Gov. Cuomo and our poor monkey brains
« previous post | next post »
My latest reader response for The New York Times Magazine's On Language column tackles a turn of phrase that has come up on Language Log many times: cannot be underestimated. The occasion is New York Gov. Andrew M. Cuomo's inaugural address earlier this month, in which the governor used the magic phrase twice (and talked about "underestimating" a third time without the cannot). I give the requisite shout-out to Language Log, of course.
There are opposing schools of thought for why “cannot be underestimated” almost always gets used in this topsy-turvy fashion. Mark Liberman, a linguist at the University of Pennsylvania who writes for the group blog Language Log (to which I also contribute), chalks up the problem to the inability of our “poor monkey brains” to handle multiple negatives properly. Similar “overnegation” can be found in such phrases as “I miss not having you around” (meaning “I miss having you around”) and “I never fail to miss an opportunity” (meaning “I never miss an opportunity”). Liberman sees such examples as evidence of an overall “temptation of overnegation” in English usage.
Really, it was just an excuse to work a reference to our "poor monkey brains" into The Times. You can read the whole thing here, and see Mark's master list of posts for further discussion on this and related misnegations.
Paul Zukowski said,
January 21, 2011 @ 11:31 am
Monkey not see, not do.
Amy Stoller said,
January 21, 2011 @ 11:53 am
Surely you mean "underovermisnegations"?
Lazar said,
January 21, 2011 @ 11:54 am
I cannot fail to disagree with Prof. Liberman less.
The Ridger said,
January 21, 2011 @ 12:00 pm
And "can" is of course a modal of permission, while "cannot" is a modal of denial of permission (despite all efforts to stamp it out, the usage 'can I…?' persists, because all modals have more than one meaning. … maybe I'd better say "most modals").
So "this cannot be underestimated" = "this must not be" / "one should not" is certainly a plausible interpretation.
James said,
January 21, 2011 @ 12:01 pm
I favor Ben's alternative explanation (deontic 'cannot').
Though come to think of it, there's no reason to think of the two explanations as exclusive. Maybe some people say 'cannot be underestimated' with deontic 'cannot', and some get confused by the two layers of negation.
What's a good test?
[(myl) I'm mildly skeptical of the modal-logic argument, which was first put forward (as far as I know) by Barbara Wallraff in the Atlantic some years ago. See "'Cannot underestimate' = 'Must not underestimate'?", 11/6/2008, for
some discussionmore discussion than you probably care to read.]Dan Lufkin said,
January 21, 2011 @ 12:10 pm
The Observer recently said that Highland Park [Scotch] is a "real classic which never fails to disappoint". Retraction here.
richard howland-bolton said,
January 21, 2011 @ 12:15 pm
Looking back on my own life I find that “I never fail to miss an opportunity” tends to mean “I never fail to miss an opportunity”.
Will said,
January 21, 2011 @ 12:38 pm
"Impossible to underestimate" gets 145,000 google hits. It clearly isn't used in any deontic sense. Admittedly, it's often used correctly–but not always.
[(myl) An excellent point — also made here.]
Marcus said,
January 21, 2011 @ 12:46 pm
And yet other languages, like japanese, handle double (or more) negatives just fine. For example, the しなければならない construction ("can't not do"), implying a positive action ("must").
The Ridger said,
January 21, 2011 @ 12:57 pm
"Can't not do" is also common in Slavic languages, though they have a different particle for "reinforcing negation" (such as "никто мне не сказал это, nobody didn't tell me"), ни [ni], than for "actual" negation ("он не мог не сделать это, he couldn't not do it"), не [ne] (examples from Russian, but similar in others). But that isn't really the same – that's two negatives acting on each other, while "can't underestimate" isn't two negatives – it's just the wrong verb (in many cases).
re Ben's suggestion, which I like: it can't apply everywhere. Things like "I miss not having you around" clearly have no modality to be ambiguous.
GeorgeW said,
January 21, 2011 @ 1:56 pm
So who says language must be logical? Isn't that the prescriptivist argument against 'double negation?' That was the explanation I was given as a child: If 'she don't want none' then logically 'she must want some.'
[(myl) Some aspects of language — morphology, for example — are typically quasi-regular, and attempts to insist that they ought to be "logical" are misguided. But on a larger scale, language works because it's compositional — the (literal) meaning of larger messages is a predictable function of the meaning of their parts and the way that the parts are combined. In languages with "negative concord" — more of the world's language than not — the interpretation of "she don't want none" as the negation of "she wants some" is how the normal compositional process works, and thus is "logical" in the terms of reference that matter.
When (what seem to be) the normal compositional processes don't work, then we have a puzzle to understand and explain. The expanation for "she don't want none" is relatively simple — English used to be a negative concord language, and non-standard varieties of it still are. The explanation for "could care less" is (I believe) that it's become an idiom, probably as the result of a process called "negation by association", discussed at length here and here.
The explanation of "impossible to underestimate" is not so clear. As I've explained at length in the linked posts, my best guess is that it's because the combination of negation, modality, and scalar predication is hard to process. Just saying that "language doesn't have to be logical" is not helpful at all, since it doesn't tell us why (for example) "impossible to underestimate" is problematic, but "impossible to understand" isn't.]
Sometimes the
MattF said,
January 21, 2011 @ 2:15 pm
I'm pleased you said 'topsy-turvy' rather than the weirdly-non-exceptional 'head over heels.' After all, your head is normally over your heels, so the standard phrase is a case of over/under-somethinging, right?
mgh said,
January 21, 2011 @ 2:39 pm
monkeys overnegate?
[(myl) Monkeys are even worse at logic problems than humans are.]
J Lee said,
January 21, 2011 @ 4:47 pm
I think I heard John McWhorter claim that every non-standard dialect of English utilizes double negatives to some extent.
And the explanation for "could care less" from The Language Instinct is that it's simply sarcastic.
[(myl) It may sometimes be for some people, but I'm extremely skeptical of this hypothesis, for reasons explained at length in the posts indexed here.]
For my part, 'cannot be underestimated' is and has always been more of an idiom (with a rhetorically complex meaning like 'it would so defy reason to underestimate X that it simply doesn't/won't happen'), so Cuomo's usage did not seem illogical.
[(myl) One reason to doubt the "idiom" hypothesis in this case is that all rough paraphrases work exactly the same way: "impossible to underestimate", "hard to underestimate", "difficult to underestimate", etc. In contrast, note that paraphrases of "could care less" generally don't work: "It's possible for me to care less", "It would be conceivable for me to care less", "I'm able to care less", etc. are at best jokes.]
Sili said,
January 21, 2011 @ 5:36 pm
Iono – I haven't seen the macaques' report on Mark Hauser yet.
Josh McNeill said,
January 21, 2011 @ 7:49 pm
Could it have anything to do with using under- as opposed to a stronger negation? For me, this barely registers as a double negative even after thinking about it for a while.
[(myl) "Under-" in underestimate isn't a negation at all. "Underestimate" is a kind of scalar predicate, indicating that an estimate of something is below the true value of the thing estimated. As explained ad nauseam in the linked posts, the confusions under discussion seem to be especially likely with expressions that involve negations, modals, and scalar predicates — and sure enough, "cannot be underestimated" involves one of each: a modal (can), a negation (not), and a scalar predicate (underestimate).]
Or maybe underestimate registers differently for some people than a word like nowhere, for instance? I've never thought of the word as 'not estimated'. It sounds strange to say 'I estimated him' but normal to say 'I underestimated him', as if they aren't opposites at all.
Ellen K. said,
January 21, 2011 @ 8:30 pm
While I kinda see the point about can and cannot being deontic (new word for me :)), still, it doesn't seem to me like when people say "cannot be underestimated" they are putting themselves in the place of saying what someone should do or has permission to do.
Especially with that passive construction. Like, I can imagine saying (though not writing) "We cannot underestimate X" to mean we should not underestimate X. But I cannot imagine "X cannot be understimated" to mean "we/you should not underestimate X".
I can see the deontic meaning having an influence though. Although, as has been pointed out here the same thing happens with phrases without that ambiguity.
James said,
January 21, 2011 @ 9:23 pm
Thanks for the link to the old cannot/must not post. I missed that one the first time.
I don't think we need a lesson in modal logic to see how deontic 'can' makes sense out of the proffered 'cannot be underestimated'. I think 'can' in general tells us that the matrix is consistent with something or other. In "nothing can go faster than the speed of light", it's the laws of nature. In "You can't drive over 65 mph on Route 95" it's the laws of Pennsylvania, etc. The latter is deontic, but not ethical, and the same might be true of 'cannot underestimate'.
Will's point about 'impossible' is a good one, I agree. If the "error" rate for "impossible to underestimate" is the same as the rate for "cannot underestimate", then I concede that the explanation for the latter is not the deontic one. But are those rates the same?
Ellen K. said,
January 21, 2011 @ 9:36 pm
James, but what about the distinction between "cannot underestimate" and "cannot be underestimated". If we said "65 miles per hour cannot be driven over", that, unlike "You can't drive over 65 miles per hour" reads as talking about what's not possible, not what's not allowed.
Chris Travers said,
January 21, 2011 @ 10:38 pm
One thing worth noting is the fact that Old English uses multiple negatives as intensifiers. So rather than a double negative making a positive, it just makes it more negative. In many other languages (including Spanish btw), agreement by negative is expected.
I think the issue here is close to what the article claims in that our brains treat multiple negatives as being in agreement rather than opposition. Moreover it is impossible to imagine a negative (try to imagine that the dog does not bark). So instead all we can do is negate a positive.
Negations of negatives make fine mathematical and logical sense, but they don't seem to match how our brains are designed to think.
Skullturf Q. Beavispants said,
January 22, 2011 @ 12:08 am
My $0.02 is that it's "can't" meaning "shouldn't", which is very common.
"You can't address the Dean by his first name." "You can't wash those whites and those colors together." And so on.
John Walden said,
January 22, 2011 @ 2:16 am
Sometimes we seem able to process double negatives successfully. I remember reading this somewhere, without it causing me any difficulty in the context, about moving to a house on the site of an old windmill; Something like:
'Now I know why there was a windmill here. It never isn't windy.'
Obviously the first sentence helps. I wondered at the time how my Spanish speaking students would process it.
PS My favourite double yolker when over-egging the negations is "unendless(ly)", and its rarer sibling "unceaseless(ly)".
Xmun said,
January 22, 2011 @ 3:49 am
There's also "irregardless", about which R. L. Trask boldly writes: "There is no such word: write _regardless_, not *_irregardless_." See his Mind the Gaffe: The Penguin Guide to Common Errors in English (2001), p. 161.
iching said,
January 22, 2011 @ 7:55 am
The Google Ngram (English corpus) trends for cannot be overestimated, should not be underestimated and cannot be underestimated are interesting. The decrease in the frequency of cannot be overestimated since it peaked around 1920 has been more than matched by the rapid increase in should not be underestimated, and to a much lesser extent by cannot be underestimated. In 1920 the frequencies were about 40, 3 and 1 per million; in 1960 they were about 19, 20 and 5 per million; and in 2000 about 16, 50 and 21 per million respectively.
[(myl) Your arithmetic is wrong. 0.000020% is 0.2 per million, not 20 per million. The original figures are percentages, not proportions, so they need to be multipled by 10^4, not 10^6, to get frequencies per million.
I blame the very unfortunate choice of y-axis scale on the part of whoever designed the Google n-gram viewer. Still, I think that common sense should warn us that "cannot be underestimated", common as it is, could not possibly have reached a frequency of 20 per million, a neighborhood where (according to the BNC and similar sources) we find words like talk, hit, summary, elected, statistical, etc.
Whatever the y-axis values, the striking rise in "cannot be underestimated" (since 1930, and especially since 1960) is, well, striking. But without knowing what the mix of sources over time is, it's hard to be certain what this means. We might be looking a change in language use (and we probably are) — but we might also be looking at a change in the mix of books in the underlying corpus.
It's quite plausible that there's a component of sampling change in the picture. A search of the COCA corpus shows that "cannot be underestimated" is about 10 times more frequent in academic writing than in fiction, and about 5 times more frequent in academic writing than in newpapers. Thus the n-gram viewer's historical growth in "cannot be underestimated" may track the historical growth in academic publishing, and a corresponding increase in the academic proportion of the underlying corpus.
There are some other quantitative puzzles here. The COCA "fiction" section shows a frequency of about 0.04 per million for "cannot be underestimated", but the "fiction" section of the Google n-gram viewer shows a frequency in recent years of about .08-.10 per million. This difference is part of a pattern that leads me to suspect that some fraction of the "fiction" items in the Google collection are mis-classified — the metadata in the collection behind the n-gram viewer is much improved over this, but it's apparently far from perfect.
As another check, take a look at the returns from a general Google Books search for "cannot be underestimated" In the first 100 hits, I find only 1 work of fiction (a fantasy novel by R.A. Salvatore).
Compare this to "talk", which is roughly twice as common the n-gram "fiction" returns as in the general n-gram viewer, and has 17 fiction hits in the first 100 returns from a general Google Books search. Or even compare "elected", which is about half as common in Google n-gram "fiction" results as in n-gram "English" results — and about 1/7 as common in COHA "fiction" vs. COHA overall frequency — but has 6 works of fiction in the top 100 Google Books searches. Something about all this doesn't add up, and I suspect that it points (among other things) to a significant fraction of mis-classified works in the Google n-gram "fiction" category.]
Alexey said,
January 22, 2011 @ 8:41 am
There's the same problem in Russian: translation of "cannot be underestimated" is "нельзя недооценить" and it's being used quite often. The correct phrase is "нельзя недооценивать" which looks alike but have the opposite meaning: "shouldn't be underestimated". There's also an idiom "трудно переоценить" meaning "hard to overestimate" which I think is a preferable choice in Russian.
[(myl) There seems to be something similar in French for "peut pas sous-estimer", and in Spanish for "no podemos subestimar", etc.. The fact that "cannot underestimate" translates so readily is another piece of evidence, I think, that it's not just a random idiom.]
iching said,
January 22, 2011 @ 8:58 am
Oops. My bad. I blame my "monkey brain". Please replace my "per million" with "per 100 million" everywhere.
MJ said,
January 22, 2011 @ 11:59 am
The n-gram viewer perplexes me–maybe someone can explain this to me. The graph for "cannot be underestimated" shows a steep turn up between 1990 and 2000, but the raw data doesn't seem to me to add up. If you look at the results for the 1992-1995 slot, it tells you there are 5020 hits, although if you page through, you see that in fact there are only about 540 hits. For 1996-1998, the results are virtually identical–it lists 4,060 hits but when you page through, you find only about 530. Given these numbers, I would expect a more gradual upward turn. Could the fact that the 1992-95 slot has one more year's worth data account for why the upward turn is so steep?
[(myl) A crucial fact is that the corpus behind the n-gram viewer is a relatively small — and not necessarily representative — fraction of the collection being searched by Google Books itself.
Because the basic Google Books date metadata was so bad, and there also were various other issues having to do with OCR quality and so on, the people behind the n-gram collection went through a long process of fixing the n-gram corpus. I believe that this was entirely done by selecting some books and eliminating others, in an iterated but always automatic way, and did not involve any human curation of works selected to be in the corpus. As a result, the n-gram corpus is not representative of the Google Books collection (which in any case has changed considerably since the process started), and of course the Google Books collection itself makes no special effort to be balanced or representative from the point of view of genre or region or whatever, but (as it should) simply includes everything they have been able to get/scan.
A second relevant fact is that the counts in Google Books searches — like the counts in Google searches in general — are not quantitatively reliable, and may well be off by orders of magnitude. These counts are typically created by an approximate process of extrapolation from the results of doing joins on the lists of top-ranked hits for sub-parts of your query. Thus when you search for "X Y", you may be get a count that is based on looking at how often the top N pages for X and the top N pages for Y also contain "X Y", where N might be something like 10,000 (but definitely not billions or trillions). If the top N pages are not representative of the rest of the collection being searched, this estimate can be arbitrarily far off.
The n-gram viewers percentages are, however, based on exact counts — but on exact counts of a collection with no specified quantitative relationship to the overall Google Books collection.]
Mark P said,
January 22, 2011 @ 1:22 pm
A previous comment must have been removed because it was offensive. If so, I apologize. My excuse is that I usually post quickly (and sometimes wish there were an option to edit a comment). What I was trying to say was that some of the phrases being discussed might be explained by some reasonable interpretation (for example, "cannot be underestimated" makes sense if it means "should (or must) not be underestimated"), but in general that seems to be a stretch to me. So what I mean by calling such an explanation a just-so story is that it seems to be misguided to try to force a literal, word-by-word interpretation of such a phrase so that it ends up the same as the speaker intended, despite the fact that it seems so clearly not to mean that. I didn't intend to imply that any attempt to explain them is misguided.
MJ said,
January 22, 2011 @ 1:32 pm
Thanks for that explanation. I specifically didn't know that there's no exact correlation between the corpus I can access by clicking on those year ranges below the graph and the percentages shown in the graph.
Ryan said,
January 22, 2011 @ 1:52 pm
Alexey,
Is it possible that the use of "нельзя недооценить" for "нельзя недооценивать" has something to do with speakers deciding to mark the semantics of aspect and modality differently? Or do you only see this confusion for this pair of verbs?
Alexey said,
January 22, 2011 @ 3:05 pm
Ryan,
the confusion actually comes from the word "нельзя" which can mean "cannot", "should not", "better not to" etc depending on the context. You can check this link: http://lingvo.yandex.ru/%D0%BD%D0%B5%D0%BB%D1%8C%D0%B7%D1%8F/%D0%BF%D0%BE-%D0%B0%D0%BD%D0%B3%D0%BB%D0%B8%D0%B9%D1%81%D0%BA%D0%B8/
Here context is made by modality of the verbs and it makes a difference. If we consider for example phrases "нельзя разрушить" and "нельзя разрушать" the first will mean "it's impossible to destroy (smth)" and the second will mean "it's not allowed to destroy (smth)".
If we get back to "estimation" example we'll see: "нельзя недооценить" considers the physical possibility and the exact translation is "cannot be underestimated". On the other hand "нельзя недооценивать" actually means "не стоит недооценивать" which can be translated as "it's not recommended to underestimate (smth) because it can cost you a lot".
My explanation could be rough (or even wrong) since I'm not a linguist, just a native speaker.
Dan Parvaz said,
January 22, 2011 @ 3:13 pm
Speaking on behalf of Jews — masters of negativity — everywhere, I think we use these overclocked negatives to mean exactly what they do mean. Hence the old saw that Arafat (or "The Arabs" or… okay, this is a whole 'nother topic. Just roll with it) "Never missed an opportunity to miss an opportunity."
Ooh Eee, Ooh ah ah…
Ryan said,
January 22, 2011 @ 6:00 pm
Alexey,
I know that нельзя is either "can't" or "shouldn't" based on context, but what I'm trying to ask is if you've seen similar confusion between perfective and imperfective with other verbs, or if it's just недооценить/недооценивать.
Horsecow said,
January 22, 2011 @ 10:43 pm
This discussion reminds me of an episode of the Simpsons:
WARDEN: How's your back?
HOMER: I can't complain. (points to sign reading "no complaining")
WARDEN: Oh, that's just for the prisoners.
HOMER: Ohhhh… My back hurts so much… And my job's so unfulfilling.
The "can't" in the common phrase "can't complain" seems similar to the "can't" in "can't be underestimated". Everyone understands it's not literally impossible to do x, it's just unwise/poor form. Is this really a confusing usage?
[(myl) If this idea interests you, please see "'Cannot underestimate' = 'must not underestimate'", 11/6/2008, linked and discussed above, where your suggestion (which has also been made by others) is considered in detail and at length.]
Alexey said,
January 23, 2011 @ 4:28 am
Ryan,
I think the same sort of confusion can occur with all transitive verbs. However I've _seen_ such mistakes mostly (or probably only) with недооценить/недооценивать likely because of the multiple negation problem. Another examples don't come into my mind at the moment.
boynamedsue said,
January 23, 2011 @ 1:03 pm
"Can't be underestimated" is right surely. Given can't can imply a value judgement.
"Men can't treat women that way."
=
"It is impossible for men to be right when they treat women that way."
"The role of Polish Airmen in the RAF can't be underestimated."
=
"It is impossible to be right when one underestimates the role of Polish airmen in the RAF."
This is actually a case of grammatically logical phrase being taken on as a set expression free of actual analytic content. The only poor monkey brains here belong to the linguists.
Ellen K. said,
January 23, 2011 @ 2:14 pm
Perhaps there's some variation? Presumably, "Men can't treat women that way", when men do treat women that way but shouldn't, really does work for Boynamedsue. For me, it doesn't. I can say "You can't treat women that way". But when the statement is less direct — talking about people instead of to them — "can't" doesn't work. It has to be "Men shouldn't treat women that way". Saying "Men can't treat women that way" sounds like it's talking about ability.
"Can't" does not equal, "impossible to be right", for me.
boynamedsue said,
January 23, 2011 @ 3:04 pm
It might be variation. I find "Men can't treat women that way." to be stronger than "Men shouldn't treat women that way", the should intimates that not treating women that way would be better, but the can't states absolute refusal to accept the treatment, similar to "must not", but from a person who has no direct authority on the situation.
But I reckon this "error" comes from people initially using "can't" in this way, and the expression has since spread, rather than being an inherent failure in the brain's logic circuits. The presence of the can't makes it stronger than shouldn't, and everybody understands what is meant so… well, no big problem?
Especially when we compare with Spanish frases like "hasta que no vuelvas" ("until you don't return")
[(myl) Please, everybody, if you're interested in this subject, try to learn a bit about the different kinds of modality, their complex alignment with the various ways of expressing modality in English, and the various modal logics that help keep straight how they interact with negation, with other operators, and with each other. This topic is somewhat complicated, and if you try to learn (at least some of) what a few hundred years of logicians and semanticists have figured out about it, you're likely to get confused. (I don't mean to suggest that the logicians and linguistic semanticists have it all figured out, or that observations from outsiders are never useful; but the fact is that there's a real body of knowledge here, and trying to sort out the phenomena without taking account of it is like trying to make sense of orbital mechanics purely on the basis of common sense.)]
boynamedsue said,
January 23, 2011 @ 5:56 pm
[(myl) "Please, everybody, if you're interested in this subject, try to learn a bit about the different kinds of modality, their complex alignment with the various ways of expressing modality in English, and the various modal logics that help keep straight how they interact with negation, with other operators, and with each other."
I don't think this question needs a massive investigation of modal logic. For some speakers of English "Must Not" can be replaced with "Can not" in all the former's range (though the reverse is obviously not true). This is simply a matter of usage, and after reading the older post you linked, the modal logic seems to confuse rather than clarify the issue.
If English speakers consider "must not" to be replaceable with "can not" then the sentence "The importance of logic can't be underestimated" is POSSIBLY logical. Of course, it may not be, as the speaker may understand the "can't" in his utterance to be related to possibility not to obligation.
Skullturf Q. Beavispants said,
January 23, 2011 @ 9:40 pm
Sometimes we use "cannot" to mean that something's impossible.
Sometimes we use "cannot" in a quasi "moral" sense, to mean something like "should not" or "must not" or "ought not to."
And — I think it's clear, as others have stated — sometimes we use "cannot" to mean, essentially, that something's not desirable. I agree with others that "cannot underestimate" is a kind of shorthand for something like "cannot underestimate without inviting bad consequences."
Maybe we're talking at cross purposes, though. I get the feeling that maybe not everyone shares my intuition about "cannot" and so the word might effectively have a different meaning for me than for you.
I'm like boynamedsue above and not like Ellen K. For me, the sentence "Men can't treat women that way" is an utterly unremarkable way of saying "It's bad for men to treat women that way." I wouldn't interpret it as "It's impossible for men to treat women that way."
Boris said,
January 24, 2011 @ 1:26 pm
If something cannot be underestimated, does that mean it is often misunderestimated? In fact, I propose we use the word from now on lest someone points out that English has no single word for "cannot be underestimated".
Mark P said,
January 25, 2011 @ 8:37 am
I heard a use of "cannot underestimate" where the "must not underestimate" explanation is at least plausible, and probably likely. It was a TV show on one of the nature channels about swimming with sharks. The narrator said the danger "can't be underestimated" in a context in which it seemed clear that he meant "should not be underestimated" or "must not be underestimated." A transcript might be useful, but I'm not sure I can find one, even if I can figure out which channel it was on.
Joke said,
February 15, 2011 @ 11:28 am
I don't see any problem with the use of "… cannot be underestimated" to mean "we should not underestimate …".
Definition 1.d. at M-W.com is "be permitted by conscience or feeling to "
Thus I re-work Cuomo's statement as:
"I also saw up close and personal the suffering that our people are facing and the devastating toll that this economy has taken. And it [is not permitted by conscience or feeling to be] underestimated."
It seems to me a common usage. Much like "Behavior x cannot be allowed".