NYT Goofup is a syntactic blend

[Laurence Horn]

At 5:08 PM -0600 3/17/02, Gerald Cohen wrote:
>On 3/17/02, Alice Faber wrote about the blend "many dioceses after
>another" (from "many dioceses" + "one diocese after another"):
>>This particular instance struck me as an uninteresting example of what
>>happens when you get interrupted in the middle of editing; you've made half
>>the changes from one structure to another and the phone rings.
>
>
>    Uninteresting? Au contraire! I spent years collecting such
>examples and then  tried to draw general linguistic conclusions from
>the collected evidence; it's written up in my article "Contributions
>to the Study of Blending", pp.81-94, in _Etymology and Linguistic
>Principles_ , vol. 1: _Pursuit of Linguistic Insight_ (ed.: Gerald
>Leonard Cohen), Rolla, Missouri: published by the editor; the volume
>was favorably reviewed in _Language_ and _Journal of Indo-European
>Studies_.
>
>     Syntactic blends are of interest for producing ellipsis,
>redundancy, grammatical change, and semantic change. They are
>relevant for possibly challenging the Law of Least Effort, the view
>that language is a code, the supposed rigorousness of distributional
>rules, the view that language has structure. They are a challenge to
>generative grammar. Some syntactic blends are apparently based on
>three (rather than just two) underlying constructions. And some
>blends may be termed "repeated blends", e.g. "time after time" +
>"again and again" blended to "time and again;" then "time and again"
>apparently blended with "time after time" to produce "time and time
>again." How would generative grammar handle this latter construction?
>

I'm not saying blends aren't interesting both intrinsically and
diachronically, but I'm not convinced they represent a challenge to
current or recent syntactic theory.  This case (e.g. the article in
today's Times) strikes me as one of the better arguments for a
competence/performance distinction, one of the cornerstones of early
generative grammar.  The fact that the later (web-based) version of
the Times article "cleaned up" the original blend indicates that it
was indeed a performance error along the lines Alice was describing,
and that it was so viewed.  This is not to say that such
"parole"-based blends can't influence the "langue", and linguists
from Jespersen on, if not earlier, have recognized this fact.  "time
and time again" is a nice example; another is "daylight savings
time", as I believe Bolinger pointed out, and the development of
comparatives in many languages is another still.  I'm not sure where
the challenge to generative grammar lies, though.  In fact, one of
the earmarks of early (and, I'm told, current) work in the Chomskyan
paradigm is to take complex structures--relative clauses, nominal
complements, etc.--to result from a "blending" of two simplex
structures through a generalized transformation.  Certainly not all
versions of generative grammar would allow this, but it's not
incompatible with the conceptual structure of generative grammar
itself.

>     "Structure," "system" imply something static. Processes in
>language (e.g. blending) put the focus on something dynamic.

Sorry, I don't buy this either.  Most current work in formal
semantics is done within dynamic models.  It's almost unfashionable
to submit an abstract that ISN'T couched within a dynamic semantics
model, and yes, they're quite systematic.

larry