Hannah Sarvasy reported back in 2020 on some suggestive research:
Languages like Japanese, Korean, Turkish and the indigenous languages of the Amazon, East Africa, and New Guinea build sentences in a way that lets them grow to enormous length. Our research shows learning one of these languages may help children create complex sentences that express multiple ideas at a younger age.
Try recounting what you did this morning, or telling a story, and chances are you’ll use a series of several sentences: “This morning, I woke early. I dressed and ate breakfast. I gathered my things, said goodbye to my family, and they waved goodbye to me. Then I drove to work.” In English, the simplest sentence, or “clause,” is just a subject plus a verb (“I dressed”). You can also join two clauses into a sentence using words like “and” or “while,” but it’s unnatural to join more than about three clauses into one English sentence.
But in many languages across Central Asia (from Turkish to Tibetan, Mongolian, Japanese, and Korean), and in many indigenous languages of the Amazon, East Africa, and New Guinea, stories can take the form of one long sentence. These sentences look more like this: “Waking up early this morning, dressing, making breakfast, eating, washing the dishes, gathering my things, saying goodbye to my family, they waving goodbye to me, I drove to work.”
These long sentences are known as “clause chains.” Unlike in English, where most of the clauses in a story would make sense if you spoke them outside the story (“I dressed”), all but the very last clause in a “clause chain” are abbreviated—they can only function in a clause chain. “Dressing” or “making breakfast” sounds unfinished on its own, and only the final verb of the clause chain tells you whether the events are happening in the past, present, or future.
Clause chains are special because they can be extremely long, pushing the boundaries of what we consider “sentences” in English. Chains of more than 100 clauses have been recorded. Non-native speakers may have trouble keeping track of who is doing what in clause chains. One linguist who studied a language of the Himalayas (where many languages use this type of sentence) color-coded clause chains in her notes to keep track of the plot.
In some languages, especially of the Amazon and New Guinea, there’s a further twist. In each clause of the chain, the speaker has to announce in advance whether a different person is carrying out the action in the upcoming clause (as in “saying goodbye to my family, they saying goodbye to me”). This is called “switch-reference marking,” and it probably means speakers of these languages have to plan further ahead than speakers of English.
So do kids learning Turkish or Japanese speak in precociously complex sentences, compared with their Anglophone peers? We investigated this for six languages—Japanese, Korean, and Turkish, plus three indigenous languages of New Guinea and Australia (Ku Waru, Nungon, and Pitjantjatjara). We used a variety of methods, looking at data from different children and from the same children over time. […] It turns out children learning these languages are first able to speak in well-formed clause chains between the ages of two and two and a half. This is around the time that children learning English and French make their first attempts at combining clauses into sentences.
But the English- and French-speaking children generally make some mistakes (for instance, by leaving out conjunctions), or actually express only a single idea across two clauses (in “look at the house that we built!”, there is only one notion: that we built a house). The children learning the clause-chain languages did not make such errors. What’s more, in most of these languages the children’s early clause chains express multiple ideas. It may be that the abbreviated verbs used in all but the last clause of a clause chain make it easy for these children to describe complex sequences of events in a single utterance. […]
It’s well known that children learning most languages go through an early phase in which their utterances are limited to two words: the “two-word phase.” After this, children don’t proceed to a “three-word phase.” Instead, the progression is from two to “more.” However, until now no-one has thought to ask when children learn to combine more than two clauses into a sentence. This may well be because research on how children learn to combine ideas into sentences has been largely shaped by speakers of English and other languages that lack clause chains. (This shows it’s important for scientists to come from varied backgrounds!)
When we investigated this idea, we found that all children learning languages with clause chains begin by speaking in two-clause chains. So no, kids don’t begin spouting sentences of 20 clauses at age two! But the “two-clause phase” lasts for as little as one or two months, and after that most children we studied advanced directly to a “more clauses phase,” in which their sentences include anywhere from two to five or more clauses. A Japanese child recorded a 20-clause chain at age three and ten months, and this may not be unusual.
Interesting stuff — thanks, Bathrobe!
I don’t understand why
“Waking up early this morning, dressing, making breakfast, eating, washing the dishes, gathering my things, saying goodbye to my family, they waving goodbye to me, I drove to work.”
would be all that different from
“Waking up early this morning. Dressing. Making breakfast. Eating. Washing the dishes. Gathering my things. Saying goodbye to my family. They waving goodbye to me. I drove to work.”
in “look at the house that we built!”, there is only one notion: that we built a house
I don’t understand this. Surely there are two notions: we built a house; please look at it.
If you only want to say that we built a house you would say “we built a house.”
Clause chaining by altering the form of clause-final verbs to a “converb” in a SOV language is actually only one manifestation of a much more general tendency.
In many languages, non-initial clauses in narrative which are carrying on the narrative thread (as opposed to forming parts of descriptive asides, flashbacks etc) formally resemble subordinate clauses. This happens in SVO Hausa and Fulfulde, and (you guessed it!) Kusaal, and numerous other West African languages,
And in VSO Biblical Hebrew, to boot.
Kusaal narrative clauses begin over and over with ka “and” (except it doesn’t really mean “and” at all in this case: the English equivalent is zero.) It’s reminiscent of the KJV Bible, with its copying of the vav-consecutive – and indeed, it’s basically the same phenomenon. Such Kusaal narrative ka-clauses are not marked for tense and lack the tonal and segmental markers of non-subordinate clauses (Kusaal marks clauses specifically for non-subordinate status, rather than the reverse, as in SAE.) [Kusaal also has its own formal equivalents of the switch-reference thing that is the “further twist” they refer to. That’s not unusual, and by no means confined to the Amazon and New Guinea.]
Hausa clauses carrying on a narrative use a distinctive form of the perfective aspect which appears elsewhere only in relative clauses and in clefting. Fulfulde does more or less exactly the same.
It is indeed “important for scientists to come from varied backgrounds.”
KJV Ephesians 1:3-23 may be more than two sentences (in Greek it was two) but still very complex and, to my ears, very beautiful, and not at all unnatural. I remember from my misspent youth that in the literary English of the later English Renaissance, this type of complexity was not uncommon.
I put a lot of this down to whatever is currently fashionable, rather than trying to say that it is something fundamental about the language.
I think that actual languages really do differ quite a bit in syntactic complexity, at least as it’s expressed in things like layers of embedding. That is not to imply that speakers of languages which don’t go much for that sort of thing don’t have perfectly good alternative ways of expressing the same concepts: they just go about it differently. Or that, just because a language does permit that sort of behaviour, speakers necessarily use that feature very much in practice.
It would be (moderately) interesting to see whether children acquiring Kusaal as a L1 go through a period where they wrongly put ka “and” in front of a clause without altering the form of the clause itself in any way, the way you can bung “and” in front of an English clause at will. (My money would be on, No, they don’t, but I could easily be quite wrong.)
I still remember my surprise on discovering this feature of Kusaal, which I hadn’t expected at all. Most of the relevant changes are tonal, too, so they don’t actually show up in written texts. It was quite a while before I was really convinced that my ears weren’t deceiving me. But Kusaal really loves syntactic embedding. Even quite ordinary informal Kusaal speech makes English seem like Pirahã by contrast.
Slightly apropos of this, I just discovered this excellent rant by Geoffrey Pullum about the persecution of Daniel Everett by the acolytes of the Mekon of MIT:
http://www.lel.ed.ac.uk/~gpullum/EverettOnPiraha_prelim.pdf
A Japanese child recorded a 20-clause chain at age three and ten months
I’m not convinced that this says meaningful at all about syntactic complexity. It means only that the child has learned exactly one morphosyntactic rule, viz, “end your verb with -te/de if you haven’t run out of stuff you want to say yet.” It’s the experimenters who interpret this as syntactic complexity.
It’s probably worth saying that the flexional morphology of modern Japanese verbs is pretty simple, too: I suspect this is rather less of an intellectual achievement than that of a Kusaasi child who at some stage will need to internalise that you say
M̀ nyɛ́ bʋ́ŋ lā.
“I saw the donkey”
but
Kà m̀ nyɛ̄ bʋ́ŋ lā.
“And I saw the donkey.”
or
M̀ nwɛ́’ bʋ́ŋ lā.
“I hit the donkey.”
but
Kà m̀ nwɛ́’ bʋ̀ŋ lā.
“And I hit the donkey.”
@George G I don’t understand why … would be all that different from …
Yeah. I fear that article has dumbed down the description to render it into English, such that the point has got lost.
Let me put the q in formal syntax fashion, as posed by programming language processors: we distinguish
* centre-embedding, in which there must be some sort of agreement/continuation between the parts preceding and following the centre (“The dog the cat the rat bit chased died.”) — @DE has layers of embedding above; vs
* iterative, in which there’s no hangover beyond some sort of interspersed connecting particle or cadence (“The rat bit the cat, which was chased by the dog, that died.”).
* The trouble (bloody Chomsky again) with phrase-structure grammars is they sweep both forms under generalised recursion. Programming languages much prefer the iterative form, (and indeed have special conventions to mark as such in their formal grammars) — for the simple reason that the discontinuous agreement makes for extra complexity and more half-finished productions to stack up and cross-match.
(wikip on centre-embedding has some examples in Japanese of “unprocessable” embeddings.)
The Kusaal examples seem at some sort of half-way point. They have a mere personal pronoun as subject. How would it go if that were a complex Noun Phrase? Would the speaker have to ‘remember’ across the NP to put a different tone on the verb? Having got past the verb, it looks like the clause-chaining is discharged(?) So no need to remember into the next clause(?)
I’m wondering if “subordinate” is misleading terminology here? “attached”? “continuation”? What is getting subordinated?
P.S. I see the Pullum article footnote 9 has plenty to say about “recursion”: “linguists’ use of it is a morass of confusion”
In my experience, it is quite natural for English-speaking children to form sentences like “I went to Central park, and I played on the swings and it was raining, but it stopped, and I saw my friend and we played tag, (breath), and we got ice cream and he got chocolate and I don’t like chocolate because it looks like poo and then I went home.”
But our budding young Kerouacs are taught from an early age to suppress this kind of sentence structure because speaking like this won’t help them get into Harvard.
Like DE, I salute the conclusion that it is “important for scientists to come from varied backgrounds,” and that one should look beyond English when studying syntax and speech development. But as a native speaker of a language that often shunts parts of the verb to the end of utterances of endless sentences with multiple embedded subclauses, I don’t see much of a difference in planning ahead between when I speak that language (you guessed it, German) and when I speak English. Just because English speakers are used to having the verb relatively early in their sentences, that doesn’t mean that speakers of German go “Oh, I need to put a pin in this, I want to use verb X at the end of the sentence”.
The clause chain languages I know are Turkic, and as DE stated for Japanese, the subclauses are actually grammatically simpler than the main clauses, because the verb is only marked for (relative) tense, aspect, and mood, not for person and number. The ordering of constituents takes some getting used to, but that’s due to speakers of European languages being used to a different order, not because it would be inherently more difficult.
@Ook: You are right that it’s perfectly possible to have endless run-on sentences in English and that it’s partly a question of stylistic fashion. But it’s my totally unscientific and personal impression that verb-final languages have a tendency to stylistically prefer longer sentences with more subclauses.
The researchers aren’t claiming that the languages that have clause chains are syntactically more complex than the ones that don’t. It’s just that complex sentences in such languages prefer or require clause chain constructions, and the study looks at how children acquire these.
The claim that children acquiring such languages do better than children acquiring English or French at the same age at clause combinations seems to be a response to the fact that the better studied languages regarding children’s ability to produce complex sentences are generally those without clause chains. They are saying that how English- and French-speaking children do in combining clauses into sentences at a given age cannot necessarily be generalized to how children acquire different languages.
Regarding the long chains produced by Japanese-speaking children, the relevant study says:
They also note that adult speakers never produce such extremely long chains to recount entire narratives.
So this has to do with a stage in language acquisition which is different from what you see in existing studies on the acquisition of better-studied, non-clause-chaining languages.
@AntC, we could mention here that programming paradigms that eschew iteration and have recursion as their preferred means of expression (often lumped as “functional programming languages) breath that implementations of these languages often special case your “iterative” case as “tail recursion”. I.e., that like Chomsky would have it, formally (“in theory” *) they are recursive and you can get very deep embeddings, but implementationally (“in practice” *) you save resources by not keeping the closures for all the spent function invocations around, or, equivalently, passing along to the tail call the continuation you were called with instead of constructing a new one (and then garbage collection does the rest innit). As a general rule, pace Stu, programmers are not much more stupid than real people.
(On the other hand, then people start using callbacks and creating their own call stacks/closure chains by calling the incoming callback from the outgoing one at a greater cost than missing a tail recursion optimization. So maybe they actually are. No language analogy to this one, I think).
_____________
(*) The difference between theory and practice being, of course, that in theory there is none.
@Lars, I was exclusively and explicitly talking about programming language syntax, not semantics nor run-time behaviour. (“language processers”, not program processors/”programming paradigms”.)
I deliberately and explicitly avoided the r-word. Linguists are not the only people who end up in “a morass of confusion”.
we could mention here …
No, please don’t. Your message muddles up two quite different things. “tail recursion” as an execution technique can be implemented in languages famously _not_ exhibiting phrase-structure grammars, such as Assembly languages, RPG, COBOL (as of last century).
The syntax (by which I mean syntactic specification) of ‘functional’ languages (such as Haskell) is not different in style from the syntactic specification of such famously not-functional languages as BCPL, Algol. All their language specs use “special conventions to mark [iteration vs centre-embedding] as such in their formal grammars”.
It is necessary to separate complexity of parsing an utterance (decoding) from complexity of the syntax used to express it (encoding). The former for me lies more in the length and (potential for) ambiguity, whereas the latter can include factors like grammatical irregularity or exceptionality. In Hans’ case of the final verb, the speaker only needs to keep track, and the listener usually knows what the verb is going to be, so both have minimal overhead,. You could arguably say English is more complex in cases where a passive construction has an identical form with an attributive adjectival one, e.g., “the seat was taken (away)”, where the German uses two different verbs.
If I am not wrong, the phenomenon referred to in the article is called ‘cosubordination’ in some circles. DE can probably enlighten us.
The Kusaal examples seem at some sort of half-way point. They have a mere personal pronoun as subject. How would it go if that were a complex Noun Phrase? Would the speaker have to ‘remember’ across the NP to put a different tone on the verb? Having got past the verb, it looks like the clause-chaining is discharged(?) So no need to remember into the next clause(?)
Actually, substituting a complex Noun Phrase for the pronoun wouldn’t change the tones of the verb or the object.
The trick with the Kusaal examples is that it’s the ones without the “and” which have a special tonal marking. Nyɛ̄ “see” is intrinsically Mid tone, whereas nwɛ̀’ “hit” is intrinsically Low tone; the sentences without “and” have a tone neutralisation to all-Low overlaid on the verb, which is the main marker of non-subordinate status: that’s how the sentences with “see” and “hit” end up tonally identical. This marker is absent in all subordinate clauses (except content clauses), and, surprisingly, also absent in clauses introduced by ka “and.” [Ka is not really “and”, and frequently introduces unequivocally subordinate clauses: I suspect that its use to mean coordinating “and” is actually an innovation, historically. Something similar has happened with Noun Phrase coordination, which uses nɛ, identical to the preposition “with”, to mean coordinating “and.”]
The Kusaal examples are then further obfuscated by tone sandhi: all words ending in MId tone, personal pronoun subjects, and nearly all full words which are not bound to the right, cause a following word-initial Low tone to change to High. However, intrinsically Low-tone verb perfectives which don’t carry the tone overlay are an exception, and don’t cause this change.
It’s all so simple …
Actually, I think there is a parallel with Japanese clause chaining etc: the subordinate forms are actually the less complex: in Japanese, this is because the verb lacks any inflection for tense or politeness etc, in Kusaal, because the verb lacks an additional tone overlay (and in narrative, also lacks tense marking), so there are fewer morphological elements present.
DE can probably enlighten us.
From me you expect enlightenment?
Actually, I’m not familiar with this term, which looks needlessly opaque anyway.
It is the case, though, that formal subordination (as manifest in internal clause structure) does not necessarily correlate with subordinate use. The whole topic of formally subordinate clauses being promoted to non-subordinate use is quite trendy: Nick Evans (the Australianist) invented the cute name “insubordination” for it.
https://www.benjamins.com/catalog/tsl.115
Once you have a handy name for the phenomenon, you start seeing it everywhere (the use of Kusaal ka-clauses in narrative certainly seems to fit; and Korean has developed a whole new politeness level, panmal, out of a “gerund” form.)
A short(ish) open-access account of it by Evans:
https://www.academia.edu/24481368/Insubordination_and_its_uses
As a general rule, pace Stu, programmers are not much more stupid than real people.
Did I really ever claim they were ? I must have had a bad hair day. On judicious consideration, I would say that programmers have more opportunities for stupidity than real people, because they have more tricky toys to play with.
In The Sorceror’s Apprentice, Mickey Mouse screwed up with only a broom and a water pail – plus a magic spell. He shouldn’t have dicked around with that spell. A lot of people never learn to spell properly.
as Ook kinda implied, i think there’s a certain kind of category-error going on here (not in the comment-chain). i’m not convinced “sentence” is a linguistically meaningful unit at all, and i’m damn sure that it’s not a meaningful unit of spoken english outside of some very specific sociolects, contexts, and registers. i’m not at all saying that the difference the article describes isn’t real, but i don’t think there’s a way to examine that difference in a meaningful way with the concept of “sentence” at its center (or maybe even present at all).
(this comment brought to you by anna deavere smith’s transcription of interviews using linebreaks and stanzas, and gertrude stein on sentences and paragraphs)
I wonder if it is possible to obtain recordings of pre-literate English speech (when the whole community is (a) illiterate (b) does not / rarely does listen to the radio, TV and so on (c) does not/rarely does communicate with literate communities).
i’m not convinced “sentence” is a linguistically meaningful unit at all, and i’m damn sure that it’s not a meaningful unit of spoken english …
Seconded. Let’s drive plenty more nails into Chomsky’s coffin.
OTOH … is <sentence> more of an abstraction? Speakers of conversational English would agree they’re somehow eliding thoughts that could be expressed as sentences in (scare-quotes) ‘careful’ speech. To take a non-English example from this thread:
The Kusaal examples are then further obfuscated by tone sandhi: …
So competent Kusaal speakers have an abstract understanding of how the utterance would ideally go without tone sandhi — even though they’ve possibly never heard such an utterance?
But of course nobody here is a reliable informant for “meaningful unit of spoken english”: we’ve all had our heads in books full of sentences for too many decades. We’re incorrigibly corrupted.
I prefer to think in terms of ‘clause’: an utterance or series of utterances each organised round a verb nucleus + participants. But then I’d have just as hard a time defining ‘verb (nucleus)’ as a “meaningful unit of spoken English”.
@antC
I would prefer to think of the sentence as having various optional/redundant components (even things like verb and subject) that can be left out by speakers where they expect this will not hinder communications. For example, I think that exchanges like
“Rain today.”
“Better tomorrow.”
“Wife OK?”
“The usual” (Russian :Normalno)
could be found also outside the sphere of Beckett plays and military bases, where they would be more frequent.
@PP, remins me this parody:
– Приборы?
– Приборы восемь.
– Что “восемь”?
– А что “приборы”?
(“instruments?” “Instruments eight” “what eight?” “and what instruments?”)
So competent Kusaal speakers have an abstract understanding of how the utterance would ideally go without tone sandhi — even though they’ve possibly never heard such an utterance?
In exactly the same sense that you have an “abstract understanding” of how you would ideally pronounce /p/ in “pin” identically to /p/ in “nip”, even though you have never done so or heard a native English speaker do so.
I’m talking grammar here. In my view, a grammar of a language is a set of theories to explain the patterns found in given specimens of speech, and to predict what patterns will occur in future (making the theories falsifiable.) A grammar is not a thing, and does not live inside a person’s head. (Except for Chomsky, who has stated outright that it’s a physical thing located in a speaker’s brain. But we need not linger over his philosophical confusions.)
What I have called “intrinsic” tones of Kusaal words occur often with nouns, including in the citation forms that speakers will give you if you ask “what is X in Kusaal?” Verbs are another matter, though there are fairly frequent contexts where you do get the “intrinsic” tones, e.g.
mán nyɛ̄ bʋ́ŋ lā
“when I saw the donkey”
mán nwɛ̀’ bʋ̀ŋ lā
“when I hit the donkey”
However, these almost certainly form a small minority of cases where the verbs actually appear in speech. There would be nothing logically inconsistent about an alternative theory in which the tones in this context were not the “intrinsic” tones of the verbs in question: the reason for not doing so is simply parsimony: you end up needing a whole lot more epicycles to make your grammar reliably predict what people will actually say.
But none of it is “real”, in the idiotic Chomsky sense.* There is no spoon.
In fact, the same applies, mutatis mutandis, to the term “sentence”, and even to the rather more useful term “clause.” As rozele implies, this is a theoretical construct; however it is certainly “linguistically meaningful”, because theoretical abstraction of this kind is what linguistics actually does.
* In what sense, if any, a theory is “real” is of course a perfectly respectable question. My (nominalist) answer is not the only reasonable one. But ANC shows no glimmering of actually understanding these issues.
Of course grammatical theories stand in some relation to contents of human heads.
On the question of verb intrinsic tone in Kusaal, it’s interesting that Buli, Konni and Nawdm, three of the four languages most closely related to Western Oti-Volta, actually don’t have lexical tone distinctions in verbs at all, and the description of the Western Oti-Volta language Farefare by Mary Kropp Dakubu (who was a very competent linguist) has even that language without lexical tone distinctions in verbs, though this does not seem to be the case for other Farefare dialects.
In the case of Kropp Dakubu’s Farefare, I suspect that the loss of verb lexical tone is a consequence of the very fact that verb tone distinctions were neutralised in main clauses (as in Kusaal), and this was generalised to subordinate clauses.
I’m not sure what’s going on with Buli, Konni and Nawdm. However, there are hints that the some Western Oti-Volta finite verb forms originated as repurposed nouns, so it’s possible that WOV tone distinctions in verbs have actually been redeveloped secondarily, and that the tone “neutralisation” in Kusaal main-clause verbs is actually a relic of the real verb tones. There are plenty of cross-linguistic parallels for subordinate clauses being organised around verbal nouns rather than finite verbs; it would mean that a language like Mooré, where main-clause verbs make the same tone distinctions as subordinate clauses, has generalised the subordinate clause pattern to main clauses. (“Insubordination”, in fact!)
[Don’t mind me, Hatters. I’m just happily free-associating here in my corner. Talk among yourselves …]
Don’t worry, 200 years later a Kusaasi girl wil be reading this comment of yours*.
Let us hope she’ll be satisfying her own curiousity and not writing a dissertation about you.
* P.S. and won’t read Chomsky:-E
I don’t mean no one will, of course.
A grammar is not a thing, and does not live inside a person’s head.
Absolutely. And I was making no such claim about the inside of Kusaal speakers’ heads. I was talking (very muddledly, I now realise) about a
speakerlistener’s ability to understand a sentence they’ve never heard before. (This isn’t some grand infiniteness-of-language metaphysics; but the prosaic hasn’t-in-fact heard before.)[Don’t mind me, Hatters. …
But I do try to pay attention and learn. I can only apologise that your valuable remarks often whizz several storeys over my head.)
Of course grammatical theories stand in some relation to contents of human heads
Sure: there must be physical processes going on in people’s brains which account* for the patterns we can observe in people’s speech.
But to identify our theories about those patterns with the physical processes is an error so egregious that only an intellect as powerful as Chomsky’s would be capable of it.
* In part: the interaction between people is central to the patterns we need to explain, and that does not reside in anyone’s brain. A private language is impossible, as St Ludwig has taught us.
you would ideally pronounce /p/ in “pin” identically to /p/ in “nip”
I don’t think you thought that example through, assuming “you” represents a speaker of standard English.
I can only apologise that your valuable remarks often whizz several storeys over my head
You are much too kind. What’s going on is rather that I haven’t explained what I mean properly (often because ! don’t actually understand it myself as well as I imagine I do.)
The question about verb intrinsic tones is actually quite interesting on both theoretical and practical levels. It is disquieting to find yourself positing underlying forms that don’t actually appear in Real Life very often.
In fact, in my Kusaal grammar, I unblushingly ascribe underlying forms to words which never appear in real life …
The most purist avoidance of this that I’ve seen in a descriptive grammar is Idelette Dugast’s grammar of Tunen (an extremely unusual Bantu language, remarkable among much else for being perhaps the only known Bantu language which is regularly SOV.)
Tunen words consistently drop underlying final vowels and undergo several tone neutralisations in isolation; Dugast is amazingly accurate in her transcription of words in context, but shows a near-pathological aversion to abstraction in her description, and never cites an individual word except as it appears in citation. It makes her works simultaneously deeply impressive and extremely annoying to read.
I don’t think you thought that example through
I should perhaps have clarified that by “ideally”, I meant “not a bit.” It was meant to be false (like the meaning I mistakenly ascribed to AntC’s comment about Kusaal tone sandhi.)
@David Eddyshaw: I read that chapter by Pullum about Everett’s work, and it was interesting in a sad way. Most (but by no means all) of the relevant linguistic information was already familiar, but the sordid details of the smear campaign against Everett were much less so. I did, however, note rather serious overstatement, apparently based on another misunderstanding that seems to be very widespread in linguistics today.
This issue was in the completely different area of phonology. Everett initially believed that there were two forms –sai, which differed through tone, but Pullum claimed that this had since been disproven. It may well be true that there are not two distinguishably different forms with different meanings, but the evidence Pullum cited was measurements of fundamental frequency—that there was no statistical difference between the two purported forms. This is certainly evidence against the existence of two forms that differ tonally, but it is not proof. Where linguistics seems to be in widespread error is in the assumption that fundamental measurements tell us everything there is to know about tone. It may be by far the most important quantity, but F0 is never the whole story.
I should perhaps have clarified that by “ideally”, I meant “not a bit.” It was meant to be false
Ah, I have once again been fooled by your Scoto-Welsh rhetorical tricksyness! I retire in defeat.
As rozele implies, this is a theoretical construct; however it is certainly “linguistically meaningful”, because theoretical abstraction of this kind is what linguistics actually does.
i’m all for theoretical abstraction as a tool of analysis, but it has to be grounded in what people actually do in/with language. i trust, for example, from his explanations of his work that DE’s “underlying forms” are built from close observation of kusaasi speech, through careful observation and comparison of pattern, change, and variation.
but i don’t think you can get the category of “sentence” (as it’s generally used, meaning what it’s generally understood to mean) out of spoken english that way. it’s not an abstraction of observed phenomena – like a “clause”, f’rinstance. it’s an imposed premise, many of whose supposed structures and constraints (no final prepositions!) are not just prescriptions but based on entirely other languages. to the extent that “sentence” is a theoretical abstraction in the field of linguistics, it seems to me that it’s in the way that phlogiston or epicycles are for astronomy – an excellent illustration of how not to do theory or abstraction.
I agree that “sentence” is not a very useful term for English grammar. CGEL actually only uses “clause”, as a deliberate decision.
The alternate-universe CGEL (the Quirk/Greenbaum/Leech/Svartvik one) doesn’t actually eschew the term “sentence” rigorously like CambridgeGEL, but it does say that it’s a pretty formless concept and doesn’t really do anything with it: the action is all about clauses there too.
Mind you, I think this to some extent reflects the unfortunate fact that the study of syntax at levels higher than clauses (or conjunctions of clauses) is still in a comparatively primitive state: not helped by the colossal wrong turn in much of linguistics induced by Chomsky. But other traditions have been much more helpful, like MAK Halliday’s Functional Grammar. And there has been work (in English, at any rate) on real spoken language in all its real-life messiness.
(Actually, I was just reading Knud Lambrecht’s Topic, Antitopic and Verb Agreement in Non-Standard French, which is all about actual spoken French and how profoundly different its syntax is from written or spoken-formal French.)
I’d say sentences are probably real in German, where the two types of clause distinguished by word order don’t occur in random sequences. English is another story.
I agree with DM.
They are more real, at least.