August 10, 2004

Single out and double up

Reader ACW reported a nice eggcorn overheard on NPR: "signaled out" for "singled out".

ACW emailed:

The eggcorn hunt must be getting old for you professionals.  But when I see or hear an unfamiliar one, I feel compelled to Report It At Once To The Authorities, and for the moment that seems to mean Dr. Liberman.  Let me know if you get tired of it.  In the meantime, as you have guessed, I have another one.

I heard it on NPR this weekend.  A family member of a soldier recently mustered to Iraq, I think it was, felt unfairly "signaled out".  Google finds a thousand or so of these, and in the twenty seconds or so that I spent getting a sense of the statistics, I got the impression that the eggcorn was outnumbered by the acorn by a factor of several hundred.  Caution: the hits for the unmarked form, "signal out", are dominated by prosaic devices sending out the usual kind of signals.

It's always tempting to guess what the attraction of an eggcorn is.  There is always something about it that makes fortuitous sense.  It doesn't have to make much: consider the number of collocations like "by and large" that we use with no discernable compositional rationale.  But to succeed as an eggcorn, a collocation has to have something going for it, a theory that licenses it and makes it seem reasonable.  In the case of "being signaled out", I'm having trouble seeing it.  Maybe it's getting support from the slightly-weird fossil expression "signal opportunity"; maybe users have an image of being selected with a pointing finger, said finger being a "signal" of that selection.

Well, actually, when it comes to eggcorns, we're pretty much all amateurs. There's no official subdiscipline of eggcornology, nor any International Journal of Eggcorn Studies. Not even a panel discussion at the LSA.

With respect to combinations like "single/signal out", it would be wrong to expect full compositionality. The combination of verbs with intransitive prepositions is one of the many pseudopods of morphological quasi-regularity that extend into the phrasal domain in English. There are lots of regular patterns, lots of idiosyncratic exceptions, and lots of small to medium-sized subregularities in between. So when you choose someone you've singled him out, but when you choose two people, you haven't doubled them out, or even coupled them out (though you might have coupled them up, depending on things work out). "Single out" seems to resonate with "point out", and so it's a little surprising that most people don't think you can signal someone out (unless you're an umpire). Of course you can't designate him out either (again with a possible exception for fancy descriptions of umpires). Though you can pick him out.

There are two problems in parallel for the distribution of verbs with intransitive prepositions. The easy part is that some particular verb-preposition combinations get idiomatic meanings. The hard part is that even the most regular and compositional-seeming combinations often don't work. See this post from last spring for a discussion of this issue with VERB+up.

ACW's post brings up some other points that are worth more discussion. For example, he invites us to "consider the number of collocations like 'by and large' that we use with no discernable compositional rationale." So I did so, and I realized that I don't have any idea what that number is. And I wonder if anyone else does either.

Now, it's not easy to count the members of a category like that. When you ask "how many words are there in English?", the answer depends on what you mean by "word", what you mean by "English" -- and even what you mean by "how many" and by "are". Different answers to these questions can change the answers by large factors. Still, you can define the question more precisely, and find some specific answers in the literature, or make some counts based on text databases, on the lemma count of specific dictionaries, or on various combinations of these.

In trying to count the number of noncompositional phrases in English, similar questions arise. For example, what is "noncompositional"? Are we talking about phrases with completely unpredictable semantics, like "red herring"? or about phrases with somewhat unpredictable semantics, like "chair lift" vs. "face lift" vs. "fork lift"? Many phrases are often used compositionally but also have a more specialized meaning in combination, like "run out"? Since complex nominals and verb-particle combinations are both quasi-regular, it's pretty hard to draw the line between what's compositional and what isn't, even before getting to more complex idioms or to phrasal terms of art.

Still, you could set a operational compositionality threshold of some sort -- or a range of thresholds -- and ask how many thus-defined noncompositional collocations there are in English. Unfortunately, I don't know any good overall treatments of this question. You could start by pulling out the multi-word items listed in dictionaries. But these lists are radically incomplete, and many of the things on them are semantically compositional in any case. It probably makes more sense to ask the question in more psychological terms -- for how many phrases do typical English speakers store information about form and meaning, independent of their general process for determining the form and meaning of phrases? You might be able to get an answer by sampling techniques.

I'm pretty sure that the following non-specific statement is true: "The number of phrases about whose form and meaning a speaker stores (at least some) information is normally many times larger than the number of words the same speaker knows".

And as for "by and large", by the way, the original semantics is fairly discernable to sailors, if not to the population at large. But even for sailors, "by and large" is one of the unknown number of terms in the phrasal lexicon.


Posted by Mark Liberman at August 10, 2004 07:51 PM