Mark Mandel thnidu at GMAIL.COM
Mon Jan 26 18:52:31 UTC 2009

FWIW, I worked out the ratios. I also did my own searches, coming up
with slightly different numbers, and discovering in the process that
"She is an actress" (capital s) has 10,700 rgh more than "she is an
actress (small s). Whenever Google said it had taken my web history
into account, I used the numbers without such customization. The
results are at
or, more conveniently:

Here they are in text, which may be more readable in a monospace font
like Courier. The percentages on the rows are first column ("I am") /
total; the percentages on the columns are first row ("actress") /

Usenet 2000-'08 Wilton
I am an... She is an...
...actress 121 313 28%
...actor 215** 43 83%*
36%* 88%

Google 20090125 Butters
I am an... She is an...
...actress 52,400 93,000 36%
...actor 93,000** 10,800 90%*
36%* 90%

Google 20090125 Mandel (non-cust)
I am an... She is an... she is an...
...actress 53,800 93,700 36% 83,000 65%
...actor 96,500** 10,300 90%* 15,500 86%*
36%* 90% 84%

*  referent's gender partly unknown
** referent's gender unknown

But what do these numbers mean? We're comparing apples with mixed
fruit here, so let's look more carefully. Each grid has eight cells:
four raw and four computed (using the "I am" column of the Mandel grid
with both cap and small "S/she is").

The variables and my symbols for them are:
Nf,Nx: noun fem (actress) or unknown-gender (actor)
Rf,Rx: referent fem (S/she) or unknown (I)
1s,3s: 1st or 3rd singular construction (I or S/she)

The raw cells:
I am an actress: Nf,Rf,1s                               S/she is an
actress: Nf,Rf,3s
I am an actor: Nx,Rx,1s                                  S/she is an
actor: Nx,Rf,3s

The computed percentages:
first row:                                                      Nf,Rf:
   1s / (1s+3s)
second row:                                                Nx:
(Rx,1s) /  (Rx,1s + Rf,3s)

first col:                                  second col:
1s: (Nf,Rf)/(Nf,Rf + Nx,Rx)      Rf,3s: (Nf)/ (Nf + Nx)

The only proportions that compare apples to apples are
- the first row: with a female noun and a therefore female referent,
how many of the pronouns are first-person: 28% - 39%, around 1/3
- the second column: with a third-person pronoun and a therefore
female referent, how many of the nouns are female: 83% - 90%, around

Neither of these numbers is very helpful by itself. The third-person
sentences give no clue to how female actors speak of *themselves*, and
the first-person ones give no clue to how *female* actors speak of

Let's play with the numbers again. I'll use Wilton's because they're
small and easy to do in the head. It should be safe to assume that all
the men quoted in first person said "I am an actor". Now let's further
assume that the first-person quotes come equally from men and from
women. There are 121+215=336 first-person quotes. 336/2 = 168, so we'd
have 168 men and the same number of women. 121 of the women said "I am
an actress", and 215-168=47 said "I am an actor": 47/168 = 28%.

The other searches yield the same proportion: significant but nowhere
near a majority. This  is somewhat higher than Ron's estimate, though
not grossly so, and (to the extent that my assumptions and procedures
are reliable) still confirms his conclusions:

While one cannot determine what percentage of hits for "I am an actor"
was actually uttered by women, a cursory check of the first 20 hits
indicates that the utterers are primarily men, not women. Still, the
fact that the lower number for "I am an actress" (as compared to "I am
an actor") could be taken to mean that as many as 20% or so of the "I
am an actor" respondents are women.

 In the end, though, 52,400 hits for "I am an actress" seems to
disconfirm Ann's perception that "women actors call themselves ...
'actor'...". The statistics seem to indicate that most do not. So, if
the criterion is self-identification, then "actress" would seem to be
more popular among female actors than "actor," though "actor"
hopefully represents a healthy (and, I personally hope, growing)

Mark Mandel

The American Dialect Society - http://www.americandialect.org

More information about the Ads-l mailing list