Picture yourself on a jury. A prosecutor stands up and makes one focused claim: "The defendant was at the scene of the crime on the night of the 14th." Then she says she'll present four pieces of evidence, and only one will prove that claim beyond a reasonable doubt. That's a Command of Evidence question. The claim is the prompt; the four quotes are the answer choices.
A weak juror gets distracted. He hears the claim and then listens to the evidence with a wandering mind. "The defendant owned a blue car" — interesting. "The defendant knew the victim" — ooh, a motive. "The defendant disliked the victim" — this guy's guilty. None of that proves he was at the scene, but the weak juror convicts on a general feeling of guilt. On the SAT, that feeling is a wrong answer. An elite juror listens to every quote with one question only: does this, on its own, prove the specific claim? Owning a car — no. Knowing the victim — no. A security camera that timestamps him leaving the building at 10:14 — yes. Case closed.
The cross-examination
Follow three steps, in order, every time.
- Read the indictment. Break the claim into its parts. "The artist's later works show a growing disillusionment with fame" has three: later works, growing disillusionment, and a link to fame. Your quote has to cover all of them.
- Question each quote on its own. Put it on the stand and ask exactly what it can testify to, and nothing more. A quote about the artist's first exhibition in 1995 says nothing about later works or fame — dismissed.
- Find the star witness. One quote will answer "yes" to the whole indictment. That's the answer.
The three flawed witnesses
Your cross-examination will show that most of the quotes are flawed in one of three predictable ways.
| Flawed witness | What goes wrong |
|---|---|
| The irrelevant one | On the right topic, wrong facts. (Claim: the policy was ineffective. Quote: the policy was signed on a Tuesday.) |
| The half-truth | Proves only part of a multi-part claim — the most dangerous trap. (Claim: the project was expensive and delayed. Quote: the cost doubled. Says nothing about delay.) |
| The hearsay | Points at the evidence without being it. (Claim: the result was surprising. Quote: "scientists later noted the outcome was unexpected" — tells you it was surprising, doesn't show the surprising result.) |
A worked case
Take a passage about the American modernist composer Charles Ives, whose dissonant work was ignored in his lifetime and only embraced decades later. The claim: the significance of Ives's work comes primarily from its forward-thinking nature. Deconstruct it — the quote has to show him being ahead of his time. Now the choices. A quote listing his experimental techniques shows he was an innovator but never says he was ahead of others (partial support). A quote about why his work was ignored explains he was misunderstood, not forward-thinking (background). A quote describing the re-evaluation of his reputation gives the result, not the reason. The one that says his music "anticipated many of the innovations of later European modernists" places him explicitly ahead of the future. That's the star witness, and it's the only one that proves the actual claim.
The support/weaken variant: isolate the variables
A related question type asks which finding would most strengthen or weaken a hypothesis. The passage buries a simple causal claim under a pile of jargon — "the haggis migrates south because it fears bagpipe music" — and your job is to ignore the jargon and treat it as one equation: A causes B.
To support the link, you want evidence that B happens when A is present, or doesn't happen when A is absent. (Sea otters decline where killer whales hunt; populations stay stable where the whales are absent — a clean control that strengthens the link.) To weaken it, sever the cord: show B happening without A, or introduce a sneaky third cause that explains B better. (A productivity study credits new software for a 15% jump — but the trial happened to fall during the company's busiest season, when productivity always rises by about that much. Suddenly the software might have done nothing.) The wrong answers in these questions are almost always interesting details about A or B that don't actually test the link between them.
Don't ask which quote is the most interesting or which one "feels" related. Ask which one, standing alone, proves the exact claim. The most tempting wrong answers are on-topic and prove nothing — and you don't even get the fancy law degree to go with the job.
Every one of these questions is a Pfungst experiment. Ignore the vibe, examine each quote, and confirm it actually proves the specific claim. This same skepticism is what Forge measures when it watches you read — whether you reliably anchor your answers in proof or get pulled toward the quote that merely sounds right, which is one of the five dimensions it maps.