The five most common mistakes in deductive reasoning are affirming the consequent, denying the antecedent, the undistributed middle, belief bias and reading "some" as "all". All five share one root: accepting a conclusion because it feels right instead of checking that the premises force it. Each has a one-line fix.
Mistakes with if-then rules
Mistake 1: affirming the consequent
"If the server is overloaded, pages load slowly. Pages are loading slowly. So the server is overloaded." The rule says that overload leads to slowness. It does not say that slowness can only come from overload, so a bad network link or a heavy page would fit the facts equally well.
Fix: an if-then rule runs one way, so a match on the "then" part tells you nothing about the "if" part.
Mistake 2: denying the antecedent
"If a candidate passes the screening, they are interviewed. Sam did not pass the screening. So Sam will not be interviewed." Perhaps Sam was referred by a colleague and skipped the screening altogether. The rule only covers what happens to people who pass.
Fix: if the "if" part is false, the rule has nothing to say about the outcome.
Mistakes with groups, belief and quantity words
Mistake 3: the undistributed middle
"All auditors are careful readers. All proofreaders are careful readers. So all auditors are proofreaders." Sharing a trait does not make two groups the same group. Auditors and proofreaders may both sit inside the careful readers without overlapping at all.
Fix: draw the circles, and if the two groups can sit apart inside the shared category, the conclusion fails. The longer treatment is in syllogisms, validity and truth.
Mistake 4: belief bias
Compare two arguments. "All vegetables are poisonous. Carrots are vegetables. So carrots are poisonous" is valid, but the first premise is absurd, so many people reject it. "All doctors are well educated. Some well-educated people are rich. So some doctors are rich" sounds sensible, yet it does not follow, so many people accept it.
In a well-known 1983 study, Jonathan Evans, Julie Barston and Paul Pollard found that judgments of syllogisms depended on how believable the conclusion was, and not only on whether it followed. Keith Stanovich and Richard West's review of individual differences in reasoning found that such slips are rarely just carelessness, and that tasks which ask you to set aside everyday context are especially demanding.
Fix: swap the content for nonsense words, such as blorps and zinks, and judge the shape of the argument alone.
Mistake 5: reading "some" as "all"
"Some managers have completed the course. Everyone who has completed the course receives a bonus. So all managers receive a bonus." Only some managers are known to have completed the course, so only some are known to receive the bonus. The conclusion that does follow is "Some managers receive a bonus".
A related slip runs the other way. "Some of the files are encrypted, so some files are not encrypted" does not follow, because in logic "some" means at least one and possibly all. A third relative is flipping "All A are B" into "All B are A", which turns "all poodles are dogs" into "all dogs are poodles".
Fix: read "some" as "at least one, maybe all", and never conclude "not all" from it.
Why the same mistakes keep coming back
These errors are not random lapses. Philip Johnson-Laird's mental-model theory predicts systematic errors in reasoning and reports evidence that supports the prediction. On that account, the more possibilities a problem makes you hold in mind at once, the harder it gets.
That is why the same few mistakes appear again and again. They are habits of the quick, intuitive route, and the cure is a slower check that you apply on purpose.
A five-second audit before you commit
Before you lock in an answer, run through this short list.
- Did I use anything that is not in the statements?
- If the rule is if-then, does my fact confirm the "if" or deny the "then"? Anything else is a trap.
- If the statements use all, no or some, can I draw a picture in which the premises hold and my conclusion fails?
- Did I judge the shape, or did the topic nudge me?
- Did I treat "some" as "all", or "all" as reversible?
The full method behind that audit is laid out in how to solve deductive reasoning questions, and the pattern behind the first two mistakes is explained in if-then reasoning.
Test yourself: name the mistake
Here are three short arguments. Name the mistake in each before you read the answer under it.
- "If the invoice is approved, it is paid. The invoice was paid, so it was approved." This is affirming the consequent, because invoices might be paid by other routes.
- "All the interns are keen. Some keen people are late. So some interns are late." This is the undistributed middle: the late keen people may all be outside the intern group.
- "Some of the files were encrypted, so some were not." This reads "some" as "not all". Perhaps every file was encrypted.
Practise spotting them
Seeing these patterns in a timed setting is different from nodding along in an article. The JobCannon Deductive Reasoning Test is built around "follows / does not follow" conclusions from a set of statements, so every one of the five traps above can turn up in it.
If you prefer a wider mix that also includes analytical puzzles, try JobCannon's older logical reasoning test. Both are written in-house, so read the results as practice feedback and not as a formal score.