This is not to say that I don't think that it isn't illogical

« previous post | next post »

In November of 2000, Ted Briscoe interviewed Gerald Gazdar about the history of "Generalized Phrase Structure Grammar". In the course of that interview, Gazdar said:

That is not to say that I don't think that corpus work can't be useful, even in theoretical syntax.

,,,by which he meant to say that he thinks that corpus work can be useful, even in theoretical syntax.

If you apply your intuitions to the problem of building this sentence up out of its parts, I think you'll find that what he said actually ought to be logically the opposite of what he meant, at least in the forms of English that lack negative concord.

For example, consider phrases of the form "This is not to say that I don't think that X". When X is an un-negated phrase, the result conveys that the speaker or writer actually does think that X. A few web examples:

This is not to say that I don't think that gold and silver are good wealth storage vehicles in this situation. To the contrary, I think they are excellent for these times.

This is not to say that I don't think that some people play video games to the detriment of their lives and relationships. I do think that happens.

This is not to say that I don't think that a good blow out looks Fabulous, because there is almost nothing sexier.

This is not to say that I don't think that children are products of their environment, I certainly think that can be the case.

That is not to say that I don't think that it's totally worth trying to do. Indeed, there has not been a better time in years to undertake this mission.

That is not to say that I don't think that the whole patent system needs fixing. In fact, I'd argue that it needs to be cancelled completely.

This convoluted litotic way of expressing "X' (or "I think that X") arises because the author is taking a position that might seem to oppose X, and wants to explicitly endorse X in order to prevent a misunderstanding.

So by logical analogy, "This is not to say that I don't think that not-X" should convey that the author is explicitly endorsing not-X. In this case, Gerald would then be endorsing the view that "corpus work can't be useful", which is exactly the opposite of what he meant (as the context makes clear — see below).

However, other real-world examples routinely agree with Gerald's form-meaning mapping. People who say or write "This is not to say that I don't think that not-X" are generally endorsing X, just as if they had said or written "This is not to say that I don't think that X", or "This is not to say that I think that not-X". Some web examples:

This is not to say that I don’t think that Jackie Robinson shouldn’t be honored appropriately…I think that April 15 should be a national civil rights holiday.

This is not to say that I don't think that chi blocking lessons shouldn't be taught to both benders and non-benders, I think that that's a great idea.

This is not to say that I don't think that the S-Tronic isn't good – to the contrary, I think it is a great transmission but I think it is better suited to a sporting vehicle rather than a family truck.

This is not to say that I don't think that Scots doesn't have a future. I believe it does and if it was to be lost completely, I would be devastated.

That is not to say that I don't think that books, reasoning, conversations and understanding aren't helpful. They are. But they don't fix things.

That is not to say that I don't think that Louis doesn't have a vested interest, I am sure he does as he works for a private company.

This is not to say that I don't think that those so described shouldn't be brought to court, as I think they are even more reprehensible than the warped and twisted sociopaths who derive pleasure from forcing themselves on the weak, the vulnerable and the powerless.

That's not to say that I don't believe that men in prison who have committed violent crimes can't turn around. I believe they can see the harm of their past deeds and embrace a better life.

That's not to say that I don't believe that Catastrophic Climate Change (CCC) isn't a real thing; it is.

In other words, this is one of the forms of misnegation where all — or nearly all — attested examples seem to be "wrong" or "illogical", exhibiting one too many (or one too few) negatives. Thus as I wrote in "Why are negations so easy to fail to miss", 2/26/2004, discussing phrases like "Although his attendance at school was still very poor, Stanley never failed to miss a movie at the local theaters":

It seems to me that there are several different psycholinguistic questions here: why do most people not even notice the problem in sentences like this? why do people stick in the extra "fail to" in the first place, given that the sentences mean what their authors intend if they just leave it out? why are uses of "fail to miss" so often accompanied by an additional negative ("doesn't fail to miss", "never failed to miss", etc.)? and why do people hardly ever use "fail to miss" to mean "fail to miss"?

There are several possible directions for answers to questions like this, as discussed here:

  1. Our poor monkey brains just can't deal with complex combinations of certain logical operators;
  2. The connection between English and modal logic may involve some unexpected ambiguities;
  3. Negative concord is alive and well in English (or in UG);
  4. Odd things become idioms or at least verbal habits ("could care less"; "fail to miss"; "still unpacked").
In the case under discussion, I lean towards a combination of (1) and (3).

The context of Gazdar's sentence: Briscoe asked this question:

Sticking with the intellectual history and attitudes to generative grammar, in the review (Gazdar 1976) of The Form of Language (Sampson 1975), you endorsed the methodology, sometimes pejoratively known as the `armchair methodology', of generative research work. Much of that review is an attack on Sampson for adopting a behaviourist but ultimately incoherent stance. Would you also still stick with that?

In a 1975 book, Geoffrey Sampson argued that the methodology of "armchair linguistics" — relying on the researcher's intuitions as data — is fatally flawed, and should be abandoned in favor of "corpus linguistics", which depends only on the distribution of attested forms in real-world language use. Gazdar's 1976 review of Sampson's 1975 book defended intuition-based methods, and attacked Sampson's position as incoherent. In 2000, Briscoe asked Gazdar for an update on those issues. Here's the next portion of the conversation, including the sentence under discussion:

GG: Yes, oh yes. That is not to say that I don't think that corpus work can't be useful, even in theoretical syntax. For example, the discovery by Partee of a crossing coreference sentence in the Los Angeles Times was a classic case where attestation provided considerable additional support for the existence and grammaticality of what would otherwise be a very exotic class of example. If you don't have an attestation for an example like that then, unfortunately, even though such examples are clearly grammatical you leave the way open for various sorts of linguistic nutter to claim that they are not and that humans have no intuitions about them. If you can find one, one ought to embrace it.

EJB: It is the more liberal view that intuition is a very good starting point but not the only way to do this.

GG:  Yes. But if you are concerned with John loves Mary then there is just no point in searching a corpus for John loves Mary or any of its counterparts – you'll find thousands of them. And if you are working on an exotic language when you know nothing about the grammar at all, you need to know about the John loves Mary sentences and the way to find out about them is to talk to an informant. My view on that is essentially the traditional one – this is how pre-Chomsky linguists would work with informants. There is nothing novel about it – Chomsky didn't invent the methodology. He simply gave a defence of it in terms of the superior status of intuitive views of one's own language and so on. My defence of it is much more pragmatic really – or it would be now.

[I was led to the EJB-GG interview, which I hadn't previously read, by Rod Johnson's comment on Geoff Pullum's post about "The language of phone numbers".]


  1. Brian said,

    November 3, 2012 @ 11:37 am

    Dang, I was really hoping that the quadruple-negative of the post's title was something that had actually been observed in the wild.

  2. Jussi Piitulainen said,

    November 3, 2012 @ 1:48 pm

    (Reacting to the end-note.)

    It's not so easy to find short stand-alone sentences like "John loves Mary". Wherever I look, people use sentences that are quite a bit more complicated, and sentences that are tied to the discourse in interesting ways. For example, the short sentence-level units in the present article are Gazdar's "Yes, oh yes." and "Yes."

    When I search the web for actual uses of "John loves her" I find it embedded in more complicated utterances like "john loves her so much that …", "So Mary's belief that John loves her is true, but …" and "Mary knows that John loves her because she inherited money." Similar with "he loves Mary". The actual "John loves Mary" appears to be a movie title and is harder to find otherwise.

    I suppose one can elicit such stand-alone sentences from an informant, and there may be good reasons to do so, but they don't seem to me to be what people (mostly) use. They are an abstraction, something you find in a laboratory.

    (I don't consider myself a linguist, though I do have some such education. I did read the GPSG book. I also read some things by John Sinclair on corpus data. That's where I learnt to notice a striking difference between "John loves Mary" and the things that people really say.)

  3. Rod Johnson said,

    November 3, 2012 @ 2:16 pm

    Glad you found something of value in that interview, Mark. I found it fascinating. Gazdar is nothing if not uninhibited (see what I did there?) about expressing his opinions.

    [(myl) Indeed. I was generally aware of his opinions, but not of this valuable and interesting interview expressing them for posterity.]

  4. Lazar said,

    November 3, 2012 @ 3:33 pm

    I think that I can just manage to process "that's not to say that I don't think" in the wild, but at the level of negation found in "that is not to say that I don't think that X can't Y", my brain just shuts down and treats it as gibberish.

  5. R. Sabey said,

    November 3, 2012 @ 4:04 pm

    I couldn't disagree with you less.

    If negative concord were alive and well in English, then "That is not to say that I don't think that corpus work can't be useful" would mean the same as "That is to say that I think that corpus work can't be useful", with its single negative reinforced by being expressed 3 times. That is, it would mean what it actually does mean anyway.

    [(myl) But in general we wouldn't expect negative concord to spread negativity out of a subordinate clause through two sucessively higher matrix clauses…]

  6. J.W. Brewer said,

    November 3, 2012 @ 4:14 pm

    Perhaps this was a transcription error and spectrographic analysis of the audio would reveal an additional Bidenesque reduced syllable that would resolve the seeming difficulty? Another possibility is that the transcriber's poor monkey brain silently added one of the negations because that made the resultant sentence seem to make more sense to him than what Gazdar actually had uttered, thus changing a spoken instance of your first set of examples to a written instance of your second set.

  7. Pharmamom said,

    November 3, 2012 @ 4:39 pm

    I thought it should be read as a negative. My brain just heard negative x3 and decided, negative. But just because examples of multiple negatives like this are seemingly always positive, I made a choice to infer positive from it. How does that make sense?

  8. Ellen K. said,

    November 3, 2012 @ 5:51 pm

    Reading that sentence, I simultaneously understood it both ways. That is, as literally saying one thing, but meant to say another. Somehow, even without context, I understood that it meant he think corpus work can be useful, even though I understood it to literally be saying he thinks corpus work isn't useful.

  9. Rod Johnson said,

    November 3, 2012 @ 8:02 pm

    Reading this, I actually spent some time trying to figure out what "negative oncord" meant.

  10. Joe Green said,

    November 3, 2012 @ 8:36 pm

    @R. Sabey

    I couldn't disagree with you less.

    This is ironic, no? Or is this a standard idiom for some people? (Which would be as confusing as "I could care less", which of course means exactly the opposite of what it appears to mean.)

  11. Lazar said,

    November 3, 2012 @ 8:37 pm

    @Rod Johnson: The idea of negative concord, of course, renders questions of "double" or "triple" quite meaningless, but I do enjoy seeing four in a row in the example sentence "I ain't never said nothing to nobody."

  12. Rod Johnson said,

    November 3, 2012 @ 8:42 pm

    @Lazar: "oncord"

  13. David Morris said,

    November 3, 2012 @ 11:13 pm

    This is why I recommend phrasing things positively, wherever possible.

  14. Faldone said,

    November 4, 2012 @ 8:02 am

    I couldn't disagree with you less.

    Or "I couldn't fail to disagree with you less."

  15. Eric P Smith said,

    November 4, 2012 @ 9:28 am

    Mark says he leans towards a combination of the "monkey brains" explanation and the "negative concord" explanation. He may be right.

    It would be interesting, however, to have the views of those who themselves use such misnegations. If it were put to Gerald Gazdar what he had said, would his response be, "Yup, I should have said, 'That is not to say that I don't think corpus work can be useful'," or would his response be that he had used a negative concord that is acceptably idiomatic? And how would his response compare with the responses of others who use such misnegations and who are not linguists (but who have enough intelligence and/or interest in linguistics to understand the point)?

  16. Rod Johnson said,

    November 4, 2012 @ 10:31 am

    (Just so nobody thinks I'm insane: my comment about "oncord" above stemmed from when the lede for this entry actually said "oncord," and I took it as a clever technical term (you know, like "sluicing") and spent some time trying to guess what it meant. Anticipatory concord maybe?)

  17. Mary Apodaca said,

    November 4, 2012 @ 12:11 pm

    D Morris: I agree. I've read somewhere that negative statements are just more difficult to understand.

  18. rhhardin said,

    November 4, 2012 @ 1:23 pm

    A Bob and Ray response, long ago

    "I hesitate to say I don't disagree with you."

    I myself have a usage peeve with "Few would disagree that…" on the grounds that it makes no sense. You can't disagree that something. The other guy has to come into it.

  19. Faldone said,

    November 4, 2012 @ 2:32 pm

    You want sense?! This is language, not sense.

  20. D.E. said,

    November 4, 2012 @ 4:16 pm

    What *is* it with the obsession with modal logic and language, anyway? We use repetition for emphasis, and repetition of negations counts as emphasis as well. That is, instead of -1 * -1 = 1, people often use -1 + -1 = -2. The more negations, the bigger the negative emphasis. The more protestations that something is totally, absolutely, no-other-ways-about-it true, the more the positive emphasis. I suppose logicians' ranges of possible truth values are only from 1 to -1, and doesn't encompass numbers of greater magnitude.

  21. Brian said,

    November 4, 2012 @ 6:04 pm

    Tone makes everything clear. There's nothing that syntax can do that tone can't remedy.

  22. Faldone said,

    November 6, 2012 @ 8:05 am

    A friend of mine once said, "Language isn't a set of rules; it's a set of behaviours."

  23. Rod Johnson said,

    November 6, 2012 @ 1:49 pm

    Parole is a set of behaviors; langue is knowledege of how to behave (if you happen to believe in that particular distinction).

  24. Svafa said,

    November 6, 2012 @ 3:59 pm

    Like Ellen K., I understood both, though not quite simultaneously. I intuitively understood the intended meaning while reading the sentence, and then parsed the literal meaning on completing the sentence.

    While I don't disagree with the combination of 1 and 3, I also lean toward some influence from 4; though I wonder if that might be more a sign of 3's subtle influence at large.

RSS feed for comments on this post