My posting yesterday on reversed substitute (one in a series on the phenomenon) moved Mike Pope to ask me about another Exchange verb, swap. I seem not to have posted here about this particular verb, but I did take part in, um, exchanges on swap on ADS-L back in 2005. The trigger was a 1/17/05 posting “”swap”: inversion of meaning” by Jon Lighter:
This is much like the odd shift in the meaning of “substitute” commented upon some weeks ago [that would be reversed substitute]:
“New dietary guidelines coming out Wednesday are expected to place more emphasis on counting calories and exercising daily, along with swapping whole grains for refined ones and eating a lot more vegetables and fruits.” — Gov’t: Calories, Not Carbs, Make You Fat (AP) January 12, 2005
This says to me (nonsensically) that if you’re eating whole grains now, you should switch to refined ones.
The news story quote has swap NEW for OLD, but Lighter expects swap to have the argument structure swap OLD for NEW. And OLD for NEW in fact is the argument structure in NOAD‘s entry for swap.
I’ll start with the NOAD entry and then cycle back to the 2005 ADS-L discussion. NOAD has three sub-entries:
[a] take part in an exchange of: we swapped phone numbers | I’d swap places with you any day | [no object] : I was wondering if you’d like to swap with me.
[b] give (one thing) and receive something else in exchange: [ex1] swap one of your sandwiches for a cheese and pickle?
[c] substitute (one thing) for another: [ex2] I swapped my busy life on Wall Street for a peaceful mountain retreat.
[a] is mutual substitution: there are two roles, or slots, in some situation, filled by two referents, OLD1 and OLD2, at an earliier time, with these switched at later time.:
OLD1 is replaced by NEW1 = OLD2;
and OLD2 is replaced by NEW2 = OLD1
The semantics is completely symmetric, with no prominence attached to either of the roles.
In [b] and [c], however, the roles differ in their discourse status: one is discourse-old (D-OLD), given in the discourse, and the other is discourse-new (D-NEW). The default strategy for ordering the parts of a sentence is for D-OLD to precede D-NEW. At the same time, ceteris paribus, when parts of a sentence denote situations that are ordered in time, the default is for the linear ordering of the parts to mirror the temporal ordering of the situations. The result is a linkage between D-OLD, (temporally) OLD, and earlier positioning in the sentence.
For a mutual-substitution verb like swap, these linkages result in the preferred argument structure
V OLD for NEW
which is in fact the standard, illustrated in the two examples in the NOAD swap sub-entries [b] and [c].
[ex2] in [c] is straightforwardly OLD for NEW, but [ex1] in [b] is a bit more complicated, because it’s a question, so that what’s old and new is to be judged from the point of view of the addressee. With that proviso, [ex1] is also OLD for NEW.
Now back to Lighter’s swap whole grains for refined ones (swap NEW for OLD) example. The example elicited a certain number of responses saying that the example was incomprehensible, an error resulting from ignorance of the standard argument structure, and so on. But also, from Gerald Cohen on 1/17/05:
This looks like a [syntactic] blend: “subsitute whole grains for refined ones” + “swap/switch refined grains for whole ones.”
And then my response (lightly edited here):
To recap the old substitute discussion [note that 13 years ago this was an old topic on ADS-L]: there are three usages:
(1) original: substitute NEW for OLD
(2) substitute OLD with/by NEW (an extension of the replace pattern)
(3) reversed: substitute OLD for NEW (David Denison suggests that this is a blend of (1) and (2). Note that both (2) and (3) have the virtue of putting OLD before NEW, iconically to the preferred sequence of old and new information)
Now, the obvious analysis of Lighter’s swap example — swap NEW for OLD, instead of swap OLD for NEW — is that it’s an extension of the pattern in (1) to new verbs with semantics similar to substitute (swap, and possibly, as Cohen suggests, switch as well; I’d add trade). This extension would be facilitated by the fact that swap, switch, and trade are verbs of mutual substitution, for which (in central uses) NEW replaces OLD and OLD replaces NEW: in I swapped/switched/traded my marbles for her baseball cards, the marbles replace the baseball cards and vice versa. Which participant is expressed by the direct and which by the oblique object could be entirely determined by matters of focus and topicality in the discourse.
So Lighter’s example could result from an extension of a construction to new head verbs semantically similar to existing ones, a phenomenon that is very widely attested.
Cohen’s syntactic blend analysis strikes me as pretty implausible. In clear examples of syntactic blends, it’s plausible to maintain that the speaker (or writer) was entertaining two competing plans for expressing the same meaning (or very similar ones) and ended up with elements of each, usually in pretty simple ways (by splicing or substitution, to use the terminology from David Fay’s 1981 paper, “Substitutions and splices: A study of sentence blends”). Cohen’s proposal is that swap NEW for OLD results from blending
substitute NEW for OLD
and
swap/switch OLD for NEW,which involves, at the surface, switches in three places, holding only the preposition for constant. It is, of course, possible that blending takes place at a more abstract level of analysis, in which the allocation of OLD and NEW to syntactic arguments is separated from the choice of for as the oblique marker.
Extensions of constructions to new head items semantically related to existing heads could always be analyzed as syntactic blends, with a certain amount of ingenuity (as above). but this seems to me like the wrong way to go, especially since people who produce these extensions so rarely treat them as inadvertent errors; in general, the extensions look like innovations in grammars, made independently by some number of speakers and then spread to other speakers by the usual means.
Leave a Reply