Deductive reasoning
Deductive reasoning is inference in which the conclusion cannot be false if the premises are true, because the conclusion was already contained in them. It runs from a general rule to a particular case, it delivers certainty rather than probability, and it delivers that certainty conditionally: certain given the premises, and no more certain than the premises are. That last clause is where most trouble with deduction lives.
The word to watch in any deductive argument is the quantifier. All, every, no and none support deduction. Most, usually and typically do not, and an argument that swaps one for the other partway through has stopped being deductive without announcing it.
The syllogism: deduction in its oldest form
A syllogism is a deductive argument with 2 premises and 1 conclusion, and Aristotle catalogued the working forms in the Prior Analytics, the earliest surviving system of formal logic. The classic shape is a general premise, a particular premise, and a conclusion about the particular case.
Every box on the loading dock carries a green seal. The box in front of you is on the loading dock. Therefore the box in front of you carries a green seal.
Nothing about boxes was learned here. The conclusion was implicit the moment the two premises were laid side by side, which is the defining feature of deduction and the reason it is useless for discovery. What it is superb at is auditing: it exposes exactly which general claim a particular conclusion is resting on, and that claim can then be checked.
Deduction goes wrong most often at the quantifier. Change the first premise to most boxes on the loading dock carry a green seal and the conclusion no longer follows at all. It becomes probable, which is a different kind of support and belongs to the other family of inference. Deductive vs inductive reasoning sets the two side by side.
Four valid forms and two convincing frauds
4 conditional forms are valid and 2 that look almost identical are not. Learning the 6 by name is the highest return investment in formal logic, because reasoning questions on the LSAT, the CAT and employer aptitude batteries are built out of them.
| Name | Shape | Verdict |
|---|---|---|
| Modus ponens | If P then Q. P. Therefore Q. | Valid |
| Modus tollens | If P then Q. Not Q. Therefore not P. | Valid |
| Hypothetical syllogism | If P then Q. If Q then R. Therefore if P then R. | Valid |
| Disjunctive syllogism | Either P or Q. Not P. Therefore Q. | Valid |
| Affirming the consequent | If P then Q. Q. Therefore P. | Invalid |
| Denying the antecedent | If P then Q. Not P. Therefore not Q. | Invalid |
Put the two frauds on a real rule and they collapse instantly. The rule: if the fire door is propped open, the corridor alarm sounds. Affirming the consequent says the alarm is sounding, therefore the fire door is propped open. It is not: something else can trigger an alarm. Denying the antecedent says the door is not propped open, therefore the alarm is not sounding. Same mistake wearing the other coat. A conditional rule licenses travel in one direction only, forward from the antecedent and backward from the denied consequent, and every other route is closed.
Deductive reasoning in geometry
Deductive reasoning in geometry is the school subject where most people meet formal proof, and geometry is used for it because the whole field is deliberately built as a deductive system. Euclid's Elements starts from a small list of definitions, postulates and common notions, and every later proposition is derived from those and from propositions already proved. Nothing is measured and nothing is sampled.
The two column proof taught in school geometry is that structure made visible: statements on the left, the reason authorizing each statement on the right, where a reason is a definition, a postulate, or a theorem already established. When a student is told to justify a step, the request is not pedantry. It is the demand that makes the argument deductive rather than merely plausible, because a step nobody can name a warrant for is a step where a picture, not a rule, did the work. A figure that looks like a right angle is an observation, and observations are not admissible in a proof.
Why a valid deduction can still be wrong
A valid deduction can carry you to a false conclusion, and this is the single most useful thing to know about deduction. Validity is a property of the form alone: it says that if the premises are true the conclusion must be, and it says nothing whatever about whether they are. Feed a perfectly valid form a false general premise and it will deliver a false conclusion with complete confidence, which is precisely why confident reasoning from an unexamined rule is dangerous rather than safe.
The vocabulary for this is validity versus soundness, and Valid vs sound arguments is the page that separates them properly. The practical upshot for anyone doing deduction: once you have checked the form, you are only half done, and the remaining half is checking whether the general premise is actually true of every case it claims.
Deductive reasoning questions and puzzles
Deductive reasoning questions come in 4 recognizable kinds, and all 4 reward the same method: extract every constraint, then eliminate rather than guess.
- Grid puzzles, which give a set of people, a set of attributes and a list of constraints, and ask who has what. The grid is a bookkeeping device: mark impossibilities, not just possibilities, because eliminations carry more information than confirmations.
- Calendar and sequence items, which ask what day a date falls on or which item comes fourth. These are arithmetic wearing a puzzle costume, and they are solved by writing the rule down explicitly instead of counting on your fingers.
- Conditional chains, which give three or four rules of the form if P then Q and ask what must follow. Draw the arrows, add the contrapositive of every rule, and read the chain.
- Assumption and inference items, which give a short passage and ask what must be true. Answer only from the stated premises and refuse anything that merely seems reasonable.
The failure that spoils grid puzzles is not arithmetic. It is settling on a candidate solution early and then reading each new clue for how it fits that candidate rather than for what it rules out. That is confirmation bias operating on a page of clues, and the countermeasure is mechanical: for every clue, write down at least one thing it eliminates before you write down anything it confirms.
Practice sets circulate widely, often as downloadable PDFs of past items with answer keys. They are worth working, on one condition: cover the answer, and after solving, write the rule you used in one sentence. A puzzle you solved without being able to name the rule you used has trained your patience rather than your logic. Critical thinking exercises works this the other way round, with the rule named for every item.
The test to run on your own deductions
Before you accept any conclusion you reached deductively, ask:
Does my general premise really say all, or does it only say most?
If it says most, nothing follows about the particular case with certainty, and any confidence you feel came from the form rather than from the facts. If it genuinely says all, the next question is who established that, and how many cases they actually checked. Logical reasoning covers the three conditions every argument has to meet before either question is worth asking.