Sexual orders

« previous post | next post »

In the comments on "The order of ancestors" (12/24/2009), there was some discussion about the possible role of gender bias in determining the preference for orders like "mothers and fathers" over "fathers and mothers".  This discussion faced a basic empirical problem: there were more plausibly-relevant principles (a long list of apparent semantic and phonological preferences) than there were facts to explain.

In this post, I'll review in more depth the evidence about the preferred orders of English binomial expressions for gendered categories of humans. This review will leave us in the same logical impasse.  Then I'll tell you about the clever solution found by Saundra Wright, Jennifer Hay and Tessa Bent in their paper "Ladies first? Phonology, frequency, and the naming conspiracy", Linguistics 43(3): 531–561, 2005.

Here are some counts from four corpus sources — transcripts of telephone conversations available from the LDC (about 25 million words); newspaper and newswire data from the LDC (about 2.6 billion words); Mark Davies' COCA corpus (about 400 million words); and the British National Corpus (about 100 million words). In each case, I've given the count for the cited order of the phrase, the count for to opposite order, and the proportion of the sum represented by the cited order.

LDC Tel LDC news COCA BNC
male and female 23 1
(96%)
3640 270
(93%)
2635 267
(91%)
446 38
(92%)
man and woman 10 0
(100%)
2748 116
(96%)
759 53
(93%)
188 8
(96%)
mother and father 120 12
(91%)
2455 917
(73%)
1640 545
(75%)
244 141
(63%)
mom and dad 289 5
(98%)
2907 98
(97%)
1885 81
(96%)
mum and dad 494 11
(98%)
boy and girl 8 4
(67%)
468 92
(84%)
170 42
(80%)
43 9
(83%)
brother and sister 98 29
(77%)
2477 800
(76%)
653 198
(77%)
162 10
(92%)
aunt and uncle 61 10
(86%)
603 125
(83%)
262 41
(86%)
40 22
(65%)
niece and nephew 23 2
(92%)
134 41
(77%)
39 5
(89%)
1 2

Here are the counts and percentages for the plural versions, including a few other pairs that are rare in the singular:

LDC Tel LDC news COCA BNC
males and females 7 0 601 60
(91%)
1264 148
(90%)
136 5
(96%)
men and women 137 17
(89%)
38165 2940
(93%)
11653 1369
(89%)
1956 251
(89%)
ladies and gentlemen 2 1 1772 ? 3513 26
(99%)
270 4
(99%)
mothers and fathers 16 2
(89%)
951 356
(73%)
475 164
(74%)
29 11
(73%)
moms and dads 9 2 702 ? 288 8
(97%)
mums and dads 30 0
(100%)
boys and girls 23 5
(82%)
5259 1068
(83%)
2287 467
(83%)
339 85
(80%)
brothers and sisters 316 31
(91%)
4876 380
(93%)
1615 217
(88%)
318 15
(95%)
aunts and uncles 114 5
(96%)
509 131
(80%)
234 66
(78%)
20 19
(51%)
nieces and nephews 117 5
(89%)
800 124
(87%)
257 42
(86%)
12 22
(35%)

The orders are remarkably consistent across the four sources and the two numbers — the only reversal is that the BNC prefers "nephew(s) and niece(s)", while the other three sources prefer "niece(s) and nephew(s)".

However, there is no obvious consistency in terms of gender. Four pairs prefer the male-first order: male(s) and female(s), man/men and woman/women, boy(s) and girl(s), brother(s) and sister(s). Four pairs prefer the female-first order: mother(s) and father(s), mom(s) and dad(s), aunt(s) and uncle(s), ladies and gentlemen. (And there are various variations on these: mom and pop, ma and pa, ladies and gents, etc.) One pair apparently goes one way in the U.S., and the opposite way in Britain: niece(s) and nephew(s).

One well-attested phonological factor (shorter words first) explains three of these orders, and is consistent with all of them (except for the British preference for "nephew and niece"). Other phonological factors (e.g. sonorant-initial words first) may explain other pairs, like "mothers and fathers". We might invoke gender bias for "brothers and sisters" and "boys and girls" — but word frequency is also known to play a role, with commoner words tending to come first, and according to COCA, brother(s) is 30% more frequent than sister(s).

Putting it all together, we have only eight word-pairs to use to test more than a dozen attested ordering principles that are plausibly involved — even before bringing in new suggestions, like a special status for motherhood. We might find a few more word-pairs, but we need hundreds if not thousands to sort out the relative contributions of ten or twenty factors and their interactions, especially because the distribution of the factors is anything but orthogonal.

Wright, Hay and Bent ("Ladies first? Phonology, frequency, and the naming conspiracy", Linguistics 43(3): 531–561, 2005) took an approach to the search for gender bias in binomial ordering that largely avoids this impasse. Here's their abstract:

In pairs of names, male names often precede female names (e.g. Romeo and Juliet). We investigate this bias and argue that preferences for name ordering are constrained by a combination of gender, phonology, and frequency. First, various phonological constraints condition the optimal ordering of binomial pairs, and findings from our corpus investigations show that male names contain those features which lend them to be preferred in first position, while female names contain features which lend them to be preferred in second position. Thus, phonology predicts that male names are more likely to precede female names than follow them. Results from our name-ordering experiments provide further evidence that this "gendered phonology" plays a role in determining ordering preferences but also that an independent gender bias exists: when phonology is controlled (i.e. when two names are "phonologically equal"), subjects prefer male names first. Finally, frequency leads to another tendency to place male names first. Further investigation shows that frequent names are ordered before less frequent names and that male names are overall more "frequent" than female names. Together, all of these factors conspire toward an overwhelming tendency to place male names before female names.

Unfortunately, Linguistics is not an open access journal, and the publisher, Mouton de Gruyter, wants you pay the usual implausible fee of $40 to read the full (32-page) article, unless your institution has a subscription. According to the relevant RoMEO page, the publisher allows authors to archive a pre-print, and may allow archiving of authors' postprint, though not of the published version — but none of the authors in this case seem to have archived any versions. That's a shame, because this article is mostly accessible to a general audience, and has many features of broad linguistic, statistical, and human interest. In addition to its value as purely passive reading material, it might provide the basis for some experiments that undergraduates could run within the context of a course. In fact, a smart high-school student could do a neat science-fair project along these general lines. (Just in case there's any doubt, let me emphasize that I intend this observation as praise, not criticism.)

[Update 1/5/2010 — Tessa Bent has now archived a copy of the paper here!]

In a later post, I'll say a bit more about this paper's interesting methods and conclusions. For now, I'll just quote a couple of passages about differences among subjects, to suggest an area where the authors themselves suggest that some additional investigation might be in order:

[W]hen phonology was controlled, subjects displayed a significant preference for male names to be ordered before female names. Overall, 61% of subjects ordered a majority of pairs in the ‘‘gender’’ condition with the male name first, 26% displayed no preference, and 13% displayed a tendency for the female name first. Moreover, we noticed that this preference was stronger for male respondents than for female respondents. Due to the composition of the subject pool in which this experiment was conducted [NWU undergraduates], we had more female respondents (31) than male respondents (15) in this study; nonetheless, when we break down our data by male and female respondents, some interesting differences are revealed. In the gender condition, 66% of male respondents showed a tendency for a male-first ordering preference, while only 58% of female respondents showed that same tendency.

So that's 10 out of 15 males who preferred male-first orders when other factors were balanced, and 18 out of 31 females — but if one male and one female had switched their overall response pattern, that would have been 9/15 = 60% of the males, and 19/31 = 61% of the females. It's plausible that other sorts of variation among subjects would show effects at least that great — undergraduates at Reed or Oral Roberts rather than Northwestern; people born in the 1930s vs. those born in the 1990s; etc.

In the authors' second experiment, they declined to speculate about the effects of subjects' sex, because there were too few male subjects:

Replicating the results from Experiment 1, we found that when phonology is controlled, subjects display a significant preference for male names in first position. Overall, 57% of subjects ordered a majority of pairs in the ‘‘gender’’ condition with the male name first, 29% displayed no preference, and 14% displayed a tendency for the female name first. Since there were a limited number of male subjects who participated in this second experiment (only six total males [out of 28 NWU undergraduates]), we were unable to determine whether there were any differences in responses according to the gender of the respondent.

But even if there had been two or three or four times as many male subjects, there's a problem here that goes beyond sampling error.

I'd like to repeat an observation that I've made several times before, echoing many others before me. There's a curious tendency for psychologists and psycholinguists to study cultural phenomena as if they were studying physiology or psychophysics. More exactly, they assume that their subject pool (typically undergraduates at their institution who take introductory courses in psychology and some other fields, and sign up for the experiment in question) are representative of the human species at large.

It's plausible that for a study of (say) two-tone suppression, a subject pool of undergraduates from a particular course at a particular university would yield results characteristic of healthy 18-22-year-olds in general (though there can be surprises even in cases like that). But it's downright bizarre to expect (for instance) that a survey of 20 UCLA medical students could tell you something worthwhile about the responses of male and female Americans in general to various presidential candidates — even if you survey them with an fMRI machine instead of a clipboard. And see here for discussion of some other similar examples.

In Wright et al.'s studies, the goal is to discover whether there's a gender bias in binomial name ordering, and how it interacts with other factors affecting binomial order. Can we take the responses of NWU undergraduates as representative? Not without some additional testing, I think.

The context and content of these experiments surely clued subjects in to the idea that they should be thinking about the role of sex in the order of names. And sex is probably a more salient dimension to undergraduates than to humans in most other life stages. On the other hand, contemporary undergraduates are fairly well sensitized to the issue of sexism. Overall, I don't feel especially confident about predicting the results of replication among elderly nuns in New York City, or 10-year-olds in the Minneapolis suburbs, or middle-aged electricians in Tulsa.

(The effect of subject sex was a secondary or tertiary point in their paper, so this is not an especially important criticism — but it does suggest an area where the previously-mentioned undergraduate or high-school science project might easily discover something new.)

One last general observation. The order preferences observed in common binomial expressions are fairly well predicted by the factors demonstrated to influence experimental subjects' choices for in novel binomials — but the order of common binomials seems to be more polarized, in general, than such factors would predict. (I haven't shown this to be true, but let's pretend for the sake of argument that I have.)

This may seem puzzling, given that we  expect a broader subject population (and a broader range of contexts) to show more diversity in response patterns. This in turn suggests that texts produced by the society at large should show less polarized (rather than more polarized) binomial orders.  If the facts are indeed the opposite of this — i.e. that the observed frequency of binomial orders is more polarized than the experimental effects would suggest — then perhaps  the tendency to remember and re-use frozen forms (in this case, frozen forms of binomials) has the effect of amplifying the "natural" ordering preferences, though the sort of "Chinese restaurant process" that plausibly operates in the case of other usage-sensitive linguistic phenomena as well.

[Update — Simon Fodden has some interesting suggestions about the distribution in Canadian legal material of the different orders of  terms in the binomials "men and women", "boys and girls", "mothers and fathers":

It seemed to be the case that where women and girls came before men it was a function of a special context: either the topic was equality or family. The simplest illustration of that might be “mothers and fathers”, which carries with it the familial context; notice here, though, that legislatures seem to be made of sterner stuff than judges, preferring “fathers and mothers” despite the stereotypical pull of family.

I don't know of any systematic studies of such contextual effects (though that is not reliable evidence that none exist).]



25 Comments

  1. Chris said,

    December 27, 2009 @ 12:56 pm

    I agree that sampling is a concern within psycholinguistics. The standard response, I'm sure, is "that's what we're stuck with." As you know, it's just not that easy to find a bunch of people willing to sit down in a small, windowless room and press the space bar a couple hundred times. Or have an eye tracker attached to their heads. Even undergraduates have to be bribed with extra credit (or, less frequently, cash). I'd love to see a chain of psycholinguistics stores pop up in every mall around the country, wedged in between the AT&T Wireless booth and LensCrafters

    … and btw, LOVED seeing someone from my undergraduate Alma mater Chico being blogged about at LL!.

    [(myl) Luckily, studies of the Wright et al. type are purely pencil-and paper (or keyboard and mouse, if you do them on the internet), and so there's less of barrier to better demographic balance. I think the main problem is that cognitive and behavioral (as opposed to social) scientists — and the referees at journals in those fields — are generally just not aware that this is an issue. Of course, it's obviously easier just to take a couple of dozen kids from the psych department's subject pool. But it would be even easier to rely on your own intuitions, and whether you go beyond that marks a different cultural distinction among researchers.]

  2. John Laviolette said,

    December 27, 2009 @ 3:14 pm

    Future research topic, for anyone who's interested: comparing ordering of male vs. female names in the general population to pairs of entertainers (comedy teams, singers/musicians, etc.) I've heard that comedy teams have a rule that the funnier person comes first and the straight man second (Laurel and Hardy, Hope and Crosby,) although this seems to be overruled by semantic and phonological principles (Abbot and Costello.) Burns and Allen breaks the pattern, but is that because of a male-first bias, or shorter-name first? In music, Sonny and Cher clearly breaks the shorter-name-first rule in preference to male-first. This is all anecdotal, of course. Has anyone done a study on naming constraints for entertainment pairs?

    [(myl) I don't think so. You could make up a list and see how it comes out. For the reasons that you cite, I think that entertainer-name binomials might not be typical of binomial patterns in general, though.]

  3. Army1987 said,

    December 27, 2009 @ 4:25 pm

    Incredibly, in all but two of these example the order is the one which sounds to me more natural in Italian, too. ("Aunt and uncle" is the exception, and that doesn't apply to "niece and nephew" as in Italian they're both "nipote".)

  4. Army1987 said,

    December 27, 2009 @ 4:28 pm

    (Except "mother and father", too.)

  5. James D said,

    December 27, 2009 @ 5:48 pm

    I do wonder if the traditional British pronunciation of "nephew" with a "v" rather than an "f" has something to do with the oddity in word order. "V" seems to be an awfully weak consonant.

  6. Lazar said,

    December 27, 2009 @ 6:18 pm

    Another one is "sons and daughters". I've mused about this question myself, including non-human cases such as "knife and fork", "salt and pepper", "eyes and ears", "arms and legs", "up and down", "north and south", "east and west", etc. etc.

  7. Kenny Easwaran said,

    December 27, 2009 @ 6:51 pm

    "Further investigation shows that frequent names are ordered before less frequent names and that male names are overall more "frequent" than female names."

    What do they mean by "frequent" names here?

    [(myl) They operationalize it in terms of census counts, as I recall, which is probably a pretty good proxy overall for most people's experience. If their raw data were available (and if I can be excused for beating this drum again, wouldn't it be nice if such data always was?), you could try substituting some corpus counts and see if it fits better.]

    Surely there are far more people named "Leticia" than there are named "Barack" (at least in the United States), but it's also quite plausible that the name "Barack" has been uttered more times (at least in the United States) than the name "Leticia". Thus, male names could be more frequent than female names either in a sense that there are fewer common male names, so that more people have any particular name, or in a sense that people are more likely to refer to males by name than females.

    [(myl) It's an old and commonplace observation that the top N male name types cover a larger percentage of the relevant tokens than the top N female names do.]

    Also, is the name "Nicole" counted as the same name when different people have it? When talking about Paris Hilton and Nicole Richie, one might prefer "Nicole and Paris", because "Nicole" is a much more common name than "Paris", but one might prefer "Paris and Nicole", because this Paris is mentioned far more often than this Nicole.

    [(myl) They modeled words, not referents. In this case, I don't think that their experiments were set up to take account of effects related to the relative celebrity of referents. It wouldn't be a shock to find that this made a difference.]

  8. Brett said,

    December 27, 2009 @ 9:05 pm

    They mentioned the specific example of Romeo and Juliet, whose order is fixed by the title of their play. I realized that it was interesting that the title is typically pronounced in such a way as to make Juliet as heavy as possible, with three distinct syllables. To my ear, this sounds much better than using the two-syllable Juliet that Shakespeare's iambic pentameter implies to be the intended pronunciation.

  9. Daniel said,

    December 27, 2009 @ 9:29 pm

    I wonder how much rhythm plays a part in deciding which order is preferred. While many of the examples given have the same rhythm in either order, there's a few that aren't very reversible rhythmically.

    "Ladies and gentlemen" has a nice triplet feel to it, whereas "gentlemen and ladies" is just a mess.

    "Male and female" is another one that has a pleasant triplet feel, whereas the reverse "female and male" becomes rhythmically clumsy. (Another curiosity: when I read "male and female", male contains a diphthong and female doesn't; when I read "female and male", female gets the diphthong, so the rhythm gets weird. If I try to read "female and male" with no diphthongs, it flows again, but feels strange in the mouth.)

    "Niece and nephew" has what I'm going to call a 2/4 rhythm (forgive the terminology: my interest in rhythm comes more from music than poetry), whereas the reverse "nephew and niece" is back to triplets. Thinking about it, I'm not sure which I prefer. The former seems more "normal", by analogy to "nieces and nephews", but the latter flows better. I don't think I've ever had occasion to say it either way.

  10. Mr Punch said,

    December 28, 2009 @ 8:07 am

    I just want to second James D's point about "nephew" — if phonology matters, it may not be coincidence that the one inconsistency in the chart involves a word that is pronounced differently by the diverging population.

  11. Simon Cauchi said,

    December 28, 2009 @ 9:58 am

    @Brett: Perhaps I'm a victim of the recency illusion, but the mispronunciation of "Juliet" that you refer to strikes me as relatively new. My (adult) son pronounces the name that way, and I correct him, yet he carries on saying the three-syllable "Juliet" (which is presumably influenced by "Juliette"). That wasn't how we pronounced the name when we read the play at my UK school in the 1950s.
    "Romeo and Juliette" doesn't sound better than the traditional pronunciation. It sounds a whole lot worse. (IMO.)

  12. Heck said,

    December 28, 2009 @ 12:29 pm

    Re: John Laviolette's comment.

    I believe the tradition in comedy teams that performed live was to give the straight role first billing (as in Martin & Lewis, Abbot & Costello, Burns & Allen. The rationale being that the straight role was effacing onstage and therefore that person would tend to be overshadowed in the public's eyes without their name coming first.

  13. J. W. Brewer said,

    December 28, 2009 @ 12:56 pm

    1. John L., I'll see your Sonny & Cher and counter with Peaches & Herb (female name, which is also longer name, first). See also, e.g., Shirley & Lee (original '50's hit version of "Let the Good Times Roll"). But then there's the Captain & Tennille . . . Leaving aside the question of whether mixed-doubles entertainment acts are representative of anything other than themselves, it might be hard but not impossible to study this in some reasonably empirical way, at least for the U.S. popular music industry. Using the massive chart-history reference works compiled and published by the one-man-industry named Joel Whitburn, it would be possible (although I don't know how you could automate the process rather than do a review by hand-and-eye) to compile a list of all male/female duos that got a record onto any of a number of specified Billboard charts (or, with additional screening in the review process, got a record at or above a specified cut-off point on the chart, e.g. top 10 or top 40) during a specified range of years, coding the resulting list for M-before-F v. F-before-M.

    [(myl) This is a good idea — and "hand and eye" is a fine method, given that the published studies out there are typically based on a few dozen to a few hundred examples. You could probably create a database of 1,000 pairs in less than a day's work, given a list to start from.]

    2. I have noticed a phenomenon in my own current little anthropological niche (reasonably affluent NYC suburbs) in formal written contexts like address labels for social events or alphabetized lists of married couples being recognized as sponsors/contributors in the written program for a charity event. If a given couple is listed with both given names, rather than as "Mr. & Mrs. Herbert Q. Snodgrass," the wife's name is quite frequently (perhaps dominantly?) given first, e.g. "Myrtle and Herbert Snodgrass," rather than the other way around. This seems perhaps especially common if the wife retained her maiden name, e.g. "Myrtle Glutz and Herbert Snodgrass," rather than the other way around. I don't know how recent or ubiquitous this pattern is, and for all I know it may be quite limited by geography, class, or ethnicity. For what's its worth, my amateur anthropological insight is that these norms of formal addressing are predominantly maintained, transmitted, and implemented by the married women of the tribe (in my immediate context generally neither extremely liberal nor extremely conservative in terms of the contemporary cultural norms of broader U.S. culture), their husbands being largely oblivious to these social nuances.

    [(myl) I expect that some of these lists are on line (e.g. here), and could be the basis for another interesting corpus.]

  14. J. W. Brewer said,

    December 28, 2009 @ 6:18 pm

    I *know* raw google hit counts are grossly unreliable in all sorts of ways, but those with access to better corpora likely to capture texts in the relevant registers/varieties of English may wish to further investigate the interesting-to-me google-hit factoid that "sistren and brethren" is (or would be if the numbers were reliable) twice as common as "brethren and sistren."

    [(myl) You share with the rest of us easy access to one of the best corpora around, Mark Davies' COCA, which now amounts to about 400 millions words of diverse American English text. It has one example of "sistern and brethren", and two instances of "brethren and sistren".

    You also have access to the British National Corpus (BNC), with about 100 million words. It has no hits for either order.

    The LDC's News corpus, with about 2.6 billion words, has 3 hits for "brethren and sistren", and no hits for the other order.

    These numbers are all believable as reports of the counts in the corpora surveyed, but they are too small to be trusted as an indication of population tendencies (if there really are any consistent tendencies in this case).

    When I try Google, I get 12,900 for "brethren and sistren", vs. 24,500 for "sistren and brethren", which is consistent with your observation. In contrast, Microsoft's Bing search engine returns 13,700 hits for "brethren and sistren" vs. 179 for "sistren and brethren". Yahoo returns 13,400 hits for "brethren and sistren", vs. 677 for "sistren and brethren". These numbers are large enough not to worry about sampling error, but I think it would be very unwise to believe any of them as accurate representations of any particular count in any particular sample, since (certainly in the case of Google) they are estimated by spectacularly unreliable techniques.

    So far, the only sources that I've found that have counts that are large enough to be free of sampling error, and also created by counting techniques that are demonstrably more reliable than a ouija board, are various large news archives. Google News Archive yields 278 hits for "brethren and sistren" (all of which are viewable at least as snippets), vs. 6 for "sistren and brethren". LexisNexis "Major U.S. and World Publications" yields 62 hits for "brethren and sistren" vs. 3 for "sistren and brethren".

    Interestingly, the proportions in these cases for the b & s order (97.9% for Google News Archive, 95.4% for LexisNexis) are within spitting distance of the Bing (98.7%) and Yahoo (88.1%) proportions, with the Google proportions as a serious outlier (34.5%). This data point certainly supports previous indications that Google counts are completely worthless as an indication of relative frequency of usage — here the ratio of counts for the orders as per Google web search (0.53) differs by a factor of 40 to 90 from the ratios estimated from apparently reliable sources of counts (Google news search 46.3, LexisNexis news search 20.7. Whether counts from Bing and Yahoo are generally more reliable is perhaps worth further study — certainly in this case, the ratios are at least well within an order of magnitude. ]

  15. Peter Taylor said,

    December 28, 2009 @ 8:13 pm

    With some of these pairs it would be tricky to account for reasons which aren't inherent in the word. For example, taking the controversial one, I referred earlier today to my nephew and niece, but I think that the reason I put them that way round is that my nephew is twice the age of my niece and so I think of him first. I believe this was referred to as the "closeness factor" in the previous post on this general topic.

    (As an aside, what's this about pronunciation? If I say "effervescent nephew" the "nephew" has the f sound of effervescent rather than the v, and I'm from south-east England. Not quite RP but low-middle to middle-middle class Home Counties).

  16. The Volokh Conspiracy » Blog Archive » Why Do We Say “Moms and Dads” But “Ladies and Gentlemen”? said,

    December 28, 2009 @ 9:24 pm

    […] fascinating post on this, with both data and theorizing, from Mark Liberman (Language Log). Categories: […]

  17. David in Brooklyn said,

    December 29, 2009 @ 12:43 am

    Since pregnancy is manifest, matrilineage is usually a surety. Patrilineage can be dicier. Perhaps this explains a preference for "ma and pa", "mom and dad", etc.

    If men had the wombs, but women the breasts — that is, patrilineage was known, but women remained special nurturers — I wonder would the preference disappear or reverse itself?

  18. The American Spectator : AmSpecBlog : Daily Must-Reads said,

    December 29, 2009 @ 11:28 am

    […] Why do we say "men and women" but "ladies and gentlemen?" Read this very long, very dry post to find out (Language Log) […]

  19. Nathan Myers said,

    December 30, 2009 @ 1:09 am

    The following appeared in an odd place: http://www.blog.sethroberts.net/2009/12/29/science-of-everyday-life-why-boys-and-girls-why-not-girls-and-boys/#comment-376256 .

    Actually, there is a simple and coherent (and testable…in this way) theory that explains this. Long ago, Roman Jakobson, the Slavic linguist who worked for years at MIT, discussed the notion of markedness in both phonological but also semantic/lexical relations. Briefly, the first term in such a pair is generally the “unmarked” term, whereas the second is the marked one. If one of them can be used as the generic, it will be the unmarked one. In Romance languages, for example, if you ask “do you have brothers?” the word for “brother” can be either “brother (m) or sibling(m/f)”. The word for sister is restricted only to female siblings. The ways that noun-phrase marking languages do this(and transform over time) is at the boundary of culture and cognition–see the signal (and early) MA thesis by Deborah Spitulnik, a senior linguistic anthropologist at Emory, on the nominal class struggle in ChiBemba (nominal here relates to nouns, not to other senses of the word), and the now classic article by linguistic anthropologist Michael Silverstein (emeritus at Chicago) called “Language and the Culture of Gender…” as well as more recent work on language and gender in ling anth and sociolinguistics by scholars such as Penelope Eckert, Kira Hall, Mary Bucholtz and the like. For specific cites see Google Scholar but these are all worth reading.

  20. Nathan Myers said,

    December 30, 2009 @ 1:14 am

    For comparison, Google counts 44,100 for "brethren and cistern", but only 4 for the reverse order. I leave research on occurrence of "mammaries and daddaries" to others.

  21. Tessa Bent said,

    December 30, 2009 @ 3:59 pm

    For those interested in reading the Wright, Hay & Bent paper, it is now posted on my website (tessabent.com). Enjoy!

  22. uberVU - social comments said,

    December 30, 2009 @ 9:37 pm

    Social comments and analytics for this post…

    This post was mentioned on Twitter by languagelog: Sexual orders: In the comments on "The order of ancestors" (12/24/2009), there was some discussion about the possib… http://bit.ly/8gEWu2

  23. Linkspam hangover (4th January, 2010) | Geek Feminism Blog said,

    January 4, 2010 @ 8:19 pm

    […] Log explores gender bias in the order of English binomial expressions through statistical analysis of four text corpus […]

  24. Sili said,

    January 7, 2010 @ 2:18 pm

    I do love the fact that LL is now so influential that 'random' authors go out of their way to please the readership.

    Thank you, dr Bent.

  25. Meirav said,

    November 11, 2010 @ 7:13 pm

    Interesting! In Hebrew we seem to have a consistent preference for male to precede female – including "father and mother", "uncles and aunts". The only exception I can think of is for "ladies and gentlemen", where we do mention ladies first.

RSS feed for comments on this post