"Can cause" vs. "may cause"
« previous post | next post »
Catherine Saint Louis, "Dessert, Laid-Back and Legal", NYT 5/14/2011:
Remember melatonin? In the 1990s, this over-the-counter dietary supplement was all the rage among frequent fliers, promoted as the miracle cure for jet lag. Now it is back in vogue, this time as a prominent ingredient in at least a half-dozen baked goods that flagrantly mimic the soothing effects of hash brownies — and do so legally. At least for now. […]
“A hangover effect has been reported” with large doses, said Anna Rouse Dulaney, a toxicologist with the Carolinas Poison Center. But she added, “I don’t want to go on the record saying this drug ‘can’ cause respiratory issues, that should be a ‘may.’ ”
Dhananjay Jagannathan, who sent in this example, wondered whether "evidentiality and modality have gotten a bit muddled" here, and suggest that perhaps the distinction between can and may has become biomedical jargon for the difference between an established statistical association ("this drug can cause respiratory issues", i.e. there's good evidence that it sometimes does) and a hypothesis suggested by some scattered or anecdotal evidence ("this drug may cause respiratory issues", i.e. there are reasons to think that it sometimes might).
A quick scan of "can cause" vs. "may cause" in New Scientist suggests that Dhanajay may be right about the pattern of use.
There are 2820 results for "X can cause Y", and the first 10 all involve cases where a published study or other authoritative pronouncement claims an association to exist between X and Y: "Henna tattoos can cause severe skin reactions"; "New roads can cause congestion"; "Oral sex can cause throat cancer"; "Trains generate pressure waves in tunnels that can cause passengers' ears to pop"; "New roads can cause congestion"; "Malaria drug can cause mental disorder"; "Electric shocks can cause motor neurone disease"; "Female genital mutilation can cause infertility"; "Lack of sleep can cause brain damage and affect memory"; "Exposure to pesticides can cause Parkinson's"; "Submicroscopic defects in chromosomes can cause mental handicap".
There are 718 results for "X may cause Y", but these seem to be more variable. Some are certainly examples of hypothetical, weakly-supported or contested connections: "Space debris may cause mysterious ball lightning"; "Lead may cause mystery male infertility"; "Weedkiller may cause amphibian loss"; "Inflammation may cause cerebral palsy"; "Marijuana may cause pregnancies to fail"; "Falling stars may cause rainy days"; "Increased CO2 may cause plant life to raise rivers"; "Expanding waistlines may cause shrinking brains"; "Bioweapons alerts may cause lasting psychological harm"; "Brain defect may cause 50% of cot deaths".
But in other cases, may in the headline does not seem to be associated with a weaker connection. Thus the story under the headline "iPods may cause pacemakers to miss a beat" starts by expressing the same idea with can:
iPods can cause implantable cardiac pacemakers to malfunction, according to a study presented at a meeting of heart specialists on Thursday.
And the evidence supporting the connection seems to have been at least as strong as in many of the "can cause" stories:
The study tested the effect of placing the portable device near the chest of 100 patients with an average age of 76 – all were fitted with pacemakers. Technicians monitored information from an electrocardiogram and from the pacemaker itself, via a telemetry screen, during the trials.
Some form of interference was detected half of the time when an iPod was held 5 centimetres from the patient's chest for between 5 and 10 seconds.
A story running under the headline "Inflammation may cause deadly blood clots" likewise presents evidence at least as good as the evidence in many of the 'X can cause Y' stories:
People who are suffering from a severe infection are more likely to develop dangerous blood clots, a new study suggests. Researchers say that patients with respiratory and urinary tract infections have a temporarily doubled risk of deep vein thrombosis (DVT), a potentially deadly condition in which blood clots block circulation in large veins.
A distinction in evidentiary strength between "can cause" and "may cause" is presumably due to selection of one of the many meanings of can (glossed by the OED as "to have the power, ability or capacity") and one of the many meanings of may (glossed as "Expressing present subjective possibility, i.e. the admissibility of a supposition"). In English and other natural languages, evidentiality has been "getting a bit muddled" with other aspects of modality such as ability and permission for a long time. What seems to be new here is the emergence, in certain contexts, of an association of can and may with graded degrees of empirical support and epistemic strength.
The cited examples are not meant to imply that reports in the New Scientist are always credible evaluations of evidentiary status — see "Envy, Navy, Whatever" (Language Log 10/27/2006), "A Plea to Save New Scientist" (The n-Category Café 9/19/2006), or Blake Stacey's discussion of the EmDrive case at Contrary Brin, for discussion of a couple of five-year-old cases in detail. For some slightly more recent examples, see "Language and personality", LL 6/28/2008; "Pop platonism and unrepresentative samples", LL 7/26/2008;
Brett said,
May 17, 2011 @ 7:58 am
I have noticed a lot of confusing locutions in discussions of risk factors associated with various medical conditions. I think this stems from a poor grasp of probability and basic epidemiological principles; sometimes, the problem is with the writers/speakers themselves, and sometimes the words are badly chosen with the hope they will not confuse a unsophisticated audience.
The example that seems to be most common (and which rankles me the most) is saying, "We have found clear evidence that X can/could/may/might/etc. increase the risk of Y," when in fact they have found clear and convincing evidence that "X does increase the risk of Y." I suspect that the problem is with many people not having a good probabilistic conception of risk. They might hear, "X increases the risk of Y," as, "If you have X, you will also get Y," or, "If you have X, you will most probably get Y." So an additional hedge is included in announcements about this discovery; even though that results in a final claim that is not logically equivalent, it is expected that people will understand what is actually meant.
As I said, sometimes the error in these cases lies with the speaker's misunderstanding of risk and probability. (This is obviously more likely to be true when the speaker is a journalist or PR representative than a scientist.) But I think that many of these unnecessary hedges (and there are certainly many others, such as the one described in this post, with various degrees of frequency and standardization) are phrased as they are purely in an attempt to avoid confusion in an unsophisticated audience. This bothers me for two reasons: It upsets me that researchers take this condescending attitude toward the people they are addressing, supposing them unable to parse a simple, logically correct statement about risk factors; and it upsets me that they may be justified in doing this, because some people probably would misunderstand "X increases the risk of Y."
Spell Me Jeff said,
May 17, 2011 @ 8:39 am
Within the last several months I read a popular-press report (sorry, I don't have a citation) of an epidemiological study that showed a correlation between second-hand smoke and something unpleasant (possibly mental health issues in kids). The article presented the findings BUT ALSO quoted the study leaders saying they have not absolutely established a causal link, even getting them to admit (what you and I consider obvious) that there may be a third, unknown factor responsible for both patterns of smoking behavior and the effect in question.
It was quite refreshing to see a popular article make such a distinction, and I think much more effective than overloading modals with more content than the average reader is likely to extract.
Boris said,
May 17, 2011 @ 8:57 am
I am not a scientist and I picked up the distinction of can vs may as soon as I read the quote. However, if I saw either construction on its own, I probably wouldn't differentiate in this way, but honestly, as a consumer of this information, the distinction is not important.
It's similar in weather forecasts. 40% chance of rain can have two meanings. Either they are 40% sure it will rain or they are fairly certain it will rain in 40% of the current area, but they don't know which 40%. The end result is the same. There is a 40% chance that it will rain here.
William Ockham said,
May 17, 2011 @ 9:06 am
I think "X can cause Y, according to this study" is equivalent to "X may cause Y". If you have a lot of studies that show the causal factor, you can say "X can cause Y, full stop." Although I suppose one very large long-term study might suffice for some people (say the Framingham Heart Study).
Mr Fnortner said,
May 17, 2011 @ 9:06 am
On a slight tangent, I have lately had a problem with "X may cause potential Y" statements, as in "Wildfire Smoke May Cause Potential Health Impacts." They seem to have one too many conditional elements. Either "X may cause Y" or "X has the potential to cause Y." I simply do not know what a potential anything is. What leads to expressions of this sort?
Jerry Friedman said,
May 17, 2011 @ 9:17 am
Maybe I'm used to this (though I don't read much about such studies), because I understand exactly the distinction between can and may that Dhananjay Jagannathan suggested. The meanings he gives strikes me as the normal, non-jargon meaning in this context.
JimG said,
May 17, 2011 @ 9:19 am
Could (Might?) it be that these poor benighted scientists don't know the difference between present and conditional forms of the verb? Could it be that they haven't any evidence to show that y at least once resulted from x, so they're guessing, er, hypothesizing that x MIGHT cause y, and because they slept through their elementary school English classes (Boooooring!), they're conflating the can/may difference with the conditional?
It's too bad that science writers (and peer review panelists) don't judge the technical merits of scientists' work by whether or not the scientist can clearly and precisely describe the subject. Of course, lawyers make a fetish of wordly precision, and look how clearly THEY as a class think about law and society.
I agree with Brett, above, re the condescencion that goes on. It's too bad that we don't demand that the general public be educated to understand precisely what words and basic science and math concepts mean. Just imagine if teachers, and TV writers, and consumers, and our leaders could …. Oh, well. Mutter mutter, grumble grumble …
Pflaumbaum said,
May 17, 2011 @ 9:39 am
Hmm, what is the 'conditional form of the verb' in English?
Henning Makholm said,
May 17, 2011 @ 10:12 am
If I had to define a semantic difference between "can" and "may", it would be be that "can" tends to assert a fact about its subject, whereas "may" qualifies the entire clause.
Under that interpretation "bananas can cause cancer" states as a fact that bananas are in general able to cause cancer (but it is possible that not all bananas actually do so, or that some bananas only cause cancer some of the time), whereas "bananas may cause cancer" states that "bananas cause cancer" is possibly, or possibly not, a true statement about the world.
But this difference (even should it exist outside my head) is clearly subtle enough that it only takes the slightest of contextual clues to make either of the words take on either of the meanings.
pj said,
May 17, 2011 @ 10:12 am
I also (without particular science-jargon background) would generally take 'X can cause Y' to mean that there's established evidence that X causes Y in some cases, and 'X may cause Y' to mean that there's some suspicion or reason to suppose that perhaps X causes Y (and incidentally not ruling out X always causing Y, which 'can' does).
But beyond 'can cause' or 'may cause' intended purely as statements of fact, 'may' also seems to be the modal idiomatically preferred for expressing a caution or warning (e.g. 'excessive consumption may cause laxative effects'), and I think that's what's going on with the iPod and inflammation stories.
John Cowan said,
May 17, 2011 @ 10:35 am
The (now retired) editor of Reuters Health told me, when I was working there, that they were very careful to use "may" in all such claims to avoid giving the impression that the new information was certain. It's quite common for newly published research, particularly surprising newsworthy research, to be not all it was cracked up to be. "New drug may cure farfalonis of the blowhole" is much safer than "New drug cures farfalonis of the blowhole". Bad reports can be fixed on line, but not in people's minds, not easily.
(All praise to the Reuters Health reporters and editors here, who consistently Get It Right.)
The Robert Redford of having to shit at totally inconvenient times and other links. « We Who Are About To Die said,
May 17, 2011 @ 11:22 am
[…] –Language Log […]
Steve T said,
May 17, 2011 @ 11:34 am
Best practices for writing RFCs (proposal documents for internet protocols and decisions) totally eliminate "can":
http://www.ietf.org/rfc/rfc2119.txt
David Eddyshaw said,
May 17, 2011 @ 11:48 am
I wouldn't have spontaneously made any distinction at all between "may cause" and "can cause"; personally if I was trying to express the sense they mean by "may cause" I would have said "might cause".
However, many people evidently have lost "might" in this sense and use "may" instead
("Hitler may have been killed by the plotters if they had been better organised")
which I'd always taken as a simple loss of the distinction. But perhaps at least some of them now use "may" always in a hypothetical sense, contrasting with "can".
Mr Punch said,
May 17, 2011 @ 12:14 pm
This seems perfectly clear and correct to me. "Can" means "is able to"; so "can cause" means "will cause in some cases." "May" is more indefinite; in fact, it implies "may or may not." The key is that melatonin has no volition – it's not as if it were deciding whether or not to cause respiratory issues, perhaps in violation of a ban on such activity.
Erik said,
May 17, 2011 @ 12:38 pm
While I was growing up, I had it pounded into my head that "can" meant causation ("This machine can be very loud") and "may" meant permission ("You may go to the bathroom"). Thus, headlines like "X may cause Y" sounded really wrong because it turned into "X is allowed to cause Y"!
Having the benefit of more experience available to me now, I still cannot give a useful distinction between "can" and "may" for conditional causation statements. I think that "may" is used when someone is trying to sound more formal (or scientific), but I can't tell any other differences for the most part.
If I were writing about X having a statistical chance of causing Y, I would probably use something like "probability of X causing Y" to hopefully eliminate the ambiguity between "can", "may" and even "will" (which often doesn't mean absolute certainty, but only a strong likelihood.) This is probably tempered by my having to write way too many scientific reports where this kind of distinction is part of the culture!
Ran Ari-Gur said,
May 17, 2011 @ 1:09 pm
Henning Makholm's point, that "may" seems to have broader scope here than "can", seems like an important one. To say that the drug may cause respiratory issues means that it may be that the drug causes respiratory issues. With can, that reading is impossible, so can is a stronger statement.
(I think the intended meaning is "the drug may can cause respiratory issues", but English syntax forbids that, and the can isn't sufficiently necessary to prompt a rephrasing as something like "the drug may be capable of causing respiratory issues".)
GeorgeW said,
May 17, 2011 @ 1:13 pm
According to "The Gregg Reference Manual" (7th Ed.), "May and might imply permission or possibility; can and could, ability or power."
I can recall being corrected as a child (mid-20th century, American South) if I asked, "Can I have some X?" I would be told, "You are capable of having X, but you MAY not now."
Nyq Only said,
May 17, 2011 @ 3:05 pm
'May' is also used as a synonym for 'might' – so it has a range of modalities. 'Permitted' (not relevant in this case), as a synonym for 'can' (is capable of) and as a synonym for 'might' (there is some chance that it could).
jc said,
May 17, 2011 @ 4:00 pm
I'd think that the examples so far most illustrate that "can" and "may" are so jumbled that, as a listener or reader, my main interpretation should be "They meant whatever the speaker/writer meant them to mean." And, since the speaker/writer usually doesn't define them, we can't know what was meant.
Of course, if you want to be cynical, you could interpret the use of either to mean "I'm trying to give the impression that there's some sort of unspecified connection here, while intentionally using words so vaguely defined that nobody will be able to pin me down to a precise meaning."
In any case, they might (;-) have had precise meanings centuries ago, but they clearly don't today. Especially not in journalistic settings.
Nick Lamb said,
May 17, 2011 @ 4:21 pm
Steve: well, not eliminate, they just don't give this word any specific meaning. Since the RFC 2119 keywords have considerable weight, you might choose to use the word "can" when none of those words is appropriate. I'd guess it would be valuable in discussion for this reason e.g.
"You can set the evil bit in this case, but that would cause a cascade failure". Writing any of "must" "should" or even "may" in place of "can" here would spark confusion (albeit tempered in written form by the preference for block capitals when using RFC 2119) because probably setting the evil bit is not encouraged by the standard. So "can" serves well to indicate only that it is technically _possible_ to set the evil bit, without itself passing judgement on whether it's permitted by the standard or a good idea.
Kay said,
May 17, 2011 @ 4:23 pm
Biomedical editor here – I've seen this tendancy to use "can" with statistical association/evidence and "may" with less solid claims, especially when discussing adverse events. @GeorgeW: I'm in the American South, and our house style guide tells us to change "may" to "might" unless there's a sense of permission!
Jonathan Mayhew said,
May 17, 2011 @ 5:02 pm
I would have taken "can" to mean "does," as in "Spring can really hang you up the most." Whereas "may" means "may or may not," we just don't know yet. After reading the comments I just don't know any more.
Paul Portner said,
May 18, 2011 @ 9:37 am
The corpus evidence is very helpful here.
On the analysis side, I don't think this has anything specific to do with the medical domain. Some semanticists (e.g., Carlson, Heim, Brennan) have suggested that there's a category of "quantificational modals" with meanings similar to adverbs of quantification, so that 'can' means something close to 'sometimes' or 'in some cases'. A good example is
– It can be filthy in New York.
This is clearly not an epistemic ("The evidence is compatible with it being filthy in New York"), deontic, or (sorry to disagree here) ability reading. Note that the "filthy" sentence entails that there has been at some time, in some place, filth in New York. This makes it different from mainstream weak modals, where the sentence can be true even though the portion under the scope of the modal is not true in the actual world. The "can cause" sentences mentioned in the post seem to have a parallel entailment.
I think these should be analyzed via quantification over situations, i.e. "in some situations in New York, it's filthy", assuming we treat classic adverbs of quantification that way too. (There's a fair amount of literature on this.) Kratzer-style situations or Davidsonian events are both plausible resources for carrying this analysis out, though I think the former are better suited (Portner 2009). If this is right, they are different from other modals, which involve quantification over possible worlds.
Dan T. said,
May 18, 2011 @ 11:37 am
We're in the right month to discuss "may", anyway.
Bob Moore said,
May 18, 2011 @ 1:09 pm
I think what we have here is a distinction between epistemic possibility (expressed here by "may") and a variety of what philisophical logicians call alethic possibility (expressed here by "can"). The first means something like "consistent with what we know," and the second means something like "consistent with scientific fact (whether we know it or not)." The article on modal logic in Wikipedia does a decent job of discussing such distinctions, including noting that English often uses the same words to express these different sorts of possibility.
ShadowFox said,
May 18, 2011 @ 1:11 pm
The distinction I'm getting is "[hypothetically] may cause" vs. "[is known sometimes to] cause" (that's the "can cause"). In other words, one is a speculation on the possibility, while the other is a distinct (however small) possibility. Yet another way to look at it is "yet undiscovered" vs. "unexplained". Both cases could be described as "possible" or "potential link", but these description miss the nuance.
On the other hand, this is close to research jargon–I doubt this distinction would be easily picked up on the street without an elaboration, at least of the kind that's been quoted in the original post.
Adam said,
May 18, 2011 @ 3:38 pm
This lack of distinction is one thing that always really befuddles me about English – my personal style is:
– 'can' for 'is able to'
– 'may' for 'is allowed to'
– 'might' for 'is possibly able to'
But as barely anyone else follows rules like these, it is very hard to discern the author's meaning. If the opening quotation used "can" then "might", it would seem a lot more clear to me.
RF said,
May 18, 2011 @ 9:20 pm
My first thought was that the word "may" is /stronger/ than "can". For instance, I might say "There's no rule that says a sitting president has to keep his Vice President as his running mate when seeking reelection; Obama can run with Palin as his running mate if he wants to." But I wouldn't say "Obama may run with Palin"; the former simply points the lack of impossibility, while the latter implies that it's a significant possibility.
And do we really need either verb? Would "This drug causes respiratory problems" be misintepreted as "Everyone who takes this drug gets respiratory problems"?
@Paul Portner
I think that "It can be filthy in New York." refers to an elliptical possibility of observance; if you go to New York, it's possible that you will observe it to be filthy. In fact, that would apply to the drug as well: if you take the drug, it's possible that it will cause respiratory problems. It's not merely "there are some cases where this is true", but "if you try to test the truth value of this statement, it's possible that you will find it to be true". Without such an analysis, the word "can" doesn't make sense. If the drug causes problems, then it causes problems. No "can" needed.
un malpaso said,
May 19, 2011 @ 12:15 am
wow. I guess I want to throw my oar in here for no other reason than I /CAN/.
In this context, they both seem to me (as a 40/m American middle class speaker) to be almost identical in meaning. I am trying to correct for prosodic or poetic or aesthetic effects that tend to want me to attribute a stronger probability to "can."
If there is any stronger sense to "can", it is VERY faint. Since most of us Americans were brought up to have a very rough sense that "may" is more polished/proper than "can" (as in, "Mother, may I"), it /MAY/ be that I am just extending that sense of "propriety" into the realm of "this person is hedging his/her bets by saying MAY, whereas CAN would be more straightforward and manly."
But basically, I see no real difference between the two here.
Peter Erwin said,
May 19, 2011 @ 6:44 am
Curiously, I think the intended distinction between "can" and "may" in this case is easier to see if you imagine negating them. For example,
"Drug X can't cause respiratory problems"
suggests a clear statement of fact: the drug in question does not cause — is indeed incapable of causing — the problems, period. On the other hand,
"Drug X may not cause respiratory problems"
merely indicates the possibility of non-causation, and leaves the true situation up in the air. (Assuming you exclude the alternate meaning of "may" as "has permission to", a sense of the verb which, as Mr Punch points out above, doesn't easily apply to a mere drug).
RF said,
May 19, 2011 @ 4:21 pm
While examining words by comparing their negations may seem logical, it is far from foolproof. Negating "may" immediately introduces ambiguity. "You may not go" can mean "Not going is something that you may do" or it can mean "Going is not something that you may do". Furthermore, the use of contractions strongly affects the relative dominance of the interpretations. "You mayn't go" would most likely be interpreted as the latter interpretation. With the right prosody, "This drug can not cause problems" can easily be interpreted as meaning "it's possible that it won't cause problems".
For a more dramatic example, "I'm smarter than any fifth grader" means the same thing as "I'm smarter than every fifth grader", but "I'm not smarter than any fifth grader" means something completely different from "I'm not smarter than every fifth grader". So your idea of comparing negations, while clever, simply is not valid.
Peter Erwin said,
May 21, 2011 @ 7:06 am
RF — I think you missed my point a bit. I'd certainly agree with you that comparing negations is not universally valid, or quite possibly not valid in the majority of cases. What I noted was that in this particular case, looking at the negations seems to emphasize the contrast that some people were arguing exists, at least in the context of biomedical jargon, between "may" and "can" used with the verb "cause".
More trivially, I chose "can't" specifically to avoid the somewhat tortured sort of ambiguity you're trying to introduce. I'd argue further that the rare "can (not do)" usage you suggest isn't actually a negation, and so it's an irrelevant objection. (As for "mayn't" — seriously?)[*]
(Questions of prosody are, in this case, largely beside the point, since most of the examples Mark discussed were written. With the right kind of prosody — and changes of subject and verb — you can make almost any auxiliary usage ambiguous. But that doesn't tell us much that's useful about the biomedical-jargon issue at hand.)
[*] Not surprisingly, while a search of Google Scholar turns up over 1500 hits for "can't cause", and over 30,000 for "may not cause", it produces exactly zero for "mayn't cause". (And certainly the first hundred or so examples of "can not cause" are all simple negations — you can replace them with "cannot cause" or "can't cause" without changing the meaning.)
m said,
May 23, 2011 @ 11:11 pm
From "Dessert, Laid-Back and Legal" New York Times May 14:
“A hangover effect has been reported” with large doses, said Anna Rouse Dulaney, a toxicologist with the Carolinas Poison Center. But she added, “I don’t want to go on the record saying this drug ‘can’ cause respiratory issues, that should be a ‘may.’ ”
Maureen said,
May 24, 2011 @ 11:14 am
It's pretty simple. Nobody has yet reported having respiratory problems after eating melatonin brownies, so the doctors can't say they have the ability to cause respiratory problems. Such respiratory problems may or may not occur in the future.
The funny bit is that all the hormone studies show that, if you take an overly large dose of melatonin, the body reacts by waking you up and making it hard for you to go to sleep again. I'd be more worried about people being allergic to valerian.
Every word carries meaning | Technical Editors: Arbiters of Quality said,
June 17, 2011 @ 3:39 pm
[…] simple enough, but consider this post from the Language Log about "may cause" versus "can cause" in scientific writing. It seems as […]