1. I don't think they can win.
  2. I know they can't win.

In the " that-clause", why does the first example use the affirmative, yet the second one use the negative? I guess the verb "can" or "can't" is bound by the subject and the main verb in the main clause, but I'm not sure.

  • 1
    @Araucaria-Nothereanymore. Well, I know that, of course! The OP asked why the verb was in the affirmative. May 6, 2021 at 7:56
  • 2
    It's interesting that you've wisely changed the cognitive-domain verb in your example pair. 'I think [that] they can't win' sounds at best unusual. But it still introduces a complication: we're not comparing like with like. While (2) is hyperbole (think Foinavon, Leicester City), it probably means estimating that one is 95+% certain of the outcome. (1) is far less confidently stated. // Using informal reckon as the cognitive-domain verb, which (unusually) works in both variants, I'd say the modal (confidence indication) levels are identical. May 6, 2021 at 11:50
  • 2
    Syntactically, there's nothing wrong with saying "I think they can't win". Nor is it "unacceptable" phrasing. It's just idiomatically less likely than "I don't think they can win". May 6, 2021 at 14:58
  • 1
    Yes, I'm quite sure that syntactically there's nothing wrong with I think they can't win. But you can just compare I claim [that] this question is not useful and I don't claim [that] this question is useful to prove to yourself that it usually makes a significant difference to the meaning whether you negate the clause before "that" or after it. May 6, 2021 at 15:20
  • 2
    This phenomenon of negating the matrix (main) clause to imply the negation of the subordinate clause only applies to verbs relating to intention, epistemic stance or opinion (thought or belief), or those which can be used performatively for advice. It is sometimes called subordinate negation implication. Your selected answer (which is goodish) deals with only one of these. However, more importantly, it doesn't describe the most important factor in the feasibility of such negations: The verbs concerned must be "medium strength" verbs. This terminology is from CGEL. (continued ...) May 6, 2021 at 23:56

1 Answer 1


There is a syntactic rule called Negative-Raising that operates on a subset of English verbs having to do with perception, thought, and belief. This rule, with these verbs only, allows equivalence between a negative in a complement clause and a negative in the main clause. Since think is one of these verbs, the following two sentences are equivalent in meaning:

  • Bill thinks that the Orioles won't win the Series.
  • Bill doesn't think that the Orioles will win the Series.

The effect of the rule is that the negation from the that-clause seems to rise up to the main clause, where in fact it doesn't apply -- Bill is thinking, but that they'll win is not what he's thinking. The literal interpretation some try to push -- that Bill has no opinions -- is not fluent English; nobody talks or writes that way, because that's not what think normally means.

However, claim is not one of these verbs, and thus the following two sentences do not mean the same thing:

  • Bill doesn't claim the Orioles will win the Series.
  • Bill claims the Orioles won't win the Series.

These do have separate senses -- in the first, Bill's belief is unknown, but his claim is denied. In the second, Bill's claim is asserted. But then claim is a different verb from think -- a claim is public and can be witnessed, but a thought is private and can be denied or imagined. That's what allows Neg-Raising in the first place. Every verb has a different grammar.

  • It's quite complicated but understandable and that's what I'm expecting. Can you possibly list out some specific verbs applied to "a subset of English verbs having to do with perception, thought, and belief "? Let me guess some; are they " think, guess, believe, suppose, expect, reckon, fancy, consider "?
    – user421993
    May 6, 2021 at 18:51
  • 1
    It's easy enough to search google for neg-raising verbs: google.com/search?q=neg+raising+verbs May 6, 2021 at 19:35
  • 2
    @ John Lawler, At present, we aren't able to reach GOOGLE due to strict censorship in PRC. They shield it. Thanks all the same.
    – user421993
    May 6, 2021 at 19:54
  • 1
    Sorry about that. Yes, the ones you suggest -- think, guess, believe, suppose, expect, reckon, fancy, consider -- all allow neg-raising. As do seem and appear, but there it's the subject complement that contains the negative; seem and appear are one-place predicates that take clausal subjects, and require either extraposition or subject-raising to move the heavy subject to the end. Any verb that refers to the same things will work the same way; but verbs of speech like claim don't. May 6, 2021 at 20:12
  • 2
    It's a very common rule; I've discussed it here quite a lot. There are other phenomena one can search for, as well as some pre-set searches like the one above. May 6, 2021 at 20:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.