Most examples

« previous post | next post »

My note this morning on "Most" stirred up some discussion:

Geoff Nunberg: I think 'most' licenses a default generalization, relative to a bunch of pragmatic factors, …
MattF: I think 'most' has a normative or qualitative sense in addition to a quantitative sense.
John Cowan: For me too, "most" has a defeasible implicature of "much more than a majority".

Those rear ends are pretty well covered — "default", "in addition to", "defeasible" — but Nicholas Waller got numerical:

I would be with John Irving – 51% of a population isn't "most" but around 60-75% would be. (90% or more would be "almost all"; well, until it hit "all" at 100%; and 75-90% would be "a very large majority")

So did T.T., although with somewhat different thresholds:

"Most X are Y", to me, means a substantial majority of X are Y—certainly more than 50%-plus-1. Even two-thirds feels borderline.

But Adrian Bailey agreed with my first reaction:

Most has always meant "more than half (but less than all)" to me. If there are 100 of us and I say "Most of us stayed behind" I mean between 51 and 99.

And there were other opinions as well. So I poured myself a cup of coffee and did a little Breakfast Experiment™, even though it's mid-afternoon.

I searched on Google for the pattern "most * percent", and picked out of the first 150 hits all the examples like these:

most Pakistanis (64 percent) believe it is important to improve relations with their powerful ally
Most (72.4 percent) said that they would consider dating someone of a different race.
Most Americans (51.4 percent) will live in poverty at some point before age 65.

There were 72 numbers in my list, and the histogram of 69 of them looked like this:

You might believe that this is a bimodal distribution, with one mode just above 60% and another just above 80% — though if you divide things up into ten-percent bins, the stretch from 60 to 90 flattens out:

In any event, it's pretty clear that the whole range from 50.1 to 99.9 is getting some action.

The three examples that I left out of the histogram were cases where "most" meant "a plurality", i.e. the subset with the largest proportion even though that proportion was less than 0.5:

There were 42,286 eye accidents reported in private industry in 2002, and the most prevalent (38 percent) type of event involved the eye or eyes being rubbed or abraded by foreign matter.
Most (44.5 percent) said that 25-49 percent of students will transfer to a 4-year college.
Most independents, or 40 percent, said they would vote for Giuliani.
According to the Third National Health and Nutrition Examination Survey, approximately one-quarter of people examined (4,801 of 17,235) had oral lesions, with most lesions (46.3 percent) located on the hard palate or gingiva where red-and-white lesions were most prevalent.

I thought that the following juxtaposition was striking evidence that most= >50% is alive and well in some quarters:

Many dancers (49 percent) were comfortable or very comfortable in revealing their occupation to others. Interestingly, most (52 percent) respondents viewed exotic dancing as a promiscuous activity.

The other 69 examples were:

Most farms — 61 percent in 2004 — do no receive any government payments.
most Pakistanis (64 percent) believe it is important to improve relations with their powerful ally
Most respondents (59.1 percent) believed that in a situation where a black and a white applicant were equally qualified, the white person was most likely to get the job.
Most (72.4 percent) said that they would consider dating someone of a different race.
Most Americans (51.4 percent) will live in poverty at some point before age 65.
Most (57.3 percent) do not keep any form of personal health records.
Most (68.3 percent) say they “strongly agree” with this statement.
Most (69 percent) of the respondents are white
Most respondents (86.3 percent) considered exotic dancing a legitimate line of work
Interestingly, most (52 percent) respondents viewed exotic dancing as a promiscuous activity.
Most (63 percent) dancers have never dated a customer.
Most (78 percent) were not exposed.
Most Californians (54 percent) say global warming is already having an impact
Most likely voters (79 percent) also view the U.S. Senate candidates' positions on the environment as at least somewhat important.
Most U.S. voters (52 percent) continue to believe that tax increases will hurt the economy
Most uninsured Nevadans (81.5 percent) are members of working families
Most (63 percent) also see blogs as an “equal playing field,” where everyone has a voice in a democratic forum.
Most, 71.5 percent ,agree with the U.S. bishops in their call for immigration reform
Most breaches (60 percent) continue to be discovered by external parties and then only after a considerable amount of time
When people search their names, most, 63 percent, find some relevant information about themselves
Most students (60.9 percent) believe it improves their learning.
Most women (71.5 percent) reported having never experienced any side effects associated with vibrator use.
Most firms (51.4 percent) had an. initial capitalization of less than $20000.
most men (71.5 percent) reported they would choose radical prostatectomy again.
Most students (63.1 percent) were not swayed away from a college by the presence of large lecture classes
Most (70.3 percent) of the fatally injured athletes were between the ages of 20 and 44
in most cases (59.3 percent), a risk category was not reported or identified.
most (82 percent) said they would be opposed to the nomination of a Supreme Court Justice who supports such abortions
most (90 percent) of these households contained elderly or disabled members
Most abortions (89 percent) in 2004 occurred during the first trimester
The injuries were, in most cases (62.0 percent), caused by the players falling or stumbling while attmepting to retrieve the shuttle.
liquid propane gas is the cooking fuel used by most households (78.5 percent)
most (71 percent) didn't even receive any World Cup spam
Most – 88 percent – had established between four and 10 anti-stress programs
Most (84 percent, or 155 of 185) of the generic drug products had no change in price during the 12-month period
Most (62 percent) work published in the leading journals was done at the individual level.
most fistulas (61.5 percent) occurred at the distal anastomosis in patients who did receive radiation therapy
most (88 percent) thought about it only occasionally or not at all
Most (57.5 percent) of the 89 protracted projects studied were between 5 and 7 years old
most (93.6 percent) were in male- headed households
most Republicans (79 percent) approve of Bush's performance and most Democrats disapprove (83 percent)
most – 62.2 percent – worked in industry.
Most of them — 51 percent — worry most about meat.
Most (74 percent) of the victims who reported violent crimes by juveniles said the offender was a male
Most U.S. farms—98 percent in 2004—are family farms
Most (85 percent) cancers in which the primary site is unknown are adenocarcinomas.
Most (89.2 percent) of the 4.4 million recent alcohol initiates were younger than 21 at the time of initiation.
of the burglaries for which the time could be established, most (62.2 percent) residential burglaries occurred during the day
Most urban physicians (81 percent) participated in Medicaid
For most (89.2 percent), the decision to go overseas was based entirely on personal interest.
which indicates that most (77.7 percent), but not all, of their responses were correct.
Most (83 percent) said that they got vaccinated to prevent themselves from getting sick
Most proxies (93.6 percent) also observed that their counterparts have no difficulty in making new friends.
Most (70 percent) have parents who are employed, many in small firms, or are self-employed.
Most — 81.5 percent — were aware of the straw poll
most (70 percent) were killed when they had car trouble or stopped to help someone on the road
Most companies — 78.5 percent and 77.4 percent for the current and most recently completed examinations, respectively — obtained a written commitment
most mayors (72.1 percent) indicated that they would cut parks, recreation, and libraries
Most (68 percent) attribute the attack to opposition of U.S. ties to Israel and U.S. policies toward Palestinians
Most (82.6 percent) of the rich women attend ANC from medically trained providers in the urban area
with most (92 percent) involving polydrug abuse
Most banks (75.1 percent) automatically enrolled customers in automated overdraft programs
Most injuries (71.3 percent) happened in the bathtub,
Most fixed annuities, 92 percent, have a death waiver
Most smokers (97.2 percent) acknowledged that smoking was harmful to them
Most producers (81.7 percent) are in the cow-calf business
Most respondents (67.7 percent) also stated that they feel pressure to reduce the law department budget

[Update — as I suppose is obvious, all of this is consistent with the view that the literal meaning of most (as applied to count nouns) is something like "the greatest number of of", which in the case of an assumed partition in two parts implies "more than half"; and that the usual Gricean considerations tend to add conversational implicatures, just as Geoff Nunberg suggested, which as usual are defeasible, just as John Cowan suggested.

This is exactly as predicted for the superlative of more. It remains plausible that some of these implicatures have been conventionalized for some people.

In a comment on Geoff Nunberg's post, Itamar points us to Martin Hackl, "On the grammar and processing of proportional quantifiers: most versus more than half", Natural Language Semantics 17:63-98, 2009, which argues for a literal formalization of this history (or something close to it).

But see also the references accumulated in "Most bibliography", 8/2/2010.]



38 Comments

  1. Coby Lubliner said,

    July 31, 2010 @ 4:34 pm

    It seems to me that most (almost all!) of the examples come from survey results that imply some sort of comparative ranking of groups, and "most" just indicates the largest group. In casual speech, perhaps "most" has a stronger meaning. I don't think people are likely to say "most of the world's people are women" if women outnumber men, say, 51 to 49.

    [(myl) The "largest subgroup" meaning is worth noting — the less-than-50% plurality cases make it especially clear — but quite a few of these examples — maybe more than half — are from situations where there's no relevant partition of respondents.]

  2. Jerry Friedman said,

    July 31, 2010 @ 4:51 pm

    I don't think I for one would say "most people" for 51 versus 49 percent, but I think I'd be more likely to for 51, 10, 10, 10, 10, and 9 percent.

    Another interesting question, in addition to what percentages "most" is used with, is what words are used with percentages between 50 and 55, say.

  3. a said,

    July 31, 2010 @ 4:54 pm

    This "Most people are men/women" example has come up a few times in the comments, and I don't know what all of the uneasiness is about. I would have no problem saying most people are male/female, if that were the statistic.

  4. a said,

    July 31, 2010 @ 5:01 pm

    For me, I think there's probably a scale of appropriateness for using the word 'most' wrt some statistic. I might feel more confident or feel that it is more appropriate to say most if the figure were something like 80% than if it were 51%, but I don't feel inappropriate saying 'most' wrt 51%.

  5. sarang said,

    July 31, 2010 @ 5:30 pm

    I think one should also keep in mind that some (perhaps most) of these examples involve journalists exploiting the ambiguity inherent in the interpretation of "most" to convert a ho-hum statistic into a story.

  6. Mark F said,

    July 31, 2010 @ 5:34 pm

    I've been in or heard a lot of conversations that could be condensed roughly to:

    Speaker A: Most people are X.
    Speaker B: What do you mean? Lots of people aren't X.
    A: "Most" just means "more than half."
    B: Well, OK.

    People also often say "probably" as if they mean "we can safely assume that", even though when pressed they may agree that "probably" means "with p > 0.5".

    Thinking as I type here, perhaps it works like this: If P(X|Y) > 0.5, and you know that Y is true and must act based on whether you think X is true, you should assume X is true. Having done so, you want to reassure yourself that you made the right decision, so you subconsciously tell yourself that P(X|Y) was probably more like 0.8 or 0.9.

    Another way of looking at it is that it's a form of rounding error. "Most" takes a p value and rounds it to 1. Any time a number has been rounded, I bet people tend to imagine the original number was closer to its rounded version than is really licensed.

  7. Jerry Friedman said,

    July 31, 2010 @ 6:43 pm

    On a related subject, there was a recent thread in a.u.e. about great majority. One participant quoted the AHD: "Majority is often preceded by great (but not by greater) in expressing emphatically the sense of 'most of': The great majority approved." He added, "The term great in this case merely adds emphasis which is unnecessary unless you are taking a poll. I contend that it is redundant in this context and that the term majority is sufficient."

    Someone else quoted Ofsted's definitions, which follow:

    97–100% Vast/overwhelming majority or almost all
    80–96% Very large majority, most
    65–79% Large majority
    51–64% Majority
    35–49% Minority
    20–34% Small minority
    4–19% Very small minority, few
    0–3% Almost no/very few

    (Ofsted looks to me like the name of a character in The Handmaid's Tale, but it turns out to be the Office for Standards in Education, Children's Services and Skills, a government department in England.

  8. Adam B said,

    July 31, 2010 @ 6:48 pm

    Where "most" is being used for less than 50% – I'd probably not say "most" to describe such a plurality, but I might well use it to describe a majority of, say, the respondents in a poll who didn't choose "don't know". There'd be an implied "Of those who expressed an opinion…" prefixed.

  9. sarang said,

    July 31, 2010 @ 7:26 pm

    Whenever most is being used for a plurality, the following statement is true: "Most people did X but most people did not do X."

  10. D.O. said,

    July 31, 2010 @ 7:51 pm

    Teacher: Kids, halves cannot be larger or smaller. Two halves are always equal.
    Student: Can you explain it more? I think the larger half of the students did not get it.

  11. Lucy Kemnitzer said,

    July 31, 2010 @ 7:52 pm

    I think the division between reserving "most" for large majorities and using "most" for any majority is quite real. I have run into misunderstandings of consequence when talking to a person wh understands "most" differently from me. (I'm of the "large majority" faction. When I want to indicate a majority that doesn't have to be large, I say "more than half," or"more often than not" depending on the context)

    I'm not sure that the thing can be codified effectively, and it may be prudent to add in extra words to be quite clear when there is any chance of confusion that matters. (clearly, if precision is needed, you'll need numbers: but if you don't need to be precise, just clear, you might add in stuff like "a large majority" in another sentence, or "more [whatever] than not," or something)

    Although it seems to me that quite a few of the examples here are uses of "most" where real clarity aren't isn't even a goal.

  12. a said,

    July 31, 2010 @ 7:52 pm

    @sarang — I don't think I agree. I think it makes a difference whether you have the article. I would say: "The most people did X but most people did not do X." Think of a runoff election: "The most people voted for Candidate X, but most people did not vote for him." BC there is no one who most people voted for.

  13. language hat said,

    July 31, 2010 @ 8:13 pm

    I think the division between reserving "most" for large majorities and using "most" for any majority is quite real.

    Apparently so, but until the subject came up here I never realized there were people for whom it meant anything other than 'more than half.'

  14. sarang said,

    July 31, 2010 @ 8:17 pm

    @ a: The article changes everything, though. Of course there's a difference between "the most votes" and "most votes." The question is whether "most" w/o the article is strange/absurd for pluralities that are not majorities. (e.g. the two examples given in the original post)

  15. Adouma said,

    July 31, 2010 @ 8:37 pm

    @a: I don't know about anyone else, but I would do a double-take if someone said to me "The most people voted for Candidate X, but most people did not vote for him," out of the context of this discussion. While it's strictly true, it just sounds like a "har har statistics" fact* that would be employed specifically to confuse people.

    *Of the same kind as "Every person in this room has more than the average number of legs," because the mean number of legs across the human population is ever-so-slightly less than 1.

  16. Nathan Myers said,

    July 31, 2010 @ 9:31 pm

    Missing from discussion up to now is the notion of decisiveness. In a vote, one vote more than 50% wins. For Irving's readership, it takes rather more than 50% to mean "most", because it's a question of expectations on meeting a reader. Therefore, there's no contradiction between taking both "anything above than 50%", "well above 50%", or "nearly all" as the same meaning of "most"; it's a question of context.

    I don't count protestations I've read here about what "most" means to readers as worth two beans.

  17. unekdoud said,

    July 31, 2010 @ 9:33 pm

    The description of a percentage should depend on how confident the person knows the value. If the percentage is close to 50% and could go either way then it is surely inappropriate to use "most" (other than in the plurality constructions). However even if the percentage is known to be >50% this can be communicated in a more explicit way, such as "just over half" or possibly "weak majority".
    One of the most interesting examples above is "Most of them — 51 percent — worry most about meat." which has "most" used two different ways. In this case, the phrase "worry most about meat" isn't constructed from a percentage, but must be inferred from a study of some sort.
    From the definition of "most" as a superlative form of "more", it should mean "more than anything else". However, when it is used to describe a majority, it means "more than everything else combined", which can be treated as a special case of the first definition. The similarity of the two is slightly confusing.
    It would be interesting to conduct a survey of how people describe various amounts, similar to xkcd's color survey perhaps.

  18. a said,

    July 31, 2010 @ 11:06 pm

    @Adouma

    That's my point, though. I was responding to sarang, who said: Whenever most is being used for a plurality, the following statement is true: "Most people did X but most people did not do X." That's certainly a trick designed to confuse people, but my point was that the way sarang phrased it, it wouldn't actually be true. At least IMO, you need the article in front of the first one.

    @sarang: re: The question is whether "most" w/o the article is strange/absurd for pluralities that are not majorities. (e.g. the two examples given in the original post)

    At least I think it is strange.

  19. D.O. said,

    August 1, 2010 @ 1:05 am

    According to the data in the post, most of the times most means 70% or more. :)

  20. John Cowan said,

    August 1, 2010 @ 1:05 am

    I was trying to be cautious, objective, and linguisticky (that's linguist-ic-y, for the record, not any of the other possible readings). For myself, I think it's actively misleading to use "most" for a bare majority, although I recognize that other people who do so aren't necessarily trying to mislead anyone. For me, "most people are men" and "most people are women" are both false tout court.

  21. Aaron Toivo said,

    August 1, 2010 @ 1:17 am

    I think it would be preferable to avoid basing quantification judgements mainly on the phrase "most people", as this often serves as code for 'everyone' or 'everyone relevant'. It can be very socially marked, for instance in "Most people chew their food before swallowing, young man", among numerous milder possibilities.

  22. John Walden said,

    August 1, 2010 @ 3:54 am

    It'd be instructive to go on a twenty-one day all-inclusive holiday which promised 'Wine or beer will be served with most meals, except breakfast' both with the 51 percenters and with the others. At the end of the holiday, having been served wine or beer with 22 of the 42 meals, I hope none of them would be in the queue complaining with the rest of us. They'd be in a huddle, saying 'Fair enough, it's what 'most' means', however ripped-off they might really be feeling.

  23. ShadowFox said,

    August 1, 2010 @ 6:32 am

    There are two basic problems both with Mark's analysis and many of the comments. Let me make a simple case for these.

    1) The histogram Mark constructed–both of them, actually–is misleading. The only obvious points are that 50+ is occasionally identified as "most", but 60+ is identified as "most" more often. This information the histogram relates very well, no matter what the interval is. But what the histogram cannot do–nor, indeed, is this something that can arise from a Google search pattern–is show the frequency for each subgroup relative to the distribution of the percent references. I'll explain below.

    2) In some instances–for example, in the examples where "most" represents merely a plurality–usage depends on the distribution among the categories being tested, not merely the percentage in the largest category alone. Suppose, for example, that the distribution is 45-5-5-5-10-10-10-10. In this case, I would contend, an author would be more likely to identify the 40% group as "most" than in a distribution such as 40-30-30. This observation is not limited, of course, to cases where "most" is the plurality, but may also have an effect on how the majority category is judged, e.g., 55-25-20 might be more likely to elicit "most", than 55-42-3.

    Now, to explain (1). A simple histogram would be suggestive if the underlying distribution of the categories was uniform or near uniform. However, it's not a difficult exercise to figure out that getting the largest category near 50% or 60% is much more likely than getting it in the 80s or 90s. This is not limited to opinion surveys, but is broadly applicable to all surveys. There are few categorizations where the largest subgroup ends up with 80+% of the population. Examples of these would be comparisons of domestic possessions between families in the 1950s and in 2000s–e.g., households with cars, 2 cars, TVs, multiple TVs, refrigerators, etc. Specifically, in many such cases, the large percentage is set up (by design) as contrast with a smaller percentage in a similar category under different circumstances (usually comparing temporarily or geographically). Opinion surveys with such high numbers for the dominant category are even more rare.

    In other words, it's not enough to know how many "most * percent" we have within a certain numerical reporting group, but how many of those we may find relative to all other expressed quantifiers for similar results (and it's more than just "many", as some have already pointed out). Under these circumstances, even if the numbers that Mark found were to decline from 60+ to 70+ to 80+ to 90+, we would have little idea of how the identification as "most" was actually distributed. Given the numbers that Mark did find, I would say that the relative frequency of most actually increases even after the jump from 50s to 60s. To put it in yet another way, the larger the percentage, the more likely it is to be referred to as "most" (as opposed to "the more frequently", which is what Mark was comparing).

    The apparent bimodality of the first histogram actually suggests something interesting–there may be two thresholds for "most", one near 60 and one near 80. It does not actually mean that the cut-off is at 60 or 80–it may well be at 58 and 75, but the underlying distribution of the actual survey results precludes us from discovering the real critical points. Still, this actually meshes well with the idea that the identification as "most" becomes more likely as the percentage in the dominant category increases.

    So the data are interesting, but ultimately misleading and insufficient. If we want to know the most/many etc. preference, we'll just have to do the survey on those preferences directly rather than relying on the Google distribution of particular phrases.

    At the same time, it would be interesting to do an underlying study of how survey results are distributed, in general, i.e., find the actual distribution of the largest subgroup percentage among all surveys. There are not a lot of options for that one, but it might be worthwhile to pick a temporal interval and record the results of all published surveys (among a selected list of publications) within that period. This should give some idea of how the numbers may be distributed. Factoring this into the Google data would help to get a better picture.

  24. James Kabala said,

    August 1, 2010 @ 9:16 am

    I can safely say that in that particular example, I would not feel ripped off at all. Now if they said "wine or beer will be usually served with meals," I probably would feel ripped off, even though they might be on a good technical linguistic ground there as well. Connotations can be a funny thing.

  25. language hat said,

    August 1, 2010 @ 9:48 am

    For myself, I think it's actively misleading to use "most" for a bare majority, although I recognize that other people who do so aren't necessarily trying to mislead anyone.

    What an odd statement. If you're going to go on to say "other people who do so aren't necessarily trying to mislead anyone" (though that "necessarily" is sneakily slanderous — "I realize you're not necessarily a terrorist/Communist/criminal…"), you should junk the "misleading," because it implies that someone is misleading you. Why not just say "For myself, I find it confusing when people use 'most' for a bare majority"?

  26. 400guy said,

    August 1, 2010 @ 10:36 am

    ShadowFox points out that there is no reason to expect the relative sizes of the subcategories of a population to be evenly distributed between 0 and 1.  Indeed, in the context of a questionnaire for example, you might well doubt the value of a binary question which everybody—or almost everybody—answers the same way.
    After the numbers are collected and tabulated, it is easy to imagine another force distorting the histogram of google results.  Even a writer firmly of the opionion that most can correctly describe any fraction from fifty-percent-plus-one on up might well decide that other fractions are more worth commenting on.  A comment on a number approaching one hundred percent might be too obvious to be worth making, while a number close to fifty percent might not suggest any brief, plausible, and interesting comment.
    Of course, the comment-worthiness of any reported number depends on whether it contradicts one's presuppositions.  That is of course beyond google's ability to analyse.  Come to think of of it, that analysis is beyond my ability too<grin />. 

  27. 400guy said,

    August 1, 2010 @ 10:49 am

    Uh-oh. I wrote my previous comment as three paragraphs, and the preview window showed it to me with blank lines between paragraphs. Where did that go?

    "Mommeeee, the computer is picking on me."

  28. Nijma said,

    August 1, 2010 @ 1:47 pm

    "Most" and "most prevalent" are not necessarily interchangeable. To me, "most" indicates a preponderance, way more than 50%, while "most prevalent" is merely the group with the most members, possibly a plurality.

  29. Andrew (not the same one) said,

    August 1, 2010 @ 4:41 pm

    Aaron Toivo: Yes, but the fact that it can be used that way is surely indicative of something. If 'most' meant simply 'more than half' and everyone agreed that that was what it meant, why should it ever have connotations of 'everyone' or 'the relevant people'?

  30. Adrian Bailey said,

    August 1, 2010 @ 6:28 pm

    I think Adouma means 2, not 1.

  31. Pseu said,

    August 1, 2010 @ 8:00 pm

    I think that the sense I get from "most" is something like "enough that you can think of that proportion as basically standing in for the whole, without losing too much precision". Which is not exactly a mathematically-precise kind of a concept.

    For your examples, I think it would be interesting to track how the %age cutoff might vary with context and purpose. In the "poverty" one, for example, it sounds like the kind of a statement of a person or group who is trying to convince readers that poverty is a really big, really widespread problem. And, so, while it's a little sketchy to take 51.4% and call it "most", doing so serves the argumentative ends of the writer.

  32. Topher said,

    August 1, 2010 @ 8:17 pm

    I think that the methodology of this survey is flawed. Context can weaken or eliminate connotation — if it couldn't then it would be indistinguishable from definition. See my comment on the previous item. For example, "greedy" has a negative connotation so that "Jack grabbed at the pile greedily" is at least somewhat disparaging to Jack, but "Jack grabbed at the pile greedily in a way that would have been unjustified for someone whose need was less great" does not disparage Jack at all (though perhaps the author of such a sentence should be disparaged).

    Using a phrase template that specifically introduces context that could mitigate or eliminate the connotation changes what is being sampled.

  33. language hat said,

    August 2, 2010 @ 9:51 am

    And, so, while it's a little sketchy to take 51.4% and call it "most", doing so serves the argumentative ends of the writer.

    It is not "sketchy" if that's what it means to you! Can we do each other the courtesy of assuming that the people who do not use words the way we do are not lying about how they use them?

  34. chris said,

    August 2, 2010 @ 11:35 am

    Yes, but the fact that it can be used that way is surely indicative of something. If 'most' meant simply 'more than half' and everyone agreed that that was what it meant, why should it ever have connotations of 'everyone' or 'the relevant people'?

    Because humans have a regrettable but widespread tendency to think sloppily about heterogeneous groups?

    Or maybe I should say that most humans have such a tendency… :)

    I think that the slippery but often very emotionally charged idea of a "normal" or "typical" representative of a group has something to do with this issue. If most X are Y, then the typical X is Y. But it's awkward to apply that to things like "the typical human is female" because there are too many exceptions, so if you tend to make that kind of mental jump, you may be likely to be uncomfortable with "most humans are female" — because, after all, you know so many who aren't.

    Can we do each other the courtesy of assuming that the people who do not use words the way we do are not lying about how they use them?

    OK, but shouldn't we be prepared to entertain, and if necessary, test, the hypothesis that they are mistaken? How people use words and how the same people think or say they use words are not necessarily the same.

  35. language hat said,

    August 2, 2010 @ 1:21 pm

    Of course, but that's a very different matter from calling someone's usage "sketchy," implying they are deliberately trying to mislead.

  36. mollymooly said,

    August 2, 2010 @ 2:32 pm

    @John Cowan:
    For myself, I think it's actively misleading to use "most" for a bare majority, although I recognize that other people who do so aren't necessarily trying to mislead anyone.

    @language hat:
    What an odd statement. If you're going to go on to say "other people who do so aren't necessarily trying to mislead anyone" (though that "necessarily" is sneakily slanderous — "I realize you're not necessarily a terrorist/Communist/criminal…"), you should junk the "misleading," because it implies that someone is misleading you. Why not just say "For myself, I find it confusing when people use 'most' for a bare majority"?

    Whoa! I interpreted John Cowan's post as something like this:

    Since John defines 'most' to mean 'great majority', John could not describe 51% as 'most' unless John was being consciously dishonest.
    In contrast, Hat who defines 'most' to mean 'any majority' could use it in good faith, but accidentally mislead John who took it to mean 'great majority'.

  37. Boris said,

    August 2, 2010 @ 3:19 pm

    The problem with looking for most (X Percent) is that the writer explicitly defines what he/she means by "most", so there is no ambiguity. It is not clear that the same people will use "most" on its own to mean "X percent". In fact, one might argue that the very existence of these examples proves that there is an ambiguity.

  38. language hat said,

    August 2, 2010 @ 7:28 pm

    Whoa! I interpreted John Cowan's post as something like this

    I suppose you could be right — in fact, I hope you are — but "I think it's actively misleading" seems to me very odd phrasing if you're talking about a hypothetical action you yourself might take.

RSS feed for comments on this post