How can we understand Human Language if we don’t know all languages?

Linguists almost never worry about this question – apparently they often assume that languages are similar enough to allow them to make claims about Human Language (g(eneral)-linguistic claims) even if they study only a few languages, or sometimes only a single language.

But isn’t this a bit like making general claims about economics after studying only socialist economy, or general claims about lakes after studying only Lake Constance? Would economists and limnologists get away with this? Why should linguists be in a position to make such general claims on the basis of a few languages?

As a comparative linguist, I’m surely biased, so I’d like to hear other people’s opinions. (And I am looking forward to a workshop at the next DGfS Annual Meeting in Hamburg, where I hope some of these issues can be discussed.) It seems to me that true claims about Human Language can be based only on two kinds of evidence:

(1) broadly comparative evidence from the world’s languages

(2) evidence from aspects of Human Language that are not learned socially

An example of the first kind of evidence is the finding from broadly comparative research that all languages have demonstratives and question words (e.g. Diessel 1999; Idiatov 2007), contrasting, for example, with the negative finding that many languages lack a distinction between present-tense and past-tense verb forms (e.g. Bybee 1985). Without this research, we would not know about the universality of demonstratives and interrogatives, and we might suspect that a present/past distinction is universal (because it seems so natural).

But not everything about Human Language is learned socially. For example, reaction-time experiments may well give us insight into Human Language in general (g-language) even if we study just one particular language (p-language), because there is nothing socially conventional about reaction times in experimental settings. (Of course, these experiments are limited by what the language in question offers, so it is still useful to do psycholinguistics cross-linguistically; e.g. Jaeger & Norcliffe 2009.)

And Chomsky (1981: 6) is of course right that stimulus poverty can give us insight into Human Language, even if we study just one language:

“A great deal can be learned about UG [= universal grammar] from the study of a single language, if such study achieves sufficient depth to put forth rules or principles that have explanatory force but are underdetermined by evidence available to the language learner. Then it is reasonable to attribute to UG those aspects of these rules or principles that are uniformly attained but underdetermined by evidence.”

However, such stimulus poverty arguments are rarely appealed to by linguists who make universal claims on the basis of a single language. And where they might be appealed to, they are sometimes clearly wrong. For example, one might say that syntactic paradigms which are productive cannot admit any exceptions because there is no negative evidence. For example, English allows even before any kind of adverbal phrase:

(1) We would have gone out even {in the rain/at night/if it had rained/had it rained/…}

German is similar, with sogar ‘even’, but it is not allowed before a complementizerless conditional clause:

(2) Wir wären hinausgegangen, sogar {im Regen/in der Nacht/wenn es geregnet hätte/*hätte es geregnet}.

How can this restriction on a very general pattern be learned? This is not immediately apparent, but what is definitely clear is that this case cannot provide evidence for “universal grammar” (= biologically determined grammatical knowledge) because the restriction is a convention of the German language, and not anything more general.

So how can it be that not all general linguists are Greenbergian typologists (or psycholinguists)?

It seems that something strange happened over the last few decades: Many (or most) grammarians became typologists in principle, but not in practice. Until the 1960s, hardly anyone even thought about cross-linguistic comparison, and until the 1980s, few linguists paid particular attention to it. Throughout the 19th (and for most of the 20th century), it was perfectly sufficient for a linguist to devote her (or his) career to the study (often the historical study) of a single language. Linguistics was part of philology (the study of written languages with a respectable literature and national culture), and the historical study of languages served the purpose of strengthening national identities (like the study of history more generally).

As the nationalist ideology was waning with post-1945/1968 globalization and liberalization, a universalist outlook seemed a good way to provide deeper motivation for studying particular languages. But how can it work, given that each particular language is a historically arbitrary set of social conventions?

This is a genuine question – and I’m skeptical. Of course, the numerous similarities between languages make it generally useful to have knowledge of other languages, but strictly speaking it is neither necessary for grammatical descriptions to be “typologically informed”, nor do they necessarily inform us about Human Language. As far as I can see, it is only Greenberg-style universals that can tell us about general properties of Human Language, and separate these from historically accidental properties of the particular language that we may be looking at. (It may be that it is the – ideologically motivated – prestige of universalist linguistics that leads linguists to claim universal relevance of their language-particular work. This is speculative, but people’s actions are often driven by prestige, for good reasons…)

References

Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins.

Chomsky, Noam A. 1981. Lectures on government and binding. Dordrecht: Foris.

Diessel, Holger. 1999. Demonstratives: Form, function, and grammaticalization. Amsterdam: Benjamins.

Idiatov, Dmitry. 2007. A typology of non-selective interrogative pronominals. Antwerp: University of Antwerp Ph.D. dissertation.

Jaeger, T. Florian & Elisabeth J. Norcliffe. 2009. The cross-linguistic study of sentence production. Language and Linguistics Compass 3(4). 866–887. doi:10.1111/j.1749-818X.2009.00147.x.


3 thoughts on “How can we understand Human Language if we don’t know all languages?

  1. Here are two points to consider.

    1/ CLAIMS ABOUT LANGUAGE UNIVERSALS
    Given that the set of languages unbounded (there are extinct languages that we have no information about and there will be new languages arising in the future), we can never claim that some feature is CERTAIN be a universal for all languages. Instead, what we can say is

    – that something is CERTAIN to be a universal with respect to a given sample;
    and
    – that something MAY be a universals for all languages.

    The first is a certainty claim; the second is a hypothesis. Hypotheses may be strengthened or weakened by the kinds of evidence Martin listed above: broad language sample, psychological experiments, and acquisitional data (“poverty of stimulus”).

    2/ CLAIMS ABOUT UNIVERSAL GRAMMAR
    What we find to be present in all members of our language sample or what we hypothesize to be present in all human languages will not porove the claim that there is a Universal Grammar – that is, a domain of human cognition specific to language. Instead, what these findings will show is something general about human cognition. Whether this something is domain-specific or domain-general needs to be separately researched.

    Edith Moravcsik

  2. We haven’t even named most species of beetle let alone studied them, yet I can tell you all kinds of facts about beetles.

    If something is true of every member of a set, you can prove that to statistical significance using a small fraction I’d the set. This is called “statistics” and you might like to look into it if you want to be a scientist.

    • Statistical sampling of the world’s population is what Greenbergians have been doing for a long time. So the question I’m asking is not about this – it is about whether one can propose universals on the basis of a single language without a clear method for putting these universals to a test (cf. Chomsky 2001: 2). As far as I know, entomologists wouldn’t do this.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.