August 13, 2004

Seeing language through a lexical lens

My recent post at AFOE on the German language reform, along with a discussion on Language Hat and It's Ablaut Time, has inspired me to do this post. My professional roots are in several highly heterodox schools of linguistics, and some of the solutions found in them seem relevant to the topic at hand.

David over at It's Abluat Time starts his post with this:

I used to think we could talk intelligently about the grammars of languages by starting with the assumption that grammars are designed for communication. The more I look at actual languages, the less I believe that this is the case. While languages obviously serve as media of communcation, they are in many ways ill-suited to this task. Grammars are too complex, too byzantine, too intricate, and indeed too beautiful, to be optimal codes for communcation.

I'm inclined instead to have a problem with the notion that grammars are designed at all. Like Language Hat, I've never been able to develop much enthusiasm for Esperanto, in part because of this very problem.

Since the rise of the structuralists, mainstream linguistics has not thought highly of artificial languages. The main reason for this is the belief that the true object of study for linguistics is the vernacular, and that any sort of role in managing language problems is a sort of unscientific prescriptivism. While I think this is no longer an appropriate attitude for linguistics, I too don't think very highly of Esperanto.

Esperantists talk at great length about how easy their language is to learn, but the simple truth is that it is very hard to learn and effectively impossible to use "correctly." Esperantists also have a long history of complaining about how poorly Esperantists speak Esperanto. But the reason Esperanto is doomed in its current form is because Esperantists believe that language is, in Steven Pinker's words, words and rules which are internalised by the speaker. I don't think that it is any such thing. A language is the set of things a community of speakers does to communicate linguistically.

It is that belief, and the claim that the above is not a circular definition, that places me far outside of the linguistics mainstream.

What this means is that English is defined not by a body of rules and a set of words, but as the protocol English speakers use to communicate when they believe they are speaking English. This shifts the definition of English to a definition first of the English-speaking community, and second an explanation of why they identify some communicative acts as speaking English and other communicative acts otherwise. This makes language not a property of individuals but a communal property. It sets the boundaries of what is and isn't English, and what is and isn't language, where it belongs: in the field of socially constructed categories.

Clearly, the individual speaker of a language possesses some internalised knowledge of something. This redefinition does not eliminate the need to identify what this knowledge is. I think that the information a language's speakers internalise is quasi-lexical in nature, and should not be understood as a seperate lexicon and a body of rules.

A number of linguists over the years have taken this approach. It is not, by itself, an especially radical thesis, although it isn't mainstream in the anglophone world. Lexically driven dependency syntax dates back a long time, and elements of this approach are found in Lucien Tesnière's work in the 50s, in the Russian lexicalist school associated with Žolkovsky, Mel'čuk and Apresjan (which has a branch office in Montreal where I got it from), with Dick Hudson's Word Grammar, and with elements of Construction Grammar and the Optimality Theory school of syntax. In more recent years, it's gotten a big boost from Rens Bod's Data-Oriented Prcessing model. It also lies at the root of the work I'm doing in statistical and example-based translation tools.

The essential idea is that humans store substructures of parse trees instead of words and rules. Parts of these structures are, in some sense, abstracted. That way, a person may store, for example, a verb and the elements of its argument structure, including possibly information about the classes of the words that can act as arguments for it. Alternatively, they might learn an adjective and the set of adverbs that most commonly modify it. Actual linguistic productions - for historical reasons the word texts is usually used to identify them in the abstract - are seen as Tinkertoy-like assemblages of ready-made structures. Judgements of grammaticality revolve around the degree to which the linkage points in this structure fit together.

This does away with the competence/performance distinction. And good riddance. I leave it as an exercise for the student to show that the generative capacity of such a formalisation is at least as large as that of traditional formal grammar theories.

For my linguistically trained readers, this is a very brief summary, designed to make a slightly off-topic point. I am aware of the problems that this theory involves, particularly with respect to word order and non-projective syntax trees, as well as to the representation of meaning. I have, in fact, an alternative formalism to parse trees which I think resolves this problem and is motivated by elements of computing theory and neurology. It may actually have some element of physiological meaning. It also comes with a theory of language acquisition, lexical representation and of linguistic meaning that moves an awfully long way away from the contents of Linguistics 101. But, talking about that involves a serious bit of math, a small number of commercial secrets, and some truly half-baked notions that are - as far as I can tell - completely off the chart for linguistics. It has the side effect of turning what started out as a fairly neat and simple linguistic theory into something else entirely. I said I was out there. You have been warned.

At any rate, this theory has huge advantages in accounting for lexical selection - for example, why what is generally a heavy rain in English is une forte pluie (lit: a strong rain) in French. It also accounts more effectively for irregular verbs (and irregular declensions), and particularly for the empirical observation that the highest frequency lexemes seem to be the most irregular. It enables us to identify how a child can build on a very limited lexical knowledge - Quine's gavagai problem, for example - by learning the combinatorial properties of words first, then learning the real world categories that those combinatorial properties refer to.

The most radical element of this theory is that it denies the existence of grammar as such. Instead, it proposes a library of templates, possessed of varying levels of abstraction. This way, we can account for, quoting It's Abluat Time, "the convoluted tone rules of some Bantu languages, or nominalizations of Kuki-Chin languages that express with a complicated systems of ablaut, consonant mutation, and tone alternations what other languages express as well with a simple affix." Each forms a part of a template, and acquiring a template is far easier than using a rule. It does an even better job of explaining linguistic gender - a feature that appears completely unnecessary in most languages. In a language like French, where gender-specific articles almost always must appear whenever there is a noun, the template that compels noun X to select article Y is naturally acquired instead of a naked word with arbitrary properties.

Some sort of version of this thesis - that linguistic knowledge is at least in part an archive of templates rather than a set of words and rules - seems unavoidable if you want to account for code switching. Multilingual people mix languages up, and do so in controlled ways. It is difficult to make this fit into a "words and rules" model of language, while it can be accounted for effortlessly in a template driven model. We can account for native bilingualism by thinking that people who grow up bilingual learn that some lexical templates select words from one language and not the other. We can also account for domain-specific and register specific language using this model.

This also correlates to most second language learners experiences - or at least mine. At first, you are presented with a set of rules, and any time you have to think about those rules, you speak poorly, incorrectly or very slowly. Later, the things you want to say just seem to be on the tip of your tongue. It just flows, naturally, and the rule is never consciously accessed. I am claiming that it's because you have not internalised a rule, you've internalised a quasi-lexical structure which you access as necessary and in the same way that you access words.

Some people would claim that that amounts to the same thing, but in terms of linguistic theory it makes a big difference. It suggests that there is no distinction between rules and words, or regular and irregular forms. Regularity merely implies the internalisation of a more abstracted structure than irregularity, with no clear dividing line to separate them into two classes.

I think that this theory better explains the "unnecessary" features of vernacular languages, and why they persist when they don't represent - and perhaps never did represent - any meaningful distinctions. David hypothesises that: "These formal filigrees, I would argue, are not there to serve some communicative function, but simply because speakers of languages assume that any pattern they can detect in their language is an essential part of the linguistic code which, if not maintained in their own speech would lead to them being outed as the language bluffers that they are." My hypothesis matches this fairly well, although it need not be so much fear of community sanction that motivates it. Speech patterns are simply reinforced by regular community and personal use, and people readily absorb the patterns that together best account for the speech patterns in their linguistic community and enable them to communicate effectively.

This is not a theory of linguistic minimalism. These inferences need not be "mistaken" at all. One of my theories is that more specific templates are easier to access, to use, and to recognise. I have some grounds in computing theory for this hypothesis, but the full reasoning is beyond my present scope. If this is true, it means that retaining the complexities of a particular structure - even if superficially unnecessary - can make it easier to speak fluently and to understand effortlessly while only slightly hampering acquisition, in contradiction to countless millenia of common sense.

This also touches on Salikoko Mufwene's claim that language change can be motivated by demographics. Rather than supposing that each new generation acquires language incompetely, we can see language as goal-oriented behaviour: modelling community speech patterns in order to communicate. Exactly what lexical templates people acquire may vary because no two people will be confronted with identical speaking communities. The construction of a creole, as Mufwene hypothesises, is then no different from other kinds of language change. It is simply a matter of speakers attempting to model local linguistic behaviour.

However, unlike Mufwene, I don't see language as a property of individuals. Language is a property of a community, a lexicon is a property of individuals. A person's lexicon may be multilingual, with each language that the individual uses defined in social terms rather than individualistic ones.

We might also identify as a cause of progressive language change the different models people develop for the speech patterns in their community. The internalisation of different lexical templates can generate identical or near-identical speech. Thus, instead of seeing language change as a matter of inaccurate language acquisition, we can simply recognise that each person has a different library of templates and that that leads to different utterances. As long as communication remains reasonably accessible, this poses no problems. However, under these circumstances, even a very isolated community must, inevitably, see its language change.

As significant a theoretical claim as this anti-grammatical theory is, the impact on practical and prescriptive language trades is even more significant. My current work focuses on using computers to assist translators to find correct translations for substructures of sentences rather than just individual words. It also has an impact on language education, suggesting that teaching lexical information in context is more important than teaching rules and vocabulary lists.

Lastly, for the language manager, it implies that no amount of force can make a language fit a grammar. Real world grammar is driven by lexical selection. This touches on the topic at AFOE.

And this is why I am very critical of Esperanto. Language is not about rules. Language is about community, and language acquisition is about internalising lexical templates. The regularisation of rules is not particularly helpful in either of these areas. I suspect that full mastery of Esperanto might well be harder to acquire than a natural language rather than easier because, first, the Esperantist community is too dispersed to provide a real community, and second, because for all its clarity about morphology and word order, it is utterly ambiguous about lexical selection.

I do, however, think that there is something to the notion of a planned international vehicular (not auxiliary!) language, but it must be constructed with an accessible user community in mind, and it must spend less effort on regularisation and more on lexicography. There are some real world examples of this that have actually succeeded, and there is one "constructed" international language that, while still inadequate by my standards, has at least made some movement in this direction. But that's for another post.

Posted 2004/08/13 10:25 (Fri) | TrackBack

"A language is the set of things a community of speakers does to communicate linguistically....This shifts the definition of English to a definition first of the English-speaking community, and second an explanation of why they identify some communicative acts as speaking English and other communicative acts otherwise. This makes language not a property of individuals but a communal property....Actual linguistic productions--for historical reasons the word texts is usually used to identify them in the abstract--are seen as Tinkertoy-like assemblages of ready-made structures....Language is about community, and language acquisition is about internalising lexical templates."

Scott, you absolutely must read some Herder. Read his Treatise on the Origin of Language, of course, but especially read a couple of articles by Michael Forster which provide a close and thoughtful consideration of Herder's whole philosophy of language ("Herder's Philosophy of Language, Interpretation, and Translation," in Review of Metaphysics 56 (December 2002); and "Some Problem Cases in Herder's Philosophy of Language," in Inquiry 46 (March 2003)). (You can also find some of Forster's arguments in his introduction to his Herder translations (Cambridge, 2002), or in his online article on Herder here.) Very basically, Herder's conception of language includes the idea that the meaning reached through lingusitic communication is a function not of any rule-bound property of the words learned, but of the collective act of expression achieved through an individual speaker's immersion in and recognition of definite structures or templates inherent to the communal context itself. In other words, he also embraced a superficially circular understanding of language, one which really isn't circular at all once you grant the role which an instinctive Denkart plays in the active learning and spreading of language. Not that you're a Herderian and just don't know it; your grasp of lingusitics is a lot more sophisticated than his (or mine, for that matter), and you very likely would have philosophical or moral objections to the way Herder connects those templates or structures to a rather metaphysical philosophy of mind. But leaving aside the ontological differences, I think you'd find some fascinating ideas and parallel's in Herder's writings.

Of course, for all I know you're already thoroughly familiar with Herder and see all sorts of differences I'm not noticing. In any case, your post excites and interests me; thanks for writing it.

Posted by: Russell Arben Fox at August 13, 2004 19:54

Why is it necessary to risk the accusation of circularity? Why not say simply that internalization (pick another term) can not be achieved by artificial means? And why isolate language? Why not say that language is part of a network of cultural and social tools/cues none of which exist in isolation either spacially or more importantly, temporally. Language is a 'thing' but it is also a description of -the way things- change. If you say that tine is a necessary function of language creation/learning, you avoid the problem. But then you end up with a philosoophical argument beginning from an idea of 'depth' and this is supposed to be a logical no-no of some sort. I'm sure someone has argued that we are connoisseurs of language (a connoisseur being someone who recognizes something before he can say why) but I wrote a couple of posts on this recently.

Posted by: seth edenbaum at August 14, 2004 3:11

I responded before I realized there was a 'fold.' Oy.
Now I'm reading the whole thing.
This may take a while

Posted by: seth edenbaum at August 14, 2004 3:33

I searched for Quine+gavai with no luck; Google suggested Quine+gavagai, which did the trick. On one level it is funny to say that a word of unknown meaning in a hypothetical language can be misspelled!

Posted by: Anton Sherwood at August 14, 2004 7:47

Russell, I am aware of Herder, but I don't know anything in depth about him. Thanks for the tip, I'll have to dig in to him now.

Seth - s'okay.

Anton - Thanks, I made the change. It's a rarely used word, half remembered. I thought I remembered it right. And in answer to your question on AFOE, there is a common suffix in Russian "-ogo" that gets pronounced "-ovo". After the revolution, many people in Russia were taught to read quickly, and since Russian is generally very phonemically written, they didn't do much more than teach them how to sound words out. So, millions of peasants saw the "-ogo" ending and assumed that that was how they ought to pronounce it. Being peasants, they figured that they had previously been speaking incorrectly. It took a number of years to put an end to that.

Posted by: Scott Martens at August 14, 2004 10:07
Being peasants, they figured that they had previously been speaking incorrectly. I took a number of years to put an end to that.

Wow, how impressive is _your_ CV :-)

Personal anecdote, that makes me viscerally agree with you on the "templates" model; I did grasp the rules in French for ordering pronominal direct and indirect objects for about five minutes, in total. I forgot them, and couldn't tell you them now. I didn't internalise them. Yet I still use indirect and indirect pronouns, and properly, and it can't be thanks to internalising this table without noticing it, because, like I said, I forgot it about five minutes after getting it into my head.

Posted by: Aidan Kehoe at August 14, 2004 15:23

"It" - so I missed a letter. I'm a crappy typist. (Actually, that one's embarrassing enough that I fixed it.)

One of the great things about Canadian French is that "lui" is pronounced exactly the same as "y", and sometimes gets used in lieu of "leur". This eliminated one more pronoun in the table.

I'm the same way. I couldn't tell you the required order any more either, I just do it. I used to be the same with German cases too, but there's too much Dutch under the bridge for me to still be able to.

Posted by: Scott Martens at August 14, 2004 16:28

I've read the post new a few times and my comments haven't changed. And actually this goes back to almost every conversation we've had Scott.
Both of us try to reconstruct a moral and intellectual defense of the collective, not by way of idealist programs as to why we should act collectively, but by describing how we already do so. But within that, we have different biases.

I'm less interested in who 'owns' language than in describing it in terms of process and time. It may be quibbliing but 'template' denotes a sort of objecthood I try to find ways to avoid such terms. Time and process have no owner.

In terms of legal debate for example, when I listen an argument between an originalist and a more liberal opponent I'm always amused that the latter doesn't just refer to the discussion the two are having as confirmation of his points. "How does law change? What are we doing now?" Law is simply an argument in time, language a set of conventions for change. Without time, there's not much going on in either.

Similarly in terms of the arts -and here of course we are in my field- I'm fascinated by the way people in the arts create patterns without knowing what they mean. Most writers concentrate on writing well and the personal aspects, interests, themes, obsessions, are made evident only as a result. For more 'intellectual' authors it's the reverse. I could go off on my usual anti-SF kick here, but that's not the point. I'm more interested in the fabricating or 'making' intelligence than in the speculative one, and it is the tendency of the speculative intelligence to construct rules and meanings and to argue that such constructions are as complex and profound as those we observe. I spend a good part of my waking life objecting to this.

The most profound orders, language and culture, are made, and the rules for them are constructed only after the fact.
I couldn't agree more.

Another argument against

Posted by: seth edenbaum at August 15, 2004 21:04

Perhaps this link is more in line w/ the previous post, but it is grenade-range close on this one, too. Not much here, and what seems a little banal to me, but I can't profess to have thought too much about it, and very likely might be missing out on loads of toothsome nuances.

Posted by: Brad at August 17, 2004 1:14

My favorite sentence from a grammar is from Kibler's Old French Grammar: "Old French doesn't have rules, but tendencies". Page 3, actually a paraphrase. This is true of most non-standardized languages.

Ivan Illych has written a lot of stuff about the way that grammars, dictionaries, and schooling have taught people that they don't really know their own native language. "Vernacular Values" is probably the best.

Posted by: zizka at August 17, 2004 6:25

Seth, I'm not sure we're that far apart. I am making a technical argument in linguistics, one about how linguistic structures are organised and acquired. I'm more inclined to see that particular argument as to some extent complimentary to issues of collectivism, and to some extent beside the point.

It's true that this notion of a lexically driven conception of language does reify the notion of a template. I have a different version of this thesis in which the idea of an abstracted dependency structure tree is replaced by a rather perverse mathematical construct, which I show is plausibly more natural to a computer with certain structural properties widely believed also to be possessed by animals with extensive nervous systems.

I have also skipped the elements of the theory of quasi-referential semantics that comes with this, which touches a litte more on what you're saying. It is one part Barsalou and two parts Vygotsky. The biggest fight I ever had at work was with a co-worker over Montague semantics.

In one sense, what I am saying is that there is no language, in the sense that there is no systematic body of knowledge which corresponds to language in the sense that people use that word. However, people persist in using the word language as if it designates a real object, and they do so because it is useful for them to do so. My philosophical biases tell me that all objects are constructed and justified on precisely the basis of whether or not people can do useful things with them. People can clearly do useful things with the concept "language", so I'm prepared to offer it status as as real a thing as any other. But, in order to make that work within my linguistics, I have to situate language as a thing defined by a community of speakers and not within the head of the individual user.

Of course, my own reified categories are subject to the same criticism. The difference is that I am perfectly willing to say that my language science categories are no less constructed, and I believe that they can be validated by proving themselves more useful than the alternatives. I take for granted that they are inadequate. I assert that inadequate categories are necessary. I reject the scientistic notion that equates progress with the development of more adequate categories, but I do equate progress with fitting categories to the problems their users have to face, and to changing categories when the old ones no longer fit the circumstances.

This is not so much about the ownership of language. I am inclined to place that in the realm of sociology and politics. Rather, I see the actual linguistic acts of people as grounded in the internalisation of quasi-lexical objects, and that new objects can be acquired - and old ones lost - as people and contexts evolve. I want to de-emphasise intergenerational language change, although I don't think it can be totally abandonned, and talk about how language can change continuously over one or more lifetimes. I think that this is necessary to account for cases in contact linguistics.

Brad, I don't see any way to align a dialectical philosophy with a belief in the eternal and unchanging nature of English. If there is one lesson to take from Hegel, it's that all things come to an end - even the notion of English as a single linguistic community. But I won't make any predictions about the future of English.

English may progress and expand its hegemony. It may be with us for another millenia or more. Or, the signs I see of a developing backlash may signify that the end is quite near, in perhaps no more than another generation or two. Who knows?

BTW, belated felicitations to you and Katrin. Give her a smooch from me. :^)

Zizka, yes, Illich is great on this subject. You don't see many willing to stand so fully for the idea that the peasants can decide for themselves. My hope is for a sort of prescriptivism that resists telling people that they are using language wrong, but that doesn't hesitate to tell them how their language situates them in the social structure around them. I want to replace prescriptive urges - and anti-prescriptive reactions - with a critical normativity. I want to acknowledge that people do - and must - sometimes change the way they use language in order to meet certain goals. But I want them to be aware that that is what they are doing, and to make judgements about whether the goals merit the efforts.

Posted by: Scott Martens at August 17, 2004 17:39

Scott: Many thanks from K. and me. I'll let you know when I'll be running around the Philosophy dept. in Leuven next month. Perhaps we can catch drinks together.

As for Hegel ... I might have a few quibbles w/ your (apparent) evolutionary reading of his dialectic. Things never so much 'end' as they are always in the process of becoming what they never realized they really were all along.

Posted by: Brad at August 18, 2004 10:21

Illych also has a lot on how prescriptivism has a policing effect even within the elite. He really argues that the modern mind was formed through enforced literacy, starting with monks studying scripture and scholasticism.

There has always been a counter-literacy too, in the form of pulp fiction type stuff. The "rise of the vernacular" meant that people able to write in Latin wrote in the vernacular instead. But there probably had always been a vernacular written literature, but it wasn't proper -- stories of adventures, monsters, tricksters, magic, adultery, etc. And it was an elite (courtly) literature too, but not monkishly orthodox or holy.

Posted by: Zizka at August 19, 2004 2:08

Illich has adapted, in his sort of Marxist-Catholic framework, the notion that the system restrains freedom of the elite to change things almost as much as it restrains the non-elite. It's one of the reasons he's always had a bigger following among the Latin American petite-bourgeoisie than anywhere else. I ought to reread him on language though - it's been a number of years.

As for the vernacular, its rise is a bit analogous to the rise of a capitalist class. The early capitalists tended to come from the already empowered nobility, and found that trade made them far richer and more powerful than being land barons. In the same way, vernacular literature was written by the people already acquainted with the classical language, who soon found that there was more fame and fortune to be had in writting tawdry romances and war stories that appealed to the lower urges of the literate class than writing classical tomes. At some point, the balance of power shifts, and trader class becomes dominant, religating the traditional order to a ceremonial role if not abolishing it outright. This is often simultaneous with the institutionalisation of the vernacular language.

Posted by: Scott Martens at August 19, 2004 10:01

"In one sense, what I am saying is that there is no language, in the sense that there is no systematic body of knowledge which corresponds to language in the sense that people use that word."

Is a novel a systematic body of language? I would say yes, but it is a system that is always one step removed from our ability to define it.
My interests come from my desire to defend the makers of things from the intellectuals who make their living trying to understand those things such people make. My parents, high modernist intellectuals and bourgeois leftists, had as much contempt for artists as they had respect for art. It was a thing to be interpreted and understood. The older it was, the further removed, the better. My mother often claimed to wish art grew on trees. I think there's something Talmudic in this, and she's not even the Jewish half of the pair.

But in response to criticism from rationalists, those in the arts who choose to defend themselves have often (always) ended up siding with hippies and theologians, and their arguments have always ended in illogic and faith. I have no interest in either, and yet I've been stuck trying to defend the same things.

But I've realized that there is another way to argue this point, and it parallels your argument about language. If we treat religion, for example, not as a series of truths but as a system of debate, than the interest in each individual is not in truth itself but skill: in argument. The topics and the rules are given, but the truths are not.
Legal debate is from the same mold, though both priests and some legal philosophers go to great pains to pretend otherwise.

What I am constructing therefore is a defense of skill as opposed to an ideal of truth. And I am trying to counter the arrogance of those who argue for scientism without making what both they and I would consider spurious arguments for faith. Your point Scott, seems to be that the genius of language is that in its complexity, it is irreducible, that its creation can not be mimiced in the abstract. Analytic philosophers et al. have a lousy time with this sort of thing. I'm still amazed when I read about Donald Davidson's rejection of conceptual schemes. So Pushkin is the same in English as in Russian? I really doubt it.

Regardless of my bias [is it anti-intellectual or merely anti-scientific?] I think there's and argument to be made, joining the arts, religion and law- but excluding faith- that says that systems of argument are better at producing things that have value as 'truth' than are many projects that have value as science. And I think you argument that language as idea, as unnamed 'thing', or as process, trumps language as 'rule,' fits well with this.

Posted by: seth edenbaum at August 20, 2004 2:34

I wish I had seen this post before yesterday. Very interesting, Scott. I have just posted the first part of a response, of sorts, over at It's Ablaut Time.

Posted by: David Mortensen at August 20, 2004 22:12
Post a comment

Remember personal info?