It seems to think women are more personally involved in what they write while men use more of the language of approximation.

- Erin

Idly sampling large blocks of text from several female authors, I found that the program thought that all of them were male, including Amy Goodman, Ellen Goodman, Molly Ivins, Jane Austen (when she's not depicting female conversation and trying hard to present them as flighty females), and Hildegarde of Bingen. The only woman I discovered who the program thought was female happened to be Katie Roiphe, but just barely. Although this identification was kind of iffy (Katie turns out to be female by only two points) the program is obviously flawed, since Katie is one of the most male-identified women I know of.

Looking over the list of words and how they're sorted, it's obvious that the program believes that almost anyone who writes professionally, or has been trained in journalism, "writes like a man," while anyone who takes any interest whatsoever in women is tainted by femininity.

The words "her," "she," and "hers" are all poison, infecting a passage with feminine sensibility. This alone casts doubt on the fairness of the test, since one would hope that either men or women might be able to write about women without prejudice. On the other hand, the word "what" is somehow masculine, so any depiction of a conversation with someone hard-of-hearing obviously brands the writer as male.

Most of the scores were actually fairly balanced, with the average difference between the scores on the order of 10%, but still...

I then looked at several women who write genre romance fiction and are very popular with female readers:

1. Christine Feehan -- Female -- Female Score: 6912 -- Male Score: 4845 over 4057 words.

2. Jennifer Apodaca -- Male -- Female Score: 1734 -- Male Score: 2200 over 1590 Words.

3. Linda Lael Miller -- Female -- Female Score: 6814 -- Male Score: 6705 over 5205 words

4. Linda Howard -- Female -- Female Score: 3665 -- Male Score: 3306 over 2589 words.

With the exception of Christine Feehan, the differences are not overwhelming, although moderately predictive over a small sample. Three out of four isn't too terribly awful.

Comparing these two groups, both made up of professional writers but the first consisting of women (with the exception of Jane Austen) who want to be taken seriously, and the second made up of women who are quite deliberately setting up to please women and write scenes that sound "girly," one does see differences, but they're not terribly consistent.

These didn't seem all that conclusive, so I looked at me:

Sample 1: Female -- Female Score: 20417 -- Male Score: 14882 over 12300 words.

Sample 2: Male -- Female Score: 28344 -- Male Score: 31287 over 21639 words.

One reckons that I'm not terribly consistent either...

The Guardian article pointed to on the page comes to similar conclusions, since *all* the female columnists were identified as male, and the only iffy identifications consisted in misidentifying male columnists as females.

All in all, it looks to me like the so-called "test" is one of gender stereotypes, and the more one attempts to conform to those stereotypes, the better the program likes it.

It's at least as much fun as reading horoscopes, and probably just as accurate.