Do field linguists really ask the “wrong questions”? A reply to Haspelmath (2012a)

In a recent paper on word classes, Haspelmath (2012a) takes field linguists to task for asking the “wrong questions,” like whether the major word classes are universal or present in a particular language. The problem with this claim is that the field linguists he cites are not asking these questions at all. While they may advocate a particular best analysis, they do not usually make either explicit or implicit universalists claims (the main exception being Chung 2012 who says that her analysis of Chamorro “supports the claim that lexical categories are universal” (p1)).

My Linguistic Typology paper on adjectives in Ecuadorian Quechua (Floyd 2011) is one of sources portrayed as universalist, despite clearly stating that it does not “not take a strong position here regarding the adjective’s universality” (p27) and that the term “adjective” in the article refers only to a “language-internal word class” (p39). Since most of the sources he cites take similar positions, Haspelmath’s critique seems unnecessary. Here I’d like to take a moment to clarify what field linguists actually are concerned with, something that is unfortunately often neglected in large-scale linguistic comparison: accuracy to the empirical descriptive facts.

Field linguists care about accuracy, not universality

Haspelmath casts the descriptivist enterprise as an endless cycle of disputes in which linguists successively counter each other’s claims about a language’s word classes; he  frames the discussion of Quechua word classes as “Weber vs. Floyd,” as if my article was in dispute with Weber’s (1989) account. Readers might be surprised to know that Weber only briefly discusses the topic over less than two pages, and with reference to a variety of Quechua that is mutually unintelligible with the variety I study (I warned about conflating distinct Quechua languages in Floyd 2011, p29). The motivation of the more detailed account that I provided was not to counter Weber, but rather – as is laid out explicitly in the article – to correct the longstanding but misleading account in Schacter (1985, p17) and other sources that misinterpreted data from different Quechua languages in the service of cross-linguistic generalizations.

The stances that field linguists take can often be better understood not as concerning disputes among descriptivists but rather to set the record straight when they see that typologists have made unfounded assertions about their languages of expertise in order to pursue some agenda. For instance, Haspelmath frames Evans and Osada’s (2005) account of Mundari word classes as countering previous descriptive accounts in Hoffman (1903), but reading their article actually reveals that their main concern is less with Hoffman’s original analysis and more with the way Mundari data has been used for typological generalizations (in Hengeveld 1992a and 1992b as well as elsewhere).

Are field linguists really “ethnocentric”?

Field linguists treat cross-linguistic terms as a useful toolkit to be pragmatically applied to language description, drawing on our accumulated knowledge of the of the world’s languages, a rich first-hand experience with diversity that hardly limits them to familiar European categories. It does not seem appropriate to characterize field linguists of the caliber of Nick Evans, author of descriptive works on a whole range of Australian and Papua New Guinea languages, including two grammars (1995; 2003), or Marianne Mithun, author of dozens of descriptive works including a comprehensive survey of North American languages (2001), as “ethnocentric” or Euro-centric, as Haspelmath implies they and others may be (2012a, 2012b). These researchers clearly have stellar records of documenting linguistic diversity, and the suggestion that they cannot think beyond Euro-centric categories is laughable. If Haspelmath is worried about universalism, field linguists seem like an odd target, especially when whole schools of linguistics remain explicitly universalist, so he might focus on critiquing those who actually make universalist claims rather than attributing the “wrong questions” to those who don’t ask them (and who actually produce the materials that make typology possible in the first place).

What good are the “right questions” with wrong data?

Typologists have been struggling for years to justify their comparative enterprise in the face of the realization that distinct linguistic systems are, in an absolute sense, incommensurable. There is no real answer to this problem, although there is there is no shortage of proposed miraculous solutions (which generally end up replicating the same kinds of categories and their associated problems despite insisting to the contrary). Even so, typology still seems to proceed as if this was not a major issue, and a look at any large-scale typological work (like the World Atlas of Language Structures; Haspelmath et al 2005) reveals how almost every point of comparison conflates cross-linguistic and language-specific morpho-syntactic criteria to some degree.

Even though this is the case, it does not seem correct to dismiss such large-scale typological efforts for asking the “wrong questions”. At their best, such comparisons actually move back and forth between more structural and more semantic or functional criteria to find similar patterns of form-meaning mappings across languages, and they are able to do this precisely due to a pragmatic and informed application of common descriptive terms. As Haspelmath himself points out in defense of typology in the face of linguistic particularism (2010), if such terms are used transparently it is possible to mediate between language-specificity and comparability. A more basic problem, however, is that such comparison is only possible to the extent that the data are accurate, something that every field linguist who has ever inspected a familiar language’s coding in a typological database knows is frequently not the case. Language specialists may attempt to do comparative linguists a favor by setting the record straight before they can make any further claims on the basis of incorrect data, which was the true project of many of the sources Haspelmath cites.

Criteria must be applied consistently, not for convenience

Good field linguists argue for what they feel is the most fitting account of a language, transparently explaining which terms and criteria from the descriptive linguistic toolkit have been applied (but giving enough detail to allow for alternative analyses), and holding their arguments fully accountable to natural speech data (not just elicited sentences and a few narratives – this is the 21st Century!). The key to this approach is being explicit about one’s criteria and then applying those criteria consistently both language-internally and cross-linguistically, regardless of the outcome. The complaint that field linguists often have is that others take the data and analyze it inconsistently and selectively in an attempt to squeeze a language into a pre-determined ‘type’ in support of some argument.

This was my point about the treatment of Quechua languages: while some authors had pointed out that in several Quechua varieties both nouns and adjectives can modify nouns, nobody had bothered to point out that adjectives cannot be modified by nouns. Consistency is not about what criteria to choose, but about faithfully applying the criteria one has chosen, even when it complicates your argument. The problem with previous comparisons of Quechua is that they had not even been able to stick to their own chosen criteria, quite apart from the issue of whether the criteria were appropriate or not. It simply does not matter whether you ask the “right” or “wrong” questions if the primary data is inaccurately represented.

Assuming lesser-known languages must be exotic is also ethnocentric

Haspelmath’s article (2012a, pp116-118) is a good illustration of the pitfalls of such inconsistency, as he unfortunately reiterates the problematic comparisons of Quechua and English that I was hoping to put to rest with my paper. He asserts that describing the Quechua property word class as ‘adjectives’ is misguided because it loses the fact that Quechua’s word classes are “interestingly different” from English, based on three explicit criteria: Quechua adjectives and nouns both occur in (1) copular predication, (2) noun modification, and (3) referentially (as independent noun phrase heads). Haspelmath cites these examples from Weber (1989 pp35-36; slightly modified here):

 (1)         a. Taqay rumi     ka-yka-n
    that      stone    be-impv-3
    ‘That is stone.’
b. Taqay hatun ka-yka-n
    that      big     be-impv-3
    ‘That is big.’
(2)        a. rumi wasi
          stone house
  ‘stone house’
b. hatun wasi
    big house
   ‘big house’
(3)        a. rumi-ta     rikaa
  stone-acc see.1sg
 ‘I see (a) stone.’
b. hatun-ta rikaa
   big-acc     see.1sg
   ‘I see (a) big.’

It is worth revisiting these examples, not to just look not at the Quechua, but also at the English translations, because these actually mirror the supposedly “interestingly different” Quechua constructions, and yet at the same time are all perfectly normal in English. Like Quechua adjectives and nouns, English adjectives and nouns also both occur with copular verbs or as noun modifiers (the latter usually analyzed as nominal compounds; see Floyd 2011, pp35-41). The only one of Haspelmath’s proposed properties of Quechua adjectives that might seem to set them apart from English adjectives is shown in (3b), where, according to some analyses, English differs from Quechua because it calls for the ‘dummy’ noun ‘one.’ Yet there has been quite a bit of research addressing adjective-headed noun phrases in English that shows that this is a simplistic account (i.e. Payne and Huddleston 2002). Recent work by Günther (2013) points out how only some specific constructions and contexts call for ‘one’ in English, while others do not permit using ‘one’ at all (and still others can vary). In fact, it only takes a moment on to find the adjective ‘big’ used as a noun phrase head with nominal morphology, just as it can be in Quechua (and to dissuade arguments that these could be non-native posters, the examples are about the very American subjects of basketball (4) and college sororities (5)).

(4)        I’d love to see two bigs on the floor for defensive and rebounding purposes.  []

(5)        Q: Sorority question…can i have two bigs? if I have a big and another girl wishes to adopt me.. how does that work?

A: Not really unless your big left the chapter for some reason. If she just graduated then ask your big if you can be adopted as a courtesy. If your big just sucks then you have to deal with it and try to be a better big yourself (that is what I did).  []

It might be pointed out that these examples are context-dependent and partially elliptical, but this was precisely the argument I made about adjective-headed noun phrases in Quechua, where I showed how particular conventions (comparable to the English convention for drink sizes: “one small, two mediums, and a large”) and contexts (like choosing different-colored candies and saying “I want a green”) can account for such occurrences (Floyd 2011, pp50-58). This is not to say that Quechua adjectives and English adjectives are exactly the same, and in fact there are many differences that could be pointed out, but none of them emerge from Haspelmath’s own specified criteria, which, if applied consistently, do not paint Quechua as “interestingly different” from English, at least not with respect to these features (as I have already made clear in Floyd 2011, p60). The flipside of ethnocentrism is exoticism, assuming that South American languages must always be radically different from European languages, when this is not necessarily the case (and English sometimes turns out to be more exotic than expected).

There is nothing ‘universalist’ or ‘ethnocentric’ about respect for accuracy

The point is this: you can call Quechua adjectives ‘a subclass of nouns’ if you must, or you can call them ‘class Z’ if you like, or you can call them ‘wugs’ for all I care, if it fits your particular project. But please stop repeating the same misleading statements about Quechua languages (e.g. ‘Quechua is different from English because English requires ‘dummy’ nouns’) if you have not consistently checked whatever criteria you are basing your statement on. Field linguists are not asking if language X has adjectives, they are asking you not to ignore the empirical facts about language X just because it is convenient for your argument. Due to the overwhelming evidence that the varieties of Quechua I study distinguish a property word class, I propose that the most straightforward way to account for the language’s word classes is to use the term ‘adjective,’ but others might choose other terms or characterizations as long as they also provide an accurate account of the data. Those interested in typological comparison obviously are not expected to simply take what one descriptivist calls “adjective” in one language and compare it to what another calls “adjective” in another language – they need to take the time to look at the data and analysis which, if the descriptivist was doing their job right, should be transparent enough to allow for different criteria to be applied depending on the questions being asked. This is was the real purpose of my article on Quechua word classes, to facilitate comparison through more accurate and complete data, and I invite readers to take a look at it and form their own opinions.


Chung, Sandra. 2012. ‘Are Lexical Categories Universal? The View from Chamorro’. Theoretical Linguistics 38: 1-56.

Evans, Nicholas. 1995. A Grammar of Kayardild. Berlin: Mouton de Gruyter.

Evans, Nicholas. 2003. Bininj Gun-wok: a pan-dialectal grammar of Mayali, Kunwinjku and Kune. (2 volumes). Canberra: Pacific Linguistics.

Evans, Nicholas & Toshiki Osada. 2005. ‘Mundari: The myth of a language without word classes’. Linguistic Typology 9(3):351-390.

Floyd, Simeon. 2011. Re-discovering the Quechua adjective. Linguistic Typology 15:25–63

Günther, Christine. 2013. The Elliptical Noun Phrase in English: Structure and Use. Routledge Studies in Germanic Linguistics. London: Routledge.

Hoffmann, John. 1903. Mundari Grammar. Calcutta: The Secretariat Press.

Haspelmath, Martin, Matthew Dryer, David Gil & Bernard Comrie. (eds.). 2005. The World Atlas of Language Structures. Oxford: Oxford University Press.

Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in cross-linguistic studies. Language 86(3): 663-687.

Haspelmath, Martin. 2012a. How to compare major word-classes across the world’s languages. In: Graf, Thomas & Paperno, Denis & Szabolcsi, Anna & Tellings, Jos (eds.) Theories of everything: in honor of Edward Keenan, 109-130 (UCLA Working Papers in Linguistics, 17.) Los Angeles: UCLA.

Haspelmath, Martin. 2012b. Escaping ethnocentrism in the study of word-class universals. Theoretical Linguistics 38(1-2):91-102.

Hengeveld, Kees. 1992a. Parts of speech. In Michael Fortescue, P. Harder, & L. Kristofferson (eds.), Layered Structure and Reference in a Functional Perspective, 29–53. Amsterdam: Benjamins.

Hengeveld, Kees. 1992b. Non-verbal predication: theory, typology, diachrony. Berlin: Mouton de Gruyter.

Mithun, Marainne. 2001. The Languages of Native North America. Cambridge: Cambridge University Press.

Payne, John & Rodney Huddleston. 2002. Nouns and noun phrases. In Huddleston, R. & G. K. Pullum, The Cambridge grammar of the English language. Cambridge: Cambridge University Press. 323-524.

Schachter, Paul. 1985. Parts-of-speech systems. In Timothy Shopen (ed.), Language typology and syntactic description, vol. I, 3–61. Cambridge: Cambridge University Press.

Weber, David. 1989. A Grammar of Huallaga (Huánaco) Quechua. Berkeley: University of California Press.

7 thoughts on “Do field linguists really ask the “wrong questions”? A reply to Haspelmath (2012a)

  1. Thanks for helping me to acquire new concepts about pcs. I also have belief that one of the best ways to maintain your notebook computer in primary condition is with a hard plastic-type case, or shell, that suits over the top of your computer. A lot of these protective gear usually are model targeted since they are made to fit perfectly within the natural casing. You can buy these directly from the owner, or from third party places if they are designed for your notebook computer, however don’t assume all laptop can have a covering on the market. Just as before, thanks for your suggestions.

  2. It is the best ttime to make some plans for the longer ter and
    it’s time to be happy. I’ve read this publish and if I could I desire to counsel you few fascinating things or tips.
    Maybe you can write next articles referring to this article.
    I wish to read more things approximately it!

  3. Hi Simeon,

    Stimulating! I just had a look at the Haspelmath piece you’re talking issue with, and I was struck by the fact that I couldn’t locate any criticism of field linguists as such, but rather of linguists more generally who presuppose the existence of cross-linguistically valid word classes. (Maybe it’s there — I skimmed.)

    I think there might be an important point here, in that I think that ‘field linguists’ — which I take to be linguists dedicated to the description and analysis of particular languages through interactions with speakers — generally work bottom up (one can think of exceptions of course — Sandy Chung may be an example), and use terms like ‘adjective’ as language-specific labels of convenience. You make the point about the bottom-up nature of this work nicely (‘field linguists care about accuracy’), as you do the point about the relatively small stakes of terminology like ‘adjective’ for labeling language-specific word classes. And I think you and Haspelmath are basically on the same page with respect to the latter issue: he says that he has no problem with this practice (e.g. on p. 115) — it’s the use of these terms in cross-linguistic comparison that he finds problematic, for reasons that we need not rehearse here.

    That said, I’m somewhat surprised that Haspelmath picked a bone with your discussion of Quechua adjectives, since I find it very hard to imagine that you asked the “wrong question”, in the sense of your approaching your data wanting to determine if Quechua (which variety, right?) has or does not have ‘adjectives’, as a universally-defined word class. What you observed was that there are two classes of words, which, on purely language-specific grounds, you decided call ‘nouns’ and ‘adjectives’, and then analyzed their characteristics. Haspelmath even observes at the end of the discussion that whether one considers Quechua to have a larger ‘nomina’ class, of which your ‘adjectives’ are a subclass, or two classes (‘nouns’ and ‘adjectives’), is not an question that admits of empirical evaluation. It’s not clear to me how this case study really illustrates the “sorts of problem [that] arise if one presupposes the existence of verbs, nouns and adjectives as cross-linguistic categories.” (p. 116)

    This is not to say that one might disagree with your criteria for distinguishing word classes, and for distinguishing major word classes from minor ones, but it seems like a bit of a stretch to lump your language-specific understanding of ‘adjective’ with Sandy Chung’s efforts to map the property of Chamorro word classes onto Baker’s universal word categories.

    • Thanks for you comments Lev. To clarify, the original article does not single out fieldworkers by name, but many of the researchers critiqued there are engaged in descriptive fieldwork, and I use the term ‘field linguists’ to refer them as a group, and to ask why they are targeted by this critique of universalism if they do not tend to take universalist positions. The term can serve to roughly delineate field-based descriptive work based on natural speech data from the different strands of linguistics that rely mainly on data from secondary sources or intuition – a huge literature with no lack of universalist claims that would be great candidates for being subjected to a critique like Haspelmath’s. This is why his focus on descriptivists is so odd, since there is so much other more appropriate work that he could have mentioned. That being said, I really do not want to dichotomize ‘field linguists’ and ‘typologists’ or other ‘theoretical’ linguists, because ultimately it is best if everyone does a bit of both.

  4. Pingback: Diversity Linguistics: Quechua adjectives and the universality of word classes | Greater Blogazonia

  5. These questions of conventionalized meanings accounting for some the examples of adjective-headed noun phrases that David Gil raises are exactly the kinds of issues that I discussed in my paper with respect to Quechua languages. Here I looked specifically for an example with ‘big’ in response to examples using the equivalent Quechua size adjective hatun, and these examples are limited to more restrictive contexts in English, but adjective-headed noun phrases are so prevalent in English that I could have easily found a more common example.

    For instance, I doubt you would have any trouble understanding examples with ‘large’, which is another possible translation for hatun: “two larges with extra cheese”. In order to understand this, you probably had to think about pizza, but this is precisely my point about Quechua: I found that Quechua speakers, when presented with adjective-headed noun phrases, typically found such examples to be so odd that they sometimes considered them ungrammatical, or at least un-interperable, somewhat like your reaction to ‘two bigs’. If they could draw on specific conventions (while ‘big-PL’ is confusing, ‘little-PL’ is not because it conventionally means ‘children’) or nominal antecedents or otherwise salient referents (choosing between shirts of two colors and saying “I like the red”) then the examples suddenly become understandable.

    This is a key difference between Quechua adjectives and Quechua nouns, because if you construct similar examples using nouns speakers will accept them without any hesitation or need for further contextualization. The same is basically true about English (both of ‘one’ constructions and of fully elliptical nouns), that adjective-headed noun phrases require some extra contextual information while noun-headed noun phrases do not; these examples with ‘big’ are only an extreme case where very specific extra information is needed. Günther (2013) shows that noun-eliding constructions in English are quite diverse, and that ‘one’ is only obligatory for some – perhaps not even most – of them. In order to compare English with other languages with respect to these questions, it would seem important to register the fact that much of the time ‘one’ is not actually obligatory.

  6. Just a data footnote on Simeon Floyd’s recent blog, and his sentences

    (4) I’d love to see two bigs on the floor for defensive and rebounding purposes.

    (5) Q: Sorority question…can i have two bigs? if I have a big and another girl wishes to adopt me.. how does that work?

    As a native speaker of English, I find both of these to be ungrammatical; yet at the same time I have no reason to doubt the accuracy of Simeon’s data. So what’s gong on here? Obviously we’re dealing with dialectal variation. However, I would like to suggest that the variation is not morphosyntactic, that is to say, in the extent to which different dialects allow for an adjective to function as a noun phrase head unsupported by “one”, but rather lexical. Specifically, these two examples strike me as representing adjective-to-noun zero-conversion of the kind that is widespread in English (e.g. when “blues” can refer to a group of Chelsea FC supporters, or to a melancholy state of mind, or to a musical genre).

    Support for the lexical-conversion analysis derives from the fact (okay, since I’m not an expert in basketball and sorority subcultures, it’s more of an educated guess) that “bigs” doesn’t just mean “big ones”; rather, it has additional idiosyncratic meaning components specific to the basketball and sorority contexts respectively. In other words, “big” [N] is more like some kind of “technical term” within these two subcultures; it’s part of their respective lexicons. Accordingly, I would predict that, in general, an arbitrarily-chosen adjective would not work the same way. For example, in the basketballing context, I would guess that many or all of the following would not be possible: “smalls”, “talls”, “shorts”, “thins”, etc.

    The English facts contrast with those in, say, Hebrew, where (the equivalents of) “small”, “tall”, “short”, “thin”, etc. are all completely fine as nominal heads, don’t require any specific licensing contexts, and have completely compositional meanings: “small one” etc. This is why, in my 2005 WALS chapter “Adjectives without Nouns”, I assigned English the feature value “marked by following word”, in recognition of the obligatory presence of the form “one”. Examples (4) and (5) do not provide reason for me to change this.

Comments are closed.