Since the 1980s, typologists have often summarized coexpression patterns (or “polysemy patterns”, or “syncretism patterns”) by semantic maps, as illustrated here for case expression (Narrog & Ito 2007: 282):
(For general introductions to semantic maps, see Haspelmath 2003; Georgakopoulos & Polis 2018). The claims about possible coexpression pattern that a semantic map makes
are the stronger, the fewer dimensions it contains. In Haspelmath (2004: 21), I suggested the following semantic map for coordinators and related functions:
This map is not unidimensional, but an important part of it is the implicational sequence in (1):
(1) V-conjunction – N-conjunction – comitative – instrumental
This map claims, for example, that there is no direct link between N-conjunction and instrumental, i.e. that forms that only express these two functions, but not comitative, do not occur.
Now since 2009, there has been an interesting development within the generative orientation in linguistics where such coexpression patterns are studied from a different perspective: the approach called nanosyntax (see Baunaz et al. 2018). I don’t know very much about this approach yet, and much of the discussion in these papers is so technical that it takes time to absorb it, but I think it’s not too early to say that the approach is pursuing some dead ends. But it’s instructive to consider where exactly the problems reside.
The initial and most influential publication (Caha 2009) was dealing mostly with coexpression (“syncretism”) patterns in Slavic case systems and built a whole theory on it, which couldn’t even begin to satisfy the typologist’s requirement of representative sampling with independent cases. Now there are two recent papers by Baunaz & Lander (2017; 2018), who take into account 21 languages from the Germanic, Romance, Slavic and Greek branches of Indo-European as well as two Finno-Ugrian languages. They are looking at coexpression patterns in demonstratives, complementizers, relarivizers, question pronouns, and indefinite pronouns. Their Table 1 (in the 2018 paper) is reproduced here:
The authors say that this pattern provides evidence for the “functional sequence” in (2):
(2) Dem > Comp > Rel > Wh > Indet
These are meant to be functional heads of the mainstream generative (“cartographic”) type, with the crucial extra idea of nanosyntax being that an exponent can “lexicalize” (= express) a syntactic structure if its features are a superset of the structure’s features. Thus demonstratives are said to have the features [Dem [Comp [Rel [Wh [Indet]]]]], complementizers are said to have the features [Comp [Rel [Wh [Indet]]]], and so on. It is not clear to me yet what is the independent evidence for these features, and whether there is a deeper difference between nanosyntax and the semantic-map view, which would simply say that (2) is a semantic map (an implicational sequence) and that each exponent must express a contiguous region on the map.
The semantic-map approach is not per se explanatory. A semantic map like Narrog & Ito’s case map above merely summarizes the observed coexpression patterns. The explanations for the cross-linguistic tendencies must come from elsewhere. In Haspelmath (1997), I tried to explain the semantic maps for indefinite pronouns by semantic-pragmatic similarities between adjacent notions on the map. But Cristofaro (2010) pointed out that the similarities can be of very different types, and she suggested that perhaps the best explanations for semantic-map adjacency come from diachronic tendencies (as also discussed in this earlier post).
Now what could be the explanation for the coexpression patterns observed by Baunaz & Lander? Positing a functional sequence in the generative tradition (especially the tradition of Cinque 1999) would seem to imply the claim that the sequence is part of the innate UG. The authors do not say so implicitly, but it is clear that such a claim would be extremely implausible.
First of all, the sample of languages on which their Table 1 (reproduced above) is based does not even begin to approach representativeness. While complementizers and relativizers are often coexpressed in the world’s languages (cf. Lehmann 1984), and indefinite-interrogative coexpression is also very widespread (Haspelmath 1997: Ch. 7), relativizers that have the form of interrogative pronouns are not found widely outside of European languages (Comrie 2005), and complementizers that have the form of demonstratives are restricted to West Germanic languages even within Baunaz & Lander’s sample.
Second, there are two fairly clear diachronic tendencies: Indefinite pronouns come from interrogative pronouns (but not vice versa; Haspelmath 1997), and relativizers in European languages come from interrogatives (and never vice versa; Lehmann 1984). For the relation between complementizers and relativizers, I know of no specialized research, but it seems to me that there is a unidirectional tendency as well: Complementizers come from relativizers, not the other way round. And relativizers occasionally derive from demonstratives (as in Germanic), and never the other way round (Diessel 1999). What explains these diachronic directionalities is a good question, but UG cannot explain them. On the contrary, given the diachronic tendencies (or mutational constraints), we can explain the coexpression tendencies.
Change trends were not highlighted in the 19th and 20th centuries, but the diachronic reasons for the similarities in Table 1 have been known for over a century. In order to convince those of their colleagues who have not completely forgotten about the diachronic knowledge that was accumulated by our ancestors, nanosyntacticians would have to show that they make testable claims that go beyond this earlier knowledge.
Unfortunately, in the 2018 paper, Baunaz & Lander go on to make even less plausible claims: They claim that German w-as and d-as, and w-er and d-er, and d-ie and w-ie can be segmented synchronically, but also that Russian što can be segmented into š-t-o, that Italian quello can be segmented into qu-e-llo, and so on. Of course, there are various similarities between these forms, and of course they are not historically unrelated. Italian que-llo is related to the article lo because it was formed from it by adding the demonstrative prefix qu(e)- (from Latin ecce eum). And at some point in pre-German, there must have been an analogical formation, where either the interrogative was was formed on the model of the demonstrative das, or vice versa. But this was a very long time ago, because these forms can be reconstructed to Proto-Germanic (hwat, þat), and indeed to Proto-Indo-European (cf. the Old Indic forms kad and tad), i.e. several thousand years ago. Could it be that this analogy is still perceived nowadays? Yes, of course, it is possible that the similarities were never “forgotten” through the dozens of generations that learned these forms.
But it is equally possible, and in fact it should be the null hypothesis, that these forms are not synchronically connected anymore. The evidence against the null hypothesis seems to be very weak: Highly frequent forms can be preserved with little change over centuries, so that the similarities were preserved in German and Dutch needs no extra explanation – in particular since some closely related languages did not preserve the similarities: English what is no longer pronounced [hwæt], and Swedish has replaced the Proto-North Germanic hvað/tað by vad/det, thus showing that the speakers did not see these forms as belonging together.
More generally, linguistic analysis requires productivity. If an apparent pattern is unproductive, there need not be any pattern at all. The similarity between ear and h-ear is striking, but since there is no way to form *h-eye (‘perceive by eye’) or *h-nose (‘perceive by nose’), we do not accept the analysis of h-ear with an h- prefix (though we cannot exclude that some speakers actually make this connection in their mental representations). This productivity condition on linguistic analysis is often forgotten (and Aronoff’s 1976 wrong claim about blocking on the basis of the unproductive suffix -ity, which cannot form *glori-os-ity from glori-ous, even became famous), so I thought that I should insert a brief reminder here. One could analyze wer/der as consisting of w+er and d+er, but the pattern is completely unproductive and hence such an analysis would be irrelevant.
If nanosyntacticians want to make true progress in understanding cross-linguistic patterns, they need to show that they are capable not only of telling nice stories, but also of making testable claims about patterns that do not have a better alternative explanation. Bobaljik’s (2012) book about comparative morphology does make such claims, and the explanations he gives are not easy to improve on (though see Plank 2016). I’ll keep paying attention to nanosyntax, because I find the cross-linguistic patterns interesting, and I hope that at some point I will be as impressed as I was by Bobaljik’s book.
Baunaz, Lena, Karen De Clercq, Liliane M. V Haegeman & Eric Lander (eds.). 2018. Exploring nanosyntax. New York: Oxford University Press.
Baunaz, Lena & Eric Lander. 2017. Syncretisms with the nominal complementizer. Studia Linguistica (vol?) 1–34. doi:10.1111/stul.12076.
Baunaz, Lena & Eric Lander. 2018. Deconstructing categories syncretic with the nominal complementizer. Glossa: a journal of general linguistics 3(1). doi:10.5334/gjgl.349. http://www.glossa-journal.org//articles/10.5334/gjgl.349/
Bobaljik, Jonathan David. 2012. Universals in comparative morphology: Suppletion, superlatives, and the structure of words. Cambridge, MA: MIT Press.
Caha, Pavel. 2009. The nanosyntax of case. Tromsø: University of Tromsø PhD dissertation. https://munin.uit.no/handle/10037/2203
Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic approach. New York: Oxford University Press.
Comrie, Bernard & Tania Kuteva. 2005. Relativization strategies. In Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The world atlas of language structures, 494–501. Oxford: Oxford University Press. http://wals.info/chapter/s8.
Cristofaro, Sonia. 2010. Semantic maps and mental representation. Linguistic Discovery 8(1). 35–52. doi:10.1349/PS1.1537-0852.A.345.
Diessel, Holger. 1999. Demonstratives: Form, function, and grammaticalization. Amsterdam: Benjamins.
Georgakopoulos, Thanasis & Stéphane Polis. 2018. The semantic map model: State of the art and future avenues for linguistic research. Language and Linguistics Compass 12(2). e12270. doi:10.1111/lnc3.12270.
Haspelmath, Martin. 1997. Indefinite pronouns. Oxford: Oxford University Press.
Haspelmath, Martin. 2003. The geometry of grammatical meaning: Semantic maps and crosslinguistic comparison. In Michael Tomasello (ed.), The New Psychology of Language, vol. 2, 211–243. New York: Lawrence Erlbaum. 10.5281/zenodo.831410.
Haspelmath, Martin. 2004. Coordinating constructions: An overview. In Martin Haspelmath (ed.), Coordinating constructions, 3–39. Amsterdam: Benjamins.
Lehmann, Christian. 1984. Der Relativsatz: Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: G. Narr. http://www.christianlehmann.eu/publ/Lehmann1984_Der_Relativsatz_OCRed.pdf.
Narrog, Heiko & Shinya Ito. 2007. Re-constructing semantic maps: The comitative-instrumental area. STUF–Sprachtypologie und Universalienforschung 60(4/2007). 273–292. doi:https://doi.org/10.1524/stuf.2007.60.4.273.
Plank, Frans. 2016. Jonathan David Bobaljik: Universals in comparative morphology: Suppletion, superlatives and the structure of words. Linguistic Typology 20(3). 687–713. doi:10.1515/lingty-2016-0038.
Yes, of course I mean synchronic analysis or description. The term “analysis” in linguistics is really confusing, and I’m trying to avoid it. I should try even harder in the future.
“linguistic analysis requires productivity”: this is perhaps valid for a strictly synchronic analysis, but certainly not for historical and comparative linguistics, right?