Scientific Reasoning

Scientific Reasoning: From Question to Defensible Conclusion

Scientific reasoning connects a focused question to evidence that can answer it. A strong investigation changes one factor deliberately, measures a relevant outcome, controls competing explanations, and reports what the data support. The goal is not to prove a favored idea. It is to design a fair test and recognize the limits of its conclusion.

Variables Define the Test

The independent variable is the factor deliberately changed. The dependent variable is the measured outcome. Controlled variables are kept consistent. A control group provides a useful baseline. A confounding variable changes along with the independent variable and offers another explanation for the outcome.

Worked Example 1: Identify the Variables

A researcher gives identical seedlings 0, 2, or 4 grams of fertilizer each week and measures height gain after six weeks.

Step 1: Ask what the researcher assigns. Fertilizer amount is the independent variable.

Step 2: Ask what is recorded as the response. Height gain is the dependent variable.

Step 3: Keep plant species, initial size, soil, pot size, water, light, and time as similar as possible.

Step 4: The zero-fertilizer group serves as a comparison baseline.

Why it works: If the groups differ systematically only in fertilizer amount, a difference in average growth is less likely to have another simple cause.

Worked Example 2: Detect a Confound

The 4-gram plants are placed by a sunny window, while the other groups remain under weaker light. Can fertilizer alone explain greater growth?

Step 1: Identify the intended change: fertilizer amount.

Step 2: Notice that light exposure also changes with treatment.

Step 3: Because light can affect photosynthesis and growth, it can produce the same outcome being measured.

Conclusion: No. Light is a confounding variable, so greater growth cannot be attributed confidently to fertilizer alone.

Better design: Randomly place plants from every fertilizer group across equal-light positions or rotate positions regularly.

Controls, Samples, and Repetition

A well-chosen control group shows what happens without the tested treatment. Random assignment helps distribute preexisting differences across groups. Larger samples reduce the influence of unusual individuals, and repeated trials reveal whether a pattern is dependable. These features strengthen evidence, but they do not rescue a measurement that does not match the question.

Worked Example 3: Improve a Weak Study

One student drinks a caffeinated beverage and completes a puzzle faster than the previous day. What changes would strengthen the study?

Step 1: Recruit more participants because one person may be unusual.

Step 2: Compare caffeine with a similar caffeine-free condition so expectation and beverage routine are less likely explanations.

Step 3: Randomize condition order and use comparable puzzles to reduce practice and order effects.

Step 4: Repeat measurements and compare group results.

Why it works: The original observation mixes caffeine with day, practice, sleep, and individual variation. Controls and repetition separate some of those influences. Even then, the conclusion should concern the tested task and population, not every kind of performance.

Choose and Read Measurement Tools

Match the tool to the quantity: a balance measures mass, a graduated cylinder measures liquid volume, a ruler measures length, a thermometer measures temperature, and a stopwatch measures time. Record all certain digits plus one estimated digit when reading an analog scale. Units are part of a measurement, not decoration.

Worked Example 4: Read a Meniscus

Water in a graduated cylinder curves downward at the center. The bottom of the curve lies halfway between 36 mL and 37 mL. What should be recorded?

Step 1: Put your eye level with the liquid surface to avoid parallax error.

Step 2: For a concave water meniscus, read the bottom of the curve.

Step 3: The halfway position is 36.5 mL.

Why it works: A consistent reading rule gives different observers the same reference point. Looking from above makes the curve appear against a different scale position and creates systematic error.

Accuracy and Precision Answer Different Questions

Accuracy is closeness to an accepted or true value. Precision is closeness of repeated measurements to one another. Results can be precise without being accurate when an instrument has a consistent offset. Scattered values can average near the accepted value while individual measurements remain imprecise.

Worked Example 5: Diagnose Measurement Quality

A certified 100.0-gram mass is measured as 104.8 g, 104.9 g, and 104.8 g.

Step 1: Compare the readings with one another. Their range is only 0.1 g, so they are tightly grouped and precise.

Step 2: Compare them with 100.0 g. Every reading is nearly 5 g too high, so accuracy is poor.

Step 3: Suspect a calibration or zeroing problem because the error is consistent.

Why it works: Repeating the same biased result demonstrates repeatability, not truth. Calibration against a known standard tests accuracy.

Interpret Data Without Overclaiming

Quantitative data are numerical; qualitative data describe qualities. A graph can show direction, size, and consistency of a pattern, but a correlation alone does not establish causation. Check axes, units, sample size, variability, and whether groups were treated comparably. Conclusions should name the tested conditions and use cautious language when alternative explanations remain.

Worked Example 6: Compute and State a Result

A treatment group's mean rises from 20 units to 25 units. What is the percent increase?

Step 1: Find the change: 25 − 20 = 5 units.

Step 2: Divide by the original value: 5 ÷ 20 = 0.25.

Step 3: Multiply by 100: 0.25 × 100 = 25%.

Why it works: Percent change compares the change with the starting amount. Dividing by 25 would use the final value and answer a different question.

Common Mistakes

• Calling the measured outcome independent: Independent is assigned; dependent is observed.
• Listing a confound without explaining it: State how it could affect the dependent variable.
• Using the largest container available: Choose the tool designed for the quantity and needed resolution.
• Reading a meniscus from above: Read at eye level using the correct part of the curve.
• Equating precision with accuracy: Compare trials with each other and with a standard.
• Claiming causation from association: Look for controlled manipulation and competing explanations.

Knowledge Check

1. Variables

A scientist changes water temperature and records how long a tablet takes to dissolve. What is the dependent variable?

Answer: Dissolving time. Temperature is deliberately changed, while time is the measured response.

2. Confounding

One teaching method is used only in the morning and another only in the evening. Why is time of day a confound?

Answer: Alertness may vary by time of day and affect scores. Because time changes with method, either factor could explain a difference.

3. Measurement

Which tool is best for measuring 42 mL of water?

Answer: A suitably sized graduated cylinder. It measures liquid volume directly and more precisely than an unmarked container.

4. Accuracy and Precision

Measurements of 10.0, 14.0, and 18.0 g have a mean of 14.0 g, the accepted value. Are they precise?

Answer: No. The mean is accurate, but the individual values are widely scattered, so precision is poor.

5. Percent Change

A value falls from 80 to 60. What is the percent decrease?

Answer: The decrease is 20; 20 ÷ 80 × 100 = 25%. The original value belongs in the denominator.

Independent Practice

1. Design an experiment testing light duration and seedling growth. Label every variable and the control group.
2. Find two possible confounds in a study comparing exercise habits of volunteers with nonvolunteers.
3. Choose tools and units for mass, temperature, liquid volume, and elapsed time.
4. Explain how eye position changes a meniscus reading.
5. Create one accurate-but-imprecise set and one precise-but-inaccurate set around an accepted value of 50.
6. Calculate percent change from 40 to 52 and write a conclusion that does not overclaim causation.

Self-check: Identify what changed, what was measured, what else could explain the pattern, and whether the measurement method supports the claim. Then match the strength of your conclusion to the strength of the design.