ACT Science Reasoning: Evaluating Evidence, Models, and Conclusions

Judging Validity of Scientific Information

Validity means “how trustworthy and well-supported a claim is by the evidence and methods used.” On ACT Science, you’re rarely asked to recall scientific facts—you’re asked to decide whether information in a passage deserves your confidence.

What makes scientific information “valid” in ACT passages?

A scientific claim becomes more valid when it is supported by data collected using a method that reduces bias and alternative explanations. In an ACT passage, you usually judge validity by looking at:

• How the data were collected (experimental setup, measurement tools, procedure)
• Whether variables were controlled (so the effect can be attributed to the tested factor)
• Whether the sample is adequate (enough trials or subjects to be convincing)
• Whether results are consistent (patterns repeat rather than appear once)
• Whether the conclusion matches the data (no overreach)

This matters because ACT questions often hide the correct answer behind a tempting, “scientific-sounding” statement that isn’t actually supported by the figure or table.

Key idea: correlation is not causation

A classic validity trap is confusing a relationship seen in data with proof that one thing caused the other.

• Correlation: two variables change together.
• Causation: changing one variable produces a change in the other.

You can infer causation more confidently when the study design is experimental (one variable is manipulated, others are controlled) rather than purely observational.

Recognizing threats to validity

In ACT Science, you don’t need fancy research-methods vocabulary, but you do need to spot when a setup makes conclusions shaky.

Confounding variable: an extra factor that changes along with the variable you think you’re testing, making it unclear what caused the effect.

Bias: a systematic influence that pushes results in one direction (for example, measuring incorrectly in a consistent way).

Measurement error: random or systematic inaccuracies in readings.

Small sample size: a result based on very few trials is less convincing, especially if data are noisy.

Lack of a control or baseline: without something to compare against, it’s harder to tell whether a change is meaningful.

How validity questions look on ACT Science

Often the passage itself gives enough information to evaluate validity. You’re expected to read with a skeptical scientist’s mindset:

1. Identify what was changed (the independent variable) and what was measured (the dependent variable).
2. Check whether the procedure keeps other factors the same.
3. See whether the data show a clear pattern rather than random scatter.
4. Decide whether the claim stays within what the data actually show.

Example (validity in action)

A passage describes an experiment testing fertilizer A vs fertilizer B on plant height. The table shows:

• Fertilizer A: 10 plants, average height 25 cm
• Fertilizer B: 10 plants, average height 26 cm

If the question asks what most weakens the conclusion “fertilizer B increases height,” a strong answer would be something like: “Plants receiving fertilizer B also received more sunlight.” That introduces a confounder—sunlight could explain the difference.

A weaker criticism would be “The plants were green.” That detail doesn’t connect to the causal claim.

Exam Focus

• Typical question patterns:
  • “Which factor would most weaken/strengthen the conclusion?”
  • “Which statement best supports the validity of the experiment?”
  • “Which change to the procedure would improve the reliability of the results?”
• Common mistakes:
  • Treating any trend as proof of causation even when no variable was controlled.
  • Ignoring the method and looking only at the conclusion sentence.
  • Choosing an answer that sounds scientific but does not address a confounder, measurement issue, or sample limitation.

Formulating Conclusions from Data

A conclusion is a claim that summarizes what the data imply. On the ACT, your job is to build conclusions that are tight: they follow directly from the numbers/graphs without adding assumptions.

Why this matters

Most ACT Science questions are essentially: “Given what you see, what can you say?” Students often miss these not because the graph is hard, but because they either:

• state too much (overgeneralize), or
• state too little (miss the main trend), or
• state something different from what the graph actually shows.

How to build a conclusion step by step

When you’re given a table or graph, use a consistent process:

1. Restate what is being compared. Identify axes/columns clearly. What is measured? Under what conditions?
2. Describe the pattern in plain language. Increasing, decreasing, constant, peak, plateau, no clear trend.
3. Quantify with evidence. Use specific points or ranges from the data.
4. Limit your claim to the tested range. If temperatures tested are 10°C to 40°C, don’t claim what happens at 80°C.
5. Avoid causal language unless the design supports it. “Associated with” and “corresponds to” are safer than “causes” unless variables were controlled.

Interpolating vs extrapolating

• Interpolation: estimating a value within the tested range (usually safer).
• Extrapolation: predicting beyond the tested range (riskier and often discouraged unless the question explicitly asks).

ACT items frequently reward the student who notices that a conclusion is only justified between the endpoints shown.

Example 1 (drawing a supported conclusion)

A graph shows enzyme activity vs pH:

• Activity rises from pH 4 to pH 7
• Peaks at pH 7
• Falls from pH 7 to pH 10

A strong conclusion is: “Enzyme activity is highest at pH 7, increasing from pH 4 to 7 and decreasing from pH 7 to 10.”

A weak conclusion is: “The enzyme works best in basic conditions.” That is not supported because the peak is at neutral pH.

Example 2 (conclusions with variability)

Suppose two lines on a graph (Group 1 and Group 2) are very close and cross each other. If the question asks which group has “greater” output, a careful conclusion might be:

• “Neither group is consistently higher across all x-values; the higher group depends on the x-range.”

Students often choose a single group because they look at one point instead of the whole relationship.

What goes wrong: “reading” the graph without reading the labels

Many errors come from forgetting units or mixing up axes. Before concluding “as x increases, y decreases,” confirm:

• which variable is on the x-axis
• which line corresponds to which condition
• whether the y-axis is linear or uses unusual scaling

Exam Focus

• Typical question patterns:
  • “Which conclusion is best supported by Figure 2?”
  • “Based on Table 1, how does Y change as X increases?”
  • “At which condition is the measured value greatest/least?”
• Common mistakes:
  • Overgeneralizing beyond the tested range (extrapolating without being asked).
  • Ignoring a peak/plateau and claiming “always increases” or “always decreases.”
  • Mixing up which line or symbol matches which experimental condition.

Making Predictions Based on Evidence

A prediction is a claim about an unmeasured case (a new x-value, a new condition, or a future observation) based on an established pattern in the data. On ACT Science, predictions must be evidence-driven, not based on what “seems reasonable” in everyday life.

Why predictions are a core ACT skill

ACT Science often asks you to extend a pattern one step:

• “If the temperature were increased to 35°C, what would happen to the rate?”
• “If Trial 5 were conducted at 2.0 M instead of 1.0 M, which result is most likely?”

These questions reward students who can detect the relationship type (linear, inverse, leveling off, etc.) and apply it carefully.

How to predict responsibly

Use a disciplined approach:

1. Identify the relationship form: linear increase, linear decrease, curved, threshold, saturation, no trend.
2. Use nearby points: base your prediction on the closest values, not the entire range.
3. Stay within the data range when possible: ACT predictions are often interpolations.
4. If extrapolating, be cautious: extend the same pattern only if the graph suggests it would continue.

Common relationship patterns in ACT graphs

You don’t need to fit equations, but you should recognize shapes.

• Linear trend: roughly equal change in y for equal change in x.
• Inverse trend: y decreases as x increases (sometimes steep early, then flattening).
• Plateau (saturation): y increases then levels off.
• Optimum curve: y increases to a maximum then decreases.

Example 1 (interpolation prediction)

A table shows solubility (g/100 g water) at different temperatures:

• 20°C: 10
• 30°C: 14
• 40°C: 18

If asked to predict at 35°C, you look between 30°C and 40°C. A reasonable interpolation is about 16, because 35°C is midway.

The key is not the exact number—it’s that your prediction respects the local pattern and lands between 14 and 18.

Example 2 (prediction with a plateau)

A graph shows reaction rate rises quickly as substrate concentration increases from 0 to 2 mM, but barely changes from 2 to 5 mM.

If asked what happens at 6 mM, the best prediction is: “The rate will be approximately the same as at 5 mM” (or only slightly higher). Extending the early steep increase would ignore the plateau evidence.

What goes wrong: predicting from a single point

Students sometimes pick the largest y-value and assume “more x means more y.” But a peak graph (optimum) breaks that logic. Always scan the full curve before predicting.

Exam Focus

• Typical question patterns:
  • “If X were increased/decreased, what would happen to Y according to the trend?”
  • “Which value best estimates the result at an untested condition?”
  • “If the pattern continues, which outcome is most likely?”
• Common mistakes:
  • Extending the wrong part of the trend (e.g., using the initial slope after a plateau).
  • Extrapolating wildly beyond the graph when only interpolation is justified.
  • Ignoring that different lines (different conditions) may have different trends.

Evaluating Competing Models and Hypotheses

A model is a simplified representation of how a system works (it can be a diagram, a set of relationships, or a conceptual explanation). A hypothesis is a testable proposed explanation or prediction. On ACT Science, “competing models” often appear in passages labeled something like “Student 1 vs Student 2” or “Scientist A vs Scientist B,” where each proposes a different explanation.

Why model evaluation is central to ACT Science

Real science often involves multiple plausible explanations. The ACT tests whether you can:

• identify what each model predicts
• compare those predictions to data
• decide which model is better supported (or whether both are consistent)

The key is that you’re not choosing the model you like—you’re choosing the one that matches evidence.

How models/hypotheses are presented

You may see:

• Two different explanations for the same phenomenon
• Different assumptions (e.g., “heat is transferred mainly by radiation” vs “mainly by conduction”)
• Different predicted trends (increase vs decrease, or different shapes)

A useful habit is to translate each model into an “If…then…” statement.

• “If Model 1 is correct, then when X increases, Y should increase.”
• “If Model 2 is correct, then when X increases, Y should decrease.”

Step-by-step strategy to evaluate competing models

1. List each model’s core claim in your own words.
2. Extract distinct predictions—look for where they disagree.
3. Find the relevant data in the table/graph that tests that disagreement.
4. Compare: Does the observed pattern match Model 1, Model 2, both, or neither?
5. Decide the best-supported conclusion based strictly on the evidence shown.

This matters because many questions are designed so that both models explain some facts, but only one fits a key piece of data.

Example (two students, one dataset)

A passage shows a graph of gas pressure vs temperature at constant volume. Two students propose:

• Student 1: Pressure increases as temperature increases.
• Student 2: Pressure decreases as temperature increases.

If the plotted data trend upward, Student 1 is supported.

A more subtle version: both students might predict “pressure changes,” but disagree on whether the relationship is linear or whether it levels off. Then you must use the shape of the curve, not just direction.

When both models can be “consistent”

Sometimes the data are not detailed enough to distinguish. For example, if the experiment tested only two x-values, both a linear model and a curved model could pass through those two points. In that case, the best answer might be:

• “The data do not distinguish between the two models”

Students often feel forced to pick one. On ACT Science, “insufficient information” can be correct when justified by the evidence.

What goes wrong: confusing explanation with evidence

A model can sound detailed and scientific, but still be unsupported. On the ACT, the winner is not the model with more words—it’s the model whose predictions match the data.

Exam Focus

• Typical question patterns:
  • “Which student’s hypothesis is best supported by the results?”
  • “Which model predicts that Y will increase when X decreases?”
  • “Which observation would support Model 2 over Model 1?”
• Common mistakes:
  • Choosing the model that matches prior knowledge instead of the passage’s data.
  • Missing where the models differ and comparing the wrong feature.
  • Forgetting that sometimes the correct conclusion is that both (or neither) fit the data.

Determining Support for Inferences

An inference is a logical interpretation that goes beyond simply restating the data. Data might show “Group A has higher values than Group B,” while an inference might be “The treatment likely increased the measured outcome.” Inferences are essential in science—but they must be justified.

Why this skill is tested

ACT Science passages often include statements from researchers or students that interpret results. Questions then ask whether the results support those interpretations. Your job is to distinguish:

• what is directly observed (data)
• what is reasonably concluded (supported inference)
• what is speculation (unsupported inference)

The “support” standard: what counts as evidence?

A strong inference is supported when:

• It matches the direction and magnitude of the observed trend.
• It is consistent across trials/conditions shown.
• Competing explanations are less consistent with the evidence provided.
• It stays within the conditions tested.

A weak inference often:

• introduces an unmeasured variable (“therefore the enzyme’s shape changed”) when the passage never measured shape
• claims a universal rule based on limited cases
• assumes causation from observational data

A practical way to test an inference

When you see an inference, ask two questions:

1. What specific data point(s) would I cite to justify this? If you can’t point to a figure/table entry, the inference may be unsupported.
2. Could the opposite inference also fit the data? If yes, the inference may be too strong.

Strengthening vs weakening an inference

ACT questions sometimes ask what additional information would support an inference. Think like a scientist:

• To strengthen: add controls, more trials, direct measurements of the proposed mechanism, or data in the disputed range.
• To weaken: introduce a confounder, show contradictory data, or show the effect disappears under controlled conditions.

Example 1 (supported inference)

A passage tests whether a filter removes dye from water. After filtration, dye concentration drops from 50 units to 5 units in every trial.

• Data statement: “Concentration decreased after filtration.”
• Supported inference: “The filter removed most of the dye from the water.”

This inference is supported because the drop is large and consistent.

Example 2 (overreaching inference)

A graph shows that, in one experiment, plants grown with classical music were taller than plants grown in silence.

An unsupported inference would be: “Music causes faster cell division in all plants.” That introduces a mechanism (cell division) and a universal claim (all plants) not measured.

A more defensible inference is: “In this experiment, plants exposed to music grew taller than plants grown in silence.” Notice how this stays close to the evidence and the tested scenario.

Inferences in “conflicting viewpoints” passages

In “Student 1/Student 2” style passages, each student often makes an inference from the same background information. To determine support:

• Identify each student’s key assumption.
• Check whether the passage provides direct evidence for that assumption.
• If the assumption is unstated and untested, the inference is weaker.

What goes wrong: mixing up “consistent with” and “proven by”

An inference can be consistent with the data without being proven by the data. ACT questions often hinge on that distinction.

• “Supported by” usually means the data make the inference likely.
• “Proved” (rare wording) would require ruling out alternatives.

When in doubt, pick the answer that makes the smallest logical leap beyond the data.

Exam Focus

• Typical question patterns:
  • “Are the results consistent with the claim that…?”
  • “Which statement is best supported by the data?”
  • “Which additional observation would most support the inference that…?”
• Common mistakes:
  • Treating a possible explanation as the only explanation.
  • Citing the wrong figure/table, or citing none at all (guessing based on topic familiarity).
  • Making a mechanistic inference (what’s happening inside the system) when the passage only reports surface measurements.