Verbal Reasoning: True, False and the Cannot Say Problem
Cannot Say is the answer most candidates get wrong. The standard of proof behind the three verdicts, and how SHL, Saville, Kenexa and Watson Glaser differ.
Verbal reasoning tests look like reading comprehension and are not. Comprehension asks what a text means. A True, False or Cannot Say item asks something narrower and stranger: whether this specific passage, treated as the only evidence in existence, settles this specific claim. Most candidates lose marks not because they misread anything but because they answer the comprehension question instead of the evidence question.
Three verdicts, and what each one actually claims
The three options are not degrees of confidence. Each is a definite claim about the relationship between the passage and the statement.
- True: the passage, on its own, makes the statement follow. Not likely, not consistent with. Follow.
- False: the passage, on its own, rules the statement out.
- Cannot Say: the passage leaves the statement open. Both a world where it is true and a world where it is false are consistent with everything the text says.
Why Cannot Say is the item people get wrong
On a well-built test, Cannot Say is correct on roughly a third of items, which makes it as common as either of the other two. Candidates still get it wrong more often, and for two opposite reasons that need opposite corrections.
Under-selection is the more common of the two. Cannot Say feels like declining to answer, and people sitting a graduate assessment want to demonstrate that they understood the passage. So they reach for the verdict that shows work. A statement that is plausible, consistent with the passage and almost certainly true in the real world gets marked True, when the passage never established it.
Over-selection appears in the last few minutes, when Cannot Say becomes a way of saying I did not have time to check. That is not what it means either. On a test with no wrong-answer penalty, a hurried Cannot Say scores the same as any other hurried guess, so the honest fix is pacing rather than a default verdict.
There is a cheap self-audit. Count how often you chose Cannot Say across a practice set. If it is well under a quarter of your answers, you are over-reading and importing knowledge. If it is over half, you are using it as a shelter. Either pattern is fixable in a couple of sessions once you can name it.
Stated versus implied: where the line actually falls
The common advice, answer only from what the passage states, is too strict and quietly misleading. Plenty of correct True answers are never stated anywhere in the text. The real line is not stated against implied. It is whether the step from passage to statement is closed or open.
Closed steps are allowed. If the passage says the Lisbon office opened in 2019 and the Madrid office opened two years later, then Madrid opened in 2021 is True. The text never says 2021. You did arithmetic, and arithmetic adds no new facts about the world. The same applies to definitions (a passage that says the board met quarterly establishes that it met four times a year) and to plain logic (if all regional managers report to Chen and Okonjo is a regional manager, Okonjo reports to Chen is True).
Open steps are not. If the passage says the Madrid office opened two years after Lisbon and the statement is that Madrid opened because the Lisbon office succeeded, you have had to supply a causal premise the text does not contain. Cannot Say. The test is mechanical: write down every premise your answer used. If any of them came from your head rather than from the page, the step was open.
The five patterns that generate most wrong answers
Test writers do not invent a new trick for every item. Nearly all distractors are one of these, and three of them are covered in more depth in the SHL Verify guide.
- Quantifier swaps. Learn the hierarchy: some and at least one are the weakest, then many, then most, then all. A statement may always travel down the hierarchy and never up. All employees attended gives you some employees attended, which is True. Many attended does not give you most attended, which is Cannot Say.
- Correlation dressed as causation. The passage reports that two things moved together or in sequence. The statement says one produced the other. Unless the text asserts the mechanism, this is Cannot Say every time.
- Comparisons and superlatives the passage never makes. A figure for one region does not make it the highest, the fastest growing, or above average, unless the passage supplies the other regions or says so directly.
- Modality and tense drift. This one is underrated and it is everywhere in business passages. Plans to, expects to, is considering, aims to and is on track to are all statements about intention, not about outcome. A passage saying the company plans to open twelve sites does not make the company will open twelve sites True, and it does not make it False either.
- Scope drift. Evidence about one year, one region, one product line or one survey group does not license a conclusion about the business, the market or the period. Compare the scope of the evidence with the scope of the claim before anything else; that single comparison resolves a surprising share of items.
Negation is worth one extra line because it catches strong candidates. The denial of all branches met target is not no branches met target. It is that at least one branch did not. If a statement negates a claim in the passage, check which of the two negations it is using.
The same three words, four different tests
True, False and Cannot Say are shared vocabulary, not a shared test. The format around them changes what good technique looks like.
Two practical consequences. First, practise in the format you were actually sent: a candidate drilled on 60-second SHL items walks into Saville and leaves three items blank. Second, check the marking rule before you decide how to spend the last two minutes. Filling every blank is correct on SHL, Saville and Kenexa. It is actively harmful on Aon.
Reading under the clock
The first thing to establish, before you read a word of the passage, is how many statements are attached to it. That number decides your strategy, and candidates who use one strategy for both situations lose time in one of them.
A rough budget for a 55-second item: ten seconds to locate, twenty-five to read around the anchor, ten to decide, ten of slack. The slack is the part people cut first and the part that stops one hard item eating the two easy ones behind it.
Practising so that it moves the score
Verbal reasoning responds to practice better than most candidates expect, but only if the practice is diagnostic. Vocabulary is almost never the barrier. Evidence discipline and pacing are, and both are trainable.
After a timed set, do not just count your score. Tag every wrong answer with which of the five patterns caused it, and tag every Cannot Say you missed with whether you under-selected or over-selected. Almost everyone finds one dominant error rather than a spread, and one dominant error is a small fixable thing. Someone who consistently upgrades many to most has a twenty-minute problem, not a reading problem.
You can sit the formats separately, with the real clocks, on the verbal reasoning practice page, and if you are facing a critical thinking appraisal rather than a standard verbal test, start from the Watson Glaser guide instead, because the standards of proof there are genuinely different.