Women stay out of Linux for many of the same reasons they stay out of computing in general, plus a few reasons specific to Linux. Many excellent books and research papers have investigated this topic in depth, but we can only summarize the top reasons why women avoid computing as a whole. We'll also debunk some common theories about why women stay out of computing in general.
Three good overall resources for the topic of women in computing are:
"Unlocking the Clubhouse: Women in Computing" by Jane Margolis and Allan Fisher
Women in Computing Keyword List
(Some of the papers referred to by this list are available online, but not all.)
"Why Are There So Few Female Computer Scientists" by Dr. Ellen Spertus
Let's start by examining two of the most common explanations for why there are so few women in computing: "Women just aren't interested in computers," and "Women aren't as smart as men." The problem with the statement, "Women just aren't interested in computers," is that it doesn't actually say anything. It's equivalent to answering the question, "Why is the sky blue?" with "The sky just is blue." The implicit argument here is that women are genetically predetermined from conception to not be interested in computers. Very few people are willing to say exactly that in so many words, but that is the message behind the "just aren't" theory. If you are unwilling to accept that women's lack of interest in computing is genetically predetermined (and I hope you aren't willing to accept it), you need to start exploring what environmental causes are involved.
A more explicit version of this theory is that "Women aren't as smart as men," or any of the usual corollaries--women aren't as good at some skill as men are, usually mathematics, spatial reasoning, or logic. Newsweek regularly trumpets studies finding gender-related mental differences while ignoring the (far more common) studies which find no difference at all. Frequently, other researchers are unable to duplicate the results or find flaws in the original researchers' methods, but those stories tend to get much less press. These studies also make no attempt to control for differences in the upbringing of men and women. For example, studies frequently show that women have better developed linguistic capability in some way. This is taken as proof, at least by the press, that women are genetically predisposed to be more verbal than men. But at the same time, studies also show that young women are rewarded more than young men for verbalization. The sheer existence of physical differences between male and female brains (an idea still in dispute) is not in and of itself proof that men and women are born with differences in mental capacity. We still need to separate out what differences are caused by genetics, and which are caused by the environment. As a result, if you ask the experts, the only consensus on gender-related mental differences is that there is no consensus. This is an area of ongoing research, where results will continue to be hotly debated for decades or centuries. (My personal opinion is that men and women do have some innate, genetically based differences which result in tendencies towards different behaviors, but I won't guess what they are or how strongly they influence behavior. Human beings are extremely adaptable creatures, so I suspect the genetic differences are minor compared to differences in environment.)
Something else to keep in mind is that similar arguments have been made about many other fields when women first began joining them, from medical science to education. For example, women couldn't be doctors because they weren't physically strong enough to set broken bones, would faint at the sight of blood, or didn't have the proper bedside manner. Those arguments were abandoned when women turned out to be just as good doctors and teachers as men were. Maybe men will turn out to be better at computer science than women, but history does not support that hypothesis.
A good reference for the general topic of measuring differences between human groups and the motivation behind those measurements is The Mismeasure of Man by Steven Jay Gould. Scientists have been "proving" differences in the brains and bodies of groups of humans for centuries, although in hindsight both their methods and their results were flawed. For example, Stephen Jay Gould reviews the methods of one scientist measuring skull capacity in men and women of different races (and by implication, brain size and intelligence). The scientist originally measured the volume of the skulls by packing them with linseed, which is somewhat compressible, and confirmed his hypothesis that white men tended to have larger skulls. When he later remeasured the volume of the skulls with incompressible lead shot, he discovered that much of the differences in volume between the skulls disappeared. He had been subconsciously stuffing the skulls belonging to white males with more linseed than the skulls belonging to women or non-white men. Keep this story in mind when you read studies claiming to find that some brain structure is a different size in men and women.
Now that we've addressed some common misconceptions about women and computing, let's look at the real reasons why women stay out of Linux and computing. I personally believe that the tendencies and behaviors I'm about to describe are the result of the way most women are raised, in other words, they are the result of gender socialization. I'm not claiming that women are born less confident, or anything else, I'm just observing general tendencies in women and pointing out how Linux culture discourages people with those tendencies. Many of the reasons I'm about to list also apply to other underrepresented groups in computing or science.
Women severely underestimate their abilities in many areas, but especially with respect to computers. One study about this topic is Undergraduate Women in Computer Science: Experience, Motivation, and Culture: http://www-2.cs.cmu.edu/~gendergap/papers/sigcse97/sigcse97.html
For example, while 53% of the male computer science freshman rated themselves as highly prepared for their CS courses, 0% of the female CS freshman rated themselves similarly. But at the end of the year, 6 out the 7 female students interviewed had either an A or B average. Objective ratings (such as grade point averages or quality and speed of programming) don't agree with most women's self-estimation. I personally encountered this phenomenon: Despite plenty of objective evidence to the contrary, including grades, time spent on assignments, and high placement in a programming contest, I still didn't consider myself to be at the top of my class in college. Looking back objectively, it seems clear to me that I was performing as well or better than many of the far more confident men in my class.
Like any other discipline, computer science is easier to learn when you have friends and mentors to ask questions of and form a community with. However, for various reasons, men usually tend to mentor and become friends with other men. When the gender imbalance is as large as it is in computer science, women find themselves with few or no other women to share their interests with. While women have male friends and mentors, it's often harder and more difficult for women to find a community and then to fit in with it. Many women leave the field who would have stayed if they had been male.
It's true that this is a feedback loop, fewer women in computing leads to fewer women in computing. It's important to understand that this feedback loop causes women to leave computing who wouldn't have left if, all other things being equal, they had been men. This is important because male classmates often assume their female counterparts leave the field because they "just aren't good enough." Women's low self-estimation contributes to this false impression.
Societal pressure for women to avoid computing begins at an extremely early age. Preschoolers already have conceptions about which jobs are men's jobs, and which jobs are women's. An excellent review of studies documenting gender role socialization from an early age can be found in Dr. Ellen Spertus's excellent "Why are There so Few Female Computer Scientists?" paper: http://www.ai.mit.edu/people/ellens/Gender/pap/node6.html
Once you realize that men and women are treated differently from, practically, birth, it becomes hard to claim that any woman hasn't experienced discrimination. Sure, if you're lucky, no one ever explicitly told you that you couldn't work with computers because you were a girl, but every time you raised your voice, an adult told you to quiet down, while the boy next to you continued to shriek. This is a handicap later on in life, when being loud and insistent is the only way to get your opinion heard--for example, on the linux-kernel mailing list.
The most striking example of a subtle bias against computing for women is that, in the U.S. at least, the family computer is more likely to be kept in a boy's room than in a girl's room. Margolis and Fisher give several telling examples of this trend and its effects on pages 22-24 of Unlocking the Clubhouse.
Working with computers is perceived to be a solitary occupation involving little or no day-to-day human contact. Since women are socialized to be more friendly, helpful, and generally more interested in human interaction than men, computing tends to be less attractive to women. I want to stress that computing is only perceived to be a non-social activity. While it is possible for a programmer to be relatively successful while being actively anti-social and programming does tend to attract people less comfortable with human interaction, computing is as social as you make it. During college, I spent most of my computer time in a computer lab at the school with several of my best friends. And recently, I changed jobs specifically in order to have more day-to-day contact with other programmers. For me, programming by myself is less fun or creative than it is when I have people around to talk to about my program.
Oddly, many occupations which are arguably less social than computing are still very attractive to women. Writing, either fiction or non-fiction, is a good example of a field that requires many hours of solitary concentration to be successful. Perhaps the answer to the paradox lies in the perception of individual writers as still being interested in social interaction, and just not having much opportunity for it.
Women in computing do exist, but most people aren't lucky enough to meet a female computer scientist. Women are socialized to be modest and avoid self-promotion, which makes them even less visible than they might otherwise be. Mothers and female schoolteachers regularly protest that they don't know anything about computers. As a result, girls grow up without examples of women who are either competent or confident with computers. I encourage all women in computing to be as visible as possible--accept all interviews, take credit publicly--even when you don't want to. You may be embarrassed, but by allowing yourself to be publicized or promoted, you might change a young girl's life.
We all know that most computer games are written by and for men. They feature non-stop gore and women with unrealistically huge breasts, but hey, if that's the market, what's the problem?
The best way I know how to illustrate the problem with the computer game industry is to tell a story from a Salon.com article (http://archive.salon.com/tech/feature/2001/05/22/e3_2001/ ) about the 2001 E3 gaming convention:
"A creative director for a leading development team cheerfully described to me how its Q.A. team made a prostitute sport a game's logo on her body during a combination gonzo video/gangbang session."
This was only one of many similar stories and events at the conference. How can an industry that views company-sponsored gangbangs as somehow appropriate *not* be driving women out of the computing arena in droves?
The next time you see a computer ad featuring a person, pay attention to that person's gender. Most likely, the person is a man. Frequently, when I do see women in a computer ad, they're wearing freakish makeup and some form of colorful skintight vinyl, or else they're acting dumb and helpless and waiting for the man to show them how to use the computer. Often, they don't appear to actually be using the computer and are just sort of decoratively posed near it. Movies and TV shows are no better. When a woman is depicted as a programmer, often more screen time is spent admiring her shapely body and kissable lips than demonstrating her competence as a programmer. Notable example: Angelina Jolie in "Hackers."
Men and women are constantly bombarded with media images which say: "Men use computers, women don't." It's difficult to overcome daily indoctrination of this sort.
Being good at computing is considered to be an activity that requires spending nearly all your waking hours either using a computer or learning about them. While this is another misperception, women generally are less willing to obsess on one topic, preferring to lead a more balanced life. Women often believe that if they enter computing, they will inexorably lose that balance, and avoid the field altogether instead. During college, I was personally very proud of not spending my leisure time playing computer games because it refuted the programmer stereotype of being at the computer all day, every day.
Linux development is more competitive and fierce than most areas of programming. Often, the only reward (or the major reward) for writing code is status and the approval of your peers. Far more often, the "reward" is a scathing flame, or worse yet, no response at all. Since women are socialized to not be competitive and avoid conflict, and since they have low self-confidence to begin with, Linux and open source in general are even more difficult than most areas of computing for women to get and stay involved in.