July 09, 2007

"Nearly/almost no" vs. "not nearly/almost"

In connection with three earlier Language Log posts ("Why is 'nearly no' nearly not?", 6/14/2007; "Nearly no: a gnarly knot", 6/16/2007; "Nearly and almost", 6/24/2007), I recently got a fascinating note from Lucia Pozzan and Susan Schweitzer.

The earlier posts observed that although nearly seems to mean just about the same thing as almost, it is much rarer before negatives:

  no one

everyone

nearly 27.6K 1.29M
almost 1.06M 1.66M
almost/nearly ratio 38.4 1.29

Lucia and Susan started by asking a simple question that the rest of us missed -- what about the interaction of nearly and almost with negation in the opposite order, e.g. "not nearly" and "not almost"? Answer: it goes the opposite way! For example:

  is|was|are|were not __ enough is|was|are|were __ enough
nearly
253,000
36,200
almost
103
184,000
almost/nearly ratio
0.0004
5.08

A guest post by Lucia and Susan about this is below.

The previous discussion on nearly and almost has focused on the different behavior of these two elements when followed by universal negative expressions such as "never", "nobody" etc...Nearly feels bad, almost feels perfect. Now, it is interesting to see that there are other negative contexts in which almost and nearly behave differently: interestingly enough, the fact get reversed and when negation precedes these elements, almost is pretty bad and nearly is prefect!

Thus "not nearly" is commoner than "not almost" (2.22M vs. 338K). Moreover, "not almost" seems in many cases ungrammatical; if not ungrammatical "not almost" has to be in some sort of echo context and seems to have a different interpretation from "not nearly" Compare:

1. Not nearly as good as it used to be
2. *?not almost as good as it used to be

Notice that "not nearly as good" seems to mean something different from "not quite as good as it used" to be, namely "way worse than it used to be".  Let's look at "NEG almost" and "NEG nearly" in an echo context (with almost in the previous context):

3. A:   Given that we are almost done with the paper, we should celebrate.
  B:   I am not almost done with my paper --
      i. I am done!
      ii: I'm still working on the results section.

B negates the almost "doneness" of the paper, by either saying that he is done or still at some earlier point than what he considers the "almost done" point.

Now suppose B said:

4. I am not nearly done with my paper

This could only mean that one is far from being done.  The same seem to happen with numbers:

5. The candidate got almost 500 votes (slightly less than 500, suppose 470)
6. ( in echo context) The candidate didn't get almost 500 votes (any number from  0 to ~470 or more than 500 votes)
7. The candidate didn't get nearly 500 votes (we got way less than 470, say 50).

So: almost and nearly, when in the scope of negation, behave differently. Not almost X is (in the relevant case) the complement of whatever we think almost X is, while not nearly X picks up a number that is on the other side of the scale.
(Something perhaps on these lines:  X= 500
Nearly X = from X-1 to 500 - (10 % 500), that is from  499 to 450
Not nearly X = from 0 to 0 + (10% 500), that is from 0 to 50)

One possibility would be to develop Jerry Sadock's idea from June 24: "Nearly n connotes that n exceeds (hence is better than) what was expected or hoped for [...]". Suppose that, whenever we use nearly, we suggest that some a value that was expected for was exceeded. Hence when we say "Nearly 10 dollars": we convey that were expecting less (suppose 5 dollars), and hence the obtained result exceeded the expectation, by having way more than 5 (suppose 9). Could we then say that when we use "Not nearly 10" we convey that a reasonable expectation nonetheless was 5. but that our amount exceeded (in worse) not only the goal but even the reasonable (5 dollars) expectation?

We think the conventional implicature hypothesis works nicely in the positive contexts, where there is an alternation between almost and nearly. The problem is that it is not clear how to build in an expectation in the negative one, where the alternation between almost and nearly is less productive and nearly seems to be able to appear in out of the blue contexts.

One other suggestion comes from Italian. Italian has only one word for translating nearly/almost, quasi. Quasi behaves exactly like almost under negation. It is weird unless an echo context is provided.  How does then the concept expressed by "NOT nearly" get conveyed? Interestingly,  Italian uses an adverb that is semantically related (but opposite) to nearly: lontanamente (which means "farly/by far")

8.   Il candidato non ha preso lontanamente 500 voti.
  The candidate didn't get farly 500 votes.

But if lontanamente means "farly", then shouldn't "not farly" mean "near"? Is this a case where one would want to claim a failure of semantic compositionality, a "close miss" situation?

Notice that in these cases we cannot say that lontanamente has wide scope over the whole sentence, given that lontanamente has to be in the scope of negation:

9.a :  *Lontanamente la miglior pizza della città
  By far the best pizza in town
9.b Non è lontanamente la miglior pizza della città
  It is not nearly the best pizza in town

Interestingly, in the Italian cases, we can have an optional even: "nemmeno" or "neanche".

10.   Il candidato non ha preso neanche lontanamente 500 voti.
  The candidate didn't get even by far 500 votes.

Now, we would like to suggets than in the English cases, when nearly is in the scope of negation, a covert even is present (or at least, nearly acts as if it were). Suppose in these cases we have an implicit ranking of propositions (in terms of most to least likely to happen or in terms of a value going from high to low) with respect to a goal (500 votes). [We use X > Y to represent the notion that X is better than Y.]

Get 500  votes   >
Get nearly 500  (say, 470 votes) >
Get not nearly 500 votes.

In turn, "not nearly 500" implies a ranking like the following 

Get 400 votes > 
Get 300 >
Get 200 > 
Get 100...

And in this case, our covert even would tell us that the proposition with the lowest ranking (let's say 0-100) holds.  Now if this is reasonable, something similar could be at play when nearly is in construction with "no one", "nothing", and so on: a covert even would get triggered by the interaction of nearly and negation. But even is incompatible with such quantifiers   (*even no one liked it); and hence such sentences would be ruled out.

Posted by Mark Liberman at July 9, 2007 08:26 AM