…
A newspaper headline states: "New tablet cures colds: 90% of people who took it recovered within a week."
- a) Identify two pieces of information missing that you would need to evaluate this claim.
- b) Explain why a control group is necessary.
- c) State what is meant by bias and give one example of how it might affect this study.
Scientists use models, such as the fluid mosaic model of the membrane or the double helix model of DNA. a) Explain what is meant by a scientific model and why models are…
Scientists use models, such as the fluid mosaic model of the membrane or the double helix model of DNA.
- a) Explain what is meant by a scientific model and why models are used.
- b) Using one example, explain how a model can be modified as new evidence emerges.
- c) Discuss one limitation of models in representing biological systems.
Back to the start: the tablet claim cannot be trusted until we know the sample, the control group and the source. A good investigation compares like with like.
Summary
-
A hypothesis is a testable explanation; it must refer to measurable variables and be capable of being shown wrong.
-
A fair test changes only the independent variable, measures the dependent variable and keeps other variables controlled. A control gives a baseline.
-
Accuracy is closeness to the true value; precision is closeness of repeat readings. Repeating improves reliability. Random errors scatter results; systematic errors shift them all the same way.
-
Use a large random sample, replicates and a statistical test; $p < 0.05$ means the difference is statistically significant.
-
Models are simplified and are modified as new evidence emerges.
-
Judge sources by peer review, method, sample size, controls, funding and bias.