Part 3 — GED ® Science · Unit 9 — Putting the Reasoning Together
Unit 9 — Putting the Reasoning Together
Inside-to-GED Bridge — Evidence Before Rumor
On a unit, two people can watch the same event and tell different stories because they noticed different details, stood in different places, or interpreted motives differently. Scientific reports can also differ because methods, samples, conditions, and measurements differ.
That does not mean every claim is equally good. Compare what was measured, how groups were formed, whether variables were controlled, and whether the conclusion is broader than the data. Confidence is not evidence, and a rumor does not become a data set because twenty people repeat it.
The final Science questions mix unfamiliar content with familiar reasoning. Read the question first, find the relevant evidence, check units, calculate only what is needed, and make the answer earn its place.
Chapter 35 — Two Investigations, Several Defensible Conclusions
GED® test targets: SP.2.a, SP.2.c, SP.4.a, SP.5.a (Science Practices chapter)
In a fictional maintenance discussion, one report says a new filter improved airflow and another says it did not. Before choosing a side, check which rooms, which starting conditions, which filter sizes, and which measurements were compared. Disagreement is not automatically fraud or failure. Sometimes the reports answer different questions while everybody argues as if they asked the same damn thing.
Now you put the habits together: read completed investigations, compare designs, translate data, and decide how far a conclusion can go. Do not worry about which chapter a word came from. Ask what the evidence wants you to do now.
Investigation one: cooling systems under different conditions
Researchers compare two coverings, A and B, for reducing heat transfer from model shelters. They use shelters of the same size and material, equal starting temperatures, and equal amounts of internal material. Each condition is repeated with five independently prepared shelters. The table reports mean temperature decrease over 20 minutes.
| Surrounding condition | Covering A | Covering B |
|---|---|---|
| Still air | 8°C decrease | 12°C decrease |
| Moving air | 14°C decrease | 15°C decrease |
The question concerns reduction of cooling. Smaller decreases represent less cooling under the stated conditions. In still air, A’s mean decrease is 4°C smaller than B’s. In moving air, the difference is only 1°C. A performs better on this measured outcome in both conditions, but the apparent advantage is larger in still air.
The two independent variables are covering type and surrounding air condition. The dependent variable is temperature decrease over 20 minutes. Shelter size, starting temperature, and internal material are controlled variables. This is a two-factor design: it compares combinations rather than changing only one factor across the entire study.
Question 1: evaluate the claim
A researcher claims, “A prevents all heat loss.” Which evidence most directly challenges the claim? A. A had an 8°C decrease in still air. B. The study used five shelters per condition. C. B also cooled. D. Both coverings have letter names.
Answer: A. The measured decrease shows that A did not prevent all cooling. The result can still support reduced cooling compared with B. B describes replication, which strengthens interpretation but does not directly refute the word all. C concerns the other covering. D is irrelevant.
This is a common reasoning move: preserve the supported part while rejecting the exaggeration. You do not have to declare the whole study useless because somebody wrote an overconfident sentence about it.
Question 2: reconcile two reports
Report One says, “A has a substantial advantage over B.” Report Two says, “A and B are much closer in performance.” Can both fit the data? Yes, if One refers to still air and Two to moving air. Their conditions differ. The table does not provide variability measures, so words such as substantial should be used cautiously; we cannot establish statistical significance from means alone.
A better combined statement is: A showed a smaller mean temperature decrease in both tested air conditions, with a larger numerical difference in still air. That sentence carries the actual evidence instead of a vague victory banner.
Question 3: choose a visual representation
A grouped bar graph with air condition on the horizontal axis and mean temperature decrease in °C on the vertical axis would compare the two coverings within each condition. Each group should contain an A bar and a B bar, with a legend. Heights should be 8 and 12 for still air, 14 and 15 for moving air.
A line graph of temperature over time would require time-series values the table does not supply. You cannot invent intermediate temperatures simply because a line graph looks more scientific. Choose a display that matches the available information.

Means from the table above (constructed teaching data). Smaller decreases indicate less cooling. The means alone do not show variability or statistical significance.
Question 4: improve the investigation
The means do not reveal how variable the five shelters were. Reporting individual values or a measure of spread would help. Testing additional temperatures and exposure times would help evaluate broader claims. It would not be appropriate to erase the least favorable trial unless there was documented evidence of a specific error.
Suppose one thermometer consistently read 2°C high at both the start and end of a trial. The measured change might remain correct because the offset cancels, while the reported absolute temperatures would be wrong. By contrast, a changing calibration error could distort the temperature difference. “Instrument error” is not enough; identify how the error affects the quantity used.
Investigation two: plants, light, and temperature
A second team investigates plant growth. They use one variety, randomly assign equal-sized seedlings, and provide equal soil, water, and nutrient amounts. Ten plants receive each combination. The outcome is mean dry-mass gain after 21 days, a measure of accumulated material rather than just height.
| Temperature | Lower light | Higher light |
|---|---|---|
| 20°C | 2.0 g | 4.0 g |
| 30°C | 1.5 g | 3.0 g |
Higher light is associated with greater mass gain at both temperatures. At 20°C, the gain doubles from 2.0 to 4.0 g. At 30°C, it doubles from 1.5 to 3.0 g. The warmer condition produces a smaller mean gain at each light level in this experiment.
Question 5: connect the science to the evidence
Which conclusion is supported? A. Every plant on Earth grows best at 20°C. B. More light increased mean gain within both tested temperature conditions. C. Temperature has no relationship to growth. D. Plants do not respire in light.
Answer: B. Both row comparisons support it. A generalizes from one variety and two temperatures to all plants. C ignores the column comparisons. D contradicts the biological principle that plants respire in light as well as darkness and is not measured by the table.
The mechanism may involve photosynthesis, respiration, and other temperature-sensitive processes. However, the study measured final dry-mass gain, not each separate process. A plausible mechanism is not automatically a directly measured result.
Question 6: calculate the relative change
At lower light, increasing temperature from 20°C to 30°C changes mean gain from 2.0 to 1.5 g. The change is −0.5 g. Relative to 2.0 g, that is −0.5/2.0 × 100 = −25%. At higher light, the decrease from 4.0 to 3.0 g is also 25%, though the absolute decrease is 1.0 g.
Equal percentage changes can correspond to different absolute amounts. The reference value explains why. If an answer says the changes are “the same,” ask whether it means grams or percent. Ambiguous language is where otherwise correct comparisons go to get themselves into trouble.
Question 7: identify an unsupported extrapolation
The data do not show what happens at 40°C or at a light level ten times higher. Predicting that growth always doubles when light doubles would require an actual quantitative light scale and evidence across that range. Here “lower” and “higher” are categories; the table never says the light intensity doubled.
An appropriate next investigation could test additional light levels while controlling temperature and measuring growth consistently. The reader’s task is to recognize that design, not perform it. All the reasoning stays on the page.
Question 8: distinguish two kinds of replication
Measuring the same plant’s mass five times can assess consistency of the measurement process. Growing five independent plants under the same condition provides information about biological variation. The two forms of repetition are useful for different reasons. Five measurements of one plant do not become five independent plants because the spreadsheet has five rows.
Carry the method forward
Across both investigations, the pattern is the same: identify the outcome, inspect the comparison, calculate only what the data support, and qualify the conclusion. When two findings appear inconsistent, check conditions before deciding one must be wrong. When an answer contains always, never, proves, or all, inspect whether the evidence really earns that word.
Science reasoning is not a contest to sound the most certain. Be as certain as the evidence allows, and no more.
Two Reports Can Differ Without One Being Bullshit
Separate the effect estimate from the certainty attached to it. An estimate describes the observed amount of change; confidence depends on sample size, variability, controls, measurement quality, and replication. A larger reported effect from a weak design is not automatically stronger evidence than a smaller effect from a credible comparison. Size and evidential strength are different properties.
Suppose report A finds airflow increases from 40 to 60 units, while B reports 80 to 100. Both increases are 20 units, but their relative increases differ: 20/40 = 50% and 20/80 = 25%. The reports can disagree in percentage without disagreeing in absolute change. State the baseline before comparing effects, or a denominator will quietly hijack the argument.
Evidence can be reconciled by identifying a moderator, a condition that changes an effect’s size or direction. A material may work better in dry air than humid air, or an intervention may matter more when starting performance is poor. That possibility should generate a testable explanation, not an excuse invented to protect a favorite result from every challenge.
Independent replication repeats the investigation with new subjects or materials, not merely repeated readings of the same sample. Repeated readings estimate measurement consistency; independent samples help describe real variation. Counting one sample ten times as ten independent subjects inflates the apparent evidence. The spreadsheet may look crowded, but biology does not give a damn how many rows you copied.
When asked which conclusion both studies support, look for the overlap in their actual findings and conditions. When asked what would resolve disagreement, choose a comparison that isolates the differing condition. Do not average incompatible outcomes or declare a universal rule. A careful synthesis says what is shared, what differs, and what remains uncertain.
Your Turn — Guided and Independent Practice — Practice Check S35
Guided: Study A finds a fertilizer increases mean growth by 3 cm with 10 plants. Study B finds no difference with 200 plants of another species in cooler conditions. Give two reasons the results need not be dishonest contradictions.
Guided answer: Species, temperature, dose, soil, duration, and sample size differ. The fertilizer may work only under some conditions, or Study A’s small sample may be unstable. Compare methods before forcing one winner.
Independent: Which conclusion is strongest when two well-designed studies under similar conditions produce different estimates? A. Say science failed. B. Choose the preferred result. C. Average everything without checking methods. D. Report uncertainty and seek replication.
Independent answer: D. Variation and uncertainty are normal; replication and method comparison clarify the pattern. B is bias, C can mix incompatible studies, and A misunderstands scientific correction. Common mistake: demanding perfect agreement before accepting any evidence.