This is no sh*t.

[Arnold M. Zwicky]

On Jul 30, 2004, at 2:32 PM, Jonathan Lighter wrote:

> Writing in 1959-60, Stuart Flexner averred that women use less slang
> than men.  This is also my impression.  I'll be happy to change my
> view (which at this point is essentially as subjective as anybody's)
> when there is good data to refute it.

[This will be long.  And, eventually, intemperate.  You've been warned.]

I'll start with annoyance at the way Jonathan has framed things here.
If he has one subjective impression and I have another, then there is
no burden of proof on one of us as opposed to the other.  For him to
demand "good data to refute" the claim is to take it as a priori more
plausible that the alternative claim (that women do not use less slang
than men).  Perhaps this is just a rhetorical move to put me down, to
wrest an advantage in argument.  Perhaps he really believes that his
subjective impressions have greater validity than other people's.  This
would be understandable.  I'm inclined to trust my instincts myself.
But I've had enough experiences in which my subjective impressions and
initial beliefs turned out to be flawed or even flat wrong -- my
research program for decades started with my attempt to demonstrate the
truth of a claim that I came to believe, on the basis of evidence, was
just false -- that I'm wary of unsubstantiated claims.

[Once again, I want to set aside taboo vocabulary, for reasons I've
already suggested and will take up again later.  Jonathan expresses
some doubt that we can distinguish this part of the slang vocabulary, a
position I don't understand at all.  Lexicographers routinely affix
usage labels to words, for instance "vulgar", "offensive", and
"obscene" -- AHD4 manages to assign all three to "poontang" -- and once
we filter out special cases like racial/ethnic slurs, we've got a list
of taboo items.  There might be a few unclear cases, but life is like
that all over, if you're a linguist.]

Actually, I think there are several a priori reasons to be dubious
about the Flexner claim (as I'll call it).  Reason 1 I've already
sketched, and Wilson Gray has picked up on it: The Flexner claim is
about women in general -- all women, in all times and places,
regardless of age, class, race/ethnicity, etc., or (especially)
*context*.  (If you're worried about my saying that the claim is "about
women", hang on.  For the moment, I'm talking about the Flexner claim
in its own terms, but I'll get to the "about women" thing eventually.)
Even if we restrict the claim to modern American women, it's absurdly
overbroad.  The problem is that it's a claim about *what people, in
particular women, know about their language use*.  It's a claim that
most women (again, hang on; I'll get to statistics later) learn to
avoid an entire stratum of the vocabulary.  How does a girl learn this?
  She doesn't have access to what women *as a class* do; her language,
like everybody's, is learned in context.  In particular, her learning
is mostly in a local context (I don't deny that mass media might have
some effect, but I do deny that it's anything like the major force), in
which associations between choices of bits of language are associated
in complex ways with the social and personal characteristics of those
who serve as models.  In this situation, a girl can certainly learn
that particular variants are associated with femininity (and others
with femininity in specific contexts, and others with masculinity, and
others with certain kinds of informal contexts, and others with
teenagers, etc.) for the people she interacts with and identifies with.
  How does this get parlayed into uniformity across women as a class?

Studies of variation are always controlled to some extent, involving
comparisons between certain social groups or contexts, with some
variables controlled and others assumed (with fingers crossed) to be
largely irrelevant.  The results are *never* in terms of (unmodified)
"big variables" like sex, age, class, region, class, etc.; instead,
they're about, say, gender differences in the usage of particular
variants for adults in certain working-class neighborhoods of Belfast,
Northern Ireland, at a particular time.  We're able to do a modest
amount of generalization when similar differences show up in different
studies.  But -- and this is hardly a surprise -- it simply never turns
out that the differences are the same for all groups and contexts.

Take a well-studied variable like <ING> in modern English, with
variants /IN/ and /In/.  The variants show a fairly robust association
with gender, women tending to have lower rates of /In/ and men higher.
But there are also other associations -- with informality and intimacy
for /In/, with preciseness and social distance for /IN/, for example --
and the gender association isn't invariable, since there are some
studies showing identical rates of /In/ for men and women, and a few
showing higher rates for women.  It all depends on the context.

Reason 2 is much like reason 1, except that it concerns the slang part
of Flexner's claim rather than the women part.  The Flexner claim is
about slang in general -- all types of slang (excepting, for the
moment, taboo slang).  Once again, this assumes that a classificatory
dimension is available for individual speakers, in this case that
"slang" has some kind of psychological reality for them.  This is a
stretch.  It's a lot like claiming that "working class" has some kind
of psychological reality for American speakers.  But "slang" and
"working class" are *analyst's* dimensions; what speakers have is
knowledge about the contexts appropriate for the use of particular
items in their social circle.  How would these bits of knowledge get
parlayed into uniformity across slang expressions as a class?

Reason 3 is different.  It derives from the fact that Flexner's claim
embodies gender dogma, a bit of folk ideology about language, and *no*
bit of folk ideology should be accepted uncritically into a scientific
analysis.  Every single bit is suspect.  (By the way, Flexner's claim
has a certain amount of fame among people who study gender and
language, since it's a free-standing, unhedged, and totally
unsubstantiated retailing of gender dogma by a respected authority.)

Flexner's claim is a sub-case of the dogma that women use fewer
non-standard and informal variants than men do.  And *that*, in turn,
is a sub-case of the dogma that women cleave more carefully to social
norms for behavior than men do -- that women are more proper,
well-behaved, polite, clean, neat, and so on, while men flaunt
convention, behave badly, act impertinent, are naturally dirty and
messy, and so on; and that women act as the conservators of social
tradition, while men are creative, exploratory, inventive.  Now,
there's no question that people sometimes act in accordance with these
folk beliefs; they play out in the classic domestic impasse in which
the wife finds herself expected to pick her husband's clothes up off
the floor, expects herself to do it, and resents these expectations,
while the husband expects his wife to pick up the clothes (while
claiming the freedom to not bother with domestic trivialities) and
resents her attempts to get him to pick up after himself (which he sees
as restraints on his freedom).  But as general claims about how people
behave, these are all travesties.

The actual research shows -- surprise! -- that gender differences are
highly variable, depending on context.  A woman who works outside her
local community, especially in contact with speakers who have more
standard or formal variants will probably show a lower rate of
non-standard and informal variants than an otherwise similar man whose
life is pretty much circumscribed by the local community.  *And
conversely*.

Meanwhile, research by Penny Eckert and others on preadolescents and
young adolescents finds that the girls tend to be the leading edge for
many linguistic variants (which, being innovations, will be by
definition non-standard or informal or both) -- the innovators and the
early adopters.  At least in some contexts, it looks like the standard
gender dogma has it exactly backwards.  And that's a good reason for
being deeply dubious about Flexner's claim.

Now to the fact that Flexner's claim is formulated as being *about
women*.  The fact is, the substantive claim is merely women and men
differ in a particular way.  Now, granted, there's no way to mention
women and men simultaneously in an English sentence; one of them has to
come first.  But it's striking that gender dogma, here and elsewhere,
is almost invariably framed as a feminine deviation from masculine
practices: women use less slang than men do, women use a higher range
of their voice pitch than men do, etc.    (I would be, like, the ten
thousandth person to make this observation, but it's so important it
needs to be said again.)  We don't hear that men use more slang than
women do, that men use a lower range of their voice pitch than women
do, etc. -- though if one formulation is true, the other one is too.

The problem is that men's practices are being treated as the standard
against which women's practices are being judged: there's a base rate
of slang use, which is that for men, and women then deviate by using a
lower rate, either because they're not exposed to as much slang or
because much of the slang they hear comes from inappropriate models or
because they tend to avoid slang as being inappropriate for their
self-perceived personas or whatever (or some combination of these
factors).  Objectively, this is just a biased way of looking at
differences, and we should simply talk about differences: each of two
groups has a range of variation in the frequency of use of some
variant; the two ranges largely overlap; but the mean frequency for one
group is significantly higher than the mean frequency for the other.
(Imagine two bars, one for the distribution of women's heights, with
the mean height indicated by a stroke, and the other for the
distribution of men's heights, again with the mean height indicated by
a stroke.  The two bars overlap hugely, but the men's mean is
significantly above the women's mean -- though not by an enormous
amount in absolute terms.  Keep this example in mind.)

But from the point of view of the speakers, for any particular
linguistic variant (vs. some alternative) and two contrasted groups,
there are four ways in which such differences can arise.  The first is
simple difference, with no discernible social value associated with the
choice of variant; this is most likely to result from differential
spread within the two groups (once a differential is established,
however accidentally, it will tend to preserve itself).

In the other three ways, one pole of a variant is associated with one
of the groups.  (It might, of course, have more than one such
association.)  In the case at hand, it could be that low use of slang
is associated with femininity, or that high use of slang is associated
with masculinity, or both.  Sometimes this can be concluded from the
relationship between the overall mean (for the two groups taken
together) and the separate group means; simplifying a whole lot, if the
two group means are roughly equidistant from the overall mean, you've
got either simple difference or a double association, while if one
group mean is significantly further from the overall mean than the
other group mean is, you've got an association to (only) one end of the
scale.  In addition, you can study the way people shift their usages
with context, their attitudes towards the variants, and so on.  None of
this is easy.

Some results are pretty clear.  Consider the way people use the range
of their voice pitch.  American men -- well, certain groups of them --
tend to use the lower part of their range, significantly more than
(roughly comparable groups of) British men do, while American women
show no significant displacement from their midpoint.  That is, lower
pitch range is associated with masculinity for American men.  As a
result, Americans tend to judge men who do *not* show the lowering
effect -- British men, and a sizable minority of American men -- as
feminized, and they believe that American women "talk in a high pitch"
(meaning: higher than the natural differential between men and women
would lead you to expect).  But it's the *men* who have the marked
system in this case, not the women.  (In other social contexts, things
are different.  Russian, Polish, and Japanese women *do* tend to use
the higher end of their pitch range, while their male counterparts have
the unmarked system.)

Why I am telling you about voice pitch?  Because the way Americans --
well, certain groups of them -- use taboo slang seems to be very
similar to the voice pitch case.  American men tend to deploy taboo
slang as a symbol of masculinity (their own interpretation, of course,
is that they're simply "acting natural"), with the result that those
who use taboo slang with roughly the overall mean frequency or below --
American women, and a significant minority of American men -- are
likely to be seen as markedly feminine.  But the fact is that it's the
men who (in this case) have the marked system.  That's why I wanted to
set taboo slang aside, since it's likely to swamp any other effect.
(Even then, as various posters have  observed, things have apparently
been changing in at least some parts of society.  For many speakers
now, it's likely that taboo slang is a stronger marker of informality
and social intimacy than of masculinity.)

Now to the serious statistical questions.  Let's start with the
relatively simple case of people's heights.  Not so simple, as things
turn out; there are researchers who spend their lives untangling this
stuff.  You'd think it would be easy: just randomly sample all the
world's people (all the world's people now?  all the world's people in
recent centuries?  back a millennium?  since the appearance of homo
sapiens?) and measure their heights (at what ages?) and calculate the
means and variances for women, for men, and for the sample as a whole
(should the number of males and females be balanced?) and see whether
there are significant differences.  As a practical matter, this just
isn't possible.  Instead, we look at female-male differences in samples
with as many other variables as possible held constant.  And again and
again, almost without exception, we find that the height difference
between (the means for) women and men is significant.  The differential
recurs across groups.

It's important to realize that this result is not especially exciting,
and not just because (this time) it's a result that "everyone knows".
There's a bit of a difference between women and men, but it would be
deeply misleading to report this in the press under the headline WOMEN
ARE SHORTER THAN MEN.  The overlap is so great, and the absolute
difference so inconsequential for most practical purposes, that it's
pretty much a big yawn.

Suppose we found a similar result for (non-taboo) slang use.  It would
also be a big yawn.  No Nobel Prizes for discovering WOMEN USE LESS
SLANG THAN MEN.

But in fact it's hard to see how you could actually investigate
(non-taboo) slang use the way people have investigated height.  Even
supposing that we were reasonably clear about which expressions were
slang and which not, and that we could decide whether it was types or
tokens we're counting, and that we could pick an appropriate sample
from the whole world of slang expressions (we can't possible check 'em
all), there's the problem of controlling the variables other than sex.
This is non-trivial for height, but bafflingly difficult for slang use,
since the number of variables that would need to be controlled is
enormous, and many of them are very specific to particular subcultures
and social purposes.  But you could imagine one very carefully
controlled study, then another, and so on.

I don't know of a single such study that attempts to investigate slang
use in general.  Does anyone have one to offer?  (Even studies of
particular expressions, with respect to the usage of men and women in
some controlled context, are thin on the ground.  But they're very very
far from Flexner's claim, which would require that *most* slang
expressions show a sex difference.)

The height result is robust.  Is there any evidence at all for the
(non-taboo) slang hypothesis?

Note that I'm not saying that Flexner's claim is false.  I'm saying
that it's a priori dubious and might be impossible to test and that I
know of no evidence for it.  It *could* (if carefully formulated and
suitably constrained) be true, though even then it's unlikely to be of
much interest.

What I truly, deeply object to is the retailing of Flexner's claim as
if told us something significant about the differences between women
and men, and as if it were proven (indeed well-known).  It totally
wipes out all the interesting nuances of gender differences and
encourages people to dig in to their ignorant prejudices about these
matters.  It's neither scientific nor humane.  I'm morally offended, in
at least two different ways.

arnold (zwicky at csli.stanford.edu) -- I *told* you it was going to be
intemperate