Cross-linguistic research is not easy because one needs to examine languages of diverse structures, but here I argue that typological classification is never “difficult”. The literature contains many references to such difficulties, but in fact, the main practical difficulty is obtaining complete data for a wide range of languages. And of course, one needs to find the right questions to ask if one wants to get interesting answers; but this is not a “difficulty” – it is the primary challenge of any kind of creative research enterprise.
Here are a few examples of passages where authors say that assigning languages or constructions to types is “difficult”. This list could be easily expanded.
Curnow (2000: 4): “there are a wide variety of issues here, such as the difficulty of … distinguishing analytically between an “inflectional” copula and a “zero” copula in languages with no verbal morphology”
Bickel & Nichols (2005): “unless tense, aspect, and mood were clearly distributed in distinct positions, we counted these as one category (thereby avoiding difficult decisions as to whether something is aspect or tense or mood)”
Dryer (2005): “Because it is difficult to distinguish verbal suffixes from clitics that can only attach to the verb, these are not distinguished here and morphemes which attach to the verb are all treated as instances of interrogative verb morphology, even if by some criteria they might be analysed as clitics.”
Iggesen (2005): “delimiting case markers from other morphological-functional entities is often particularly difficult in such languages”
Rijkhoff (2007: 711): “words were already categorized into distinct groups by ancient Greek philosophers, but it has proven to be a rather difficult task to come up with definitions for the various lexical word classes that have cross-linguistic validity”
Velupillai (2012: 121): “Because of the difficulty of distinguishing it from cliticization, it is beyond the scope of this book to discuss pronoun incorporation.”
Degand et al. (2013: 2): “What makes it difficult to tell apart discourse markers from modal particles?”
Song (2018: 79): “the distinction between a language and a dialect is a notoriously difficult one to draw”
Norris (2019: 5): “There are examples of languages that fit very clearly into each of those categories [i.e. types of gender], but there are also borderline cases that are difficult to classify”
Kilian (2021: 30): It is “difficult to determine whether localizers are verbal predicators“.
It may be that some of these authors actually mean that the difficulty lies in obtaining complete information – but this is not what they say. They all talk about the difficulty of drawing distinctions, and the purpose of this blogpost is to point out that drawing distinctions is not difficult. Of course, one needs to know what the distinctions are – but if one talks about a conceptual contrast in a scientific context, one can be expected to know what the contrast is.
So what is the problem here? I suspect that it is that the authors are not sure that they know what the distinctions really are (as comparative concepts). Bickel & Nichols seem to be unsure what the distinction between tense and aspect is; Dryer seems to be unclear about the precise difference between affixes and clitics; and Degand et al. may not be certain about how to distinguish between discourse markers and modal particles.
We often take our terms from the earlier literature, which provides salient examples but no clear definitions of many technical terms (maybe because it assumes that the concepts are innate and thus their definition is a matter of research, not of convention). Thus, when we take our terms “off the shelf” (i.e. the earlier tradition of linguistics), we may not have good characterizations of what they mean. For example, there simply is no clear delimitation of “affix” (from other bound forms) in the earlier literature (though I made a concrete proposal quite recently in Haspelmath 2021).
Once this uncertainty in the legacy literature is recognized, there is an easy solution, in principle – comparative grammarians (“typologists”) can define the concepts they use themselves, without (necessary) reference to the earlier tradition. If they create their own distinctions, they no longer have any reason to be unsure about the nature of the distinctions. And if they know exactly what their comparative concepts mean, they can apply them without difficulty to the languages they compare (as long as there is no missing data problem).
Consider the example of polar question marking that is studied by Dryer (2005). He says that
“The distinction between separate question particles and interrogative affixes is not always clear. It is common in verb-final languages for question particles to immediately follow the verb, and they often loosely attach to the verb as clitics. Because it is difficult to distinguish verbal suffixes from clitics that can only attach to the verb, these are not distinguished here”
Dryer does not say what exactly he means by “affix”, “clitic” and “particle” here. But in his chapter on definite articles, he says that definite markers which can occur on nouns or adnominal modifiers are regarded as clitics. Thus, for example, the Basque definite article -a occurs after the last element of a nominal, which may be the noun (e.g. etxe-a ‘the house’) or an adjective (e.g. etxe gorri-a ‘the red house’). Thus, the decisive criterion is not the spelling or any kind of “phonological cohesion”, but the promiscuity: Definite markers can occur on roots of different root classes (I adopted this criterion in Haspelmath 2021). (There was some discussion of this back in 2008 on the WALS blog.)
Using this definition, it would not be “difficult to distinguish verbal suffixes from clitics” – one just needs to determine whether the marker in question can occur on verbs only or also on adjectives or nouns. In an analogous fashion, Bickel & Nichols could specify what they mean by “tense” and by “aspect”, and then apply their definition to their data; and likewise for “inflectional copula” (Curnow), “case” (Iggesen), “modal particles” (Degand et al.), and so on.
(Maybe Dryer actually means that it is difficult to know whether a marker can occur on other elements, because grammars often give just one or two examples with a question marker occurring on a verb and do not say whether the element also occurs elsewhere. Then this would be the data incompleteness problem, which is indeed a difficulty of typology, as I noted earlier.)
Apart from the data problem, there is a terminology difficulty: We generally expect our terms to be self-explanatory, and we expect our colleagues to use familiar terms. Reading long sections of definitions is tedious, and even data appendixes (which have no space limitations) rarely contain fully explicit definitions. However, this expectation is often too optimistic – our terms are not always self-explanatory. What is “noun incorporation”? What is an “impersonal construction”? What is a “converb”? What is a “serial verb construction”? We typically agree on core cases, but where are the boundaries? Is it better to propose new terms for one’s comparative concepts, or should one propose new meanings for old terms? Either way, our decisions may make our readers unhappy.
So the data difficulty and the terminology difficulty are very real, and not easy to overcome.
But there is no classification difficulty: Once a comparative concept is defined clearly in such a way that its defining criteria are applicable uniformly to all languages, the classification is straightforward if we know the facts of the language.
Bickel, Balthasar & Nichols, Johanna. 2005. Inflectional synthesis of the verb. In Martin Haspelmath, David Gil & Bernard Comrie, Matthew S. Dryer (ed.), The world atlas of language structures. Oxford: Oxford University Press. (https://wals.info/chapter/22)
Curnow, Timothy Jowan. 2000. Towards a cross-linguistic typology of copula constructions. In Henderson, John (ed.), Proceedings of the 1999 Conference of the Australian Linguistic Society, 1–9. (http://www.als.asn.au)
Degand, Liesbeth & Cornillie, Bert & Pietrandrea, Paola. 2013. Discourse markers and modal particles: Categorization and description. Amsterdam: Benjamins.
Dryer, Matthew S. 2005. Polar questions. In Haspelmath, Martin & Dryer, Matthew S. & Gil, David & Comrie, Bernard (eds.), The world atlas of language structures, 470–473. Oxford: Oxford University Press. (https://wals.info/chapter/116)
Haspelmath, Martin. 2021. Bound forms, welded forms, and affixes: Basic concepts for morphological comparison. Voprosy Jazykoznanija 2021(1). 7–28. (doi:10.31857/0373-658X.2021.1.7-28)
Iggesen, Oliver A. 2005. Number of cases. In Haspelmath, Martin & Dryer, Matthew S. & Gil, David & Comrie, Bernard (eds.), The world atlas of language structures, 202–205. Oxford: Oxford University Press. (http://wals.info/chapter/49)
Killian, Don. 2021. Towards a typology of predicative demonstratives. Linguistic Typology (doi:10.1515/lingty-2021-2078)
Norris, Mark. 2019. A typological perspective on nominal concord. Proceedings of the Linguistic Society of America 4(1). 12-1–15. (doi:10.3765/plsa.v4i1.4515)
Rijkhoff, Jan. 2007. Word classes. Language and Linguistics Compass 1/6: 709-726.
Song, Jae Jung. 2018. Linguistic typology. Oxford: Oxford University Press.
Velupillai, Viveka. 2012. An introduction to linguistic typology. Amsterdam: Benjamins.
Perhaps the difficulty comes from working with prototype style concepts? Various grammatical phenomena have clear-cut, prototypical examples where basically everyone agrees on how to classify them. The problem comes when people attempt to classify less clear cut examples. When this happens, it is difficult to decide what the “correct” classification is. It seems like a lot of people at least implicitly work with prototypical concepts, hence why this problem is common.
Yes, people are generally aware that the traditional terms do not always carry over straightforwardly to other languages, and their reaction is often to think of the concept in vague terms (or “prototypucal”, which sounds better than “vague”). Instead, I think it’s better to define the terms with sharp boundaries. Prototypes exist in cognition (and in the meanings of everyday words), but technical terms should have sharp boundaries. I may write a separate paper about this at some point.
I definitely agree that it would be better practice to use clearly defined technical terminology, rather than prototypical/vague terminology. I think the problem is that most practicing linguists don’t realize that there are problems with using vague terminology, or at least, don’t think it’s an important enough problem to require changing their usual practice of using prototypical terms.
Many textbooks explicitly mention some ambiguities with existing terminology, yet they continue to use it, which might be one reason why people think that this is not a very important issue.
I was amused. Of course the very notion of “complete data” makes no sense. And I sympathize with authors who, at least, admitted the difficulty. It seems to me that maintaining that one is able to provide a complete description, in order to satisfy the worker at his desk, has a kind of kafkaian turn, that a few of us, certainly are able to feel. Thanks,
François Jacquesson, CNRS, France
Well, “complete” is relative – for some kinds of questions, it’s not so diffcult to get reasonably complete data. But of course, there is no limit to the kinds of questions that one may ask, and to the kinds of additional data that one might want to gather… Thanks for your comment!
Thanks Martin for the response! I think the situation you are describing is applicable also to at least some of the cases you described in your posting: linguists find that there isn’t really any good definition that could be used to draw a distinction, especially if we want it to fit people’s intuitions about the concepts involved. For some distinctions, I think the right thing to say is that they are based on incoherent assumptions about how things work — that would apply to the distinction between languages and dialects, for instance. In other cases — and I think the notion of auxiliary belongs to them — the phenomenon is real enough, but the fuzziness of its borders makes any delimitation arbitrary.
Yes, I agree: People often have intuitions about the meanings of traditional terms, so the challenge is to provide a definition that fits largely with the intuitions. This is a terminological challenge, not a “difficulty of classification”. Maybe this difference is not so important, but I would like to emphasize it, because the general perception seems to be the opposite: That terminology is easy/trivial, while classification is difficult. In fact, I think it’s the other way round. As for “auxiliary”: I don’t have a problem with arbitrary decisions, but here I cannot think of a general way of delimiting auxiliaries from particles that are associated with verbs. For example, is Mandarin progressive zài an auxiliary?
On March 23, you wrote on the Lingtyp list:
“The English Passive includes an Auxiliary, but there is no good cross-linguistic definition of “auxiliary”, so we don’t want to say that auxiliaries can be criterial for passives.”
How is this different from saying that it is difficult to distinguish auxiliaries from everything else? If there isn’t a good definition of “auxiliary”, why don’t you go ahead and decide on one, rather than excluding auxiliaries from your analysis?
Thanks for this pertinent comment, Östen – indeed, why don’t I go ahead and define “auxiliary”, as I have done (super-boldly) with other terms (such as “morph” or “affix”)? The reason is that I couldn’t think of a definition that matches the legacy extension with 80% accuracy. The term “auxiliary”, which is well-established for English, has also been used for many other languages, but especially for those whose structure is (imprssionistically) similar to English. For languages like Chinese or Dyirbal, it’s unclear what would count as an “auxiliary”, because there are no truly general definitional criteria. I think we should give comparative-concept definitions for well-known legacy terms only if these definitions cover 80% of the traditional extension. For similar reasons, I wouldn’t try defining “supine” or “prepositional case” as comparative concepts. In the case of “auxiliary”, this is bad news, because so many people think that it is a good general concept. (And maybe they are right and I’m missing something. But I read Anderson (2007) and couldn’t find a definition, or come up with one myself…)