We know almost nothing about the earliest language(s) of humans, because humans had language(s) over 100,000 years ago, and there are no records or other good methods for learning about those languages. But there is a lot of interesting speculation, and some of this is potentially relevant to understanding similarities among present-day languages. In particular, one might ask whether some similarities or universals are due to inheritance from Proto-World (the language from which all modern languages descend).
If this were the case, then there would be an additional possible source of linguistic uniformity – above and beyond those that are most commonly discussed: universal functional-adaptive constraints, universal biocognitive constraints, and universal mutational constraints (Haspelmath 2019).
So if one wanted to argue for a functional-adaptive explanation of some general properties of languages (e.g. processing efficiency explaining word order universals, Hawkins 2014), one would have to show not only that a biocognitive explanation of the generative sort (Baker 2010) is not clearly superior, but also that the feature in question is not simply an inheritance from Proto-World.
In biology, this consideration has been known and discussed for quite some time. Sterelny & Griffiths (1999: 246) describe the problem as follows:
“The systematic study of convergence [supporting an argument for adaptation] requires an extensive use of the comparative method. For without a proper phylogenetic tree, it is not even possible to tell whether something is a convergence… A phylogenetic tree may reveal that [a particular piece of neural hardware] evolved just once, in the ancient common ancestor of all species that display the trait. If so, then the existence of this cognitive trait in many species is no convergence at all. It has not evolved repeatedly in response to some repeated feature of the environment… it may serve many different adaptive functions in different species, and exist in others merely by “phylogenetic inertia”. [For example, it would be misguided to] seek an adaptive explanation of why humans, birds, and seals all have such similar bones in their forelimbs.”
A similar argument was made for anthropology by Mace & Pagel (1994) in their widely cited paper:
“A cross-cultural association between two or more elements of culture does not necessarily indicate that the two elements tend to evolve or to be acquired together. They may both have been acquired for unrelated reasons in an older or ancestral culture and then subsequently transmitted to many later descendant cultures.” (Mace & Pagel 1994: 550)
This logic is compelling, and it has long been recognized by linguists, at least implicitly. I heard the term “Galton’s Problem” only in 2014, but I have been aware of the issue ever since I got into typology in the early 1980s. In fact, there was probably never a point when linguists were not aware of it. We know, for example, that we cannot claim that verb-second order is causally related to the existence of weak adjectives, because both phenomena are very unusual in the world’s languages, and we observe them together in German, Dutch, Danish and Swedish only because they were inherited from Proto-Germanic.
What linguists have been doing to address this issue is to study languages from a wide variety of families and geographical regions (e.g. Bell 1978; Bakker 2011), where the chances that the similarities are due to common inheritance are minimal. In effect, what we get through such samples is “independent instances of cultural change” (Mace & Pagel 1994: 550), because by hypothesis, the languages are not descended from the same protolanguage, and are distant enough from each other to rule out borrowing or at least make it unlikely).
However, this sampling approach presupposes that there are a substantial number of different language families (dozens, or maybe hundreds). This corresponds to the general view among linguists (Glottolog.org includes 429 families and isolates), but we cannot be quite sure. Maybe there are actually far fewer families? Maybe there are some widespread features that were are not functional adaptations but were simply inherited from Proto-World?
This possibility was suggested, for example, by Cysouw (2011: 417):
“It is possible that there are still founder effects available in the current distribution of the world’s languages, i.e., that there are preferences in the current world’s languages that go back to incidental events during the spread of languages over the world (Maslova 2000).”
This is indeed a remote possibility, but it has not been taken very seriously by comparative grammarians. Is this a mistake? Levinson et al. (2011: 512) urge us to take is very seriously, and they argue for the correlated-evolution method (instead of the sampling approach) to circumvent Galton’s Problem:
“Many typologists may assume that the dangers of covert phylogenetic dependence are remote. But given the apparent genetic bottlenecks at the beginning of the modern human diaspora out of Africa (Amos & Hoffman 2009), something close to language monogenesis seems a reasonable assumption, rendering Galton’s problem insurmountable. Furthermore, typological features can be remarkably stable over the life of whole language families (witness the case of Bantu in our study), so typological settings are likely in part to outdate known families (see Dunn et al. 2005, Reesink et al. 2009, Greenhill et al. 2010, Dediu 2011). We therefore think there is every reason to take covert phylogenetic dependence seriously.”
But is it really true that “typological features can be remarkably stable”? Which feature has been shown to be super-stable? The Bantu languages are known to be very closely related (and geographically compact), so it is quite odd that they should be mentioned in this context. The other papers mentioned in this paragraph are all quite speculative, and researcher’s outside of Levinson’s group have not tried to build on them (Dunn et al. 2005’s conclusions were questioned by Donohue et al. 2008 and Donohue et al. 2011). It may also be worth noting here that in a more recent (2017) study, the same authors (Levinson, Dunn, Greenhill, Gray and others) find that grammatical features changes more rapidly than words and are especially influenced by contact with unrelated languages (on the basis of Austronesian data; Greenhill et al. 2017). In the press release, Levinson is quoted as saying:
“This is a bit of an unexpected finding, since many have thought that grammar might give us deeper insight into the linguistic past than vocabulary, but there is still some reason for caution: we compared highly conservative vocabulary with an unfiltered range of grammar variables, and the language family is unusual for the way it diversified during colonization of successive islands. But what is clear is that grammar and vocabulary changes are not closely coupled, even within branches of a family, so looking at them both significantly advances our ability to reconstruct linguistic history.”
This is surprising only if one thinks that the critiques by Donohue and colleagues are wrong, because they had argued that grammatical features are more susceptible to borrowing and areal effect than lexical features (in line with a long tradition going back at least to Greenberg 1948: 24-26; see, e.g., Croft 2004)
So most of the evidence seems to be pointing to low stability of grammatical features, i.e. no reason to think that some typological trends were inherited from Proto-World. But I kept hearing the Proto-World argument, e.g. last autumn from Gerhard Jäger, who told me that he used the correlated-evolution method for his ALT 2019 talk because
„…wir bei synchronen Verteilungen nicht ausschließen können, dass wir Reflexe von Proto-Welt sehen (oder Reflexe eines population bottlenecks). Dieses Problem kann man durch keine Sampling-Methode der Welt lösen. Dazu muss man in die Diachronie gehen.“ [from a personal e-mail]
So earlier this year, I posted a query on the LINGTYP list, asking whether anyone knew of a general tendency that has been attributed to Proto-World.
The most substantive answer came from Michael Cysouw, who wrote (the full post is here)
“Whatever/When “Proto-World” was, it surely had a founder-effect in the sense that the original population of speakers was small, and whatever language-structures these people starting using were surely just a small selection of the many different possibilities that human language can have. .. Just to put things in perspective:
• Immediately post-glacial (10kya) the total worldwide human population was in the order of 1-10 million.
• Even today the median number of speakers per language is in the order of 10-100 thousand speakers.
• Taking upper (10M humans) and lower (10K speakers per language) estimates for an upper boundary, this means that 10kya there were maximally 1000 languages, possibly much less.
• Estimates for human populations before the last glacial maximum are much less clear (but see e.g. Atkinson et al. 2008), but 100kya we are probably talking more about 10K humans in total, i.e. just a handful of different languages. By 50kya there are probably still clearly less than 50K humans in total in the worlds, i.e. a few dozens of languages.
In my opinion there is intriguing (though surely not conclusive) evidence that the few languages that started it all off would show rather different typological profiles than a sample of today’s languages (e.g. Cysouw 2002; Cysouw & Comrie 2012; 2013). Some possibilities that might be considered for EHLS (“early human language structures”) are: much less fixed order, OV-type order (when fixed order is used), possibly some verbal morphology (with case only coming later?), no tone, no voicing oppositions. …
Given the small number of humans and the small number of languages, my guess would also be that the rate of language change would have been much smaller for most of the history of human languages. There was simply not much pressure to introduce much change (less contact, less need for social separation, in total less interactions because simply fewer people).
What seems rather clear is that the development of languages with more than 1M speakers is recent (I would guess that there were no languages with more than 1M speakers before the last glacial maximum), and that the development of such large speaker communities has had a profound impact on the typological profile of these languages.”
So these are some intriguing possibilities, but they do not seem to be more than Cysouw’s hunches. True, the idea that earlier languages had more OV patterns and the more recent, larger families have more VO orders has been around for some time (Givón 1979; Newmeyer 2000; Gell-Mann & Ruhlen 2011). But both OV and VO languages are found on all continents, and there is no plausible explanation connecting VO patterns with larger speaker populations or more recent time periods. And the typological profiles of hunter-gather languages have been studied recently by Bickel & Nichols (2020), who do not find any specific structural features associated with hunter-gatherer languages and conclude:
“Until such relations [between food procurement types and linguistic types] are demonstrated, typological generalizations drawn from modern languages can be assumed to be valid for all of the history and prehistory of language… frequencies and distributions, but not principles or defaults or constraints, have changed since the Paleolithic.” (2020: 73)
Now what about Cysouw’s suggestion that the rate of change may have been small in “early human language”? This is similar to a remark by Evans & Levinson (2009):
“The structural properties of language change on a near-glacial time scale. … a structural feature within a single large language-family like Austronesian changes on average just once about every 50,000 years. What that implies is that all the languages we now sample from are within structural spitting distance of the ancestral tongue! It is quite surprising in this light that typologists have been able to catalogue so much linguistic variation.” (Evans & Levinson 2009: 477)
But in diametric contrast to this, Nick Evans writes in a more recent paper (2017):
“widespread multilingualism should increase the rates of language change, in particular the rate at which new typological features appear.” (Evans 2017: 930)
In the LINGTYP discussion, Evans said that he would no longer agree with the 2009 passage (which may have come from Levinson anyway). In any event, the contrast nicely illustrates how little we know about rates of change of grammatical features, and how speculative any suggestion of retentions from Proto-World are. There is no doubt that such retentions are possible in theory, but there are very few indications that such retention play a role in practice.
Another interesting aspect to this was mentioned by van der Hulst (2008: §10). He mentions the possibility of a Proto-World explanation of universals, but he is generally skeptical, noting that
“the problem with this explanation is that languages seem to change so profoundly through time that it doesn’t seem clear why certain properties should have remained constant… We would need a theory of language change that could explain why certain apparently arbitrary properties of language (word order, morphology, sounds, and syllable structure) are subject to change, whereas other properties remain constant.”
And then he makes an observation that I had not heard before:
“A flaw in this line of explanation is that one would expect to find that the homologies in the languages of the world are not shared by many African languages, since, presumably, only a single African language underlies the languages outside Africa… I don’t think that there is evidence for making such a sharp distinction between the majority of African languages and all other languages.”
It may be too strong to assume that humans who left Africa spoke only one language, but clearly, the linguistic diversity of the first non-African founder population would have been much lower than the diversity within Africa. And if typological features were preserved from Proto-World to a noticeable extent, we would also expect a special “non-African founder” profile that is shared by all the non-African languages.
So I conclude that there are no particular reasons to think that Proto-World retentions are reflected in the general properties of the world’s languages. For a large number of general features, we have excellent reasons to think that they are more recent functional adaptations (e.g. triangular vowel systems or Zipfian brevity effects), so there are powerful known mechanisms that can explain many language universals. The theoretical possibility of Proto-World explanations does not seem to be sufficient to make phylogenetic tree estimation superior to the traditional sampling approach.
References
Amos, William & Joe I. Hoffman. 2009. Evidence that two main bottleneck events shaped modern human genetic diversity. Proceedings of the Royal Society B: Biological Sciences 1678. 131– 137.
Atkinson, Quentin D., Russell D. Gray, Alexei J. Drummond. 2008. mtDNA Variation Predicts Population Size in Humans and Reveals a Major Southern Asian Chapter in Human Prehistory. Molecular Biology and Evolution, Volume 25, Issue 2, Pages 468–474, https://doi.org/10.1093/molbev/msm277
Baker, Mark. 2010. Formal generative typology. In Bernd Heine & Heiko Narrog (eds.), The Oxford handbook of linguistic analysis, 285–312. Oxford: Oxford University Press. (6 July, 2015).
Bakker, Dik. 2011. Language sampling. In Jae Jung Song (ed.), The Oxford handbook of linguistic typology, 100–127. Oxford: Oxford University Press.
Bell, Alan. 1978. Language samples. In Joseph H. Greenberg (ed.), Universals of human language, vol. 1, 123–156. Stanford: Stanford University Press.
Bickel, Balthasar & Johanna Nichols. 2020. Linguistic typology and hunter-gatherer languages. In Tom Güldemann, Patrick McConvell & Richard A. Rhodes (eds.), The language of hunter-gatherers. Cambridge: Cambridge University Press.
Comrie, Bernard & Michael Cysouw. 2012. New Guinea through the eyes of WALS. Language and Linguistics in Melanesia 30. 65-95.
Croft, William. 2004. Typological traits and genetic linguistics. Albuquerque: University of New Mexico, ms. http://www.unm.edu/~wcroft/Papers/Typ-Gen.pdf.
Cysouw, Michael & Bernard Comrie. 2013. Some observations on typological features of hunter-gatherer languages. In Balthasar Bickel, Lenore A. Grenoble, David A. Peterson & Alan Timberlake (eds.), Language Typology and Historical Contingency, 383-394. Amsterdam: Benjamins.Cysouw, Michael. 2002. Interpreting typological clusters. Linguistic Typology 6(1). 69-93.
Cysouw, Michael. 2011. Understanding transition probabilities. Linguistic Typology 15(2). 415–431.
Dediu, Daniel. 2011. A Bayesian phylogenetic approach to estimating the stability of linguistic features and the genetic biasing of tone. Proceedings of the Royal Society B: Biological Sciences 278. 474–479.
Donohue, Mark, Simon Musgrave, Bronwen Whiting & Søren Wichmann. 2011. Typological feature analysis models linguistic geography. Language. Linguistic Society of America 87(2). 369–383.
Donohue, Mark, Søren Wichmann & Mihai Albu. 2008. Typology, areality, and diffusion. Oceanic Linguistics. University of Hawai’i Press 47(1). 223–232.
Dunn, Michael, Angela Terrill, Ger Reesink, Robert A. Foley & Stephen C. Levinson. 2005. Structural phylogenetics and the reconstruction of ancient language history. Science 309. 2072–2075.
Evans, Nicholas & Stephen Levinson. 2009. With diversity in mind: Freeing the language sciences from universal grammar. The Behavioral and Brain Sciences 32(5). 472–492.
Evans, Nicholas. 2017. Did language evolve in multilingual settings? Biology & Philosophy 32(6). 905–933. https://doi.org/10.1007/s10539-018-9609-3.
Gell-Mann, Murray & Merritt Ruhlen. 2011. The origin and evolution of word order. PNAS: Proceedings of the National Academy of Sciences of the United States of America 108(42). 17290-17295.
Givón, Talmy. 1979. On understanding grammar. New York: Academic Press.
Greenberg, Joseph H. 1948. The classification of African languages. American Anthropologist 50.24-30.
Greenhill, Simon J., Chieh-Hsi Wu, Xia Hua, Michael Dunn, Stephen C. Levinson & Russell D. Gray. 2017. Evolutionary dynamics of language systems. Proceedings of the National Academy of Sciences 114(42). E8822–E8829. https://doi.org/10.1073/pnas.1700388114.
Greenhill, Simon J., Quentin D. Atkinson, Andrew Meade & Russell D. Gray. 2010a. The shape and tempo of language evolution. Proceedings of the Royal Society B: Biological Sciences 277. 2443–2450.
Haspelmath, Martin. 2019. Can cross-linguistic regularities be explained by constraints on change? In Karsten Schmidtke-Bode, Natalia Levshina, Susanne Maria Michaelis & Ilja A. Seržant (eds.), Competing explanations in linguistic typology, 1–23. Berlin: Language Science Press. http://langsci-press.org/catalog/book/220.
Hawkins, John A. 2014. Cross-linguistic variation and efficiency. New York: Oxford University Press.
Hulst, Harry van der. 2008. On the question of linguistic universals. The Linguistic Review. De Gruyter Mouton 25(1–2). 1–34. https://doi.org/10.1515/TLIR.2008.001.
Levinson, Stephen C., Simon J. Greenhill, Russell D. Gray & Michael Dunn. 2011. Universal typological dependencies should be detectable in the history of language families. Linguistic Typology 15. 509–534.
Mace, Ruth & Mark Pagel. 1994. The comparative method in anthropology. Current Anthropology 35(5). 549–557.
Maslova, Elena. 2000. A dynamic approach to the verification of distributional universals. Linguistic Typology 4. 307 – 333.
Newmeyer, Frederick J. 2000. On the reconstruction of “Proto-World” word order. In Chris Knight, Michael Studdert-Kennedy & James Hurford (eds.), The Evolutionary Emergence of Language: Social Function and the Origins of Linguistic Form, 372–388. Cambridge: Cambridge University Press.
Reesink, Ger, Ruth Singer & Michael Dunn. 2009. Explaining the linguistic diversity of Sahul using population models. PLoS Biology 7(11). e1000241. http://www.plosbiology.org/article/info:doi/10.1371/journal.pbio.1000241
Sterelny, Kim & Paul E. Griffiths. 1999. Sex and death: An introduction to philosophy of biology. Chicago: University of Chicago Press.
I agree with everything you say about the alleged Proto-World leftover in present day languages and I also don’t think that phylogenetic tree estimation is superior to the traditional sampling approach. However, I find the idea of languages developing by way of functional adaptation somewhat difficult to comprehend. In biology species are developing by functional adaptation to the changing environment which is easy to understand. Change in what is responsible for functional adaptation in languages?
In biology functional adaptation doesn’t trigger mutations in organisms but is merely selecting the fittest variants generated by random change. What is the trigger of language change?
I think everybody agrees that contemporary synchronic states of languages are due to millennia of language change. But is the traditional sampling approach the right way to investigate this change? You seem to assume that cross-linguistic preferences established by synchronic sampling accurately reflect the targets of adaptive change. This is, however, far from evident. It is known that cross-linguistically common structures in one part of grammar can emerge as by-products of change in different parts of grammar.
If this is acknowledged, then how to establish the functional motivation? Sometimes this functional motivation seems obvious. However, what is obvious to us is not necessarily true, nature and language are known to sometimes behave against our intuition.
Just one example: for decades people were talking about a clear, intuitively obvious functional/cognitive motivation for overtly marking the 1 and 2 person on verbs and leaving the 3 person unmarked (so-called Watkins’ Law). The alleged preference for unmarked 3 person (contrasting with 1 and 2 persons) seemed supported by cases of recent reanalysis of a marked 3 person as bare stem (and subsequent restructuring of the whole system as happened, for instance, in the past tense of Insular Celtic).
However, a comprehensive synchronic sampling (Bickel and his group) revealed that systems with overt argument indexing on all three persons are more common. Isn’t this clearly contradicting the functional/cognitive preference for unmarked 3 person which people were finding so obvious?
Now, if those people were wrong and we now say that languages actually prefer systems with overt argument indexes in all three persons, and provide an obvious functional motivation, then how do we explain the cases of reanalysis (mentioned above)?
Yes, the parallels are not complete. Indeed, there are no different environments to which different languages adapt – they all inhabit (more or less) the same environment. But this is true of some biological species as well – e.g. marsupial rats, marsupial moles and marsupial wolves inhabit very similar environments as their placental counterparts, so they look very similar. You seem to suggest that biological change is driven by different environments, but I think that change is primarily random, both in biology and in language. Linguists tend to focus on identifying causes of change, but I think that that focus is misguided (though it’s not totally pointless to think about possible causes). Most changes are random, and languages typically diverge in unpredictable directions. But some changes lead languages into similar directions, presumably because of functional adaptation. And it’s these (rare) changes that lead to observed universal tendencies.
Yes, intuition is not (always) the best guide, so we should make sure to make predictions that can be tested rigorously (= objectively and with clear methods). Causal factors cannot be established for sure with correlational studies, but sometimes there are few competitors. For example, triangular vowel systems are apparently due to efficient dispersion – there is no alternative explanation that would work well (compare continental drift and plate tectonics – there’s simply no serious alternative explanation). Experimental studies are useful, too, e.g. artificial language learning experiments.
As for the two negative Bickel et al. 2015 papers, I don’t think they have shown what they claim to show. There is a very clear preference for differential object marking to follow referential prominence scales (see Schmidtke-Bode & Levshina 2018), and there is also a clear worldwide preference for zero in paradigms to occur in 3SG slots. So I think the old explanations are still valid.
I think your position on change and adaptation in language would be logically impeccable if it could be assumed – as you do – that language change is random (like random mutations in the DNS which generate variation in biology on which selection can operate).
But language change is not random in this sense of the word. On the contrary, different kinds of language change have well-established properties and we also know something about the cause(s) of change. It is, for instance, a fact that phonological change operates on segments and strings of segments (morae, syllables etc.), not on morphs or syntactic units. In morphological change, it is established that particular markers often emerge from particular sources (person indexes from personal pronouns and/or inflected copula). One well-established cause for change on different levels (phonology, morphology, syntax) is language contact.
I think it would be very strange if these properties pertaining to different kinds of change wouldn’t affect its outcome, i.e. the synchronic distribution of structures. But if it’s regognised that they probably do, then we have to take them into account. For instance, since change in phonology operates on phonological units (being blind to morphology), it’s randomly generating all possible kinds of morphological structures which can be replicated/can spread by language contact. The most frequent sources for bound person indexes necessarily generate many systems with 3-zero because many languages lack a 3-person personal pronoun and/or 3-person form of the copula. And so on.
Yes, we do need to take apparent regularities of change into account – as I say in my 2019 paper, there may well be mutational constraints explaining synchronic regularities in addition to functional-adaptive constraints. So it’s not just that frequent paths of change lead to frequent outcomes, but it may also be that some paths of change are for some reason impossible and so some universal trends are due to mutational constraints. So indeed, 3rd person zero paradigms may come from languages with no copula for 3rd persons. But where does that come from? I think it’s itself an efficiency effect. (And I don’t think that languages lacking 3rd person pronouns will give rise to zero paradigms more often – these languages generally do have special anaphoric forms based on demonstratives. I’d be curious to hear about any specific proposal in this direction. It seems implausible.)