Tuesday, June 17, 2014

Why Yiddish is not Slavic, and language families are not families

Recently I came across a popular article, Where Did Yiddish Come From?, discussing Paul Wexler's eccentric claim that Yiddish is a "relexified" Slavic language (and Modern Hebrew, in turn, "relexified" Yiddish). To make any sense of this claim, we have to stop and consider what historical linguists mean when they talk about language origins.

If you want to learn a language perfectly, the best way to start is to pick it up as a child from your family and the community they're part of. That way, you and your generation end up speaking the same language as your parents and their generation, modulo a few little innovations you threw in just to annoy them. As those little innovations pile up, generation on generation, sooner or later you end up speaking something that the first generation wouldn't have been able to understand. In such a scenario, everyone agrees, the latest generation's language – let's call it B – is descended from the first generation's (A). If some of the children of that first generation moved far away early on and went through the same process of gradual change, their descendants speak another language, C, which speakers of B can't understand, but which is also descended from A. So we say that B and C belong to the same language family, just as their speakers belong at some remove to the same extended family.

If you're reading this, it's probably too late to learn a language that way. (Sorry.) You can still learn another language, say B, but the odds are that, at best, you'll always speak it with a bit of a foreign accent, and keep using expressions that make sense in English but sound weird to native speakers. If you're just an individual migrant learning it to fit in, that won't matter in the long run – your kids will learn the language in the playground and come back speaking it better than you do. But what if it's not just you that's learning it, but also your spouse, and your brothers, and almost everyone you know? What if your whole community is starting to prefer to speak this language with their kids, instead of the one they grew up with? In that case, the kids will still end up speaking it – but instead of speaking it like natives, they'll probably end up speaking it with your foreign accent and all those expressions of yours that native speakers laugh at. In that scenario, does the kids' language (let's call it D) belong to the same language family as B and C, or not? That's the ambiguity that Wexler is playing with.

The obvious answer – and the one most linguists would give – is yes*. For one thing, assuming you did a half-decent job of learning B, it's the same language – speakers of D can understand speakers of B, and vice versa, even if they laugh at each other's crazy accents. The influence of Gaelic may pervade Irish English, but Irish English is still English, not some Celtic language. It's the vocabulary and the morphology that really make English understandable – a weird accent or a funny way of putting things is just not that big an obstacle on its own. Wexler proposes exactly the opposite criterion: "Yiddish – in contrast to its massive German vocabulary – has a native Slavic syntax and sound system – and thus must be classified as a Slavic language" (1993:5). The origins of Yiddish syntax and phonology I can't comment on, but there's a good reason why historical linguists normally prioritise the vocabulary and the morphology over the syntax and phonology, even apart from the one just given. Vocabulary and morphology are eminently reconstructible, using the comparative method. Phonology, on the other hand, can only be reconstructed from vocabulary, and syntax is notoriously hard to reconstruct at all. If language families were to be defined based on phonology and syntax, it would hardly be possible to define them, much less reconstruct them or state regular correspondences between them.

In short, saying that Yiddish (much less Modern Hebrew) belongs to the Slavic language family is just a word game – in the sense that historical linguists normally use the concept of "language family", it doesn't, and wouldn't even if every last Yiddish speaker happened to be of Slavic ancestry and to speak Yiddish with a heavy Slavic accent. But such word games do not vitiate Wexler's work. After a large enough community has shifted to a different language, it is usually possible to find traces of their former language – although identifying them as such, rather than as later borrowings, may be hard. That's what Wexler is trying to do for Yiddish, and that's how he supports his claim that Yiddish speakers' ancestors used to speak a Slavic language.

* However, the question can easily be made more controversial. Suppose you and your community didn't learn it that well to start with, and aren't trying to imitate native speakers anyway? In that case, the kids will end up speaking something that sounds utterly ridiculous to native speakers; the basic words are recognisable, but the way they're put together seems all wrong. Whatever Tok Pisin is, most people would agree that it's not English. A few people would defend the claim that Tok Pisin belongs to the same family as English, on the basis that that's where the vocabulary comes from, but most would say that it doesn't belong to a language family. The language family model presupposes that the language is being passed on reasonably well as a whole, including not just vocabulary but also some amount of grammar; if all that's learned is a bunch of words, the model breaks down. The border must be drawn somewhere between the extremes of Irish English and Tok Pisin, but linguists can and do disagree on where exactly to draw it.


Kevin Hughes said...

Thank you for an interesting post. I was wondering if you might be able to provide a bit of clarification. One aspect of your discussion that I don't understand is your statement that

"Historical linguists normally prioritize the vocabulary and the morphology over the syntax and phonology ... Vocabulary and morphology are eminently reconstructible, using the comparative method. Phonology, on the other hand, can only be reconstructed from vocabulary ..."

As I understand it, phonology is one of the main areas where the comparative method is applied, grounded in the neogrammarian hypothesis (regularity of sound change). Maybe I'm not clear on your meaning of "prioritize" in this context, but is it not the case that comparative phonology has been one of the most lucrative ventures in historical and comparative linguistics? One case, for example, would be Saussure's laryngeal theory with regard to Proto-Indo-European.

I'm also unclear on how you are differentiating between comparison of vocabulary and comparison of phonology. Examining the vocabulary is how one comes to reconstruct proto-phoneme inventories, as well as proto-Lexical items; it seems like you can't compare one without comparing the other. I don't think it's inaccurate to say that language families are often established based on comparison of phonological systems. Although these comparisons are restricted to subsets of lexical items, you are still comparing phonemes.

Perhaps I've misunderstood, and you intend "phonology" to reference synchronic phonological rules and phoneme inventories, without consideration of regular sound change or reconstruction in this context.

Lameen Souag الأمين سواق said...

Yes, what I (like Wexler) mean by "phonology" here is the synchronic phonological system of a language – its phoneme inventory, its allophones, its phonotactics, etc. The concept of a phonological system is of course very useful in reconstruction, but the term "comparative phonology" is misleading here. From its structure, you would think it meant applying the comparative method to phonologies. But what it actually means is finding a phonology to account for data produced by applying the comparative method to vocabulary. A complete phoneme inventory for every Romance language would tell us absolutely nothing about Vulgar Latin phonology on its own, for example, whereas a phonetically transcribed set of Romance 200-word lists would be enough to reconstruct most of both Vulgar Latin's phonology and each individual Romance language's phonology.

David Marjanović said...

The origins of Yiddish [...] phonology I can't comment on

I think it's the lowest common denominator of German and Slavic. For instance, there are 5 vowel phonemes – the Slavic languages of the region have about 6, German dialects have maybe 10 to maybe 20.

bulbul said...

All true, but it's not just about funny accent or weird expressions. Wexler (and Zuckermann) argue that in Modern Hebrew, it goes deeper than that. For example Zuckermann argues that the inchoative meaning of nif'al and hitpa'el was expanded under the influence of Yiddish prefixed inchoative verbs, that Modern Hebrew is a habere language like Yiddish was and Wexler has give a number of examples of sentences in Modern Hebrew where the syntax is strikingly Yiddish- and Slavic-like and very unlike Biblical or Rabbinic Hebrew, hence relexification. Now don't get me wrong, I'm not saying they are right, it's just that the process is bit more complicated. Especially with Modern Hebrew, considering its origins.
FYI, if I recall correctly, the Slavic language Yiddish was allegedly built upon was Sorbian and in fact, Wexler often refers to Yiddish (at least in its early stages) as Judeo-Sorbian.

John Cowan said...

Here's what I posted at Language Hat on the subject:

The purpose of the Beider article is to reconcile the historic disagreements about Yiddish between the “Germanist” Stammbaum model and the “Yiddishist” relexification model of the origin of Yiddish. Beider rightly says that the relexification model, whereby Yiddish starts to exist as soon as Jews begin to speak German, and Yiddish itself is a language of mixed origins, doesn’t meet the standards for historical linguistics theories.

Unquestionably Yiddish and Modern High German are the descendants of Middle (or Early Modern) High German. Yiddish is not, as Wexler would have it, a relexification of Judaeo-Sorbian or some other Judaeo-Slavic language, nor of Judaeo-French as Weinreich thought, with German vocabulary. No relexification is so thorough, nor is the morphosyntax of Yiddish anything but a normal descendant of MHG/ENHG morphosyntax.

(Beider takes some trouble to show how the use of the name German for both the ancestral language and one, but not both, of the daughter languages tends to confuse people, though we are stuck with it. The simple act of using the name Old East Slavic for the ancestor of modern Russian, Ukrainian, and Belarusian instead of Old Russian has apparently quieted a host of nationalist ambitions and fears.)

However, by a mere reinterpretation of terms, the “Yiddishist” view can be made to say something very interesting, not about the genetic descent of the Yiddish language, but about the language shifts of Ashkenazi Jews. [...] Rather than saying that Yiddish descends from Judaeo-French, we can say that in Jews who spoke Jews’ French began to speak Jews’ German, an ethnolect of Old High German, retaining from the substrate language a number of words of French, Aramaic, and Hebrew origin. It was not until some centuries later that Jews’ German separated sufficiently from the German of non-Jews to be called a separate language, Yiddish, which then divided into Western and Eastern Yiddish. Almost all speakers of Western Yiddish then shifted languages again, to a new version of Jews’ German, based on NHG this time. So the line from Jews’ French to Jews’ German (I) to Western Yiddish to Jews’ German (II), although in no way a genetic line of descent, does represent a reality, namely a sequence of substrates each of which leaves a layer of vocabulary in the languages which follow it.

Jim said...

"Hebrew where the syntax is strikingly Yiddish- and Slavic-like and very unlike Biblical or Rabbinic Hebrew,"

Nuu-chah-nulth is suffixing but verb initial, just like the Salishan languages around it and the other Wakashan languages. It is structurally similar enough the whole Sapir thought these langauges formed a family he called Mosan. It turned out to be a Sprachbund. Unfortunately we don't have any record of what proto-Wakashan looked like before this contact started having these effects.

"It was not until some centuries later that Jews’ German separated sufficiently from the German of non-Jews to be called a separate language, Yiddish..."

And not all non-Jewish Germans, by a long shot. Yiddish sounds a lot like Nuernberg Franconian and shares a lot of the same sound changes that distinguish Rhineland dialects form High German - and these similarities track the geographically spread of Jewish communities in the early Middle Ages,

Lameen Souag الأمين سواق said...

Bulbul: Certainly it does go deeper than I was suggesting in the case of Modern Hebrew. Even for Yiddish, Wexler makes an interesting argument that the meanings of particle+verb combinations are almost entirely calqued from Slavic, rather than reflecting German semantics. But all his comparisons seem to be with standard German and modern Sorbian, so I don't know how seriously they can be taken. In any event, I would insist that syntactic-semantic patterns like those that Wexler and Zuckermann highlight cannot be reconstructed with anything like the accuracy of vocabulary and morphology, and that it would be perverse to insist that a language's family membership should be conditional on the natural transmission of such patterns. (Wexler, indeed, gleefully points out that his definition would logically make it impossible "to identify genetic relatedness on the basis of linguistic criteria alone", which to me is precisely why it should be rejected.)

Jim: I wonder what Wexler's claims would have looked like if he had compared Yiddish with Franconian rather than with Standard German...

Jim said...


I hadn't thought of that until you mentioned it and now I have to wonder why he even thought there was any point in using Standard German for his comparison.

And now that you mention Sorbian, I have to ask the same question about the point of using it. Surely it was a rural language, spoken often by low status migrants into formerly Germanic territory. Jews were predominantly urban dwellers, and cities in central and eastern Europe in the Middle Ages were pretty much uniformly German-speaking. How and why would Jews have learned any Sorbian anyway?

Lameen Souag الأمين سواق said...

That's simple: he postulates that most Ashkenazi Jews are descended from Slavic converts to Judaism, who supposedly converted as a way to evade persecution and enslavement by Christians for being pagan while preserving their identity. (Not the most effective long-term strategy on either count, then.) In fact, there is a rather large non-Jewish genetic component among Ashkenazis, but it seems to be linked more to Italians than to Slavs.

marie-lucie said...

Vocabulary, in the sense of the mass of words in a language, allows linguists to establish not only the phonological system but also the morphological system. Words are usually classifiable into nouns, verbs and other categories, but that membership is often expressed by some modifications such as nominal case endings, verbal stem modifications and/or endings indicating such things as tense, aspect, person, etc (more or less, according to the language). A language may borrow a lot of vocabulary from another without borrowing the associated morphology, instead modifying the borrowed words according to its own morphological system. English has borrowed heavily from French and Latin without borrowing the morphology associated with the words in the original languages: although educated English maintains some Latin noun plurals, borrowed verbs (or verbalized items) are modified with English verbal affixes, not French or Latin ones. Conversely, English verbs borrowed into French are fitted as stems into the much more complex French verbal morphology (eg the recently adopted verb "tacler" "to tackle", first adopted in the context of sports, which can have all the conjugation forms of verbs ending in -er in the infinitive). Supposing that Yiddish is actually a relexified Slavic language, why is its morphology not Slavic but Germanic?

David Marjanović said...

In fact, there is a rather large non-Jewish genetic component among Ashkenazis, but it seems to be linked more to Italians than to Slavs.

Italians? That's a surprise.