anacoluthon, anyone?

[Arnold M. Zwicky]

I've been working with Stanford graduate student Liz Coppock on
(inadvertent) syntactic blends, a project that requires us to
distinguish syntactic blends from a number of other error types.  The
tricky part in categorizing errors is that you have to make plausible
speculations about what went on in the mind of the person who made the
error.  An example: I've posted a number of times here about the
distinction between (inadvertent) Fay/Cutler malapropisms -- speaking
metaphorically, you aim for a word and get a phonologically associated
word instead -- and classical malapropisms -- you get the word you're
aiming for, but the phonology of what you have in that slot isn't what
most other people have.  Once we sort the examples into these two
categories, setting aside some number that we can't categorize with
much assurance, it turns out that the two error types have strikingly
different properties.

The processing metaphor for syntactic blends is that you start out
trying to express some semantic content, entertain two alternative (and
competing) expressions of that content, and end up producing a kind of
amalgam, something that has parts of both: "It really becomes down to a
question of..." (NPR commentator, Morning Edition, 2/26/04), involving
"It really becomes a question of..." and "It really comes down to a
question of...".  But there are lots of things that can look sort of
like this on the surface, but probably arise from other causes.

For example, many of the things labeled as pleonasms or (unnecessary
and unwelcome) redundancies.  MWDEU, under "redundancy", notes some
commonly proscribed locutions of this sort that nevertheless might be
useful: refer back, final result, collaborate together, continue on,
end result, past history, general consensus, personal friend, off of,
"and many many more".   Similarly, "You'll both have a lot in common",
or even "You'll both share a lot in common".  Or "The parallels are
amazingly similar" (Karen Liebreich, author of Fallen Order, comparing
scandals in the Roman Catholic Church at two different periods, on NPR
Weekend Edition Sunday, 9/26/04).   Here, the intention is to pile on,
to repeat for emphasis and clarity; you want both expressions, and
they're not in competition with one another syntactically.

Sometimes you want two or more expressions with similar meaning, each
contributing a slightly different shade of meaning: "small little
present" and "little tiny present", where one adjective might
contribute primarily size information and the other primarily affective
(hypocoristic) content.  Again, the expressions aren't in competition
with one another syntactically, and there's no reason to see these as
blends.

Somewhat more subtly, a word (or sometimes larger expression) can
intrude within an including construction, replacing an element of the
larger construction with something that strikes you as especially
appropriate or sensible: syntactic eggcorns like "harp back" (for "hark
  back", "free reign" (for "free rein"), "bunker down" (for "hunker
down").

And then there are extensions of some lexical class associated with a
construction: Object-to-Subject Raising lexically generalized to allow
things like "I think Carlos to be a spy" ("think" belonging
semantically to a lexical class that includes OSR verbs like "believe"
and "suppose").  While it would be possible to view "I think Carlos to
be a spy" as a blend of "I believe/suppose Carlos to be a spy" and "I
think (that) Carlos is a spy", it doesn't seem insightful to do so;
actually, one of these putative contributing constructions provides the
semantics, the other the syntactic form, and neither of the
contributors is uniquely determinable.

And then -- I'm almost done here -- there's defaulting to some unmarked
element, the unmarked preposition "of", for instance: "I'm bored of
linguistics", which *could* be seen as a blend of "I'm tired/weary of
linguistics" and "I'm bored with/by linguistics", though, as in the
previous type, nothing is gained by that.

Finally, we come to a type that is much more difficult to distinguish
from clear examples of syntactic blends: cases involving, not
competition between alternative plans, but either a failure to plan
ahead far enough syntactically (though you have at least a rough idea
of the semantics you want to convey) or the abandonment of one plan and
adoption of a different one, in medias res.  Until a few days ago, I
thought that the appropriate label here was "anacoluthon", and that
still seems like a pretty good choice, though probably the two
subtypes, lack of syntactic commitment and syntactic shift, will need
to be distinguished.

What happened a couple of days ago?  I graduated from the discussion of
anacoluthon, in the works of the famously stumbling U.S. presidents
Warren G. Harding, Dwight D. Eisenhower, and George W. Bush, in Michael
Silverstein's Talking Politics: The Substance of Style from Abe to "W"
-- where I thought I pretty much understood the concept -- and actually
looked up the technical term "anacoluthon" in various specialized
dictionaries, only to discover a high degree of fuzzy thinking and
variability in use.  I started with several on-line dictionaries but
was so dismayed and puzzled by what I found there that I turned to
heavier hitters.

The Penguin Dictionary of Literary Terms and Literary Theory (4th ed.)
defines the term succinctly as "Beginning a sentence in one way and
continuing or ending it in another" and supplies as its one example
something that fit my prior understanding perfectly: 'You know what I
-- but let's forget it!'

Alas, Beckson/Ganz, Literary Terms, has a woolly definition -- "A
sentence which does not maintain a consistent grammatical sequence" --
and gives as its one example a classic sentence-initial dangling
participle.  You can sort of see what they had in mind: a
sentence-initial free adjunct is supposed to be (for some value of
"supposed to be") followed by a main-clause subject that supplies (the
reference of) the subject that's missing in the adjunct, so if it
isn't, the "grammatical sequence" is "inconsistent".  But looking at
such examples, it's hard to see what they have in common, from the
processing point of view, with such anacoluthic delights as this
material (punctuated as a sentence) from Eisenhower's West Point
address on 6/5/60 (from Dwight Macdonald's Parodies, p. 450): "This is
what I hope to do myself, so far as it is proper and the people who
will meet within a few short weeks to take over the direction of
campaigns--I am ready to do my part."  So-called dangling participles
might be savaged by usage manuals, but it's very clear that people who
produce them, as I did just above, intend to produce just what they
come out with; maybe their grammar isn't yours, but they're not messing
up their sentences in the course of production.  (Remember Fay/Cutler
vs. classical malapropisms?)

You're thinking that if we start down this road, than all sorts of
ungrammaticalities or purportedly non-standard usages are going to get
labeled as anacolutha, and you're right.  Dupriez/Halsall, A Dictionary
of Literary Devices, dives right in.  Their definition is about as wide
as you can get -- "a breakdown in the syntactic construction of a
sentence" -- and their examples are all over the map: dangling
participles, resumptive pronouns (colloquial, and perhaps non-standard,
but usually not any kind of production error), left dislocations
(ditto), run-on sentences ("He couldn't go, how could he?"), Molly
Bloom's monologue (now *this* is anacoluthic), recastings and
restartings of sentences, and (technically) unfinished sentences ("If
you only knew").  Grammarians, dialectologists, sociolinguists,
discourse analysts, all of you, weep.

Fowler the Great, in DMEU, produces two examples of considerable
interest.  In both, what's going on, intuitively, is that you haven't
planned far enough ahead syntactically, so that at a later point you're
no longer aware that choices you made early on constrain your choices
now, and you make a later choice that has the right semantics but fails
to conform to the earlier determining material:

(1) Can I make you understand that if you don't get reconciled to your
father what is to happen to you?  [The earlier "that" should have
committed you to a declarative rather than interrogative complement.]

(2) Pliny speaks of divers engaged in the strategy of ancient warfare,
carrying tubes in their mouths & so drew in necessary air to their
lungs.  [The earlier present participle "carrying" should have
committed you to a parallel "drawing" rather than the past tense
"drew".]

These are failure-of-commitment anacolutha.  Various types of
determination by the nearest fall into this subcategory: the failed
subject-verb agreement in "Everything from doorknobs to live alligators
are for sale" (NPR reporter, All Things Considered, 6/9/04) and "Let's
see which one of the two of are are next" (Beverly Hills 90210
episode); the failure of verb-form parallelism in "She had never and
was never going to wear it" (NPR reporter, All Things Considered,
8/16/04) and "They would delay, or not get tested at all, for sexually
transmitted diseases" (NPR reporter, Morning Edition, 8/14/02).

Failure-of-commitment anacoluthon is certainly to be distinguished from
syntactic blending (though it's almost always possible to concoct a
blend analysis) and possibly from shift anacoluthon of the Eisenhower
type.

In any case, it's not hard to see blending and anacoluthon as subtypes
of on-line glitches in sentence planning (distinct from positioning
errors: doubling errors, either anticipatory or perseveratory, and
exchange errors).  If we can find a way to classify these planning
errors as either blends or anacolutha, then we can ask whether there
are symptoms of their difference -- in particular, whether blends show
a significantly higher amount of overlapping material than anacolutha.
The overlap effect is very strong in morphological blends like
"chunnel" /C^n at l/, which doesn't merely begin like "channel" and end
like "tunnel", but has both of the contributing words sharing almost
all of their content with the blend: /C_n at l/ and /_^n at l/
(see Stefan Gries, "Shouldn't it be _breakfunch_?", Linguistics 42
(2004)).

Actually, I suspect that the classification of planning errors as
blends or anacolutha will turn out not to be drastically hard -- once
we have a corpus of anacolutha, in the proper understanding of that
term.  We *do* have a wonderful corpus of blend slips, thanks to Gerald
Cohen's 1987 Syntactic Blends in English _Parole_, which Coppock has
scanned in for further analysis.  Anyone have any ideas about sources
of anacolutha?

arnold (zwicky at csli.stanford.edu)