Why dictionaries don't supply meaning : Miller on communication

Aug 26 2010 Published by under Forget What You've Read!

At the moment, I am taking a (temporary) break from my furious critiquing of peer review, and have begun working busily on a new series about the workings of human languages.  Writing about this is for a general audience is hard, particularly because I suspect that many people have unexamined intuitive views about language that might be very different from the view I am trying to put forth. Additionally, if you're a linguist, an analytic philosopher, or a psychologist studying language, you will likely have a long-held world view that my writings may challenge.  (It's all rather intimidating, really...)

But in any case, one of the serious puzzles that I'll be piecing together in upcoming posts is how on earth we are able to communicate about the wonderful complexity of the world through a noisy, lo-fi channel (speech).  Some of the most important questions I'll be asking are : How do we understand what someone means through words?  How do we communicate meaningfully through words when we speak?  What is the relationship between words and the world?

To ease us all through the process, I've decided to begin with passages from a range of well known voices on this topic.  The questions and the ideas these thinkers pose motivate much of the work that we do in our lab (and they may motivate you to spend some time puzzling on your own).  Today's passage is written by George Miller (1951) and raises two intriguing questions :

First)  We think of words as having meanings.  But is the meaning of a word concrete and static?  Or is it dynamic and contextual?

Second)  Are words even the right linguistic unit to be looking at when it comes to measuring 'meaning'?  What about phonemes? syllables?  chunks?  sentences?  whole conversations?

Let's take a look at what Miller has to say on this:

"It is widely recognized and usually forgotten that modern dictionaries of English are based on usage rather than fiat.  The lexicographer collects many sentences by writers of literary or historic importance who have used a certain word.  He then proceeds to classify these sentences into groups that seem to use the word similarly.  For each group he writes a short phrase that can be substituted into the sentences in place of the word itself.  A dictionary definition is, therefore, an alternative verbalization derived from the contexts in which the word has been used previously.  A given word may have a variety of definitions inferred from the variety of contexts in which it appears.

Consider the word ‘take.’  In the Oxford English Dictionary there are 317 definitions of this word… This is a confusing state of affairs.  To find an analogy, suppose that ‘da-di-dit’ in the International Morse Code usually represented ‘D’ but under certain conditions might represent almost any other letter in the alphabet.  How could one tell which letter was intended?  A set of rules saying that ‘da-di-dit’ is ‘D’ unless used before one thing or after another, etc, would be required.  The decoding would get much more complicated.

Why do people tolerate such ambiguity?  The answer is that they don’t [1].  There is nothing ambiguous about ‘take’ as it is used in everyday speech.  The ambiguity appears only when we, quite arbitrarily, call isolated words the unit of meaning.  It is possible to draw up a dictionary using some other verbal unit – the syllable, say, instead of the word.  …Most people [would] object to such a syllable dictionary.  It is useless, they feel, because syllables standing alone have no meaning.  Or if they do have meanings, the meanings are so many and so ambiguous that it is impossible to know what any syllable means unless you know what other syllabels precede it and follow it.  Words standing alone have no meaning either or, more precisely, have no single meaning.  The dictionary definitions are derived from the contexts in which the words occur.

There are no meanings in the dictionary.  There are only equivalent verbalizations, other ways for saying almost the same thing.  There is a common belief that to define a word is to give its meaning.  It is healthier to say that by defining the word we substitute one verbal pattern for another." (Miller, 1951, p. 111-2)

Fascinating, yes?  There's developmental evidence, by the way, that young children don't initially distinguish between individual words as we do as adults.  The psychologist Elena Lieven has found, for example, that fully half of the first 400 identifiable multi-word utterances that 2-year olds produce are 'frozen' -- meaning the children have no working knowledge of the meanings of the individual words.  Lieven has also found that even with apparently 'non-frozen' multi-word expressions, children tend to be pretty conservative in their productions -- fully 75% of the 'non-frozen' utterances 2-year-olds produce only differ by the addition, subtraction or substitution of a single word.  These findings suggest that children aren't automatically segmenting speech into individual 'words' as we might expect that they would -- which drives home Miller's point that words aren't necessarily the right units to be inspecting in the first place, when we're looking at how we arrive at 'meaning.'

Even more unsettling: there is some thought that without a writing system to emphasize the distinctions, adults might not think of words as distinct 'entities' either. (And even with a writing system in place, we need only think about how often people misspell "a lot" as "alot" to realize that the question of where we should draw the boundaries between words isn't always clear [2]).  There's way more where that came from -- this, I think, is just a taste of what's to come!

Quick Note

[1]  This is the one statement that I take issue with.  I come from an 'information theoretic' perspective on language and think that that a certain level of ambiguity --called 'uncertainty' in the literature-- is a standard part of the linguistic 'signal.'  Also -- it's simply not the case that we tend to understand each other with 100% certainty all the time.  Just think -- if you're sitting in a lecture on a subject you know little about, even if you know all (or even most) of the words, there's still a good chance you're not going to be as sure about what's going on as the expert sitting next to you.  You're both hearing the same words, but you're not processing them in the same way.

[2]  Some linguists think that the unit of meaning depends on the language -- you can read about the proposed differences between 'fusional' and 'agglutinative' languages on Wiki and elsewhere.

Further Reading

Miller, G.A.  (1951) Language and Communication.  McGraw-Hill Book Company: New York. p. 111-112.  [Please note:  This book is available for free online at Questia]

Lieven, E., Pine, J., & Baldwin, G. (1997). Lexically-based learning and the development of grammar in early multi-word speech Journal of Child Language, 24 (1), 187-219

Lieven E, Behrens H, Speares J, & Tomasello M (2003). Early syntactic creativity: a usage-based approach. Journal of child language, 30 (2), 333-70 PMID: 12846301

Lieven, E., Salomo, D., & Tomasello, M. (2009). Two-year-old children’s production of multiword utterances: A usage-based analysis Cognitive Linguistics, 20 (3), 481-508

15 responses so far

  • Paul Murray says:

    I had a bit of a read of a translation of Beowulf, and one thing I noticed was the length of the sentences. I suppose that the original was unpunctuated - but the question is, how does writing without punctuation affect the way you *think*?

    • melodye says:

      It probably affects our ability to learn second and third languages as adults --you should look at Inbal Arnon's work on this.

      Found a reference, by the way : "Greek and Roman writing did not consistently separate words, and it was not until about the tenth century that the printer's pride in the legibility of his work led him to dignify word units by leaving space between them." --from Language and Communication, by G.Miller

  • What about if I've studied a little bit of linguistics, analytic philosophy, and psychology of language? Will I be in danger of having my mind blown completely? (English lit major with a joint minor in psychology and philosophy)

    Seriously, I'm looking forward to this series.

  • I totally buy the argument that words are not the basic unit of meaning in speech, even if they are in formal grammar.

    I look forward to the rest of the series!

  • Lieven's work (and much of what comes out of that tradition) is over-interpreted. Most words are used in relatively limited ways in adult speech as well, but that could be for any number of reasons (e.g., the world-is-a-boring-place hypothesis). I suggest you check out Charles Yang's discussion of the Tomasello research group's work: http://www.ling.upenn.edu/~ycharles/papers/zipfnew.pdf

    This isn't to say that the work coming out of Tomasello's group isn't interesting -- just that it probably doesn't mean what they think it does.

    • melodye says:

      I can't stand Charles Yang's work!! Oh my goodness. (No offense). I remember getting a review paper on a conference paper that suggested I didn't understand the possible scope of 'syntactic theories' and reading Yang and thinking : wait, yes I do, and they're still just as bad as I thought.

      There are many, many things that upset me about that paper (because I think it's theoretical garbage, generally), but this in particular:

      "Yet the Zipfian distribution of linguistic combinations... ensure that most “pairings of form and function” simply will never be heard, never mind stored, and those that do appear may do so with sufficiently low frequency such that no reliable storage and use is possible."

      Has no one read the literature on the other side of the data sparsity debate? In particular -- has no one read Dan Yarlett?

      Similarity-based generalization, my friends!

      • Hmm, but how would it work with Kayardild (http://www.ling.yale.edu/~erich/Home_files/Round%202009%20PhD%20Dissertation.pdf), where nominals have somewhere between 30K and 90K inflectional forms each.

      • melodye: I think you've missed perhaps *the* key aspect of the Yang critique: on every reasonable account of language, most words will appear in only a small number of constructions. That's even true of the most wild-eyed universal grammar theory you can dream up. So evidence that most words only appear in a small number of constructions isn't evidence for construction grammar. It isn't really evidence at all.

        The rest of that Zipf paper also tries to argue that you can, within those confines, argue about exactly what the distribution of constructions should be. Yang presents some pretty compelling evidence that UG + Zipf = existing distributions. Since Tomasello has never really produced a formal theory (according to Yang, and to the best of my knowledge), it's impossible to know for sure how well his account would do (Yang argues it would do very poorly), but in any case, at a first pass, it's not going to do better than UG + Zipf.

        What's sort of interesting about this from a history of science point of view, is part of the Chomskian revolution was realizing that while relatively simple theories can explain what is actually produced, you need much more complex theories to explain what *can* be produced. I'm not saying that a sufficiently complex construction grammar account couldn't pull this off (though I doubt it). I'm just reiterating that evidence that complex grammatical theories aren't necessary to explain routine production isn't evidence against those theories, since that's not the data they were designed to account for. It's a little like complaining Schrodinger's equation isn't necessary to explain classical mechanics. It's not -- but that's not the point.

  • arnsholt says:

    This does sound like an interesting series. It'll definitely be interesting to see where this goes.

    On the ambiguity thing, I think I agree. There is definitely ambiguity in language (quite a bit of it, in fact) but we filter out most of it without thinking too much about it. Also, it's often used intentionally for effect: wordplay, double entendres, and the like.

    On the fundamentalness of words I'm a bit more unsure. The syntactic theories I consider to be more or less viable (or at least less bad than Chomskian grammar), LFG and HPSG, do tend towards considering the word a fundamental entity. But to properly settle such a question we really do need a solid definition of "word". For your example of "a lot" vs. "alot", for example, I think it could be argued that in modern spoken English (I'm not a native speaker, mind you) it can be considered a single word (except not quite, since you can still say things like "a whole lot"), but due to written language being a conservative force, the written form can be quite different. You can probably find more egregious examples if you look for them.

  • For words, an interesting thing to think about is Ancient Mediterranean writing systems, which tended to use 'word dividers' quite consistently when writing on cheap media such as stone, but Ancient Greek (the later script, origin of ours, not Linear B) was an exception, being quite inconsistent, as discussed in Barry Powell's "Homer and the Origin of the Greek Alphabet".

  • Dexter Edge says:

    Could you give an example of a "frozen" multi-word utterance?

    Earlier this summer, a two-year-old girl who lives across the street from me went through a phase for a couple of weeks where every time she saw me, she'd yell: "Hey Dexter. I have an idea!"

    If I asked "What's your idea?," it would be pretty clear that there wasn't anything in particular. But since I'd asked her a question, she's say "Come here!" And when I went over and asked her again "What's your idea?," she'd say "Chocolate."

    Which always sounds like a good idea to me.

    I'm assuming that she had by this point some idea of the word "I" as a separate functional unit. But I'm guessing that "have an idea" might be an example of a "frozen" utterance?

  • Comment on note 2: most linguists would probably deny there was *any* single size of ''unit of meaning", in any language, so that for example English has meaningful affixes (word parts), like 'de-' in 'decontaminate', and multi-word combinations ranging from things like 'put ... down' in 'Bob put Maggie down' (one compositional and at least two idiomatic senses', depending on what kind of creature Maggie is) to whole sentences such as 'Houston, we have a problem', sometimes produced by annoying computer programs as a way of telling you that they are unable to do what they are supposed to do. I've got a story about how the simplest cases of things like this might work in my LFG08 paper; Andy Egan's 'Pretense for the Complete Idiom' (Nous) covers many harder aspects of the problem, in a more general way.

    Languages certainly differ in their strategies for packing things into 'words', but for this to be a meaningful issue we need to understand what we mean by 'word', which is not easy to figure out.

  • Jess says:

    1) Are you familiar with Systemic Functional Grammar?
    2) What do you think about the role of corpus linguistics in addressing questions of 'meaningful units' (i.e. using collocations etc)?
    3) You seem to be setting up a dualism here: I wouldn't ask questions about how words connect to the world - I'd ask questions about how members of a speech community, or community of practise, negotiate meanings through language. For me at least it all seems to fall apart when the focus is on individual production and utterances (especially when contrived), rather than on language as a social artefact and tool.
    4) (on a more personal note) Why in the world do you think that a linguist would have 'long-held worldviews' about language which you risk blowing apart with a few blog posts? We do research, read research, and daydream about crazy, mind-blowing, status-quo challenging, life-changing research just like people in any other discipline... and we are probably more aware than anyone else about the problems of 'folk linguistics' and issues pertaining to the huge, dynamic, and often counter-intuitive, gulf between spoken language and written language that come into play here.

    • Grep Agni says:

      On point (4):

      I don't think melodye is worried that anyone's head will explode from the sudden overload of insight contained in her blog post series. Rather, she is noting that there is a lot of controversy in this area and implicitly requesting that her readers consider her thoughts as an invitation to discussion rather than an attack on a particular theory or (god forbid) individual.

  • WOW just what I was looking for. Came here by searching for
    pools brisbane