Judge whether ACT Science data actually supports a stated conclusion, tell a supported claim from an overreach, and name the new evidence that would confirm a result.
Accept a conclusion only if the printed data directly shows it.
True-in-real-life is not the same as supported by THIS experiment.
Tested variable? Within tested range? Matches the data direction?
A claim that fails any of the three is an overreach.
A result is consistent with a claim if it moves the way the claim predicts.
Match the direction of one result to the direction the hypothesis expects.
Variables moving together = associated, not proven cause and effect.
Claim causation only when every other factor was held constant.
Strengthen by closing an untested gap; weaken by exposing a lurking variable.
Match the new evidence to the exact hole in the current argument.
Once you can read a table and name a trend, the ACT asks the harder question: does the data support this conclusion? A "Drawing Conclusions" item hands you a claim — sometimes in the stem, sometimes as four competing claims in the choices — and asks which one the experiment actually backs up. The trap is that several choices will be true-sounding or even true in the real world, yet unsupported by this experiment. Your job is to accept only the claim the printed data can defend, and to reject every claim that goes beyond it.
A conclusion is supported when the data in the passage directly shows it. A conclusion is unsupported (an "overreach") when it stretches past what was measured — even if it sounds reasonable. Consider a study of how a coating affects how fast a metal bar rusts:
| Trial | Coating thickness (µm) | Rust after 30 days (mg) |
|---|---|---|
| 1 | 0 | 240 |
| 2 | 10 | 155 |
| 3 | 20 | 88 |
| 4 | 30 | 41 |
The data supports: "Thicker coatings were associated with less rust after 30 days." Every step confirms it. The data does not support: "A coating would prevent rust entirely" (never tested, and the numbers approach zero without reaching it), nor "The coating stops rust at every temperature" (temperature was never varied). Both extra claims add a variable or a value the experiment never touched.
Before accepting a conclusion, run it through three scope questions:
Only a claim that passes all three deserves your answer.
Many stems ask which result is consistent with a hypothesis, or which the data supports. Read this as: does the number move the way the claim predicts? If a hypothesis says "more sunlight yields taller plants," a trial where the sunniest pot grew tallest is consistent with it; a trial where the shadiest pot grew tallest contradicts it. You are matching the direction of a single result to the direction a claim predicts — nothing more.
Two variables can move together without one causing the other, and the ACT tests this restraint. If a table shows ice-cream sales and drowning rates both rising over the summer months, the data supports "both increased together," not "ice cream causes drowning." When the experiment did not isolate a cause — did not hold other factors constant while changing only one — the supported conclusion is the modest one: the variables are associated, and you cannot claim one drove the other.
A frequent stem asks what new data would make a conclusion more (or less) convincing. The logic is always the same: strengthen a conclusion by testing it where it has not yet been tested, or by ruling out an alternative explanation.
Match the proposed evidence to the gap in the current argument, and the right choice is the one that fills exactly that gap.
The recurring lesson: the ACT rewards discipline. The strongest-sounding conclusion is usually an overreach, and the answer is the careful statement the data can actually carry.
Table 1 shows how a plant food affected tomato yield.
| Trial | Plant food (g per week) | Tomatoes harvested |
|---|---|---|
| 1 | 0 | 8 |
| 2 | 5 | 14 |
| 3 | 10 | 21 |
| 4 | 15 | 26 |
Which conclusion is supported by Table 1?
Using the same tomato study (0–15 g of plant food, yields 8–26), a student concludes: "Adding 25 g of plant food per week would produce at least 40 tomatoes." Is this conclusion supported?
A scientist hypothesizes that a certain fish grows faster in warmer water. Table 1 reports the results.
| Tank | Water temp (°C) | Mass gain in 30 days (g) |
|---|---|---|
| 1 | 14 | 9 |
| 2 | 18 | 15 |
| 3 | 22 | 22 |
| 4 | 26 | 20 |
Which tanks give results consistent with the hypothesis?
Over one summer a town recorded, each week, the number of fans sold and the number of heat-related ER visits. Both rose steadily from May to August. A blog claims "selling fans causes heat illness." Does the data support that causal claim?
A study found that seeds soaked in warm water germinated faster than dry seeds, using one batch of 20 seeds each. A reviewer wants to strengthen the conclusion that soaking speeds germination. Which additional evidence would most strengthen it?
Table 1 shows that a new insulation reduced a house's heating energy use as its thickness increased. The builder concludes: "This insulation reduces heating energy regardless of climate."
| Thickness (cm) | Heating energy (kWh per month) |
|---|---|
| 5 | 620 |
| 10 | 505 |
| 15 | 430 |
Which new finding would most weaken the builder's conclusion?
Choosing a conclusion because it is true in the real world, even though the experiment never measured it.
Judge only against the printed data. If the table did not test the variable or the value, a real-world fact cannot rescue the claim.
Accepting a specific prediction far beyond the last tested row as if the trend is guaranteed to continue.
Interpolation between tested points is usually safe; a precise number well past the tested range is an overreach the ACT will not reward.
Reading two variables that rise together as proof that one causes the other.
Without a controlled trial isolating one factor, the supported claim is "associated." Look for a shared cause before asserting causation.
For "what would strengthen/weaken" items, picking evidence that repeats what is already known instead of closing the argument's gap.
Name the current weakness first (small sample, untested condition, lurking variable), then choose the evidence that targets exactly that hole.
Table 1 shows how the loudness of an alarm affected how quickly sleepers woke.
| Alarm volume (dB) | Time to wake (s) |
|---|---|
| 40 | 52 |
| 55 | 34 |
| 70 | 19 |
| 85 | 8 |
Which conclusion is best supported by Table 1?
Table 1 shows the distance a toy car rolled after being released from ramps of different heights.
| Ramp height (cm) | Roll distance (m) |
|---|---|
| 10 | 1.4 |
| 20 | 2.9 |
| 30 | 4.3 |
A student concludes that releasing the car from a 60 cm ramp would make it roll exactly 8.6 m. How should this conclusion be judged?
A scientist hypothesizes that a fungus grows best at moderate humidity, not at very low or very high humidity. Table 1 reports colony size after one week.
| Humidity (%) | Colony diameter (mm) |
|---|---|
| 30 | 6 |
| 50 | 14 |
| 70 | 21 |
| 90 | 12 |
Which statement about the data and the hypothesis is correct?
A city report notes that neighborhoods with more coffee shops also have higher average incomes, based on a one-time survey. A columnist writes, "Opening coffee shops raises a neighborhood's income." Which judgment of this claim is most accurate?
A single trial found that plants given classical music grew 2 cm taller than silent-room plants over one week. A researcher wants to strengthen the conclusion that the music caused the extra growth. Which additional step would most strengthen it?
Table 1 shows that a diet additive lowered the resting heart rate of test mice as the dose increased. The lab concludes: "This additive lowers resting heart rate in all mammals." Which new finding would most weaken that conclusion?
Identify independent, dependent, and controlled variables in an ACT Research Summary, spot the control group, and explain why an experimental step was included — all from the passage.
Test a model or prediction against ACT Science data: decide whether results support it, which data would strengthen or weaken it, and how to refine a model that partly fails.