Quite a few people have already responded to Keith Chen’s paper, but as far as I can see, none have highlighted a rather glaring problem with Chen’s claims: Chen completely ignores the hard-won insights about language structure that have emerged from research on language change, especially grammaticalization studies.
If I understand Chen right, he’s arguing for a direct, causal (‘Whorfian’) link between language structure and ‘habitual behavior.’ As he writes, ‘my findings are largely consistent with the hypothesis that languages with obligatory future-time reference lead their speakers to engage in less future-oriented behavior.’
On the other hand, Greenbergian/Givónian typologists have consistently come to the conclusion that language structure is best explained as the result of language change, e.g.:
‘… the study of grammatical typology is meaningless as a purely synchronic enterprise. This is so because the sum total of the various structural types that can code a particular target functional domain is nothing but the sum total of the various diachronic grammaticalization pathways from possible – functionally similar – source domains’ (T. Givón 2001: 23).
‘All explanations for linguistic phenomena, both universal and language-specific, must necessarily have a diachronic dimension, since all linguistic phenomena have histories which determine their present conventionalized state. [….] [T]he focus for establishing the explanations for cross-linguistic similarities should be on the mechanisms of change. […] (J. Bybee 2008: 108-109).
But language change, as nearly everyone agrees today, is ultimately the result of synchronic constraints on performance, which can be physiological, perceptual, or socio-cultural. This makes it likely that Chen has it backwards: typologically-oriented work on language change indicates that habitual behavior – including habitual discourse practices – can end up indirectly shaping language structure.
Let’s look at the domain of future reference. Typologists, as Östen Dahl points out, have focused on particular ways of encoding future reference. But whether a particular meaning is encoded morphosyntactically is not the only question. As Joan Bybee, Martin Haspelmath, and many others have argued, frequency of usage is also crucial for understanding how language change plays out. And here’s where socio-cultural constraints come into the picture.
For example, in some cultures, while there are clear morphosyntactic ways to encode future time reference, a significant number of speakers may refrain from making committed predictions, habitually or just some of the time. This can have a considerable impact on the grammatical encoding of this domain, as Kate Burridge (1995) showed in her work on Pennsylvania German. Burridge shows that socio-cultural mores shared by members of a speech community – a reluctance to talk very definitely or commitedly about future eventualities – led to the loss of the inherited future tense and the grammaticalization of an innovative future tense based on a verb meaning ‘to count.’ For speakers of this language, the socio-cultural constraint is salient and even conscious to an extent:
Mer sett net ganz definitely saage – oh mir dien so und so, weil mor wees nie net. Mer wees nie net was happene kennt.
‘We shouldn’t definitely say – oh we’ll do such and such, because we never know. We never know what will happen.’
In Modern Hebrew and Arabic, many ‘traditional’ or religious speakers refrain from making a committed prediction, using expressions like ‘no oath’ or ‘God willing’ or ‘with the help of God’ to attenuate their own responsibility. Here’s a taste of what an actual conversation can look like:
Q: Dan ve-Yonatan yavo’u l-a-yomuledet?
Dan and-Yonatan will.come.pl to-the-birthday?
‘Will Dan and Yonatan come to the birthday (party)?’
A: bezrataʃem
‘God willing (lit. ‘with God’s help’).’
This is very salient and sociolinguistically relevant for Hebrew speakers, as the following dialogue somewhat jokingly shows (given here in translation):
A: Okay, so we’re on for Monday?
S: Yeah, sure, bli neder (‘without oath,’ roughly ‘no promises’) I’ll come (FUT).
A: Wait a sec, what do you mean bli neder?
S: It’s a date. I said I’m coming (PRES). Bli neder I’m coming (PRES).
A: Can you just tell me if I can count on you or not? I just can’t understand you.’
S: Abigail, I’m coming (PRES), bli neder!
A: Can you stop for one moment with this religious craziness and tell me if you really intend to come. I need help and I don’t have anyone to ask except you.
S: Yes. I already told you that I’ll help you (FUT). So, bli neder, I’ll come (FUT).
A: You know what? I don’t need you or your help! Is it so hard to just say ‘yes’?
S: It isn’t hard at all, it’s just forbidden for it to sound like an oath (neder), so I just say bli neder but sure I’ll come (FUT).
In Egyptian Arabic, a typical exchange might look like this (example from Michal Marmorstein):
Q: ašūf-ik bukra?
Will I see you tomorrow?
A: in-šāʾa-llā(h)
Inshallah (God willing).
Both languages are spoken by fairly heterogeneous populations, not all of whose members subscribe to the world-views that underlie these speech practices. This is unlike the communities described by Burridge. This is probably what leads to the different outcomes of culturally-imposed constraints on talking about the future in a very definite or committed way: while in Pennsylvania German, a language with a small and socio-culturally homogeneous speech community, the effects of this constraint were very significant, in Hebrew and Arabic, on the other hand, there are enough speakers who do not ‘obey’ these culturally-imposed constraints, and as a result (multiple) dedicated future tenses are alive and well.
Heiko Narrog (2010) has made comparable arguments about the pathway “strong obligation > epistemic certainty” (e.g., English ‘must’). Narrog points out that this pathway is not as cross-linguistically frequent as the literature has led us to believe, which is in turn related to the fact that strong obligation markers are not that common, cross-linguistically. He says “[e]ven if obligation markers and constructions are grammaticalized, they may still be used less frequently than in English, or be subject to various constraints, thus making further grammaticalization and semantic extension less likely. I suggested that the non-development of obligation markers in many languages may be related to the social and cultural background in which these languages are spoken. Obligations are pragmatically impositions. They are thus potentially face-threatening, and their direct expression might be avoided.” In other words, socio-cultural constraints, including pragmatic maxims like ‘don’t make your interlocutor lose face by imposing an obligation on him/her’ can end up influencing linguistic structure.
In a similar vein, Peter Trudgill has been arguing for quite some time that sociolinguistic factors (e.g., size of the speech community, degree of isolation vs. contact, degree of intimacy of speakers) influence grammatical structure; see his recent (and fascinating) book Sociolinguistic Typology. He suggests (citing Givón, Perkins, Kay and others) that ‘informational homogeneity’ (large amounts of communally shared information), which is related to the size of the speech community, has an impact on grammatical structure. This is probably related to what can be taken as presupposed information in a given discourse situation. Aikhenvald (2004) offers a possible explanation for the relationship between small community size and the presence of elaborate evidential systems, and in a personal communication to Trudgill, writes that small tightly-knit communities tend to employ discourse which is very elliptical, ‘which is understandable, since in a small language community – say, one or two villages – people usually know what is being talked about’ (Trudgill 2011: 127). None of this should be very surprising, since Milroy & Milroy have showed again and again that the size, density, and other aspects of social networks has a lot to do with the seemingly intractable ‘actuation question.’
All this points to the idea that socio-cultural constraints might influence the relative frequency of use of the kind of constructions that are involved in expressing future time reference and various modal meanings. If these constraints are strong enough, or the relevant speech communities are small enough and/or homogeneous enough, they can end up having a strong impact on grammaticalization, and hence, on grammatical structure. In short, socio-cultural constraints can inform discourse practices (including relative frequency across a speech community) that in turn shape grammatical structure. This seems a more likely, well-founded – and for those interested in such things, a more economical – account than the other way around.
As a coda, one might mention that it is not so obvious that the use of a ‘present tense’ form to encode future tense reference means that a temporally ‘unmarked’ verb form is being used, pace Chen. As Haspelmath (1998) has shown, present tense forms can develop into futures or subjunctives. Given that in some languages, the so-called ‘present tense’ can be used to express or co-express future time reference but not reference to an ongoing event in the present, some of these ‘unmarked’ forms might well be on the road to becoming incipient futures.
Some references
Kate Burridge, 1995. ‘Grammaticalization in Pennsylvania German,’ in: H. Andersen (ed.), Historical Linguistics 1993. John Benjamins.
Joan Bybee, 2008. Formal universals as emergent phenomena: the origins of structure preservation. In J. Good (ed.), Language Universals and Language Change. Oxford: Oxford University Press. 108-121.
T. Givón, Syntax, vol. 1. John Benjamins.
Martin Haspelmath, 1998. ‘The semantic development of old presents: new futures and subjunctives without grammaticalization,’ Diachronica XV.1: 29-62.
Peter Trudgill, 2011. Sociolinguistic Typology. Oxford University Press.
Hi David,
It might be that the main thrust of what I wrote was unclear. I wasn’t arguing that grammaticalization is the only explanation for language structure. But I would argue that little about language makes sense without historical explanations. Synchronic constraints (cognitive and otherwise) are definitely part of the picture, mainly, in this line of thought, because they constrain language change. Anyway, it’s hard to claim that either Givon or Bybee has an anti-cognitive bias.
What I *was* trying to say is that examining this whole question without taking into account the insights of research on language change is probably unfortunate, and examining a correlation between the grammatical expression of future reference and extra-linguistic factors without taking into account the kind of things that have emerged from historical linguistics is likely to be a partial picture. A lot of this work has turned up interesting ways in which socio-cultural factors influence language structure via language change. An obvious issue (other than those I mentioned) is language contact.
As for the question of pro- or anti-Whorf, I admit that it isn’t the main thing that interests me here. I think you’re right that taking into account language change doesn’t rule out a ‘Whorfian causal chain.’ Let’s see what Boroditsky’s work turns up!
Re Eitan Grossman’s response to Keith Chen’s paper: It seems to me that this response suffers from the same fallacy as a large proportion of both the “anti-Whorf” and the “pro-Whorf” argumentation, namely, assuming that it’s an all-or-nothing story, that either language affects thought or thought affects language (but not both); or that either grammaticalization is irrelevant or it’s the only show in town (but not that it’s one among a number of diverse factors resulting in language being the way it is). These all seem to me to be manifestations of that primal (but scientifically unsound) human urge to have one single unitary theory that accounts for everything (or at least everything that one is concerned with).
To the case at hand: I would dispute Givòn’s claim that “the sum total of the various structural types that can code a particular target functional domain is NOTHING BUT [my emphasis, DG] the sum total of the various diachronic grammaticalization pathways from possible – functionally similar – source domains’; I would prefer to say that it is that PLUS all sorts of other things: synchronic cognitive constraints, accidents of history that may have wiped out other structural types, etc. etc. But even if Givòn were right and it was nothing but grammaticalization pathways fueled by discourse preferences, well this would at best provide support for a causal chain along the lines of thought/culture > discourse > grammar, in certain domains — but would hardly rule out the possibility of an opposite “Whorfian” causal chain in any number of other domains. Surely it is not incoherent to suppose that language might affect thought AND vice versa, in different cases and domains. Personally, I find it rather difficult to imagine a world in which this were not the case.
Of course, the devil is in the details. One of the merits of Lera Boroditsky’s work (and why I was surprised to learn in Östen’s latest post that she is collaborating with Kieth Chen) is that she devises subtle experimental methods aimed at teasing out the directionality of the causation. I’m curious to see what they’ll come up with here!