Part 3 — GED ® Science · Unit 1 — Read the Evidence
Chapter 2 — Fair Tests and Honest Conclusions
GED® test targets: SP.2.a, SP.2.b, SP.2.c, SP.2.d, SP.2.e, SP.3.d (Science Practices chapter)
Imagine a fictional laundry detail reporting that a new drying schedule worked faster. Then you learn the second load also contained fewer towels and used a different machine. Well, shit: three things changed. A fair experiment exists to untangle that mess. It does not require fancy equipment in your hands; this chapter asks you to reason about a complete investigation already described on paper.
A study can collect a mountain of numbers and still ask its question badly. If one group gets a different temperature, different light, and different water, you cannot confidently credit the outcome to just one of those changes. More numbers will not magically clean up a confused comparison.
An experiment deliberately changes a condition and measures a response. An observational study records conditions without assigning the change being studied. Observational evidence matters enormously in astronomy, ecology, and health research. Its limitations differ from those of a controlled experiment; it is not automatically worthless.
Name the variables by their jobs
A variable is a characteristic that can differ. The independent variable is the condition changed or compared. The dependent variable is the response measured. Controlled variables are conditions intentionally kept alike so they do not muddy the comparison. A control group provides a baseline; it is not the same thing as a controlled variable.
Completed study: researchers assigned similar bean seedlings randomly to two light conditions. Both groups received the same soil, water volume, container size, and temperature. After 14 days, they measured the increase in height. Light condition was the independent variable. Height increase was the dependent variable. Soil, water, container size, and temperature were controlled variables.
| Condition | Number of seedlings | Mean height increase in 14 days |
|---|---|---|
| 6 hours of light daily | 20 | 4.2 cm |
| 10 hours of light daily | 20 | 6.8 cm |
The data support greater average height increase under the longer-light condition in this study. They do not establish that height always increases as light increases, or that the taller seedlings were healthier in every respect. Height is what the researchers measured. Do not quietly replace it with a different outcome.
A hypothesis needs a measurable prediction
“Plants like light” is vague. “If daily light exposure increases from six to ten hours while other conditions remain similar, average seedling height increase over 14 days will be greater” identifies the comparison, response, and time period. It could turn out to be wrong. That is a strength: evidence can actually challenge it.
A sound design states the question, identifies variables, chooses comparable groups and measurements, collects data consistently, and evaluates the result. Real science may revisit earlier steps rather than march through one sacred sequence. But measuring an outcome before establishing the comparison, then inventing a convenient prediction afterward, is a different and weaker process.
Replication, randomization, and sampling
Replication means repeating observations or trials to see whether a result holds up. Twenty seedlings are more informative than one, especially when living things vary. Repeating a measurement on the same seedling checks measurement consistency; it does not create twenty independent seedlings. Counting repeats as new subjects is a sneaky way to make evidence look stronger than it is.
Random assignment distributes subjects among experimental groups by chance. It helps prevent preexisting differences from consistently favoring one group. Random sampling selects subjects from a larger population by chance. It helps make a sample representative. Assignment and sampling solve different problems. A well-randomized experiment involving one plant variety may still not represent every plant variety.
To estimate vegetation across a field, sampling only the wet corner is biased. Taking randomly located equal-area samples across the field is better. A larger biased sample can remain biased. Interviewing a thousand people from one narrow group does not make them represent everyone.
Error is not always carelessness
Random error produces scatter: small unpredictable differences among measurements. More independent measurements can help estimate an average. Systematic error shifts measurements in a consistent direction, such as a balance reading two grams too high. Repeating that reading a hundred times does not remove the two-gram problem. Calibration, checking an instrument against a known reference, addresses it more directly.
A confounding variable changes along with the factor being studied and offers an alternative explanation. If the long-light plants also receive more fertilizer, fertilizer is a confound. Repair the comparison by keeping fertilizer alike or designing separate groups that let its effect be evaluated. Do not simply announce that light won because that was the original plan.
Worked design judgment
Question: A researcher tests whether container color affects warming. A black metal container and a white plastic container receive equal lamp exposure. Which change best improves the design? A. Change both exposure time and container size. B. Use containers made of the same material and size, differing only in color. C. Measure only the black container. D. Stop recording the starting temperature.
Answer: B. Material is a competing explanation, so matching it isolates color more effectively. A introduces additional differences. C removes the comparison. D hides whether the containers started equally warm. Also repeat the comparison and measure temperature changes rather than relying only on final values.
Measurement tools must fit the quantity: a balance measures mass, a thermometer temperature, a graduated cylinder liquid volume, and a timer elapsed time. You need to recognize those roles in a written investigation; nothing in this book requires you to obtain or use the equipment.
What carries forward: ask what changed, what was measured, what stayed alike, how subjects were selected, and what else could explain the result. A fair test does not need to be fancy. It needs to make the comparison make sense.
Make the Comparison Do the Work
Start with the causal claim: airflow affects drying. The independent variable is the assigned airflow condition. Define the dependent variable precisely, such as grams of water lost in 20 minutes, rather than the vague phrase “better drying.” Keep cloth type, starting water, exposed area, and temperature comparable. Those controlled variables block alternative explanations. They are not the control group, which is the baseline condition used for comparison.
Suppose three separate cloth samples under ordinary airflow lose 8, 10, and 12 grams. Their mean loss is (8 + 10 + 12) ÷ 3 = 10 g. Three samples under increased airflow lose 12, 14, and 16 grams; their mean is 14 g. The mean difference is 4 g. This models the calculation; it is not an instruction to alter laundry procedures. The overlapping value of 12 g also shows why a group tendency is not a guarantee for every sample.
Random assignment helps distribute preexisting differences without someone deliberately putting the driest samples in the favored group. Replication shows how much outcomes vary. Neither one is magic. Randomly assigning only two samples can still produce an unlucky imbalance, and repeating a biased procedure a hundred times just makes a very organized pile of bullshit. Consistent measurement and a credible comparison still matter.
Reliability asks whether a method produces reasonably consistent results. Validity asks whether the investigation actually measures or tests what it claims. A scale that reads every sample 5 g too heavy can be highly consistent but inaccurate. A study of towel color cannot establish the effect of airflow if airflow was never varied. These are different weaknesses, so the repair has to match the problem.
Transfer that reasoning to unfamiliar seeds, metals, or solar panels: locate the manipulated factor, the measured outcome, the comparison, and the remaining confounds. When a design-improvement answer changes everything at once, reject it. More activity is not better science. The best repair makes the competing explanations harder to sustain.
Sources of Error and Refining a Hypothesis
A source of error is anything in the design or the measuring that pushes results away from the true value. The GED® test usually shows it to you as a discrepant result: one trial or one instrument that does not match the rest. Your job is to name the likely source and change the design to reduce it (SP.2.a).
Worked example (SP.2.a). A fictional study times how long ice cubes take to melt in four identical cups. Three cups take 31, 32, and 30 minutes. The fourth takes 22 minutes, and the notes say that cup sat in direct sunlight by a window. That odd result points to a source of error in the design, an uncontrolled condition: the cups did not share the same surrounding temperature. Better design: put every cup in the same spot, record the air temperature, and repeat with several cups. Do not just delete the weird trial and act like it never happened. Explain it, fix the design, and run it again.
A hypothesis is a testable, possible explanation. Refining a hypothesis means making it more precise once the data show where it holds and where it breaks (SP.2.b).
Worked example (SP.2.b). Hypothesis: “More fertilizer makes bean plants grow taller.” Fictional results: 0 g → 10 cm, 2 g → 14 cm, 4 g → 17 cm, 8 g → 12 cm. Height rose through 4 g and fell at 8 g, so the original hypothesis is only partly right. Refined hypothesis: “Fertilizer increases bean plant height up to a moderate dose (about 4 g in this setup), and larger doses reduce growth.” The refined version fits all four results and can be tested again with doses between 4 g and 8 g.
Practice A (SP.2.a): Five people measure the same table with the same meter stick and get 152, 151, 152, 160, and 151 cm. The person who got 160 cm lined up the stick’s 8-cm mark, not its zero end, with the table edge. Name the source of error and the fix.
Practice B (SP.2.b): Hypothesis: “This plant grows faster in warmer air.” Fictional data: 15°C → 2 cm per week, 25°C → 4 cm per week, 35°C → 1 cm per week. Write a refined hypothesis.
Answers: A. The 160-cm reading is discrepant because the measurement started at the 8-cm mark, which adds 8 cm (152 + 8 = 160). Fix: line up the zero end with the edge, repeat each measurement, and report the mean. B. “This plant’s growth increases with temperature up to about 25°C, and higher temperatures reduce growth.” A refined hypothesis keeps what the data support and adds the limit the data revealed.
Your Turn — Guided and Independent Practice — Practice Check S2
Guided: A fictional laundry study compares drying time for equal cloth samples. One group gets moving air; the other does not. Both use the same cloth, starting water mass, temperature, and observation period. Identify the independent variable, dependent variable, and two controlled variables.
Guided answer: Moving air is the independent variable. Drying time or water mass lost is the dependent variable, depending on what is measured. Cloth type, starting water mass, temperature, and sample size are controlled variables. The comparison works because one main condition changes.
Inside guided: Two cups of the same drink mix are left out to compare how fast ice melts. One cup sits in a sunny window spot and the other on a shaded shelf, and the cups are different sizes. Name the design problem and fix it.
Answer: Two things changed at once (sunlight and cup size), so either could explain a difference. Use identical cups, change only the location, and repeat with several cups in each spot. Change two things at once and you’ve got no idea which one did the damn work.
Independent: A researcher tests sanitizer concentration but also changes contact time for every concentration. What is the best criticism? A. The dependent variable should be changed on purpose. B. Concentration cannot be measured. C. Repetition is always harmful. D. Contact time is a confounding variable.
Independent answer: D. Concentration and contact time change together, so either could explain the result. B is false, C misunderstands replication, and A confuses the measured response with the manipulated factor. Common mistake: spotting the variable you care about while ignoring the extra variable sneaking around beside it.