Literature & Language Arts: Counterexample hunting

What you'll be able to do: Find the line in your own text that your reading can't explain, before a marker or a rival critic finds it for you.

Close reading is mostly this

A generalisation about a text feels true the moment you notice the pattern that suggested it. "Hamlet is a play about indecision" feels obviously right the first time you notice Hamlet delaying. What it doesn't survive is a search for the moments where he doesn't delay at all: he stabs Polonius through the arras without a beat of hesitation, he rewrites the commission that sends Rosencrantz and Guildenstern to their deaths with cold efficiency, he leaps into Ophelia's grave on impulse. A reading that has to explain those scenes, rather than quietly ignore them, is doing more work than the cliché, and is usually more interesting: not "Hamlet can't act" but something narrower about which kinds of action he can't take.

The generalisation you haven't tested against a counterexample isn't a reading yet. It's a first impression that happened to survive because you weren't looking for what would kill it.

How to run it

1. State the claim plainly. Don't soften it, a hedged claim is hard to test because it's hard to be wrong about.
2. Ask the tool to find the hardest passage for it, not passages that confirm it. Find the line in this text that's most difficult to reconcile with my claim.
3. Check every suggestion against the actual text , this method inherits the confabulation risk from article 07 in full, because "finding" a counterexample requires the tool to have the passage right in the first place.
4. Decide: does the counterexample break the claim, or sharpen it? Usually the latter, the claim gets narrower and more defensible, which is the whole point.
5. Write the narrowed claim, with the counterexample named and addressed in the essay rather than silently avoided.

My claim is: [claim], about [text]. Don't tell me whether it's right. Find the passage or passages in [text] hardest to reconcile with this claim, and tell me specifically why they're a problem for it, not evidence that confirms it. I'll check the exact wording against my own copy before I use anything you find.

Across the disciplines

Literature. "Dickinson's dashes signal hesitation." Test it across several poems, not one: in "Because I could not stop for Death -", does the dash before "Death" read as hesitation, or as a stately, measured pause that slows the whole poem to the pace of the carriage it describes? If the second reading fits better, the claim narrows from "hesitation" to something about controlled pacing, a stronger, more specific thesis than the one you started with.

Poetry more broadly. A claim that a poet "always" uses a device one way is the easiest kind of claim to falsify and the most common kind students make uncontested. Test it across the poet's other work before it goes in an essay.

Rhetoric. "This speaker never appeals to emotion, only logic." Hunt for the counterexample in the speech itself, a single loaded phrase, one appeal to shared fear or pride, is usually there, and finding it produces a more accurate and more interesting claim about a speaker who mostly avoids pathos and departs from that once, deliberately.

Composition. Test your own thesis the same way before you submit. If you can't find a paragraph in your source material that complicates your claim, either the claim is too safe to be worth arguing, or you haven't looked hard enough.

Linguistics, where this method does its most important work. "English adjectives always precede the noun they modify" is a rule taught with total confidence and falsified by ordinary usage: the president elect, attorney general, the devil incarnate, something wicked (indefinite pronouns regularly take postpositive adjectives). A linguist doesn't defend the rule by ignoring the counterexamples, they narrow the rule until it correctly predicts both the regular cases and the exceptions, which is what a descriptive grammar actually is, as opposed to a prescriptive one that just asserts the regular case and calls the rest mistakes.

Pitfalls

1. Accepting a counterexample you haven't checked. The tool can invent a "counterexample" as fluently as it can invent a supporting quotation. Verify before you build an argument around it.
2. Treating one counterexample as refutation. Usually it narrows the claim rather than killing it, the skill is knowing which.
3. Cherry-picking the counterexample search itself. Ask for the hardest case, not a middling one you can dismiss quickly.
4. The tell: every counterexample offered is one you already knew about, or none is offered at all. The second usually means the claim was too vague to be wrong in the first place, go back to article 02.

Try this today

Take a generalisation you're confident about, something you'd write in an essay without a second thought. Ask only for the hardest passage against it, and check it word for word before deciding whether your claim survives, narrows, or breaks.