What corpus analytics has to offer (women’s ) history
Last week I gave my first talk based on doing history with digital tools  methods at the wonderful conference Women’s History in a Digital World.  Emboldened by Laura Mandell’s kind query about whether I planned to publish these initial results, I’m going to offer a preliminary report.  If you want to know how I got here, the journey is blogged. However I would be remiss if I started this account without thanking two people, people Heather Froelich and Whitney Esson as well as the many digital historians and humanists mostly on twitter who helped me along the way.  
During my sabbatical in 2011-2012 I began working on a book, The Politics of Women’s Culture.  I’ve been working, in one fashion or the other, on the issue of women’s culture since the early 1990s. As I worked on my book, which I describe as an intellectual history of the concept of women’s culture, I revisited the issue of the origins of this particular phrase.  Of course 20 years ago I answered it using the tools methods had, hand searching printed matter, but I began wondering what digital means might be available to aid in a wider search.
Months latter, and many tools methods tried and discarded, I’d embarked on a far more involved project that addressed historiographical questions using digital history tools methods.  In particular I worked with AntConc, thus far on the full run of Chrysalis a magazine of women’s culture(1978-1981) and date matched issues of  Off Our Backs (OOB), a nationally circulated women’s liberation newspaper.  My main goal was to determine if the widely cited interpretations of the women’s movement from the late 1980s, that something called cultural feminism derailed the women’s movement, could be validated digitally.[1]
According to historian Alice Echols in Daring to be Bad “cultural feminism insisted on women’s essential sameness to each other” “vilified” male “values” and “valorized” female bonding (244) and “evaded” rather than ‘engaged” patriarchy.”(1) The year before, Linda Alcoff argued cultural feminism “is an ideology of a female nature or a female essence used to “revalidate undervalued female attributes.”  Both Alcoff and Echols are widely cited in both women’s history and women’s/gender studies despite the fact that their conclusions have been challenged by a wide range of scholars. 
I chose these two periodicals for pragmatic as well as illustrative reasons.  JSTOR contained digitized copies of OOB.  Ideologically, scholars identify OOBidentified as emblematic of radical feminism. Echols cites OOB as the location of the first published critique of cultural feminism.  Chrysalis, which my wonderful graduate student assistant Whitney Esson digitized for me by scanning and converting using adobe (but not cleaning yet), was chosen to represent women’s liberation activists with an interest in culture.  In Women’s Periodicals in the United States: Social and Political Issues, Kathleen N Endres describes As Off Our Backs ” the best-known newspaper of the ‘radical feminist’ ” family while Sammaye Johnson positions Chrysalis a magazine of women’s culture conceived of by Rennie and Grinstead as a “new national  feminist magazine” (58). Endres quotes Grinstead and Rennie’s criticism that OOB had left behind radical feminist movement to focus on intra-movement schisms from a “predominantly male left perspective” (269).
Enter twitter, which allowed me to dive head first into the digital humanities and it that entails.  I was guided to a tool AntCOnc and a field corpus analytics as well as a methodology, Critical discourse analysis, defined by Van Dijk as “the way social power abuse, dominance, and inequality are enacted, reproduced, and resisted by text and talk in the social and political context.
I then began to reconceive my query about the validity of interpretations into a more complex question, could I take an established way to look at discourse in linguistics and apply it to the women’s movement?   Could I find a grammar of women’s liberation by figuring out if various activists were even speaking the same language? Are there/what are similarities & differences in discourses of OOB & Chrysalis.[2]
AntConc allows me to parse frequency out from collocates, word that are likely to appear together (with likelihood measured against in greater than random chance and together being measured by various parameters in my case words adjacent to one another), and clusters, words that do appear together in the corpus.  It also provides a concordance view which allows for close readings, as well as a measure called keyness that allows the comparison of words in one corpus to another, in this case measuring OOB against Chrysalis and vice versa.
If we follow just the word “culture” through the OOB & Chrysalis, first see that in terms of word frequency, neither word particular frequent in either corpus.  As expected form the subtitle of Chrysalis, appeared at higher frequency (356thv 463rd in OOB). Of the top 10 clusters for culture in both periodicals, 50% are the same, and that rises to a 60% overlap for collocates.  Of the twenty-seven most frequent lexical words that appear in both periodicals, 50% are common.     

But is 50% or 60% enough to say speaking the same language in both periodicals?  The differences are suggestive.  For example, the collocates from Chrysalisthat do not appear in OOB include “women” and “our,” reflecting the focus on women’s culture, as well as “dominant” perhaps suggesting that women’s culture is defined in opposition to the dominant (male) culture.  Less easily explained are the collocates in OOB (that, which, to, as).  Clusters reveal some fascinating divergences as well.  For example “feminist culture” appears in the top 10 of OOB while “female culture” appears on list from Chrysalis.  Conversely, “Male culture” appears in OOB results while “dominant culture” (recall the collocates) is featured in Chrysalis.
What about how culture is used as revealed by concordances? Both periodicals show “s” as a common collocate/cluster for culture, because the digitization process renders women’s as three separate “women” ‘ ‘” “s”.  Using the concordance view for those results allows for a closer analysis of the occurrences.  The cluster appears 53 times in Chrysalis.   32% pertain to the title of the magazine itself.[3]
The cluster appears 54 times in OOB, roughly the same as in Chrysalis, which is half the size of the OOB corpus. However closer analysis reveals that while in Chrysalis “‘s culture” always correlated with women’s culture, in OOB, that is not the case.[4]In the 40 instances of “women’s culture,” two announce Chrysalis itself.[5]
OOB n= 38
Chrysalis n=53
Difference
Letters
5%
11%
6%
Editorials
7.8%
17%
9%
Advertisements
15%
3.7%
12%
Reports on women’s movement
18%
Substantive articles
20%
Looking at the similarities first, in term of reports on the women’s movement, the mainstay of OOB as a newspaper as compared to the longer articles in Chrysalis, the two periodicals look relatively similar in terms of discussion of women’s culture, until close reading reveals that in OOB the occurrences all derive from reports on the 30th anniversary conference of the Second Sex. 
The divergence in letters to the editor and editorial content, over double in Chrysalis as opposed to OOB maybe accounted for due to the content of Chrysalis and the usage of “women’s culture” as a descriptor of the content.  Advertising however is the most divergent, more prevalent in OOB by a factor of three.  While historians may view the periodicals as representing different ideological perspectives on culture, it appears vendors voted with their dollars to reach potential customers for products they described using “women’s culture” in a OOB periodical that historians have considered hostile to the idea?
next up, looking at keyness, and contemplating other “words”



[1] Using quantitative methods to overturn prevailing beliefs in women’s history also

 done by Newman and Block “Among many other findings, we argue against two popular beliefs about women’s history: that women’s historians are overly-focused on recent history and that women’s and gender history is an ever increasing proportion of the profession.” JWH 2011 spring
(1) Echols has warned against the overly rigid application of her thesis, but to little avail in Shaky ground: the sixties and its aftershocks (69).

[2] In a sense I’m comparing apples and oranges.  While date matched, OOB consists of far more individual items because it was a newspaper rather than a magazine.  The size in terms of word count also varied dramatically with OOB almost twice as large as Chrysalis (1525881 v 734,479 words).
date matched Oct 1978 to April 1980
[3] Another 11% come from letters to the editor from readers.  Another two occurrences are in advertisements 3.7%. Articles account for about 20% the times women’s culture is used, with 17% (9) coming from editorial content.  
[4] “womyn’s culture” occurs 4x, one’s culture5x, people’s culture 3x and “woman’s culture” 1, “man’s culture”
[5] Of the remaining 38 items, the highest number, seven (18%), occur in reports about the women’s movement, from a single issue that covered the 30th anniversary of the Second Sex conference. 15% (n=6)  occurred in advertisements, 7.8% editorial, 5% letters from readers.

Advertisements