I spent quite a lot of time as an undergraduate analysing Hungarian syntax with my generative head on and using the minimalist framework. Bear with me. This post is the result of me trying to marry all of them hours spent reading “The Minimalist Program” (Chomsky, 1995) and starring at Hungarian with what I’m currently doing¹ and ultimately trying to convince myself that I wasn’t wasting time.
So here’s a condensed summary of what my dissertation was about:
Hungarian has a massive case system which, as well as structural cases, has many items which have locational, instrumental and relational uses (lexical case markers). Because of this many constructions which feature prepositions in English, when translated into Hungarian can be translated as case markers or postpositions.
It struck me as odd that these 2 things; case markers and postpositions, despite having the same position in the structure (as a right-headed modifier to the noun) and very similar semantic function, would have different analyses in the syntactic framework, simply due to the fact that one was morphologically attached (case markers) and the other not (postpositions).
Postpositions fall into 3 categories in Hungarian and each of these shows varying degrees of similarity to lexical case markers. I spent most of the 10,000 words discussing these similarities (and differences) at length and drawing a lot of trees. I’m not going to go into all of that here but if you’re interested you can email me (hanachronism at gmail dot com) and I’ll send it to you (it really is dull dull dull).
Skipping to the conclusion because that’s always the exciting bit:
Historically a lot of case markers (at least 10) were once postpositions (Spencer 2008). I concluded that “type 1” postpositions are on their way to becoming fully fledged case markers and through language change will gradually become morphologically attached to the nouns they specify and undergo vowel harmony. Due to the weak prosodic status of both postpositions and case markers (Kiss 2002, Kabak 2006) and there being no determiner category that intervenes between the postposition and the noun they meet the conditions that allow the linear fusion hypothesis to occur. (Kabak 2006). (The Linear Fusion Hypothesis states that items which are used together are fused together (I love that this rhymes). (Bybee, 2002)).
“Type 1” postpositions also do not assign case to the nouns they modify (other postposition types do). Kabak (2006) states in his paper on the morphologisation of postpositions that:
Postpositions following an uninflected form of complement have a greater chance of turning into case suffixes or clitics than those following a case-inflected form.
All case suffixes that were developed out of postpositions in Hungarian had once assigned no overt case suffix, just like type 1 postpositions!
The Hungarian case system is growing and has been for some time.
Now I’m going to get to the part that’s relevant to Language Evolution.
Bentz and Christiansen (2010) in a talk at EVOLANG8 put forward the hypothesis that an SVO language with a small amount of case markers is what all languages are tending towards due cognitive constraints on learnability. They also make the assertion that the case system shrinking in romance languages demonstrates the presence of a constraint which may have shaped linguistic adaptation over evolutionary time. This is the sort of claim that, regardless of what the general direction of language change in the majority is, one language (or two, I’ve heard Estonian is going the same way as Hungarian) will disprove it because cognitive constraints on learnability are universal.
The assertion that a smaller case system contributes to the system being simpler (and therefore more learnable) as a whole isn’t very convincing. In the case of Hungarian, as the case system is getting larger, the postpositional system becomes smaller, and so the analogous systems together are staying pretty stable in terms of complexity.
So what can Hungarian Postpositions tell us about Language Evolution? Just that smaller case systems aren’t more learnable than larger ones I guess.
References
Bentz, C and M. H. Christiansen 2010. ‘Linguistic adaptation at work? The change of word order and case system from Latin to the Romance Languages.’ In The Evolution of Language. Eds. A. D. M. Smith, M. Schouwstra, B. de Boer and K. Smith. World Scientific. pp. 26-33.
Bybee J. 2002. Sequentiality as the basis of the constituent structure.’ In Givón, T. & B. F. Malle (eds.) The Evolution of Language out of Pre-Language, 109-134, Amsterdam: John Benjamins
Chomsky, N. 1995. The Minimalist program. MIT Press.
Kabak, B. 2006. ‘An obstacle to the morphologization of postpositions.’ In Studies in Language 30: 1. pp. 33 – 68. John Benjamins.
Kiss, K.E. 2002. The syntax of Hungarian. Cambridge.
Spencer, A. 2008. ‘Does Hungarian have a case system?’ In Case and grammatical relations: studies in honor of Bernard Comrie. Eds. B. Comrie, G. G. Corbett and M. Noonan. Pp. 35-56
———————————————————————————————–
¹ Studying language evolution in an environment where almost everybody believes that structure in language is mostly a result of a cumulative result of cultural evolution, as opposed to the result of an innate universal grammar.
You’re entirely right – I’d also say that if their claim was true, we would *not* have languages with case and we would *not* have languages which weren’t SVO. We’ve been around long enough to weed out such features if they were indeed less learnable…
I agree. I think to say something along the lines of “case systems are hard to learn” is massively overgeneralising. Case systems cannot be separated from the context of the languages in which they’re found. While it might be true in the case of at least some Romance languages that case systems make things harder to learn, to say this is a universal constraint ignores the cognitive reality that learnability constraints are balanced across systems as a whole. If by expanding the case system you diminish the influence of some other part of the grammar that is irregular and hard to learn (as may be the case in Hungarian, but I’m no expert) it seems entirely natural that this could happen.
Are we really still at the stage where properties of romance languages are generalised across all other languages leading to vastly convoluted and psychologically improbable analyses of any languages that don’t fall neatly into the resulting unfalsifiable paradigm?
The inadequacy of unidirectional proposals like Bentz & Christiansen’s was already known more than a century ago. In 1891, for example, von der Gabelentz said “[inflectional] affixes become fused and finally they disappear without leaving any trace begind, but their functions remain, and strive once more after expression. In the isolating languages they find it in word-order or formal elements, which again succumb in the course of time to agglutination, fusion and eclipse.” (as quoted in The Loom of Language, p. 205.) In addition to Hungarian and Estonian, such examples as Bengali may be cited, which lost its original case marking then developed a whole new set of cases.
In general, changes that make production simpler make processing (and learning) harder, and vice versa. Any account of language evolution that takes into account only half of this dynamic will end up unrealistically predicting that all languages should end up looking pretty much the same.
Also, come to think of it (you’re probably ahead of me here), there’s a big difference between Hungarian cases and Latin ones – Latin cases are morphologically incredibly irregular, with at least 5 very different declensions, while Hungarian ones keep pretty much the same form, modulo vowel harmony, whatever you attach them to. On the face of it Latin cases ought to be harder to learn than corresponding adpositions, simply because their form is so much more variable – but that argument wouldn’t carry over to Hungarian well at all.
Firstly, I’m glad we all agree.
L – You’re right, there is a big difference between Latin and Hungarian cases. I’d say the main difference is that the majority of case-markers within Latin were structural and in Hungarian most are lexical (and definitely the ones derived from postpositions).
In the case of Latin, adpositions generally can’t take the place of or be replaced by structural case, instead it has to be replaced by a strict word order system (which is what happened in romance languages). So the 2 systems of change aren’t really directly comparable beyond a simple statement about the size of the case-system which is what Bentz and Christiansen (2010) (and me) are making claims about.
With regards to the amount of variation in the realisation of case markers within Hungarian, I would say that for the most part it is very predictable and regular due to vowel harmony. I think that may be oversimplifying it a bit too much though because due to the vast number of cases there is significant cross over in meaning in places which might create a barrier for children when learning because they generally don’t like 2 things to mean the same thing. This is clearly also a problem within Latin because synonymy avoidance in children should drive out irregular case markers which mean the same thing, unless of course they have a salient categorical system which keep these things separate (such as the declension system in Latin)(Carstairs-McCarthy 2010). It’s hard for people learning Latin in a classrooms now to get their heads around the declension system (I’ve tried to do it myself) but it was probably quite easy for children with Latin as their L1.
If it was the irregularity which lead to the decline of case within romance language, it’s odd that the case markers wouldn’t just become more regular over time instead of disappearing altogether, and they certainly weren’t replaced by adpositions with more regularity but instead by a rigid word order.
Have I addressed the problem there? I’m slightly hungover and aware that I may not be making too much sense.
Refs.
Carstairs-McCarthy (2010) The Evolution of Morphology. Oxford: Oxford University Press.
Good point about the structural-lexical distinction. However, I would say at least 2 out of 5 Latin cases got replaced by prepositions – ad for the dative, de for the genitive. The only thing rigid word order replaced as far as I know is the nominative/accusative distinction, which was also the last distinction to be dropped (it survived into Old French), and even that got partly “replaced” by the preposition a in Spanish. It does seem a bit odd that there wasn’t any trend to simply regularise the case system though.
By virtue of being Hungarian, I feel supremely qualified to comment. 🙂
I agree with your conclusion. There’s nothing surprising in a growing case system. After all, Hungarian is an extensively agglutinative language, and most probably it didn’t spring to the world fully armored with all those layers on layers of suffixes like a linguistic Athena. Rather, its words have probably gradually engulfed other words that became “paralyzed” as postpositions do today. Just as mitochondria were once free-living bacteria, what are bound morphemes today must have been full-fledged words. Indeed, postpositions are in a vulnerable position due to their relative immobility. There are signs that some are already being fixated.
While this process drains the pool of postpositions, their numbers don’t necessarily dwindle as they may be replenished. Some words may ossify into postpositions.
On the other hand, the pool of case markers also has a drain. Witness the ongoing merger of ‘-ba’ (into) and ‘-ban’ (in) suffixes. In spoken language hardly anybody makes the distinction any more (‘-ba’ is used for both). In written language it’s still considered a grave error, though.
The flip side of the cognitive capacity argument is that there is at least some evidence from neuroscience that the brain is hardwired, or at least predisposed to hardwire itself, to put certain kinds of ideas in certain parts of the brain, and those general categories seem to bear a striking resemblence to the case systems of languages that have them, with the full set matching the largest case systems found in places as distant from each other as West Africa and Australia and the Caucuses.
There is likewise evidence that case systems are, on average, older, and one would think that a long, long time ago, when society was changing less rapidly and populations were more stable, that languages would be closer to an equilibrium state.
Put another way, case markers are the sort of thing that second language learners have trouble with and for them, the are an extra burden and get dumped in creolization contexts, but what is hard for second language learners, may actually be more fundamentally tuned to our brains and natural for first language learners.
ohwilleke: what sort of evidence? I’d be very interested to read references.
This is the sort of claim that, regardless of what the general direction of language change in the majority is, one language (or two, I’ve heard Estonian is going the same way as Hungarian) will disprove it because cognitive constraints on learnability are universal.
I haven’t got access to their whole paper but I see that they mentioned Wray & Grace(2007) who argue that differences in the makeup of the speaker population will influence the learnability pressures. So, for instance, those in an exoteric population are more likely to have a higher degree of non-native (L2) speakers, subsequently creating a pressure for the language to become more learnable for these individuals, which, in Bentz & Christiansen’s case, will result in a shrinking (or simplified) case system. If, on the other hand, the population is esoteric then it is more reliant on L1 acquisition.
My point being that the existence of languages which run counter Bentz & Christiansen’s observation does not disprove the hypothesis. Rather, it suggests that those languages such as Hungarian and Estonian have different speaker demographics that add a pressure to retain, or even increase, the case system. Now, I know that changes in language, if influenced by demographic conditions, will lag behind changes in demography, but I think something worth investigating would be the relationship between the case system and net migration. The hypothesis being: those languages with simpler case systems will have higher net migration (reflecting a pressure for the language to become more learnable for L2 learners) than those with more complex case systems (reflecting a pressure from L1 learners). Quickly looking at some examples:
United Kingdom: 2.16 migrant(s)/1,000 population
Italy: 2.06 migrant(s)/1,000 population
Hungary: 0.87 migrant(s)/1,000 population
Estonia: -3.26 migrant(s)/1,000 population
Four languages are far from proof, but here is a link to the CIA World Fact Book on Net Migration worldwide: https://www.cia.gov/library/publications/the-world-factbook/fields/2112.html. It would certainly be better to look at historical migration rates, although I’m not sure whether there are any useful sources for this…
Also, as I said earlier: I haven’t read Bentz & Christiansen’s paper, so I’m not exactly sure as to whether they made their argument in direct reference to Wray & Grace or not.
L – Ok, Do we know if the nom. and acc. changing to rigid word order happened before or after the dative etc. changed to prepositions? It’d be interesting to see what changed first/which pressures were stronger.
Karoly – Thanks, I didn’t know about ‘-ba’ and ‘-ban’. That’s really interesting!
Distinction between no. of L1/L2 speakers –
You make very very good points and after going back over Bentz and Christiansen they make the same point (without the comparisons to Hungarian and Estonian as I was the one who brought them into this.)
Talking about (vulgar) Latin becoming a lingua franca in the Roman Empire (and hence having large numbers of L2 speakers) they say the following:
Sooooo… yes, ignore what I said completely. They make a good point and we should start drawing up correlations between world immigration rates and case system growth/decay. Sorry, I’m an idiot. I’m glad we’ve talked this through and all reached the same conclusion though.
Wintz,
While it may be that that low migration figure has been valid in Hungary for the last ~hundred years (with borders drawn after WW I that became impermeable and created an ethnically homogeneous country), it cannot possibly have been in earlier times. Hungarians have absorbed various peoples and come under the influence of others. Although they came from the east (Urals?), genetically they’re indistinguishable from their neighbors today. (They still managed to keep their non-Indo-European language.)
@Karoly: I did say: ‘It would certainly be better to look at historical migration rates’. Plus, there are other factors underlying an exoteric population (population size, geographic spread etc).
Romance still has a fairly free word order, because they are by now to a large extent head-marking languages, using enclitic pronouns attached to the verb to refer to direct and indirect objects. These elements are more like affixes that words now, and thus the verb has grown in complexity. The rules for using these elements differ somewhat between the languages. In Italian you use them if you take the object out of its most usual place just after the verb. In Spanish the indirect object is always referred to in the verb, an the direct object referred to as in Italian.
Another language with a growing case system is Finnish, and this language is a lot more Latin-like than Hungarian, with more fusion of elements and agreement in number and case of adjectives. In Finnish all postpositions govern a case, most often the genitive, but still they tend to fuse. Also in Finnish case forms of nouns with a derivation ending tend to become new case forms.