SC3 Evaluating Scientific Arguments & Models

Lesson

Science is 40 questions in 40 minutes, and calculators are not permitted in this section. That is not a trivia fact about test policy - it tells you what is being measured. No item here needs arithmetic you cannot do in the margin of the booklet. What every item needs is a fast, defensible judgement about what a piece of evidence does and does not establish.

This reporting category - evaluating scientific arguments, models and conclusions - accounts for a large share of scored Science items, and two passage types dominate it. One is Conflicting Viewpoints: two or three named positions on a single question, sometimes with no data at all. The other is a described model presented next to data, where the questions ask what the data supports, what it does not, and what would break the model.

Claim, evidence, prediction - keep them apart

A viewpoint is a claim, plus the evidence offered for it, plus what should follow if it is true. Items are built to catch students who fuse the three. A scientist may write "the ocean warmed by 4 °C" (evidence, which the rival also accepts) and "that warming drove the extinction" (claim, which the rival denies). Asked what the two disagree about, a student who has not separated those sentences will pick the temperature.

• Claim - the sentence the scientist would still defend if everything else were stripped away.
• Evidence - the observation offered in support. Frequently shared, frequently uncontested.
• Prediction - what should be seen if the claim is right and would not be seen otherwise. This is where the questions live.

Strengthen, weaken, and why "consistent with" is weaker than "predicted by"

An observation strengthens a hypothesis when that hypothesis predicted it and the rivals did not. It weakens the hypothesis when the hypothesis predicted the opposite. An observation every viewpoint would happily accept does neither, however impressive it sounds.

That is the gap between consistent with and predicted by, and the ACT tests it directly. Both chemists in a set may be consistent with "the reaction produces an alcohol." Only one predicts "the rate does not change when you double the hydroxide concentration." The second is worth something. The first is worth nothing, and it will be offered to you as a choice.

Two further shapes are worth recognising. A prediction with two halves - lower the damage, and lifespan will rise - is tested by the second half; delivering the first half and stopping proves nothing. And a hypothesis is never weakened by the failure of a prediction it did not make, so before you mark a viewpoint down, check that the prediction in question was actually its own.

Disagreement versus different emphasis

Two scientists writing about the same event are not automatically contradicting each other. Ask: is there a sentence in each passage that cannot both be true? If Scientist 1 says the impact caused the extinction and Scientist 2 says the volcanism did, they disagree. If Scientist 1 writes at length about the impact and Scientist 2 writes at length about the volcanism, they may not - one may simply be describing a different part of the same story.

"On which of the following would all of the scientists agree?" is usually settled by one of two things: a number in a shared table, or a fact stated in the introduction before the first viewpoint begins. Look there first. What they agree on is almost never a conclusion; it is the problem they are all trying to solve.

Is the stated conclusion supported by the data?

A conclusion can fail in ordinary, recognisable ways. Run this list:

• Too strong. The data show two things moving together; the conclusion asserts that one causes the other.
• Outside the range. The experiment ran from 20 °C to 60 °C; the conclusion is about 200 °C.
• A rival explanation. Two things changed at once and the design cannot separate them.
• The wrong half. The prediction had two parts and only one was tested.
• A counterexample in the table. One row breaks the trend the conclusion needs. Look for it - the test writers usually put it there deliberately.

Evaluating a model

A model is a deliberate simplification, so "the model is not exactly right" is never the point of the question. Three questions carry you through every model item:

• What does it explain? Which observations follow from its assumptions with no further help.
• What does it not address? Not "gets wrong" - is silent about. A model of a gas as non-interacting points has nothing to say about why a gas condenses, because there is no attraction anywhere in it.
• What would break it? The single observation its assumptions forbid. If a model says predation drives the prey cycle, prey still cycling with every predator removed is that observation.

When a model is patched - a correction term added, a layer inserted - the patch carries a testable claim of its own. If a constant is introduced to stand for the volume the molecules themselves occupy, then a fit that needs a different value of that constant at every pressure is telling you the interpretation is wrong, not merely that the number needs adjusting.

Pacing

Conflicting Viewpoints passages are the longest reading in the section and the least answerable by glancing at a figure. Budget accordingly: read the viewpoints properly once, annotate claim and prediction beside each, then answer. Skimming them and hunting for keywords afterwards costs more time than it saves, because nearly every item turns on knowing which scientist said what.