Citation is one thing. Discussion is another. You can drop a reference without really engaging someone's work (e.g., Snerdfoot 2011). But as Helen de Cruz has emphasized, discussing a possible Bechdel test for philosophy papers, citation analysis is insufficient as a measure of serious engagement with someone's work. I propose two rough measures of "discussion".
"Discussion" itself I operationalize as follows: A person is discussed if that person's name appears in the abstract of an article. Looking at the Philosopher's Index database, I have examined discussion arcs over time for various well-known philosophers in a series of blog posts (e.g., here, here, here, here).
"Extended discussion" I operationalize as follows: A person receives extended discussion if that person is referred to at least twice in the abstract of the article, by either name or pronoun. The nominative pronoun might be especially telling, since its presence suggests that the person is being referred to repeatedly in independent clauses. For example:
Later, Nussbaum gradually reconsidered the notion of patriotism in texts that remained largely unknown and rarely discussed. This article begins with a brief account of her shift from cosmopolitanism to what she terms 'a globally sensitive patriotism,' and the task assigned to education within this framework....
This suggests a possible rough and simple measure of the relative rates at which women receive extended discussion in philosophy articles compared to men: Compare the ratio of "he" to "she" in philosophy abstracts, then remove cases in which those words are used with generic intent (e.g., "If the agent wouldn't have done otherwise whether or not she could have....") or otherwise not referring to an individual philosopher whose work is being discussed (e.g., reference to historical leaders, or third-person references to the author herself for abstracts written in the third person).
Method:
I searched Philosopher's Index for all appearances of "he" or "she" in abstracts from 1970 to the present in a sample of ten ethics journals and ten general philosophy journals. [See Note 1 for journal details.] This yielded a total of 2321 abstracts. I then skimmed each abstract to remove all cases in which the pronoun was not used to refer to a specific philosopher whose work was being discussed. [Yes, I looked at over 2000 abstracts! Obviously, my determinations had to be quick, but in almost every case it could be made confidently within just a few seconds.] To examine temporal trends, I grouped results by decade. I also separated citations of pre-20th-century historical figures from 20th and 21st century figures.
Results:
Percentage of recipients of extended discussion (as measured by nominative pronoun use in abstracts) who are women:
1970s Ethics: 8/92 (9% women) [so 91% of those receiving extended discussion are men] General journals: 4/134 (3%)
1980s Ethics: 3/77 (4%) General: 0/137 (0%)
1990s Ethics: 20/147 (14%) General: 9/189 (5%)
2000s Ethics: 16/184 (9%) General: 16/229 (7%)
2010s Ethics: 19/120 (16%) General: 27/244 (11%)Merging the ethics and the general journals, so far in the 2010s, approximately 13% of philosophers receiving extended discussion in these journals are women.
In contrast with my earlier data on authorship in the most elite journals, this does appear to be a statistically significant increase since the 1970s (5% vs. 13%, z = 3.2, p = .001).
If we remove discussion the pre-20th century figures (Kant, Plato, etc.), then numbers look like this:
1970s Ethics: 8/64 (13% women) General journals: 4/97 (4%)
1980s Ethics: 3/54 (6%) General: 0/106 (0%)
1990s Ethics: 19/94 (20%) General: 9/144 (6%)
2000s Ethics: 15/140 (11%) General: 16/184 (9%)
2010s Ethics: 18/95 (19%) General: 26/210 (12%)Still only 14%!
(Since ethics is a minority of the discipline, it makes sense that the center of gravity would be closer to the general journals.)
These numbers are consonant with two other measures I've done that suggest that at the very highest levels of prestige our discipline is still predominately male.
For the discipline as a whole, percentages of faculty in the 21st century are typically in the low 20%'s (U.S. data here).
The outlier analysis here is my analysis of American Philosophical Association meetings, where women were 35% (144/413) of the invited symposium speakers on the main program, and 32% of main program participants overall.
----------------------------------------------------------
Note 1: Ethics and non-ethics were analyzed separately because previous analyses have found differences by area, and because journals divide fairly naturally into those specializing in ethics/political, "general" journals that publish proportionately little ethics, and other types of specialty journals (like philosophy of science).
Ethics journals were the top ranked journals in surveys by Brian Weatherson and Brian Leiter (excluding the non-ethics journals appearing in the latter) and include Ethics, Philosophy & Public Affairs, Journal of Political Philosophy, Utilitas, Social Philosophy and Policy, Journal of Ethics, Ethical Theory & Moral Practice, Journal of Social Philosophy, Journal of Value Inquiry, and Journal of Moral Philosophy.
The comparison list was a stratified sample of "general" philosophy journals drawn from Leiter's surveys here and here and included Nous, Midwest Studies in Philosophy, Synthese, Mind, Philosophical Studies, Proceedings of the Aristotelian Society, European Journal of Philosophy, Dialectica, Philosophical Topics, and Theoria. The sample was stratified so that the selected journals would not differ too much in overall prestige from the ethics journals. ----------------------------------------------------------
For a related post, also including discussion of the use of "he" and "she" and generic pronouns, see Use of "She" and "He" in Philosopher's Index Abstracts (Sep. 16, 2014). See also Kieran Healy's nice analysis of gender citation patterns in four leading philosophy journals.
13 comments:
Interesting way to look at this issue.
Did you calculate percentages for men? Obviously it can't be done by simple subtraction: we can't deduce from the fact that 10% of papers in journal x discuss women that 90% discuss men (because many philosophy papers don't discuss any person's extensively enough to warrant mentioning them in the abstract). This fact makes interpretation of changes across time difficult: do they reflect changes in the status of women, or differences in how topics are approached (though discussion of particular persons' views).
Sorry if it wasn't clear, Neil. It is percentage of women of all who receive discussion, so x% women implies 100-x% men. I'll revise the post to make this point clearer.
Thanks for the clarification, Eric. The data is still hard to interpret without any context. Papers that devote substantial discussion to particular people may be a small minority of papers. They may come from particular subdisciplines (perhaps shifting over time). And so on. That doesn't make the measure useless, of course, but context might indicate how important a measure it is.
Gosh. On that metric, I think I never discuss ANYONE. Names just don't generally show up in the abstracts of more technical papers.
So assuming there were say 2000 general papers published in the 2010s on pre-figures, that means only 12% engage in extended discussion of an individual at all, and of those 10.5% extendedly discuss men and 1% extendedly discuss women, is that right?
Thanks for the continued discussion Neil, Sara, and Anon Jan 4: Yes, this is a high bar for "extended discussion" which only a minority of papers meet. I still think it's an interesting category, but you're right that I should more explicitly contextualize it by reference to the fact that most abstracts outside of history of philosophy do not involve multiple references to any individual. Dropping "she" or "he" from the search criteria yields about 3900 non-ethics hits and 3100 ethics hits in the sampled journals, so only about 4% of articles meet the criteria. It is intended as a very high bar for prominence, as exemplified in the the abstract about Nussbaum, who is treated there as a philosopher sufficiently important that a leading ethics journal would have a paper that is primarily about her work.
So really its an indirect measure of what proportion of women are high status - high enough to warrant articles being written about them. A small minority of articles in analytic philosophy are about contemporary individuals (so small that I bet you never noticed that the AJP has a policy of not publishing such discussions).
Yes, and yes, I hadn't noticed that about AJP! (I'm tempted to look at the data to see if any AJP articles meet our criteria.)
It's not clear to me that x% women implies 100-x% men. Some articles might contain substantial discussion of women *and* men, in which case (the number of articles that contain substantial discussion of women) + (the number of articles that contain substantial discussion of men) might exceed (the number of articles that contain substantial discussion of humans).
Anon Jan 6: We counted up the total of articles with women and the total with men, then looked at the percentage of the total that were women. In the case you describe the article would be double-counted. So x% women does imply 100-x% men, given our method. Double-counted articles will presumably be rare, since such a small percentage of articles meet the criterion for either women or men. To be clear: It's not percentage of *articles* that extensively discuss women. It's percentage of *extensively discussed individuals* who are women. Yes?
The method of double counting might provide a reasonable approximation of what you are measuring provided that the number of double-counted articles, which is "presumably" small, really is small. Otherwise, you could potentially get some very misleading results. If there are exactly 100 articles that count as discussing someone, and all 100 discuss both men and women, then (I think) that via your method of double counting you end up saying that women (and men) are discussed in only 50% of the articles, when in fact they are discussed in them all. Obviously that's an extreme and unlikely case, but it indicates that what you say you are calculating (% of total number of discussions that are discussions of women) is not what you are in fact calculating.
Bernard: What I said I was reporting was "Percentage of recipients of extended discussion" who were women. I don't think that overlaps change this. Suppose every single article overlaps with one man and one woman. Then my measure will report that 50% of the recipients of extended discussion are women, which would in fact be accurate. I hope I did not speak in a sloppy way at some point that led to a different impression, but the phrase I quoted is the official way I describe my results.
I have skimmed every single abstract. Though I didn't systematically look for overlap, my impression based on that skimming is that there is zero overlap. It could well be that there were a few that overlapped that I didn't notice, but in any case it should make little difference to the overall percentages.
Post a Comment