Science bible stories, take 27

« previous post | next post »

Yesterday I wrote about a recent scientific paper that looks for evidence of the cultural effects of American urbanization in word counts from the Google ngram viewer. The paper was Patricia Greenfield, "The Changing Psychology of Culture From 1800 Through 2000", Psychological Science 8/7/2013, and my post about it is "The culturomic psychology of urbanization",  8/18/2013.

I learned about Greenfield's paper indirectly, when a reader sent a link to a daytime TV discussion, "Selfish U.S.?: Study says country becoming more self-centered", on CBS This Morning, 8/15/2013.   Charlie Rose and Norah O'Donnell were the show's co-hosts, and their guests were  John Tierney and Anne Fulenwider.

It didn't surprise me to find that neither the show's video nor its online context provided a reference to Greenfield's paper, or even the name of the author.  Judging from the content of the discussion, I suspect that none of the four talking heads had read anything except a press release — in any case, they mostly ignored the paper, and instead offered various associated ideas of their own.  For them, the role of the paper was  to add scientific gravitas to their opinions about the selfishness of Americans today, the importance of self-esteem, or the role of women in society.

As I observed a few years ago, "scientific studies"  have taken over the place that bible stories used to occupy. It's only fundamentalists like me who worry about whether they're true. For most people, it's enough that they can be interpreted to be morally instructive.

Also on August 15, I read Paul Krugman's assessment of the quality of traditional journalistic judgment ("The Good Web", NYT 8/15/2013):

Pundits like Samuelson seem to long for an age when wise men, from their platforms at major news orgs, sifted truth from falsehood and delivered sound judgment to the masses. The trouble is, that age never existed. I read a lot of economics reporting in the pre-Internet era, and by and large it was terrible. In part this was because the reporters and pundits often knew little economics — in fact, there was a sort of bias against having reporters with too much expertise, on the grounds that they wouldn’t be able to relate to the readership. In part it was because there wasn’t an effective mechanism for checking facts and interpretations: a reporter or pundit could say something that everyone who knew anything about the subject realized was all wrong, but those with better knowledge had no way of getting that knowledge out in real time.

I'd add a third important factor: by and large, the "wise men" (and now the "wise women") don't really care about whether the empirical and theoretical foundations of their opinions are sound . They care about readers, ratings, and reputation — and in some cases about political outcomes or cultural values —  with truth relevant only insofar as it affects those goals.

The CBS This Morning discussion strikes me as a good example of this situation. Here it is, so you can judge for yourself (sorry about the introductory ad):

Those are four Very Serious People: You can draw your own conclusions about what they're doing.

What about the rest of the response to Greenfield's paper? I've found a few thoughtful evaluations. For example, Lisa Suhey ("Do shifts in the words we use define our culture's personality?", The Christian Science Monitor 8/9/2013) made some sensible points about changes in the meaning and popularity of words that make simple lexical frequencies problematic as proxies for Greenfieldian dimensions of "the psychology of culture":

If the use of the word “friend” picks up around Facebook’s founding, then that implies two things: One, that culture affects the words used in literature; and two, that we may be using words in more than one way – that “get” doesn’t have to mean “come to have or hold.”

Using Ngram, the Google tool Professor Greenfield used to conduct her study, I charted the frequency of “friend” in American books published over the past two centuries.

The chart shows “friend” peaked in 1836 and then began a decline that would last until 1980. It made a slow resurgence over the next two decades until, after Facebook’s founding in 2004, the frequency of “friend” spiked. […]

In some cases I think this UCLA study needs a companion study of general word usage and decline to help us out in drawing conclusions.  For example, I think of the word “obliged” and “folks” in the same way – colloquialisms, verbal and written antiques.

Philip Hensher makes a similar point ("There’s nothing out of date about duty – read this year’s Man Booker longlist", The Telegraph 8/8/2013):

I hope Professor Greenfield’s analysis took account of the fact that Jane Austen usually spells “choose” “chuse”. Examination of the collected works on my Kindle shows that Austen uses the word “duty” 120 times and “chuse” 61.

These studies fill one with a sense of envy for the analysts, who apparently feel no requirement to read a book before asking computer programs to come to conclusions about it. The assumption, that duty and obligation in a novel are only expressed by the use of the words “duty” or “obligation”, is staggeringly naive. An age which speaks ceaselessly about duty may value it highly, or have a nagging anxiety about it.

And he goes on to make a broader claim:

Duty, imposed from outside, is a key topic of the novel. The conflict between inclination and obligation, between what someone must do and what he wants to do, has generated literature right from the beginning. […]

The term “duty” may have passed out of common usage in novels, but is the useful conflict between inclination and requirement really disappearing in favour of a world of wilfulness?

A good test is this year’s Man Booker longlist. It is, admittedly, hard to find the word “duty” in most of them; it occurs most frequently in Eleanor Catton’s pastiche 19th-century novel, The Luminaries, which may seem to prove Professor Greenfield’s point. But when one actually reads the novels, rather than running a word search, a rather different conclusion emerges. They are as obsessed with duty as ever. Jhumpa Lahiri’s The Lowland turns on the felt obligation of a man to marry his murdered brother’s pregnant widow. Colum McCann’s TransAtlantic contemplates the requirement to act well in a public sphere, including that of an anti-slavery campaigner in 1840s famine-ridden Ireland. Ruth Ozeki’s narrator in A Tale for the Time Being feels a plain duty to track down the author of a manuscript for her safety’s sake. Donal Ryan’s hero, in The Spinning Heart, feels a mysterious duty not to deny that he killed his father.

And so it goes on; duty and obligation are balanced against choice and personal freedom in almost exactly the same way that they are in Persuasion or Our Mutual Friend. The vocabulary may have changed, and perhaps has become more occluded, but there is no suggestion that the raw materials of duty and expectation have disappeared.

And Tom Chivers puts similar concerns in a different frame  ("'Society is going to hell in a handcart and the English language is decaying', pt 2,445 in a continuing series", The Telegraph 8/8/2013):

Have we really become more selfish, and can we tell it from studying our word use? According to news reports of a study by Patricia Greenfield (no relation, I assume), we have, and we can. Apparently, using the publicly available and utterly brilliant Google Ngrams tool, Prof Greenfield examined more than 1.5 million books and found that the use of words such as "obliged" and "give" has dropped in frequency in English-language books, while words such as "get", "child", "unique", "individual" and "self" have become more common.

I can't get hold of Prof Greenfield's study so I have to be a bit careful about this, but I'm sceptical. I should note that the stories don't seem to quite represent her own position: she is quoted as saying that her findings suggest an increase in "individualism" and "materialism", which isn't quite the same as "greed" and "self-interest" and "self-centredness", which is how it has been reported. But even so, I think we should take the findings with something of a pinch of salt.

(It's too bad to see so much mass-media discussion of a study that few people can access, so here's a link, for Tom Chivers and everyone else in his position.)

He continues:

[S]he suggests that the fact that "obliged" and "duty" have become less common reveals that we are less bound by these concepts than we used to be. But "much obliged" was once a commonly used phrase for "thank you", in a way that it rarely is now; that could account for a significant amount of the drop. Likewise, while "duty" has dropped in frequency, "responsibility" has increased (and "obligation", interestingly, has stayed steady). I think a large amount of the change could simply reflect random changes in word use.

Similarly, the use of the word "get" has indeed increased, but so has the use of idiomatic phrases like "get away", "get in", "get out" and "get up", over roughly the same period. The idiomatic phrase "give over", meanwhile, has dropped significantly more quickly than has the word "give". This sort of change in speech fashion could well play a large part in observed changes in the use of a word, without having any reflection of how often we mentally employ the underlying concepts.

As I say, I'm loath to criticise a scientific paper having only read the press release version, and sincere apologies to Prof Greenfield if she's already addressed these concerns. But I know how much people like to jump on the "the English language is going to hell" bandwagon, and if they can hitch it to the equally popular "our society is falling apart at the seams" bandwagon, so much the better. I wouldn't be at all surprised to learn that society is more individualist, for better or worse, than it used to be, but I'd be extremely careful about using this as evidence to support the hypothesis.

The rest of the media uptake (or the part that I've found so far) is substantially more credulous and less thoughtful:

UCLA press release, "Changes in language and word use reflect our shifting values, UCLA psychologist reports" 8/7/2013; "Changes in Language and Word Use Reflect Our Shifting Values", Science Daily 8/7/2013; Richard Gray, "Language in books shows how we have grown more selfish", The Telegraph 8/7/2013; Emily Badger, "200 Years of Books Prove That City-Living Changes Our Psychology", The Atlantic 8/12/2013; Fiona MacRae and Holly Gregory, "The selfish generation: How modern books reveal we've become more self-centred", The Daily Mail 8/8/2013; Hannah Devlin, "Jane Austen's sense of duty and obligation has been lost in modern literature, study reveals", The Times 8/8/2013; Lecia Bushak, "Language Reflects Changing Values In An Urban Society: ‘Get’ Used More Frequently Than ‘Give’", Medical Daily 8/8/2013; Catherine Armitage, "The 'selfie' is just the tip of online narcissism's iceberg", Brisbane Times 8/18/2013; "‘Give’ Gives Way as Word Usage Reflects Shift in Values", Pacific Standard 8/7/2013; "De jane Austen à Bridget Jones, toute l'histoire de l'individualisme: La preuve est dans les livres", Acutalitté 8/8/2013; Lina Bensenouci, "Littérature: La nouvelle génération d'auteurs est-elle égoiste?", Grazia 8/8/2013; "Libros revelan que somos más egoístas", Sexenio 817/2013; "Wie sozialer Wandel Worte macht", Die Presse 8/7/2013; Dirk Förger, "Wortwahl in der Literatur belegt Egozentrik von Stadtbewohnern", Wissenschaft Actuell 8/8/2013; Virginia Groenendijk, "Engelstaligen 'voelen' en 'kiezen' meer, ze 'moeten' minder", Volkskrant 8/12/2013; "书中用词反映现代人变自私", Xinhua 8/9/2013; etc.



  1. Science bible stories, take 27 | said,

    August 19, 2013 @ 12:25 pm

    […] Article FROM < Yesterday I wrote about a recent scientific paper that looks for evidence of the cultural […]

  2. Alex Bollinger said,

    August 19, 2013 @ 12:41 pm

    I notice the word "socialism" spiked in the late 1970's in English lit according to google. Does this mean that the English-speaking world was particularly enamored with socialism in the late 70's?

    "Work" has been in decline since the 1920's, which makes sense because people in the 1950's were just loafing around and getting paid for it unlike their parents' generation, who had to walk uphill both ways in the snow just to get to work.

    "Ginger" has been rising since the 1970's, clearly a sign that red hair is more valued now than ever.

    We could do this all day.

  3. Jonathan Badger said,

    August 19, 2013 @ 1:28 pm

    I know you are being facetious (at least about the latter two; there might actually be something to the first), but the correct way to use n-gram data would seem not to take the results as an end-point but as a starting point. I'm a computational biologist, and we like to look at databases of genetic sequences for trends, but once we find an interesting one, typically it isn't enough to just point it out but rather to do more investigation to see what is causing it.

  4. MB said,

    August 19, 2013 @ 2:12 pm

    Didn't David Lodge deal with this in his early campus fiction back in the seventies?

  5. Steve Kass said,

    August 19, 2013 @ 3:29 pm

    It's hard not to be cynical about this junk, so I won't try.

    Greenfield writes,

    To show that this chronological patterning of “choose” and “obliged” was a function of underlying constructs rather than idiosyncratic to particular words or parts of speech, I selected two noun synonyms—“decision” (noun synonym of the verb “choose”) and “duty” (noun synonym of the adjective “obliged”). The pattern of frequency change was similar to that shown in Figure 2.

    Ok, my turn.

    To show that this chronological patterning was not a function of underlying constructs rather than idiosyncratic to particular words or parts of speech, I selected two noun synonyms—“preference” (noun synonym of the verb “choose”) and “requirement” (noun synonym of the adjective “obliged”). The pattern of frequency change was completely different from that shown in Figure 2.

    How about Greenfield's pair (give, get)? You know, the one where

    "Get" starts to rise again in the 1970s, perhaps because of the onset of the women’s movement.

    Well, perhaps not. The Ngram for (give, receive) doesn't look much the same, does it?

    Enough, though. I like games with rules better.

  6. naddy said,

    August 19, 2013 @ 4:18 pm

    by and large, the "wise men" (and now the "wise women") don't really care about whether the empirical and theoretical foundations of their opinions are sound . They care about readers, ratings, and reputation — and in some cases about political outcomes or cultural values — with truth relevant only insofar as it affects those goals.

    So, to use H.G. Frankfurt's sophisticated philosophical term, they are spouting bullshit.

  7. MB said,

    August 19, 2013 @ 5:27 pm

    _Small World_ (1984). Robin Dempsey of Computational Stylistics, and his discovery of poor Persse's favorite words, beginning with "grease."

    [(myl) Indeed.]

  8. bks said,

    August 19, 2013 @ 11:31 pm

    Jonathan Badger, yes more investigation, but enough?

    One notable finding from the contest was that different assemblers – and the same assemblers in the hands of different teams – did not give consistent results. That echoes the results of Assemblathon 1, which wrapped up in 2011. But the problem itself may be more significant now than it was then due to the democratization of genomics, with many more labs now using many more methods to assemble many more genomes from scratch.


  9. Bert said,

    August 20, 2013 @ 6:33 am

    "… the utterly brilliant Google Ngram tool" – I wonder what's so brilliant about it compared to well-prepared linguistic corpora. That it's publicly available?

    German readers might be interested to read "Sprachlog" blogger Susanne Flach's take on the subject, where she shows that at least the data for "choose" is non-significant:

  10. D-AW said,

    August 20, 2013 @ 7:15 am

    "I wonder what's so brilliant about it compared to well-prepared linguistic corpora. That it's publicly available?"
    Yes, and that its design is intuitive (aside from the Y axis scale) and functions easy to use and understand. This means that lots of people use it to poke around the corpus, which is also a good thing. Not that it can't lead to bad things as well if poorly used or understood.
    I recently posted about something I think Google Ngrams might be showing about "lets you" ( The post is meant to be exploratory in the way Jonathan Badger suggests, and cautious about drawing conclusions. Critiques of the uses of Ngrams on LL are all very educative, but perhaps a primer on making inferences and drawing conclusions from Ngrams would be even more productive, since Google has removed basically all the technical barriers to making the graphs.

  11. Alex Bollinger said,

    August 20, 2013 @ 8:13 am


    The general history of the word socialism shows a rise during the Cold War and then a decline starting at around 1980. My best guess is that's because the English speaking world was so engaged in fighting socialism during that time period. While "my best guess" isn't a substitute for proof, it was the first thing that came to mind as a word that got commonly used because opposition, not support, of the idea it represented.

    The same could be said about Greenfield's or other studies: what if people just got really, really critical of individualism instead of really supportive? It's impossible to know how each use of the word was meant without actually reading the context it's presented in.

  12. Corey B said,

    August 20, 2013 @ 9:08 am

    It's even more like Bible stories in that one doesn't even need the whole "story" to be authoritative; just a reference to the fact that "Researchers Say X, Which Proves Y" is enough to end the debate.

    "I read in the Times somewhere, or heard somewhere, that RESEARCH SAYS society is definitely becoming more selfish."

    Ah, the Voice of Authority.

  13. J.W. Brewer said,

    August 20, 2013 @ 11:09 am

    I think this comparison may be unfair to Bible stories. Since the underlying dataset (i.e., the text of the Bible) is widely available to non-specialists, data that is difficult to reconcile with the edifying moral lesson of the day is ready to hand for students and others who wish to ask impertinent questions, thus requiring a high degree of exegetical legerdemain by preachers and Sunday school teachers who are trying to stay on-message. See, e.g., (discussing the problematic response to kids having made fun of Elijah: "And there came forth two she bears out of the wood, and tare forty and two children of them."). Do tv pundits pontificating about sloppy corpus linguistics claims ever get confronted with that sort of challenge?

  14. J.W. Brewer said,

    August 20, 2013 @ 11:47 am

    Oops. I muddled Elijah with Elisha. But, again, the number of people out there with the ability to notice the error (without even necessarily having to check a reference work) and catch me out is much greater than the number who can figure out what's wrong w/ Greenberg's methodology.

  15. Bob Ladd said,

    August 20, 2013 @ 12:04 pm

    @J. W. Brewer:
    Thanks for the phrase "exegetical legerdemain". Made my day.

  16. Paul said,

    August 20, 2013 @ 12:44 pm

    That video is truly horrifying. It reminds me of this Onion article:,33500/

  17. Dvora said,

    August 20, 2013 @ 11:53 pm

    It's spelled Bible, not 'bible.'

  18. Rodger C said,

    August 21, 2013 @ 8:36 am

    Nothing in the Bible was intened as "history as we know it," nor was any contemporary text. Nor was any of it intended as a pretext for easy moral lessons. As for Elisha and the bears, surely its genre is Shaman Wonder Story, and the way to read it (IMHO) is to imagine we just found it in Dioszegi or somebody and go, Whoa.

  19. A Primer of Genome Wissenschaft, Third Edition | INFOWEBLOG.NET said,

    October 10, 2013 @ 4:40 pm

    […] Language Log » Science bible stories, take 27 […]

RSS feed for comments on this post