Basic Principles and Processes · 4 min read
Scientific Inquiry and Experimental Design
Types of Investigations
Scientific inquiry includes several kinds of investigations. A descriptive study observes and records (mapping bird species in a park). A comparative or correlational study looks for relationships between variables that are not manipulated (comparing asthma rates in cities with different air quality). A controlled experiment deliberately changes one variable and measures the effect while holding others constant; only a controlled experiment can establish cause and effect directly.
In practice the cycle loops: unexpected results lead to new questions and revised hypotheses, and a step may be repeated many times.
Questions and Hypotheses
A good scientific question is specific and answerable with measurable evidence ("How does water temperature affect the dissolving time of a sugar cube?" rather than "Why is sugar good?"). A hypothesis is a testable, falsifiable proposed explanation; it is often written as a prediction in if–then form: "If water temperature increases, then the sugar cube will dissolve faster, because particles move faster at higher temperature."
In statistical testing, the null hypothesis states there is no effect or difference ("temperature has no effect on dissolving time"); the alternative hypothesis states there is an effect. Data lead scientists to reject or fail to reject the null hypothesis.
Variables and Controls
| Component | Definition | Sugar-cube example |
|---|---|---|
| Independent variable | Factor the investigator changes | Water temperature |
| Dependent variable | Factor measured in response | Time to dissolve (s) |
| Controlled variables | Factors kept constant | Water volume, cube size, stirring |
| Control group | Baseline for comparison | Room-temperature water |
| Experimental groups | Receive the changed condition | Water at 40 °C and 60 °C |
A positive control is a group expected to show the effect (it confirms the procedure can detect a result); a negative control is expected to show no effect (it reveals false positives or contamination). A confounding variable is an uncontrolled factor that varies with the independent variable and could explain the results.
Worked Example: Identifying Variables
A student tests whether the mass of a pendulum bob affects its period. She uses bobs of 50 g, 100 g, and 200 g, keeps the string length at 0.75 m and the release angle at 10°, and times 10 swings for each bob, three trials apiece.
- Independent variable: mass of the bob (50 g, 100 g, 200 g).
- Dependent variable: period (time for one complete swing). Timing 10 swings and dividing by 10 reduces reaction-time error; for example, 17.4 s ÷ 10 = 1.74 s.
- Controlled variables: string length, release angle, location, timing method.
- Expected result: for small angles the period depends on length and gravitational acceleration, not on mass, so the periods should be nearly equal (about 1.74 s for a 0.75 m pendulum).
Reliability, Validity, and Bias
Designing a Fair Test
- Correlation studies cannot prove causation; controlled experiments can test it.
- Field studies trade some control for realism; lab studies trade realism for control.
- A written procedure detailed enough for replication is part of good design.
- Results that do not support the hypothesis are still valid, useful data.
Finished reading?
Lock it in with a quick Basic Principles and Processes quiz.