ACT Evaluation of Models, Inferences, and Experimental Results
Last updated: May 2, 2026
Evaluation of Models, Inferences, and Experimental Results questions are one of the highest-leverage areas to study for the ACT. This guide breaks down the rule, the elements you need to recognize, the named traps that catch most students, and a memory aid that scales to test day. Read it once, then practice the same sub-topic adaptively in the app.
The rule
On ACT Science evaluation questions, your job is to judge whether a claim, hypothesis, or model is supported, contradicted, or untestable given the data shown. Test every statement against the actual numbers, trends, and experimental conditions on the page — not against your outside knowledge. The right answer always rides on a specific feature of the data: a direction of change, a magnitude, a comparison between two groups, or a condition that was (or wasn't) tested.
Elements breakdown
Identify the claim
Pin down exactly what the model, hypothesis, or scientist is asserting.
- Restate the claim in your own words
- Note which variable predicts which outcome
- Note the predicted direction of the effect
- Note any conditions or boundaries on the claim
Locate the relevant data
Find the specific table column, figure axis, or experiment that speaks to the claim.
- Match claim variables to table or figure labels
- Read units carefully before comparing values
- Check which experiment manipulated the variable
- Ignore data that addresses a different variable
Test direction and magnitude
Compare what the claim predicts to what the data actually shows.
- Check whether the trend goes up, down, or flat
- Verify the size of change matches the prediction
- Watch for trend reversals at extreme values
- Confirm the effect appears in every relevant group
Check the experimental design
Decide whether the study can answer the question being asked.
- Identify independent and dependent variables
- Identify what was held constant (controlled)
- Note the range of conditions actually tested
- Flag missing controls or untested conditions
Decide: supported, weakened, or untestable
Pick the verdict that matches what the data can actually demonstrate.
- Supported: data shows predicted pattern across cases
- Weakened: data shows opposite or absent pattern
- Untestable: variable was never manipulated or measured
- Beware of partial support across only some conditions
Common patterns and traps
The Untested-Variable Distractor
A wrong answer asserts that a hypothesis is supported (or weakened) by data, but the variable the hypothesis depends on was never actually changed in the experiment. Students fall for this when they recognize the topic and assume any data on the topic counts as evidence. The fix is to check what was manipulated versus what was held constant.
An answer that says a claim about light intensity is supported by an experiment in which all trials used the same lamp and only temperature varied.
The Out-of-Range Extrapolation
A wrong answer applies a trend observed within the tested range to a value beyond the data — usually a much higher or lower setting that was never measured. The ACT rewards students who treat the tested range as a fence: claims about points outside the fence are unsupported, even if 'extending the line' seems natural.
An answer claiming that a reaction rate at 100 °C will be double the rate at 50 °C, when the experiment only tested 10 °C through 40 °C.
The Partial-Pattern Overreach
A wrong answer treats a pattern that holds in some conditions as if it holds in all. Often the trend reverses or flattens in one condition, and the wrong answer simply ignores that condition. Always scan every row, every group, every experiment before declaring a pattern universal.
An answer asserting 'X always increases Y' when one of three test groups shows X having no effect on Y, or even decreasing Y.
The Reversed-Causation Trap
A wrong answer correctly notes a correlation but flips which variable causes which, or claims causation when only correlation was measured. Evaluation questions reward students who notice that observing two variables move together does not establish that one drives the other, especially without a controlled manipulation.
An answer claiming that changes in soil pH cause changes in plant height, when the experiment only measured both variables across natural sites without manipulating pH.
The Plausible-Sounding Outside Fact
A wrong answer states something that is true in real-world science but is not demonstrated anywhere in the passage. The ACT specifically tests whether you can stay inside the four corners of the experiment. If the data don't show it, the data don't support it — even if you happen to know it from biology class.
An answer about photosynthesis requiring chlorophyll appearing in a question about a study that measured only oxygen output, without ever discussing pigments.
How it works
Imagine a scientist claims that a fertilizer increases tomato yield only above 20 °C. To evaluate this, you'd look for a table that varied both fertilizer presence and temperature. If the data show fertilized plants outperforming controls at 25 °C and 30 °C, but not at 15 °C, the claim is supported. If fertilized plants outperformed controls at every temperature equally, the claim is weakened — the temperature condition didn't matter. And if temperature was held constant at 22 °C across all trials, the claim is untestable from this experiment, no matter how compelling the yield numbers look. The ACT loves this last category: an answer choice that sounds right but rests on a variable the experiment never explored. Always ask, 'Did they actually vary the thing the claim depends on?' before you mark anything supported.
Worked examples
Researcher Marta Reyes studied how a freshwater algae species, Chlorenia minuta, grows under different light and nutrient conditions. She prepared identical 500 mL flasks of pond water and varied two factors: light intensity (low: 50 µmol/m²/s; high: 200 µmol/m²/s) and nitrogen concentration (low N: 0.5 mg/L; high N: 5.0 mg/L). All flasks were held at 22 °C and stirred identically. After 7 days, Reyes measured algae cell density (cells/mL). Reyes hypothesized that nitrogen availability is the limiting factor for C. minuta growth in this pond, and that light intensity has little effect once nitrogen is sufficient.
Do the results in Table 1 support Reyes's hypothesis that light intensity has little effect once nitrogen is sufficient?
- A Yes; at low nitrogen, cell density was nearly identical at both light levels.
- B Yes; cell density increased from low N to high N at both light levels, showing nitrogen is the only factor that matters.
- C No; at high nitrogen, cell density at high light (9.6 × 10⁵) was twice the cell density at low light (4.8 × 10⁵), indicating light still has a substantial effect. ✓ Correct
- D No; cell density at high light + low N (1.4 × 10⁵) was lower than at low light + high N (4.8 × 10⁵), proving light has no effect.
Why C is correct: Reyes's hypothesis predicts that once nitrogen is sufficient (high N), changing light should not much affect growth. But Table 1 shows that at high N, switching from low light to high light doubles cell density (4.8 → 9.6 × 10⁵). That is a large effect at the very condition where the hypothesis predicts a small effect, so the data weaken the hypothesis.
Why each wrong choice fails:
- A: This observation is true (1.2 vs 1.4 at low N) but doesn't address the hypothesis, which is specifically about what happens when nitrogen is sufficient (high N), not low. (The Partial-Pattern Overreach)
- B: Nitrogen does increase cell density, but the claim that it's 'the only factor that matters' contradicts the doubling at high N when light is increased — light clearly also matters. (The Partial-Pattern Overreach)
- D: The comparison is between two different nitrogen levels, so it doesn't isolate the effect of light. To test light's effect, you must hold nitrogen constant, which the correct answer does. (The Untested-Variable Distractor)
Two students, Fei Liu and Jordan Park, debate why a metal rod expands when heated. Liu's Model: Heat increases the average vibration amplitude of atoms in the rod. Larger amplitudes push neighboring atoms farther apart on average, lengthening the rod. Liu predicts that any metal will expand by a fixed fraction per degree Celsius regardless of which metal it is, because all metals are made of vibrating atoms. Park's Model: Expansion depends on the strength of the bonds between atoms. Metals with weaker bonds expand more per degree because their atoms can move farther apart for the same vibration energy. Park predicts that softer metals (like lead) expand more per degree than harder metals (like tungsten).
Do the results in Table 1 support Liu's model, Park's model, both, or neither?
- A Liu's model only, because every metal expanded with heat.
- B Park's model only, because the expansion coefficient increased as the metals became softer. ✓ Correct
- C Both models, because the data show expansion that varies among metals.
- D Neither model, because the data do not include bond strengths.
Why B is correct: Park's model predicts that softer metals will expand more per degree than harder ones. The table lists metals in order of increasing softness, and the expansion coefficients rise from 4.5 (tungsten, hardest) to 28.9 (lead, softest), matching Park's prediction. Liu's model predicts a fixed fraction for all metals, but the values differ by more than a factor of six, contradicting Liu.
Why each wrong choice fails:
- A: Liu predicts the same expansion coefficient for every metal, not just any expansion. Wildly different coefficients (4.5 vs 28.9) directly contradict Liu's model. (The Partial-Pattern Overreach)
- C: The data contradict Liu specifically because his model predicts uniform expansion. Variation among metals is exactly what his model rules out. (The Plausible-Sounding Outside Fact)
- D: Bond strength itself was not measured, but Park's model uses softness as a proxy, which the table does provide. The data are sufficient to test Park's prediction even without direct bond-strength values. (The Untested-Variable Distractor)
Ecologist Henrik Sato investigated whether a parasitic wasp, Aprostocetus pallidus, reduces populations of an invasive beetle, Galerita oryzae, in rice fields. Sato selected six rice paddies of similar size, soil, and rice variety in a single region. In three randomly chosen paddies (treatment), he released 500 wasps at the start of the growing season. In the other three (control), he released no wasps. He counted beetle larvae per square meter at three time points: Day 0, Day 30, and Day 60. Sato hypothesized that wasps would reduce beetle larvae density by Day 60.
Which of the following statements is best supported by the data in Table 1?
- A The wasps eliminated all beetle larvae from the treatment paddies by Day 60.
- B Releasing wasps was associated with a lower beetle larvae density at Day 60 compared to control paddies. ✓ Correct
- C The wasps would continue to reduce beetle larvae for at least another 60 days after the experiment ended.
- D Beetle larvae density depends on rice variety more than on wasp presence.
Why B is correct: At Day 60, control paddies had a mean of 25 larvae/m² while treatment paddies had 9 larvae/m² — a clear difference between the two groups. Because rice variety, soil, and size were held constant, the difference is associated with wasp release. The statement claims an association at Day 60, which is exactly what the table shows.
Why each wrong choice fails:
- A: Treatment paddies still had 9 larvae/m² at Day 60, not zero. 'Eliminated all' overstates what the data show. (The Partial-Pattern Overreach)
- C: The experiment ended at Day 60. Any claim about what happens beyond Day 60 extends the trend past the tested range without evidence. (The Out-of-Range Extrapolation)
- D: Rice variety was held constant across all six paddies, so the experiment cannot compare its effect to anything. The claim addresses a variable that wasn't manipulated. (The Untested-Variable Distractor)
Memory aid
C-D-D: Claim, Data, Decide. Restate the Claim, find the Data that varies the right variable, then Decide supported / weakened / untestable.
Key distinction
A claim being plausible is not the same as a claim being supported. Support requires that the experiment actually varied the variable in question and produced the predicted pattern — anything else is, at best, consistent with the claim, not evidence for it.
Summary
Match the claim to the variable that was manipulated, check direction and magnitude in the data, and reject any answer that relies on conditions the experiment never tested.
Practice evaluation of models, inferences, and experimental results adaptively
Reading the rule is the start. Working ACT-format questions on this sub-topic with adaptive selection, watching your mastery score climb in real time, and seeing the items you missed return on a spaced-repetition schedule — that's where score lift actually happens. Free for seven days. No credit card required.
Start your free 7-day trialFrequently asked questions
What is evaluation of models, inferences, and experimental results on the ACT?
On ACT Science evaluation questions, your job is to judge whether a claim, hypothesis, or model is supported, contradicted, or untestable given the data shown. Test every statement against the actual numbers, trends, and experimental conditions on the page — not against your outside knowledge. The right answer always rides on a specific feature of the data: a direction of change, a magnitude, a comparison between two groups, or a condition that was (or wasn't) tested.
How do I practice evaluation of models, inferences, and experimental results questions?
The fastest way to improve on evaluation of models, inferences, and experimental results is targeted, adaptive practice — working questions that focus on your specific weak spots within this sub-topic, getting immediate feedback, and revisiting items you missed on a spaced-repetition schedule. Neureto's adaptive engine does this automatically across the ACT; start a free 7-day trial to see your sub-topic mastery climb in real time.
What's the most important distinction to remember for evaluation of models, inferences, and experimental results?
A claim being plausible is not the same as a claim being supported. Support requires that the experiment actually varied the variable in question and produced the predicted pattern — anything else is, at best, consistent with the claim, not evidence for it.
Is there a memory aid for evaluation of models, inferences, and experimental results questions?
C-D-D: Claim, Data, Decide. Restate the Claim, find the Data that varies the right variable, then Decide supported / weakened / untestable.
What is "The outside-knowledge trap" in evaluation of models, inferences, and experimental results questions?
picking an answer that is true in real life but unsupported by the passage data.
What is "The trend-flip trap" in evaluation of models, inferences, and experimental results questions?
assuming a trend continues past the highest tested value when data near the top already bends.
Ready to drill these patterns?
Take a free ACT assessment — about 15 minutes and Neureto will route more evaluation of models, inferences, and experimental results questions your way until your sub-topic mastery score reflects real improvement, not luck. Free for seven days. No credit card required.
Start your free 7-day trial