The Short Definition
Logical reasoning is the ability to work out what follows from what. Given a set of statements, it is the capacity to say which conclusions are forced, which are merely possible, and which are ruled out.
The word doing the work is follows. Logic is concerned with the relationship between premises and conclusion, not with whether the premises happen to be true.
That separation is what makes it testable. A valid argument from false premises is still valid, and recognising this is precisely what a logical reasoning item is built to detect.
The Four Things a Logical Test Actually Asks
Almost every logical battery in circulation draws on four question families.
- Syllogisms. Two premises about categories, one conclusion. "All auditors are graduates; some graduates are accountants" — and the work is seeing what that does not establish.
- Conditional logic. If–then statements, and what can be inferred when the antecedent or consequent is affirmed or denied. This family generates the most confident wrong answers.
- Analytical puzzles. A set of constraints — seating, ordering, scheduling — from which specific facts must be deduced. Constraint satisfaction with a clock attached.
- Argument analysis. A short passage, then a question about its assumption, its flaw, or what would strengthen or weaken it. The most job-relevant family and the least formal.
Validity Is Not Truth
This is the distinction on which most logical reasoning items turn, and it is genuinely counter-intuitive.
An argument is valid when the conclusion cannot be false if the premises are true. It is sound when it is valid and the premises are actually true.
Why tests exploit the gap
Items are routinely built with conclusions that are obviously true in the real world but do not follow from the premises given. The candidate who checks the conclusion against their own knowledge marks it correct and loses the point.
The discipline being measured is the willingness to reason from the page rather than from what you already believe — the same habit a "cannot say" verbal item tests, in formal dress.
The experiment that made the point
Peter Wason's 1968 selection task is the cleanest demonstration. Four cards read A, K, 4 and 7. Each has a letter on one side and a number on the other. Which must you turn over to test the rule if a card has a vowel on one side, it has an even number on the other?
The answer is A and 7. Turning A can expose an odd number and break the rule; turning 7 can expose a vowel and break it. The 4 is the tempting card and it settles nothing — the rule makes no claim about what sits behind an even number.
Most educated adults pick A and 4. That is the same conditional error that costs marks on a test item, stripped of the prose that usually hides it.
Three Modes, and the One Tests Rarely Reach
Reasoning in the wild comes in three shapes, and a commercial battery samples them very unevenly.
- Deduction. From general rules to a forced conclusion: if the premises hold, the conclusion cannot fail. Syllogism and conditional items live here, and it is the only mode that can be marked unambiguously right or wrong.
- Induction. From cases to a general rule that is probable rather than forced. Sequence and pattern items sit here, which is why they are scored against the intended rule rather than against any rule that happens to fit.
- Abduction. From an observation to the explanation that best accounts for it. A clinician reading symptoms, an engineer reading a failure, an analyst reading fragments.
Almost every battery measures the first two and leaves the third alone, because "best explanation" resists a single defensible answer key.
Worth knowing when you read a score: the mode most professional reasoning actually runs on is the one the test did not look at.
Where It Sits in the Map of Abilities
Logical reasoning is not a single factor in the structural models. It sits across two.
The formal, novel-pattern end — matrix items, unfamiliar constraint puzzles — loads on fluid intelligence, because nothing stored helps. The argument-analysis end draws on verbal comprehension, since the reasoning arrives wrapped in prose that must first be parsed correctly.
John Carroll's 1993 hierarchical synthesis placed sequential and inductive reasoning as narrow abilities under the broad fluid factor, which is roughly where a modern logical battery aims.
Is It the Same as Critical Thinking?
No, and the difference is practical rather than academic.
Logical reasoning asks whether the conclusion follows from the premises as stated. Critical thinking asks the further questions: are these premises credible, who is asserting them, what has been left out.
Someone can be excellent at the first and poor at the second. Formal validity says nothing about whether the source is reliable — and in real work, the premises are usually where the problem is.
Why Employers Test It
Logical screening is standard in law, consulting, software, and analytical graduate schemes, and the reason is that most professional errors are inferential rather than factual.
Cognitive ability measures have been among the more consistently predictive tools in selection research. Schmidt and Hunter's widely cited 1998 meta-analysis placed general mental ability at the top of the predictor list; Sackett and colleagues' 2022 re-examination produced substantially lower corrected estimates while leaving it a real predictor.
What an employer is actually asking is narrow: when this person is handed a set of facts and a proposed conclusion, will they notice if the conclusion does not follow?
Where it actually pays off
The professions that screen hardest are the ones where a bad inference survives long enough to become expensive.
- Law. Whether an oral assignment breaches a clause requiring written consent is a question about the structure of the clause, not about the facts.
- Software engineering. Does the loop terminate, does the invariant hold at every step, is that branch reachable — reasoning about states nobody can observe directly.
- Consulting and strategy. Splitting a question into parts that neither overlap nor leave gaps, then building a recommendation that survives being argued with.
- Research. Spotting the confound, the unwarranted generalisation, and the alternative explanation the author waved away too quickly.
- Analysis and intelligence work. Assembling fragments into the account that fits best, while keeping track of how likely each account was before the fragments arrived.
Can It Be Improved?
Yes, and the training literature here is unusually encouraging compared with other fluid measures.
Learning the named forms — the valid inference patterns and the two conditional fallacies that mirror them — produces real gains, because most errors come from applying an invalid pattern confidently rather than from failing to reason at all.
Where the ceiling is
Formal-pattern gains arrive quickly. Beyond them, performance depends on working-memory capacity — holding four constraints in mind while testing a fifth — and that component responds poorly to practice.
The practical version: learning the fallacies is worth several hours, and drilling harder puzzles after that is worth much less than it feels.
Reading Your Own Result Honestly
A logical score is taken under conditions unlike real reasoning: no time to sleep on it, no colleague to check the inference against, and no opportunity to go and question the premises.
Treat a strong score as evidence that formal structure does not trip you up, and a weak one as a prompt to find out which family cost the marks — the remedies differ completely.
Conditional errors are the most fixable thing in cognitive testing, because they are a small set of named patterns. Argument-analysis errors are usually a reading problem in logical clothing, and they respond to slowing down on the passage rather than to studying logic.