Novelty vs. Expandability (of What?)

[Wallace Chafe]

I'd like to express some support for Doug Oliver and Eleanor
Batchelder.  Sorry this has to be long.

It's a good idea at the beginning, I think, to separate the issues of (1)
novelty, (2) expandability (how long or complex "sentences" or whatever
can be), and (3) the question of what kinds of units are involved anyway
(should we always be talking about sentences?).

So far as novelty is concerned, in the middle of May I heard
someone say "Frank Sinatra died."  I don't believe I had ever heard
that sentence before, and yet  I knew what it meant (up to a point).
And I suppose that the person who uttered it had never said it
before (at least before that day), and yet she was able to produce it
without difficulty (or at least without grammatical difficulty).  It was
novel, but hardly very long or complex.

There doesn't seem to be any marvelous new insight here.  Why
would anyone find anything surprising about this kind of novelty?
If I understand what Esa Itkonen was getting at with his original
question, it was that the idea expressed as "Frank Sinatra", the idea
of dying, and the way those ideas were put together to say "Frank
Sinatra died" were all totally familiar to speakers of English.  The
only thing that was new was their combination.

There's one more thing that should be said regarding novelty.
There's an important sense in which just about everything people
say is novel, simply because every human experience is unique.  If I
say, for example, "I saw a deer this morning," I may have used
exactly that sentence (as a sequence of words etc.) before, but on each
occasion the experience communicated was different:  a different
mental image, a different deer, perhaps a different emotion,
probably a different location, and certainly a different time.  We
have to plug each unique experience into our limited linguistic
resources in order to talk, and to some extent in order to think (and,
yes, we do that differently in different languages).  The point is that
the experiences underlying language are always in some respects
novel, even though the form of the language may be the same.

Such considerations aside, if examples like "Frank Sinatra died"
were the whole story, one could think of language as providing a
number of patterns and a (very large) number of lexical items that
could be inserted into those patterns.  It's not that one learns
language just by memorizing sentences, although I'm sure that
happens too.  People don't get enough credit for their huge
memories, which certainly extend beyond individual words.  But of
course one also learns patterns and lexical items and is able to
combine them.  Although those patterns and lexical items can
change and be augmented over time, at any one moment in an
individual's life the number of possible combinations is vast, but
finite.  Plenty of "novel sentences" are easily available without
much fancy footwork.

What Chomsky came up with in the 1950s was the idea that the
patterns could be expanded without limit, and not diachronically
but synchronically.  At that point there was already a certain gap
between theory and observation.  While the kind of novelty
illustrated by "Frank Sinatra died" is intuitively obvious and can in
fact be observed all around us, the kind of novelty provided by
infinite recursion is by definition impossible to observe.

What do we find when we examine how people actually speak?
We find for one thing that sentences are something of a problem, if
nothing else because prosody and syntax don't always coincide.
Ignoring that problem, we do find that sentences, whatever they
may be, vary from very short to very long.  My own finding (and
I've looked at a lot of ordinary speech with this in mind) has been
that people insert sentence boundaries whenever they decide (on
the fly, often for some passing reason) that some kind of closure has
been reached in the flow of ideas.  It's an on-line decision and,
judging from repetitions of the same content by the same person on
different occasions (a very worthwhile kind of data to examine),
sentences don't seem to, or need not, reflect units of mental storage.
On-line decisions about closure are interesting, but there's more to
language structure than that.

Sentences are intermediate in length between smaller prosodic
phrases (expressing foci of active consciousness) and expressions of
larger discourse topics (with material in peripheral consciousness),
both of which are subject to interesting cognitive constraints that
don't apply to sentences per se.  Prosodic phrases ("intonation
units") are subject to what I've called the one-new-idea constraint,
which keeps them from getting very big.  I think it has a much
more important effect on the shape of language than George
Miller's 7 +/- 2 constraint, as I've tried to show in numerous places.

Topics may be short or long, but the interesting thing is that, once a
topic has been opened in a conversation, there's an expectation that
it will sooner or later be closed, after which another one can begin.
Opening a topic is like creating an open parenthesis that demands
eventual closure.  Topics are what keep language moving.  There's a great
deal to be said about this, but here I might just point out here that the
ludic analogy is more relevant to topics than to sentences.  The length of
a tennis game seems especially apt.  Leaving aside the prolongation of a
game through repeated deuces, how many times can the ball cross the net
before a point is scored?  Limits on skill and stamina would seem to keep
the number within asymptotic bounds, but any arbitrary limit might in
theory always be extended by one.

Topics are like that.  No topic goes on all day, but it's impossible to
assign anything but an arbitrary limit to topic size.  Sentences are
usually properly contained within a topic, but on rare occasions they
may expand to be coextensive.  In terms of clauses, that can happen
in a trivial way through the use of "and" to link every clause.
Prosodically it can be done by postponing a falling pitch until the
topic is concluded.  I've observed this with 10-year-old boys, when
they repeat the currently popular question intonation at the end of
every phrase before finally letting their pitch fall when I'm about
ready to go home.  I found it also with a couple of our "pear stories",
where the film was described with what sounded like a shopping
list of events that didn't end until the narrative was finished.

My general plea is that we distinguish novelty from expandability,
and that we move beyond the rather special and sometimes
puzzling strings of words that have been called "sentences," as if
they were all that language had to offer, to a broader concern for the
richness of what happens when people actually speak.

--Wally Chafe