Experimental Design and the Scientific Method

From question to testable prediction

Science starts with an answerable question. A hypothesis is a testable proposed explanation or prediction. It should connect a change to a measurable result.

Example: “If light time increases, then bean plants will grow taller over two weeks.” This can be tested. “Plants like light” is too vague because “like” is not clearly measurable.

A hypothesis must be able to be supported or rejected with evidence. It is not a final proven fact. A prediction states the expected result if the hypothesis is correct.

Identify what changes and what responds

The independent variable is the factor the researcher changes. The dependent variable is the response measured. Controlled variables are kept the same.

In the plant example, light time is independent and plant height is dependent. Plant type, soil, water, pot size, and study length should stay the same. Changing several factors at once makes the cause unclear.

Operational definitions state exactly how variables are changed or measured. “Growth” might mean change in height in centimeters over 14 days. This makes the test repeatable and prevents people from measuring different outcomes.

Samples, replication, and bias

A control group gives a comparison without the tested treatment. Random assignment helps spread hidden differences among groups. A larger, representative sample usually gives stronger evidence than a small biased sample.

Repeated trials show whether a result is consistent. Replication means other researchers can repeat the study. Blinding can reduce the effect of expectations on measurement.

Worked example. Divide similar plants randomly into no-fertilizer and fertilizer groups. Give all plants the same light, soil, water, and time. Measure height the same way. Several plants per group and repeated trials reduce the influence of chance.

Random sampling concerns who enters the sample and helps generalization; random assignment concerns which treatment sampled subjects receive and helps causal comparisons. Neither guarantees perfection. Repeated readings of one plant are not the same as independent replication with several plants.

Trust the right conclusion

Reliability means getting consistent results. Validity means the design truly tests the intended question. A tool can give very consistent but wrong readings, so reliability alone does not prove accuracy or validity.

A confounding variable changes with the independent variable and could explain the result. Good controls reduce confounding and support a cause-and-effect conclusion.

If a study changes fertilizer and water together, it cannot tell which change caused growth.

Summary

• A useful hypothesis is testable.
• The independent variable is changed.
• The dependent variable is measured.
• Controls make comparison possible.
• Replication improves reliability.
• Validity asks whether the design tests the intended claim.

Change one main factor, measure the response, and keep other conditions the same.

Sources

• OpenStax Biology 2e: The Science of Biology
• NIST Metric (SI) Prefixes