On pronoun typology and economic measures

« previous post | next post »

Below is a guest post by Bob Kennedy.


This post is adapted from a letter I wrote to the editors of the journal Kyklos, in response to the recent publication of "Do Linguistic Structures Affect Human Capital? The Case of Pronoun Drop", by Prof Horst Feldmann of the University of Bath.

I shared this letter in a public Facebook post that also circulated via email and Twitter, hoping to gain the support of peers and colleagues willing to co-sign in agreement. Through social media, emails, and other means, I have received such support from nearly 500 individuals, a number that far exceeded my expectations.

I attribute this mass of support to the observation that Prof Feldmann's paper is laden with empirical issues with drastically problematic implications, hinging on a series of false and faulty premises. Along with the endorsement of these cosignatories, I have thus requested that the editors take the time to understand the perspectives I share here, and to consider retracting the paper. I hope this degree of reaction is seen as a result of serious problems we detect in the analysis, and that a constructive outcome will be possible.

In his paper, Prof Feldmann levels the claim that nations whose languages have the linguistic feature of "Pro-Drop", a phenomenon in which subject pronouns are optional, are also more likely to score lower on a series of measurements of individual freedom and economic indicators of prosperity. He presents an argument nested in complicated statistical analysis to support the claim that this relationship between a language's grammar and the economics of its community is a robust one. In addition, he makes the claim that Pro-Drop languages preserve an ancient cultural value that non-Pro-Drop languages have evolved beyond. Such a claim is unfortunately ethnocentric.

There are numerous empirical issues to be pointed out that immediately refute the arguments in this paper, and it is not merely a matter of a different theoretical perspective. I expand on several of these here.

First, most confounding for this paper is its statistical analysis. While Prof Feldmann does not list all the languages and nations included in his analysis, he does note that his data are drawn from the World Atlas of Linguistic Structures (WALS). Although WALS is a reliable source for understanding the geographic extent of grammatical features such as Pro-Drop, its data are not appropriate for the types of statistical analysis that Prof Feldmann uses.

Prof Feldmann's analysis explicitly attributes higher scores in economic and individualistic measures (causally) to the absence of the Pro-Drop feature. As such, it seeks to identify factors which may ensure a particular society enjoys more individual freedom as opposed to collectivism. Indeed, it so happens that most of the non-Pro-Drop languages included are in his study are spoken in wealthy northern European nations. From that, I infer that the author seeks to argue that the languages of northern Europeans beget individualism over collectivism more so than the languages of other communities do. The aim of the hypothesis thus seems be to attribute wealth and individualism (in part) to the structure of the language of a community – I consider this stance ethnocentric, and it risks being interpreted as implicitly racist. I would have the same interpretation if the statistical analysis were not so confounded, but the fact that the analysis errs in favor of supporting this hypothesis is what motivates writing this letter.

In particular, his analysis is fundamentally and fatally flawed because it fails to account for the fact that languages share grammatical features because of shared history, and communities and nations share economic traits because of cultural diffusion. For example, English, Dutch, German, and the languages of Scandinavia share the trait of being non-Pro-Drop because they are related to each other; French also arguably has this trait because of its areal proximity to these other northern European languages. The economic traits of these nations are also highly correlated. Yet Prof Feldmann's analysis requires an assumption that each language has developed the non-Pro-Drop feature independently, and that the economic systems of these nations have also developed independently of each other.

This represents a profound abuse of typological linguistic data. Statistical analysis of such data should account for language relatedness in its sampling methods; otherwise, the size of each group in comparison is conveniently inflated, distorting the power and effect size of the analysis. I would assume this is a result of the Prof Feldmann's lack of awareness of linguistics as a field, rather than a deliberate manipulation of data, but regardless, his paper's misrepresentation of linguistic facts is irresponsible. It seems much of the prior literature that Prof Feldmann cites is fraught with the same underlying issue, going back to Kashima and Kashima (1998). Moreover, other publications already note these same issues; for example, Lee (Journal of Language Evolution 2, 2017) demonstrate that work including Kashima & Kashima as well as subsequent research on Pro-Drop effects use models that include a large number of closely related (and thus correlated) Indo-European languages in their samples. When the analysis is properly controlled for the inter-relatedness of languages, the effects disappear.
Second, the characterization of Pro-Drop languages as fundamentally different in grammatical character from non-Pro-Drop languages is highly problematic; indeed, assuming it to be a binary dimension is itself inaccurate. Many so-called non-Pro-Drop languages (notably among them, English) tolerate a certain degree of subject pronoun deletion. Furthermore, the vast majority of Pro-Drop languages indicate the identity of their grammatical subjects through other means, such as verbal agreement systems or switch-reference systems, among others. In addition, Prof Feldmann implies that non-Pro-Drop systems evolve from Pro-Drop systems; however, historical linguistic analysis demonstrates that Pro-Drop systems develop from non-Pro-Drop systems and vice versa.

Third, Prof Feldmann's argument rests in part on an assumption that language boundaries are co-extensive with cultural, national, and economic boundaries. While this may be necessary for the implementation of its statistical tests, it does not match the reality of how linguistic and cultural phenomena transcend national boundaries.

It is frustrating for linguists to see such papers published in the journals of other fields, particularly with this type of argumentation. Linguistics as a field has been working for more than a century to refute the common misconception that some languages are more suited than others at expressing complex ideas. In fact, the original claims of Sapir and Whorf (cited in Feldmann's paper) were of a very different nature: that a language can always meet the needs of its speakers. Papers like Prof Feldmann's instead advance an ethnocentric narrative that some languages are more conducive to technological and economic advancement. This in turn risks coming across as implicitly racist, even if this were not the author's intent. That this narrative is presented in your journal under the guise of empirical science, but on the basis of a series of false or falsified claims, is especially troubling.

A large and internationally diverse number of linguists and anthropologists have voiced their agreement with the stance I present here. With their support, I express my sincere disappointment at the publication of this paper, and have requested that the editors of Kyklos consider retracting the paper. In addition, I make a general request to editors in other fields, that should you encounter submissions in the future that rely so heavily on complex understanding of the findings of linguistics, I implore you to choose reviewers who read such submissions with more attention to their claims about language.

Despite the substantial support I received for making this stance public, I am aware of concerns that the actions here amount to censorship. I think such concerns are over-wrought; censhorship is an act of an authority silencing the voices of artists, dissenters, journalists, or academics. We as a community have no sway or recourse over this or any other journal to silence its choice of papers for publication – all we have is an opportunity to voice our concern, collectively, about any analysis that uses linguistic data in such a problematic way.

Last, it is also clear to me that it is just as appropriate to respond to this paper via standard routes of peer-reviewed publication instead of via letters to editorial boards. I believe both avenues are useful directions to pursue, but only a public letter can fully capture the wide-ranging disappointment among our colleagues.


Above is a guest post by Bob Kennedy.

As background, you may find it worthwhile to review our exchange a few years ago with Keith Chen.

Geoff Pullum started the discussion: "Keith Chen, Whorfian economist", 2/9/2012.

I added some simulations relating to the problem of cultural diffusion and statistical degrees of freedom: "Cultural diffusion and the Whorfian hypothesis", 2/12/2012.

Keith Chen responded in a remarkably collegial way: "Whorfian economics", 2/21/2012. His paper was published as "The Effect of Language on Economic Behavior: Evidence from Savings Rates, Health Behaviors, and Retirement Assets", American Economic Review 2013. To his credit, he engaged a large number of linguists in discussion, as indicated in that paper's acknowledgements:

I am indebted to Judy Chevalier, Östen Dahl, Ashwini Deo, Bob Frank, Shane Frederick, Emir Kamenica, Mark Liberman, Elisa Long, John McWhorter, Sharon Oster, Ben Polak, Geoffrey Pullum, Frances Woolley, an anonymous referee and seminar participants at Berkeley, Boston University, Harvard, Penn Linguistics, Stanford Economics, Stanford Linguistics, UCLA, UCSD, Wharton, Yale Economics, and Yale Linguistics, for generous feedback and suggestions. Special thanks are due to Nicole Palffy-Muhoray, who provided extensive feedback on multiple drafts of this paper, and to Jane Bang and Ryan Caro for their research assistance. All errors are my own. I have no relevant or financial interests related to this project to disclose.

The whole thing struck me as a model of productive interaction.

 



37 Comments

  1. AntC said,

    December 10, 2018 @ 3:25 pm

    historical linguistic analysis seems to suggest no particular developmental precedence for whether a language has always-required vs sometimes-optional pronouns.

    Then which idiot called it "Pro-drop"? (As if I need to ask: it was the usual bigoted English/IE-oriented monoglots.) If Linguistics terminology has already slewed the bias that pronouns are "really" there in 'deep structure', no wonder amateurs are going to run with that conceit. There could be no evidence that humans' Language Facility mentally produces a pronoun then drops it. If some model of grammar wants to go through those motions for sake of 'efficient description', that's an artefact of the theory: only.

    Let's get some neutral terminology before calling the kettle black.

  2. Victor Mair said,

    December 10, 2018 @ 4:45 pm

    Pro-drop language:

    =====

    The term "pro-drop" stems from Noam Chomsky's "Lectures on Government and Binding" from 1981 as a cluster of properties of which "null subject" was one (for the occurrence of pro as a predicate rather than a subject in sentences with the copula see Moro 1997).

    =====

    Source

  3. J.W. Brewer said,

    December 10, 2018 @ 8:17 pm

    Professor Feldman has another paper from a few years ago (also published in the journal Kyklos) with this abstract:

    "Both Spanish and French colonial education included several features that restricted education. Many of them persisted long after independence. Against this background, this paper econometrically studies whether in the recent past the colonial legacy still affected schooling in the ex-colonies of these two former colonial powers – and, for comparison, in the ex-colonies of Britain, the third of the former big three colonial powers. Using a large sample of countries and numerous controls, it finds substantial negative effects on both secondary enrollment and average years of schooling in former French and, especially, in former Spanish colonies. The negative effects on females are particularly large. By contrast, there are no effects in former British colonies."

    This paper may or may not be well-done, and might or might not be thought by specialists in the history of the educational systems involved to handle the evidence well, but it strikes me ex ante as more empirically plausible than a claim that any pop-Whorfian side effects of morphosyntactic differences among English, French, and Spanish had similar effects. But I'm not sure that the allegation of ethnocentrism and/or implicit racism should have any more weight for one paper than the other. It would be rather distressing if it is the case that either: a) Prof. Kennedy and his supporters do not think the alleged mishandling and misunderstanding of linguistic data would be worth bringing to the attention of the editors of Kyklos absent the alleged ethnocentrism/etc angle; or b) Prof. Kennedy and his supporters do not think the editors of Kyklos will take seriously allegations that their author mishandled and misunderstood the linguistic data unless the alleged ethnocentrism/etc angle is also thrown into the mix. Of course, it could be b) and they could as an empirical matter be correct, which would also be distressing.

    FWIW, I tend to agree with AntC that the Chomskyan labeling of the phenomenon as "pro-drop" (with its implicit presuppositions about what is the norm and what is a deviation from that norm) might be more fairly called ethnocentric!

  4. chris said,

    December 10, 2018 @ 10:48 pm

    I agree that "pro-drop" seems like a weird way of describing the phenomenon. You can't drop something that was never there in the first place. Wouldn't "zero subject" be a more natural way of describing it? Some languages permit a zero subject and some don't?

    Are there any languages that regularly allow a zero subject *and* a zero copula, so that a bare adjective is a grammatically complete sentence (e.g. "Hungry" = "I am hungry", or depending on context, "The person we are talking about is hungry")?

    Although now that I've asked that, I realize that "regularly" is doing a lot of work and is a rather slippery concept. I would think nothing of either asking or being asked something like "Hungry?" or "Tired?", and the indicative equivalents require only a minor amount of context dependence to be perfectly normal. If those aren't "complete sentences", then completeness is manifestly optional in live usage, so why care about it?

    If the boundaries of what a language "permits" are that fuzzy, then neatly separating languages into "pro-drop" (or whatever you want to call it) and non seems to be a bit of a fool's errand, which is unfortunate when it's also an indispensable prerequisite for the type of analysis you are trying to do. Or to put it a little more bluntly, you can't demonstrate a correlative, let alone causal, relationship involving a variable that isn't well defined in the first place.

  5. Victor Mair said,

    December 11, 2018 @ 12:50 am

    Readings

    "Keith Chen, Whorfian economist" (2/9/12)

    "Cultural diffusion and the Whorfian hypothesis", 2/12/2012)

    "Whorfian Economics" (2/21/12)

    "Thought experiments on language and thought" (2/22/12)

    "Keith Chen at TED" (2/20/13)

    "Keith Chen animated" (9/7/13)

    "Futurity in Chinese and English and Its Supposed Economic Consequences" (April, 2013)

  6. R. Fenwick said,

    December 11, 2018 @ 1:08 am

    @chris:

    Are there any languages that regularly allow a zero subject *and* a zero copula, so that a bare adjective is a grammatically complete sentence (e.g. "Hungry" = "I am hungry", or depending on context, "The person we are talking about is hungry")?

    The closest I know of is Turkish: hasta "(it is) ill, sick". This is largely due to the almost total erosion of the older Ottoman Turkish i(mek) "to be". In modern Turkish, the older root *i– only surfaces as a vestigial –y– when non-final after a vowel-final adjective (or noun) stem: hastaydı (← *hasta idi) "it was sick", hastayken (← *hasta iken) "while being sick".

    But perhaps it's not quite what you're looking for, since zero subject-marking isn't semantically null in Turkish: although the bare adjective may be a grammatically complete sentence (uzun "it is tall"), it's obligatorily understood as third-person singular, present tense. Turkish bare adjective prompts to Ubykh-speaking informants invariably elicit Ubykh forms overtly conjugated for person (Turkish uzun → Ubykh a-wá "it is tall", rather than the root Ubykh adjective wa "tall").

  7. Philip Spaelti said,

    December 11, 2018 @ 3:50 am

    Starting from a discussion which complains about taking things out of context, it's more than a little ironic that people are complaining about "pro-drop" being "ethnocentric". The original term was coined in an extremely technical sense – where "pro" stood in contrast to "PRO" (affectionately known as big-PRO and little-pro) as well as other empty categories. Note that "pro" in this context stands for a grammatical feature, and is not just some sloppy name for "pronoun". There was never any ethnographic intention to these terms.

    But more to the point, I do think it is appropriate to see the languages which allow omission of pronouns to be "dropping" something. I'm pretty sure that all languages allow overt pronouns under the right conditions. And languages that have generally obligatory pronouns do often allow omission in certain cases. For example Swiss-German systematically omits the 2nd person singular pronoun (in questions), even though pronouns are obligatory everywhere else. Does that make it "pro-drop" or not?

  8. maidhc said,

    December 11, 2018 @ 3:54 am

    So if a country became more autocratic, its citizens would start dropping pronouns? And if it became more democratic they would start adding them back? Like the way it's said that women's hemlines match the state of the economy.

    Spanish is a language where you can put the subject pronoun in or not, because it is implied by the form of the verb. So you could do an experiment over a corpus of publications in Spain during the 20th century, determining whether pronoun usage reflected the historical progression of the Republic, the rise of Franco, the death of Franco, etc.

    I believe that Japanese is a language where the subject can be omitted if it is obvious from context. And Japan has had some fairly major changes in governance over the 20th century, so a similar study could be done.

    Thus it seems fairly easy to test this hypothesis using available data.

    If one wants to argue that the effects take place over a much longer timescale, it is more problematic to test, since democratic government has only appeared fairly recently on the historical time scale. This did occur first among IE-speaking countries, but the reasons are fairly complex, and it would be hard to argue that is has to do with the nature of language. The progression of the Industrial Revolution might have more to do with it.

    Pronoun-using Tudor England, for example, was what we would now consider a fascist state, complete with a secret police and torture chambers for producing confessions from suspected traitors.

    There is at least one case that sort of supports the idea, although not exactly. In Vietnamese, when you talk to someone you have to include a status relationship expressed as a family member, like grandma, uncle, nephew, etc. But I read an article, I'm not sure if it was here or elsewhere, that in immigrant communities in the US and other countries, these relationship terms are frequently omitted. So I suppose one could argue that this shows the influence of a more informal and egalitarian culture.

  9. Philip Taylor said,

    December 11, 2018 @ 7:10 am

    Chris — "Are there any languages that regularly allow a zero subject *and* a zero copula" — to the best of my belief, Chinese/Mandarin/MSC/Putonghua, where in response to a question such as 你是英國人嗎
    ("Nǐ shì yīngguórén ma ?", = "Are you English ?", a valid response would be 是 ("Shì", = "[I] am"), although more idiomatic would be 是 的 ("Shì de", = "[I] am"). VHM will, of course, be in a far better position than I to judge whether this is correct.

  10. Thaomas said,

    December 11, 2018 @ 7:38 am

    The legitimate complaint is not whether Feldman's paper is "ethnocentric" or even possibly interpretable as "racist," as but rather that it's null hypothesis is (probably) not validly rejected because the assumptions of the statistical test (that each observation is independent of the rest) is, as Kennedy points out, incorrect. The correct response should not be to request withdrawal of Feldman's paper, but publication of a comment pointing out the statistical flaw in the argument.

  11. Philip Taylor said,

    December 11, 2018 @ 7:43 am

    "The correct response should not be to request withdrawal of Feldman's paper, but publication of a comment pointing out the statistical flaw in the argument". I agree.

  12. J.W. Brewer said,

    December 11, 2018 @ 12:36 pm

    Presumably the journal in which the piece was published and/or the general run of journals in the relevant subdiscipline has/have their own standards for how bad a methodological law in an already-published article needs to be before the article gets withdrawn versus supplemented with the publication of a comment focused on the flaw versus left alone (with the notion being that critics are free to get another journal to publish their critique but the original journal has no affirmative obligation to do so). I don't know the terrain well enough to know which of those is the right outcome here, and I imagine there's enough unpredictable variation between different academic areas when it comes to the standards for this sort of thing that this set of critics may not really know either.

    It sounds like the journal in question may be unusually interested in the sort of interdisciplinary scholarship that frequently means scholars from economics backgrounds will be drawing on prior scholarship or datasets from other disciplines and thus probably be frequently at risk of misunderstanding or misapplying what they are borrowing. Whether that means the journal takes a fairly relaxed attitude toward that sort of misunderstanding because it is inherent in the sort of work they're trying to encourage more of or is to the contrary unusually vigilant about it because it is a characteristic risk of the sort of work they're trying to encourage be done well is unknown to me.

  13. J.W. Brewer said,

    December 11, 2018 @ 12:37 pm

    damnyouautocorrect – "methological law" in the prior comment should of course be "methodological flaw."

  14. Doug said,

    December 11, 2018 @ 1:58 pm

    Victor Mair said:
    'The term "pro-drop" stems from Noam Chomsky's "Lectures on Government and Binding" from 1981…'

    I was very surprised to learn that the term is so recent. Since the pro-drop vs. non-pro-drop distinction is an obvious one even if you look only at major European languages, people must have discussed it earlier.

    Does anyone know if there was an earlier term before "pro-drop"?

  15. Jerry Packard said,

    December 11, 2018 @ 2:29 pm

    My memory is the same as Philip Spaelti's, and also Victor's 1981 reference, where the term pro-drop was coined in the Chomskian technical sense and then it was popularly adopted in a more general sense to describe languages that more commonly drop pronouns than others.

  16. Philip Taylor said,

    December 11, 2018 @ 2:35 pm

    The Google n-grams viewer appears to suggest 1974–1976 as the earliest attested usage.

  17. Cervantes said,

    December 11, 2018 @ 2:38 pm

    maidhc — The subject pronoun is not fully implied by the form of Spanish verbs, because the gender is not specified. Usually a speaker will use a subject pronoun either to resolve ambiguity, or to emphasize the subject for some rhetorical reason. But regardless of whether the pronoun is used, the person and number are for the most part specified, yes; although third person can actually function as second person and the second person plural can be singular in some dialects, but these are matters of social formality and recognize the relationship between the speaker and the interlocutor. So indeed, functionally the presence or absence of a pronoun doesn't signal some sort of collectivism.

    On the contrary, a feature of the Spanish language that might do so is the presence of a first-person indirect object in reflexive constructions. E.g., a mother may say of her child "Se me enfermó," which means that the child got sick but implies that this event befell the mother as well. This does not translate into English because you need a preposition, and "He got sick on me" does not mean the same thing.

  18. Jerry Packard said,

    December 11, 2018 @ 2:58 pm

    @Philip Taylor

    It's hard to tell, then, whether 1974-1976 would be the technical or non-technical usage.

  19. J.W. Brewer said,

    December 11, 2018 @ 4:00 pm

    The '81 Chomsky book was based on a set of talks given in '79, and it wouldn't surprise me if a lot of the jargon and defined technical terms unveiled to the world at large in that book had been floating around more informally in the Chomskyan inner circle for several years prior to '79, with the possibility that some hanger-on might have actually used it in print in that technical sense significantly before the Master Himself did.

    It's not clear to me whether the "technical" sense is really describing a different phenomenon from the non-technical sense, or simply giving a specific account (probably one that is only interesting and/or intelligible within a particular Chomskyan framework) of how/why the same phenomenon works.

  20. Michael Watts said,

    December 11, 2018 @ 4:18 pm

    "Are there any languages that regularly allow a zero subject *and* a zero copula" — to the best of my belief, Chinese/Mandarin/MSC/Putonghua, where in response to a question such as 你是英國人嗎
    ("Nǐ shì yīngguórén ma ?", = "Are you English ?", a valid response would be 是 ("Shì", = "[I] am"), although more idiomatic would be 是的 ("Shì de", = "[I] am"). VHM will, of course, be in a far better position than I to judge whether this is correct.

    But 是 is the copula, so that's not an example of a zero subject and a zero copula.

    In general, you can answer a yes-or-no question like that by repeating the verb:

    问:你喜欢英国菜吗? [Do you like (喜欢) English food?]
    答:喜欢 [Yes.]

    Or you can answer any yes-or-no question with 是的 "[that] is [so]" regardless of whether the question used the verb 是. The question 你是英国人吗 blends those two cases.

    a feature of the Spanish language that might do so is the presence of a first-person indirect object in reflexive constructions. E.g., a mother may say of her child "Se me enfermó," which means that the child got sick but implies that this event befell the mother as well. This does not translate into English because you need a preposition, and "He got sick on me" does not mean the same thing.

    If "he got sick on me" doesn't mean the same thing, how did you pick "on" as the preposition in your example?

    Of course it does mean the same thing; English uses "on" to mark the "beneficiaries" of negative events. "My car died on me" would not normally mean that I'm trapped under the car.

  21. Michael Watts said,

    December 11, 2018 @ 4:22 pm

    Expanding on the "are there any languages that regularly allow a zero subject and a zero copula", I think the question needs to be more fully specified. As stated, English is such a language — an exchange like this is fully idiomatic and routine:

    Q: How are you feeling?
    A: Hungry!

    But rather than saying that the answer is missing its subject and copula, I would say that they are present in the question and the answer is parasitically relying on the explicit context.

  22. James Parkin said,

    December 11, 2018 @ 7:03 pm

    The Interior Salish languages of the PNW are regularly pro-drop in the 3rd-person, with no copular verb. "ʔilxʷt" (he/she is-was hungry) is a valid expression in Colville-Okanagan Salish.

  23. cameron said,

    December 11, 2018 @ 9:49 pm

    I agree with Michael Watts above, I'd interpret "he got sick on me" only in the "my car died on me" sense. There is another sense of "get sick on" but I don't think it's likely any native speaker would relate an unfortunate regurgitative event with that exact wording – there are too many better alternatives.

  24. Daniel N. said,

    December 12, 2018 @ 9:44 am

    So it happens that Germanic peoples are generally richer than Slavic peoples. But Japanese are rich too, and Japanese is "pro drop". Chinese are slowly getting rich too.

  25. Coby Lubliner said,

    December 12, 2018 @ 9:58 am

    "French also arguably has this trait because of its areal proximity to these other northern European languages."

    No, it's because of the general fusion of vocalic word endings into a schwa (and eventually to no vowel), so that, where Italian has parlo, parli, parla, French has the indistinguishable parle, parles, parle, hence requiring pronouns.

    With the few verbs in which the first-person-singular form differs from the others, it's possible to drop the pronoun, as in Verlaine's famous rien ne suis, rien ne puis.

  26. J.W. Brewer said,

    December 12, 2018 @ 12:12 pm

    Japanese seems at first glance an excellent example of a counterexample, but to be fair to Prof. Feldmann, his argument seems to have two steps: (i) explicit pronoun marking is correlated with (and maybe causes?) a more individualistic and less collectivistic cultural outlook; and (ii) a more indiv. and less collectiv. cult. outl. in turn is correlated with (and maybe causes?) economic prosperity. The prosperity of Japan is a pretty good demonstration that step (ii) is definitely not an exceptionless rule but at most a general statistical tendency. (Maybe it is; maybe it isn't — I expect there's a scholarly literature out there on the topic.) But once you accept that at least *some* prosperous countries do not have that individualistic cultural thing, I'm not sure that the fact those same countries speak a Pro Drop language really independently undermines step (i) of Feldmann's argument — although step (i) seems likely to have lots of other problems.

    To take another recurrent pop-Whorfian claim that comes up on LL from time to time, if it were true (which it's probably not!) that higher-than-average rates of usage of first-person pronouns correlated with narcissism or some such other DSM diagnosis, one would then expect that linguistic phenomenon also to correlate with whatever life outcomes narcissism in turn correlates with, and maybe to correlate with life outcomes where a contested claim of correlation with narcissism is the subject of a separate and unresolved ongoing scholarly debate.

  27. ktschwarz said,

    December 12, 2018 @ 1:47 pm

    @Cervantes: The subject pronoun is not fully implied by the form of Spanish verbs, because the gender is not specified. Yes, and that's a huge obstacle to machine translation, as discussed here from time to time.

    Everybody discussing the "pro-drop" terminology: Go read Martin Haspelmath's attack on this. The most common system worldwide is the Spanish one, where first/second/third person is expressed on the verb and a separate subject pronoun is not required. The German system, where the verb is inflected and a subject pronoun is required, is the least common.

  28. J.W. Brewer said,

    December 12, 2018 @ 2:01 pm

    On a PPP/GDP per capita basis, Spanish-speaking countries are largely absent from the richest end of the worldwide standings and likewise absent from the poorest end of the list, placing them mostly somewhere in the typical middle. Coincidence? (Someone else is going to need to code all the countries into Haspelmath's four categories per their dominant language(s) and see what if any pattern emerges — obviously lots of languages other than Spanish have the "Spanish" system.)

  29. Chandra said,

    December 12, 2018 @ 3:22 pm

    @Michael Watts –

    I had the same thought about zero subject and zero copula in English, but I assume the distinction with Japanese (and other similar languages) has to do with register? You can get away with a single adjective response like "Hungry!" in conversational English, but such a construction would be considered a sentence fragment error in formal writing, for example.

  30. RP said,

    December 12, 2018 @ 3:24 pm

    Thanks @ktschwarz.
    Haspelmath's piece is very interesting.

    I wonder, though, if we were to adopt his four types…
    "Spanish-type languages with cross-indexes that can be conominated (3a), Swedish-type languages with no bound person forms (3b), German-type languages (3c), where an index on the verb has to be combined with an independent pronoun, and Japanese-type languages, where the subject can indeed be null when it can be inferred from the context (3d)."

    How do we count English, which bound person forms but only in the 3rd person singular present tense? Is that a variant of 3c or is it a variant of 3b? If it is a variant of 3b, at what point along the spectrum does a language become a variant of 3c? Also, if we look at French, where as has been noted above the singular forms of the verb are very often all pronounced the same, what if you had a language where the indexing in both singular and plural was only written down and not spoken – would that be 3c or 3b?

    Well, further on, he implies English is 3c, German-like: "Siewierska (1999: 239) mentioned a few other languages of the German type: Standard German, Swiss German, English, Icelandic, Faroese, Romansh, French, plus Anejom". Yet this seems odd because these are meant to be examples of languages where the pronoun is required even though the verb form expresses the differentiation. But it would be odd to think that English verb forms express sufficient difference to allow the verb forms to be routinely unexpressed (although contextually it can happen in certain cases, e.g. "am feeling tired", "tried calling you earlier", "went shopping" – but it would be neither 3b nor 3c if we were counting these). So English actually shouldn't be lumped together with German here.

  31. Christian Weisgerber said,

    December 12, 2018 @ 5:33 pm

    I have a hard time understanding the supposed psychological significance of "pro-drop" or "zero subject" when the first example that is bandied about is Spanish where the subject is largely encoded in the verb ending. The subject doesn't seem to be so much missing there as it is simply encoded in a different manner.

    The modern European non-pro-drop languages have suffered erosion of the verb endings due to sound shifts. Without distinctive verb endings, an overt subject became obligatory. The most obvious example is French, but I this applies to the Germanic languages as well. (And you could argue that the French clitic pronouns are now well on the way to turning into personal prefixes.) Presumably PIE was pro-drop… but then its personal endings suspiciously look like they have arisen from the fusion of pronouns to the verb. So, uhm, what is the fundamental difference?

  32. 번하드 said,

    December 12, 2018 @ 7:45 pm

    Has somebody thrown North Korean and South Korean at Prof. Feldmann already?
    Or if, whyever, he really wants to stay in Europe "East German" and "West German"?

  33. Mick O said,

    December 13, 2018 @ 2:56 pm

    https://www.google.com/search?q="saw+this+guy"
    Is this an example of pro-drop English?

    Addendum: "Love this!" is not an uncommon comment on Instagram.

  34. Alec said,

    December 13, 2018 @ 8:52 pm

    On the statistical question about non-independence of related languages, biologists have been dealing with this issue for a very long time. Check out Felsenstein (1985) Phylogenies and the Comparative Method, cited almost 8,000 times since then. Not saying that the biological methods can just be taken straight over into linguistics, but there must be something to be learned. I didn't see any sign of recognition in the Feldmann paper that this is even an issue, never mind that there are actually methods to deal with it.

  35. AntC said,

    December 14, 2018 @ 4:16 am

    @VHM thank you for the reference, but I had little doubt the "idiot" I referred to would turn out to be ANC.

    @ktschwartz Everybody discussing the "pro-drop" terminology: Go read Martin Haspelmath's attack on this.

    Thank you! 'Null subject'. Exactly.

    If 'Pro-drop' refers specifically to marking of the subject (or agent?) specifically, then it should include 'subject' in the term. Otherwise it might be taken to be talking about pronouns (or marking thereof) in other positions. And what about ergative-absolutive languages? Did I say English/IE biased?

  36. Christopher E. Soli said,

    December 16, 2018 @ 7:31 am

    > Prof Feldmann's paper is laden with empirical issues

    With you so far…

    > with drastically problematic implications

    How you feel about its implications has no causal bearing whatsoever on the correctness of the argument. This word "problematic" comes up a lot in arguments of this style – it's like a sleight-of-hand where the illusionist tries to trick the audience into believing a sort of emotional *modus tollens*. "If-p-then-q, and q makes me sad, so not p."

    > Such a claim is unfortunately ethnocentric.

    Again, negative emotional connotations don't count as evidence against an argument.

    > There are numerous empirical issues to be pointed out

    I wish you would do more of this and less hand-wringing.

    > [WALS] data are not appropriate for the types of statistical analysis that Prof Feldmann uses

    Could you please expand on this?

    > I consider this stance ethnocentric, and it risks being interpreted as implicitly racist

    And again, your "considerations" have *literally* zero bearing on whether the original argument is valid.

    > it fails to account for the fact that languages share grammatical features because of shared history

    I think this is your most salient point, but I don't think it's especially damning. Sure, there are shared ancestral nodes in the causal network – all this means is that we should a priori discount this *particular* aspect of shared history and spread the total causal weight over all aspects of shared history, weighted by some plausible metric of economic importance. However, among the set of "important" shared causal factors, language seems fairly significant. Perhaps the only aspects of shared history that I would weight as comparable to language are genetics, religion, and philosophy. I agree, of course, it would be instructive to repeat the analysis with these confounders in mind.

    > I would assume this is a result of the Prof Feldmann's lack of awareness of linguistics as a field, rather than a deliberate manipulation of data,

    The fact that Feldmann isn't a member of (and, therefore, is not subject to the taboos of) an academic high school clique is evidence in *favor* of his research, not against.

    > his paper's misrepresentation of linguistic facts is irresponsible.

    If you could focus more on these purported facts and less on moralizing about what's problematic, or ethnocentric, or responsible, or whatever, that would be great.

    > When the analysis is properly controlled for the inter-relatedness of languages, the effects disappear.

    Could you link to this revised analysis? It would be nice to know *how* they controlled for "inter-relatedness"; depending on what you mean by this, it might be a foregone conclusion that the effects disappear.

    > assuming it to be a binary dimension is itself inaccurate. Many so-called non-Pro-Drop languages (notably among them, English) tolerate a certain degree of subject pronoun deletion

    Factoring that "certain degree" into the analysis would be instructive, but would probably have a small effect given any reasonable quantification of how strong the pro-drop effect is.

    > It is frustrating for linguists to see such papers published in the journals of other fields

    Because you can't impose your field's social proscriptions on them?

    > Linguistics as a field has been working for more than a century to refute the common misconception that some languages are more suited than others at expressing complex ideas

    And, despite working backwards from this goal (destroying any causal relationship between the claim and reality), it has not succeeded…

    > a language can always meet the needs of its speakers

    This statement is almost a tautology. This whole paragraph really seems like it's more intended as a social signal than to actually convey any information.

    > some languages are more conducive to technological and economic advancement.

    This is an entirely unsurprising claim to anyone who isn't blinded by dogma. Obviously a descriptively rich and quantitatively capable language is more suitable to describing and reasoning about mathematics than e.g. a trade pidgin or Gumulgal.

    > This in turn risks coming across as implicitly racist

    And, again: How something "comes across" has no bearing on its truthhood.

    > That this narrative is presented in your journal under the guise of empirical science

    An impressive display of hypocrisy!

    > I implore you to choose reviewers who read such submissions with more attention to their claims about language

    I.e. you want other fields to subject themselves to whatever politically-motivated censorship plagues your own.

    > censhorship is an act of an authority silencing the voices of artists, dissenters, journalists, or academics. We as a community have no sway or recourse over this or any other journal to silence its choice of papers for publication

    Authority can be collective, and right now whatever strain of anti-intellectualism spawned this post has collective authority in academia.

  37. ktschwarz said,

    December 17, 2018 @ 11:09 am

    Elsewhere, marie-lucie has commented on French pronoun dropping:

    This omission of the pronoun il is quite frequent in colloquial speech, with phrases where il does not have a referent, as in (il) faut 'must', (il) vaut mieux '(better, rather (to do …)', and some others including (il) y a 'there is'. For example, in the folk song (or rather rhyme) Y a un' pie dans l' poirier … 'there is a magpie in the pear tree …'

    I'm getting the impression that any language will drop pronouns in some registers and contexts and add them in others. Colloquial and familiar registers are the last thing you can ignore if, like Feldmann, you want to make claims about cultural values that are transmitted through the family.

RSS feed for comments on this post