Modals of Deduction
Level B2 · Modals
These modals show how certain you are about something, based on evidence or logic — not obligation or permission. must = you're almost sure it's true; can't = you're almost sure it's not true; might/may/could = you're not sure either way. The same set of modals works for present situations and, in a different form, for past ones.
For a present situation, use modal + be (or modal + base verb):
- must + be for a near-certain positive conclusion: He must be tired — he's been working all day.
- can't/couldn't + be for a near-certain negative conclusion: She can't be home — the lights are off (never mustn't for this meaning).
- might/may/could + be for an uncertain possibility: She might be at work; he could be stuck in traffic.
- The same pattern works with a base verb instead of be: She must know the answer; he can't like this weather.
For a past situation, add have + past participle after the modal — the modal itself never changes form:
- must have + past participle for a near-certain positive conclusion about the past: She must have left already — her coat is gone.
- can't have/couldn't have + past participle for a near-certain negative conclusion: He can't have finished — it's only been five minutes.
- might have/may have/could have + past participle for an uncertain past possibility: They might have missed the train.
- have is essential here — it can never be dropped: must have left, not must left.
The most common mistake is using mustn't for a negative deduction: mustn't be tired is wrong — mustn't means forbidden, not 'probably not.' Use can't/couldn't instead: can't be tired. A second frequent error: dropping have in past deduction — he must left is wrong; it needs must have left.
Look for must/can't/might/could/may followed by be or have + past participle used to draw a conclusion rather than state a fact directly, and context clues like judging by, it looks like, apparently, or a piece of evidence mentioned just before or after — signs that someone is guessing, not simply reporting.