June 16, 2007

Nearly no: a gnarly knot

My post "Why is 'nearly no' nearly not?" struck a chord with readers -- more than 30 of you sent in ideas and/or evidence. Given the number of notes, I need to apologize in advance for not referencing everyone's suggestions in this morning's narrow blogging window, before I go take the 8:50 JR train to Kansai airport.

The puzzle is why nearly seems not to consort well with negatives, as indicated by the following table of Google counts:

  no one

everyone

everyone/no one ratio
nearly
27.6K
1.29M
47.8
almost
1.06M
1.66M
1.57
almost/nearly ratio
38.4
1.29

I considered and rejected one hypothesis, namely that nearly prefers the positive ends of scales.

Many readers suggested that the problem is alliteration: thus Aaron Toivo asked "Could it be a clash of some sort between the two /n/s?" But several readers also provided evidence to undermine this suggestion: thus Aaron offered Google counts from eight vs. nine, which lend no support to the "avoid alliteration" theory:

  eight

nine

eight/nine ratio
nearly
866K
694K
1.25
almost
623K
557K
1.12
almost/nearly ratio
0.72
0.80

And I can supplement this with data from October vs. November, which suggest that if anything, alliteration is sometimes a slightly attractive force:

  October

November

October/November ratio
nearly
273
3.3K
0.082
almost
10.3K
10.1K
1.02
almost/nearly ratio
37.7
3.06

 

Other readers suggested that "nearly NEG" is not just dis-preferred, it's really entirely outside the norms of the language, with apparent counterexamples involving a different scope for nearly. Thus Steve Carter offered the examples:

"Thomas nearly never got to ride in his hometown."
"Thomas almost never got to ride in his hometown."

and commented

In the first example, it feels to me more ready to be interpreted as "Thomas narrowly escaped being prevented from ever riding in his hometown." -- though it may easily be followed by "As it turned out, he had a long and successful four seasons' riding and became a local hero."

The second example, "almost never," seems to bind more closely, more strongly suggesting "Thomas seldom got to ride in his hometown."

Two observations on the above:

1) It probably doesn't contribute to the original thrust of your post
2) The cause of the above perception is probably the very fact that "nearly never" is seen more rarely than "almost never".

This may well be right -- perhaps nearly has wide scope in a certain fraction of the web's apparent "nearly NEG" examples, making that pattern ever rarer than it seemed to be. Unfortunately I don't have time this morning to check.

Joseph Pentheroudakis wrote:

I thought I’d check what happens with similar modifiers like “just about”, “virtually”, and “literally.” I tried them in the context ‘___ everything/everyone/everybody’ and ‘___ nothing/no one/nobody’. Here are the results:

  every(thing/body/one) no(thing/body/one) pos/neg ratio
just about
3.318M
97K
34.2
nearly
2.55M
272K
9.3
virtually
2.288M
1.51M
1.5
almost
3.98M
2.78M
1.38
literally
280K
320K
0.8

Two items very strongly associate with positive polarity then (‘just about’ and ‘nearly’), and the rest sort of swim in the same end of the pond, compared to the first two anyway.

That's interesting. And most of the examples of "nearly NEG" in his compilation are from "nearly nothing" (226K out of 272K), which may reflect the fact that Norma Loquendi doesn't seem to object much to "nearly zero":

  zero

one

zero/one ratio
nearly
873K
1.26M
0.69
almost
1.07M
1.27M
0.84
almost/nearly ratio
1.23
1.01

If we eliminate the -thing examples from his table, the split gets even bigger:

  every(body/one) no(body/one) pos/neg ratio
just about
1.778M
27K
65.8
nearly
1.43M
46K
31.1
virtually
1,078M
400K
2.70
almost
2.46M
1.56M
1.58
literally
96K
50K
1.92

Alan Wechsler suggested:

I have a theory, though I don't know how to test it.  Theory: in any context that implies a numerical assessement, "nearly" means "a number slightly less than".  This leads to discomfort in contexts where the numerical measure is zero, because it implies a value less than zero. 

Oh, I guess on second thought I do have a test case.  I don't like "nearly free", in cases where "free" means "available for no cost".  This is hard to count on Google because there are other senses of "free" which accept "nearly" quite nicely: "nearly free from deal", "nearly free of dabblers", and so on.

This makes sense -- except that the counts cited earlier suggest that "nearly zero" is not much of a problem. (Though maybe some of these are cases in which a function is approaching zero from below?)

Ben van Heuvelen suggested a cognitive explanation, with a bit of a theological twist in the tail:

Here's a little conjecture about what's going on:

"Nearly" is a more concrete word than "almost." Both adverbs are used as degree modifiers, but "nearly" entails a slight metaphor, since the adjective and preposition forms of the word ("near") suggest physical proximity. It's impossible to use "nearly" without subtly invoking physical space. For example, my understanding of the sentence "They were nearly happy" is informed by my previous understandings of sentences like "You are near the supermarket."

(True, you can also use "almost" when talking about physical space -- "You are almost at the supermarket" -- but to do so you have to add a preposition. There isn't a metaphor built into the word.)

My theory is that we tend to rely on "almost" when the idea we're conveying is more abstract, something we can't easily picture. "Nearly everyone" is easy to picture (a big crowd), as are the "nearly worthless" things I encounter every day. "Almost no one" is almost impossible to picture (an empty space? a space filled with a couple semi-translucent bodies?), while an "almost priceless" object is a logical impossibility. In the latter two cases, we rely on the more abstract adverb, "almost."

In response to your admonition -- "Don't bother to tell me that priceless is like unique and shouldn't get any sort of degree modification at all" -- why not? It's a relevant question: after all, a word that illogically receives degree modification becomes harder to picture.

Consider the case of "good" versus "evil." I know many people who are mostly (but not entirely) "good." I also know some people who tend more towards "evil." But if I'm describing the latter, I'll probably choose a narrower word (like "thoughtless" or "cruel") rather than "evil," which suggests something absolute. Call me a hopeless optimist, but I think it's much easier to picture someone who's "nearly good" than someone who's "nearly evil." A quick Googling suggests I'm not alone:

  good evil good/evil ratio
nearly
144K
2.19K
65.7
almost
178K
39.9K
4.46
almost/nearly ratio
1.24
18.2

Finally, as for the occasional uses of "nearly no one" by "otherwise competent" writers: I think they're probably just enjoying a little alliteration.

This might be right -- but it seems easy enough to me to come up with a concrete visualization of a room with almost/nearly no one in it -- it's a situation very near to that of an empty room, but with a few people sprinkled sparsely around. Isn't the sparse sprinkling of people a concrete-enough visualization of "nearly no one"? And neither I nor the net seem to have any real trouble with "nearly all gone"; "almost all gone" has only twice as many, compared to more than 38 times as many for "almost no one" vs. "nearly no one".

The facts about evil, seem similar to whatever's going on with other negatively-evaluated end-of-continuum words like hungry and thirsty and tired. These don't seem to be especially abstract or hard-to-visualize concepts, but they show a somewhat similar pattern, though it's hard to find clear opposing terms to test them against (e.g. full means lots of things besides not hungry):

  full hungry full/hungry ratio
nearly 992K 204 4863
almost 1.51M 3.47K 435
almost/nearly ratio 1.52 17.0  

So we've placed the pattern in a larger context, but I'm not convinced that we have an explanation yet

Posted by Mark Liberman at June 16, 2007 05:54 PM