the thin line between error and mere variation II

Arnold M. Zwicky zwicky at CSLI.STANFORD.EDU
Sat Jul 3 16:56:01 UTC 2004

On Jul 3, 2004, at 4:57 AM, Dennis R. Preston wrote:

> as arnold well knows, this mismatch has been looked at formally in
> so-called "linguistic insecurity" studies...

yes, i should have cited this literature.  and also made it clear that
there are two rather different sorts of examples in what's been posted
so far; they differ in the degree to which speakers are aware of an
alternation.  the classic linguistic insecurity studies involved
variables that people were aware of; usually, they'd been instructed,
formally or informally, that one variant was "wrong" or "bad" or
"incorrect".  g-dropping and the possessive antecedent proscription are
two such cases.  (i have yet to find anyone who has a problem with
things like "Mary's mother admires her" who *wasn't* taught a "rule".)
in such cases, we get misreporting, sometimes hypercorrection,
sometimes (as with split infinitives) avoidance.

but the "skreek" (for "street") and "quater" (for "quarter") cases are
somewhat different.  the level of social awareness of the non-standard
variants hovers near zero; it's the sort of thing only a linguist would
notice.  still, the spellings provide a model for how the words
"should" be pronounced, so people believe that they *do* pronounce them
that way.

there are many other types of cases, of course.  even when there's
instruction, some non-standard variants are much more accessible to
awareness than others; for example, it's hugely easier to get
AAVE-speaking kids to notice (and standardize, for classroom purposes)
multiple negation than habitual "be".

and when there's no instruction, people who use the non-standard
variant often simply take it to be perfectly ordinary; this is my
position on the GoToGo construction ("She's going to San Francisco and
talk on firewalls").  our grad student laura staum estimates that about
20% of american speakers (at least those on the internet) use such
examples and judge them to be acceptable; the others view them as
speech errors.  i'm a GoToGo speaker -- the example above was from me
-- and my first response on having the construction called to my
was to ask, "how else would you say that?"  (the answer is: the longer
"She's going to go to San Francisco and talk about firewalls", or the
semantically non-equivalent "She's going to San Francisco to talk about

well, there's more still.

and, as dInIs noted, trained linguists are no better at judging which
variants they use, how often, than normal people.  how could we be?
we're too busy talking and writing to monitor the details of our
productions.  hell, we're not even very good at estimating, off the
cuff, the frequency of features in corpora, since some instances escape
our notice and others weigh too heavily.  transcribing data and
hand-counting instances of features are both tremendously difficult
(and tedious) -- despite the fact that we're trying to be as aware as

ok, this is preaching to the choir...

arnold (zwicky at

More information about the Ads-l mailing list