March 12, 2004

Overnegation supererogation

Following up on this followup to our long list of overnegation posts, Steve (Language Hat) emailed a link leading to the sentence

"As for the site, I'm going to try to get back on track with updating soon, but don't be surprised if the new story doesn't debut as late as April."

Steve wrote: "I'm pretty sure he (Derek Kirk Kim) meant '...if the new story debuts...'"

I agree.

However, this example may be different from "fail to miss" and "impossible to underestimate" and the other overnegation examples we've discussed. In those cases, the source seems to be natural processes like construction generalization and negative concord, revealed because the processing difficulty of multiple negations prevents the results from being noticed and edited out. In this case, the source itself was probably editorial second thoughts. To make a long story short, I think the the writer first thought (or even wrote) something like "... don't be surprised if the new story doesn't debut until as late as April", and then took out "until" without taking out the associated negation. The difficulty of dealing with multiple-negation sentences helps explain the error, but (if this analysis is correct) the process is a different one.

Here's a longer "pop semantics" analysis of the case.

If we say that something changes state at time X, we often mean that it doesn't change state until time X, and vice versa. Whether to say "CHANGE-OF-STATE-VERB TIME-ADVERBIAL" or "not CHANGE-OF-STATE-VERB until TIME-ADVERBIAL" is then a matter of nuance. For example, a description of Delayed-Onset Muscle Soreness says that "It typically begins a day or two after the activity that caused it"; but it could as well have said "It typically doesn't begin until a day or two after the activity that caused it." A page on fetal development say that "The expression of the sex characteristics doesn't begin until about 6 weeks after gestation", but it could also have said "The expression of the sex characteristics begins about 6 weeks after gestation." It's easy to find similar examples with stopping instead of starting, or with changes like opening or closing.

In such cases, the issue seems to be whether you want to focus on the period before the change of state, or the period after it. For example, the DOMS explanation was followed by another clause focusing on the following period: "... and lasts several days to a week".

When you put the whole thing in the complement of a propositional attitude verb like "surprise", this time-period focus can become part of what the attitude applies to. Here's a sentence from a page of poison-ivy information: "[D]on't be surprised if the rash doesn't begin until a day or two after the child has touched the plant." Here the rashless period is precisely what we're not supposed to be surprised about. Rewriting the sentence as "Don't be surprised if the rash begins a day or two after the child has touched the plant" may cause confusion (though in speaking, intonational emphasis on "a day or two after" would clarify things).

You can get the same effect of focusing on the rash-onset delay by adding a qualifier like "as late as" to the time adverbial: "Don't be surprised if the rash begins as late as a day or two after the child has touched the plant." It also works to use both the belt and the suspenders -- "Don't be surprised if the rash doesn't begin until as late as a day or two after the child has touched the plant" -- at the cost of adding complexity to an already intricate little syntactico-semantic gizmo. However, until and the negation are a package deal -- they have to come and go together.

In the example that Steve cited, the writer may have gone down this road, and then decided to simplify things by letting "as late as" help "don't be suprised" find its focus, eliminating "until" as an unnecessary bit of mechanism. The extra negation was left behind by mistake.

Posted by Mark Liberman at March 12, 2004 12:33 PM