I here present a map of the values associated with the noteworthy, courtesy of my ongoing collaboration with Andrew Higgins (University of Illinois, Urbana-Champaign) and Jacob Levernier (University of Oregon). This map was generated from data that I collected from obituaries in the New York Times. These obituaries are much different from the ones we’ve looked at previously. Unlike obituaries in local papers, which are typically written by kin or next-of-king, these are commissioned from professional writers. Also unlike obits in local papers, which are about ordinary folk, these about noteworthy individuals. Additionally, they tend to be quite a bit longer than local obits and somewhat more critical. There’s also a huge gender divide: for each obit about a woman, there are roughly six about men. That makes the gender-comparisons less helpful. The map in this post is based on about 70 obituaries. To get a really robust map, I’d probably have to look at 200 or more. Anyway, without further ado, here’s the map:
As before, you’ll want to open this image in a separate tab and zoom in to see what’s really going on. In this map, edge color indicates gender (blue for me, red for women). As I mentioned, there weren’t that many women, so this is probably not that informative. As usual, edge width represents the number of times the connected pair of terms co-occurred in a single obituary. Also as usual, the size of a term represents the number of times it co-occurred with other terms. In this case, the color of a term indicates its modularity. As a reminder, modularity is a kind of cluster analysis. Terms of the same color tend to co-occur with each other and not with other terms.
In Andrew’s words, very lightly edited, “This network is very different, both in terms of structure and content. In terms of structure, we have more clusters with high clustering coefficients, less overall interconnectivity. But that’s an obvious result of (1) the small number of obits, and (2) the fact that lots of descriptions are given in each. The more interesting feature of this network is the content. If there were any evidence against the hypothesis that we only speak well of the dead, this would be it. There are so many seemingly vicious traits ascribed to some of these people. It seems like we’re getting a more complete picture of the person, perhaps because of the extra space available for describing the person in full. Here are the top descriptions, in order of weighted degree (i.e., the number of terms that co-occurred with the given term across the whole sample): honored (269), author (246), leader (237), veteran (162), civil rights advocate (125), teacher (100).”
A few more thoughts from me: I didn’t explicitly note how many of these people had been divorced (some more than once), but I’d guess it was something like 80%; it seems like being famous enough to get an obit in the Times is not good for your family-life.
It’s interesting to compare this map with the Schwartz theory of basic human values. Here’s a quick-and-dirty summary of his model:
Each sector in the figure above represents a value that many many people say they endorse. Sectors that are adjacent to each other tend to correlated positively. Sectors that are opposite each other tend to correlate negatively. To what extent do our modules map onto this model?
Here’s my first-blush read on the modules:
purple (on the left): The biggest term is ‘author’. The cluster around it seems to have to do mostly with being unapologetically critical of tradition, institutions, etc. This is a very intellectual set of traits. Intuitively, this cluster should correlate negatively with tradition, conformity, and security, which would place it among the hedonism, stimulation, and self-direction values. Hedonism doesn’t seem to fit, but the other two do to some extent. Schwartz glosses stimulation in terms of excitement, novelty, and overcoming challenge; he glosses self-direction in terms of independent thought and action.dark green (top left): I’d summarize this one as a kind of autonomy. Like the purple cluster, it involves breaking the rules, but it’s not breaking the rules in order to change them or criticize them. It’s breaking the rules because you don’t care about the rules. Thus, this cluster seems to involve elements from both the self-direction and the achievement sectors.kelly green (middle and top right): This is the closest, I think, to our family/friends/christian category in the local maps. It’s about commitment to community. That involves some improvement of community, like the purple cluster, but seems to be more taking it for granted that the community is already good and worth supporting. This cluster seems to match pretty well the security, conformity, and tradition sectors of the Schwartz model.yellowish brown (far right): This is clearly the lawyer category. Lots of intelligence and smartness, not much morality. It’s not clear to me whether this matches any of the Schwartz values.grey (bottom): This is another political category. Unlike the purple cluster, it’s not about cutting into the soul of one’s community. Unlike the kelly green cluster, it’s not about leading the dominant part of society. It seems to be more about leading the oppressed. This cluster seems to involve elements of both the power sector and the universalism & benevolence sectors.There are a few other, smaller clusters, which I’m reluctant to try to interpret.
We thus get some partial overlap with the Schwartz model but also some conflict with it. We’ll need to continue thinking about this contrast as our research develops.
In our continuing exploration of the words we use to talk about the dead, Andrew Higgins, Jacob Levernier, and I have created an “omnibus” map of the traits and other values associated with men and women across the country. Our sample draws from Eugene, Flint, Wasilla, and Amherst. (Eventually, we will be adding lots of other towns… it’s hard work reading hundreds of obits!) As usual, size represents interconnections, edge width represents co-occurrences, and centrality represents, well, centrality. In addition, color in this map represents gender: the bluer the term, the more its associated with men; the redder the term, the more it’s associated with women. Here it is:
If you zoom in, you’ll see a number of unsurprising gender-differences. For example, men are much more likely to be described as veterans, while women are much more likely to be described as cooks. We don’t need to mine obituaries to realize that World War II happened and that woman still disproportionately work as homemakers. But there are also some surprising differences, given the prevalence of traditional gender roles in American society. Women are more likely to be described as courageous. Men are more likely to be described as helpful. Women are more likely to be described as independent and spirited. Men are more likely to be described as understanding and affectionate. There are also some surprising lacks of difference. Most notably, men and women are equally associated with family, with volunteering, with having a sense of humor, and with leadership.
Here’s another version of the same map, with the terms replaced by nodes of various sizes: