How…does this even work?

Here is a curious fact.

My wife Cathy is using Duolingo to learn German; she wants to be able to read sources on Iron Age and Viking costume in the original.

Duolingo takes her through a lot of pronunciation drills.

I’ve learned something by listening to her – which is that somehow, somewhere, I have internalized a very precise understanding of German phonology and phonotactics. As in, I not only know right pronunciation from wrong, I give her detailed advice on how to match Duolingo’s model speaker that we can both tell is correct.

What makes this weird is that I don’t speak German. At all. Nor have I ever lived where it’s spoken; I’ve visited Germany once, German-speaking Switzerland once, and that’s it.

This raises questions in my mind:

1. How the fuck? I mean, I suppose it’s related to my knack for generating names in the style of any specified language, and I could handwave about Markov-chain models, but…how the fuck?

2. What dialect of German have I templated on? Could there be any way to tell?

3. What other entire language phonologies have I swallowed … without … me … actually … noticing …

4. Does this happen to other people?

The human brain is a very odd thing.

79 comments

  1. I’m a native german; If you can link to a speech examle somewere, i can tell you the dialect.

    1. >I’m a native german; If you can link to a speech examle somewere, i can tell you the dialect.

      There’s a problem. What you’d have to do to identify the dialect I’ve locked on to is get me to speak German. Which I don’t. A linguist would say that I have specialized recognition knowledge, but not production knowledge.

      What is the German equivalent of “standard American”? That is, is there a regional variety that is generally considered neutral and correct and tends to dominate TV newscasting, movies, etc? That is most likely what I have absorbed.

      In the U.S., “neutral” dialect is spoken in a region which begins where I live on the Atlantic Coast, runs west through Pennsylvania, then through Ohio and Indiana and into the upper Midwest. Philadelphia, Pittsburgh, Chicago, and Minneapolis are the major cities. However I should also note that dialect variation in American English is much narrower than in German; only a very few extreme Southern dialects vary from “neutral” as much as any random two German dialects do from each other.

  2. My brother lives and works in Germany and I actually have visited Germany. It’s a softer language than I initially thought. But then I visited South Germany, more precisely Baden-Württemberg, somewhere near Stuttgart, where I think the accent is softer and more influenced by neighbouring Switzerland and possibly Austria, than North Germany.

    Like my native tongue Tamil, German also has a lot of complex words, formed out of joining several words together.

  3. From a friend of my brother, I believe many Germans in this region can speak both Swabian German as well as standard German well, but north Germans have trouble with this dialect and accent. They are justly proud of their local accents and dialects.

  4. I give her detailed advice on how to match Duolingo’s model speaker that we can both tell is correct. What makes this weird is that I don’t speak German.

    On the evidence given, I would guess that you don’t have a model of German mysteriously kicking around in your head, but that you’re simply really listening to the Duolingo audio in a way most people aren’t trained to. That is, you’re hearing distinctions that Cathy’s brain is trying to ignore because they’re linguistically irrelevant in English. I would expect that sort of ability to correlate with a keen ear for music, which I gather you have.

    I can do similar things — but I also have linguistic training, so it feels a lot less mysterious!

    1. >On the evidence given, I would guess that you don’t have a model of German mysteriously kicking around in your head, but that you’re simply really listening to the Duolingo audio in a way most people aren’t trained to.

      This is very possible. Your guess that I have a keen ear for music is correct, and I have noted before on this blog that I’m good at hearing phonetic distinctions that don’t exist in any language I speak.

      There are a couple of reasons it seems like I have a model of German, though. The biggest one is that I am aware that I would not pronounce German exactly the way the model speaker does, but that Cathy should. The model speaker is female, and German is one of many languages where female pronunciation slightly differs from male – consonants are slightly softer and points of articulation tend to shift towards the front of the mouth. As I am sure you know, many dialects of English have this pattern too, though it’s not usually strongly expressed in standard American.

      (Some years ago I wrote about hearing Liv Tyler pronounce Elvish in the first Rings movie and realizing that (a) she was doing it slightly wrong, and (b) I knew why her speech coach wasn’t fixing it. She was female-izing Sindarin a bit too much. Could you hear this?)

      Another is that I don’t have to be able to hear the Duolingo speaker to correct Cathy’s pronunciation. A few days ago, while the tutor wasn’t running, I taught her how to pronounce the medial sound in “schön”.

      So at minimum not just sound-matching in real time is involved; it’s more complicated than that, something about the way I build representations in auditory memory.

  5. Schaffe Butze Sbare: des isch a Schwob.
    Wennst oiwei im Biagartn hockst und vom Weissbia gsaffa biss, des is Boarisch.
    Ois anderes, des san dia Sau Preissn.

    From a Hungarian fan of German dialects. Germans do not exist. There are only Bavarians, Shwabians and Prussians. That’s the essence of what I’ve learned while working with Germans for more than 15 years. :-)

  6. The dialect-neutral version of German is indeed called Standard German, and it’s clearly dominant in national-wide communications, so that’s probably your dialect.

    I think the large diversity of German dialect comes from its history; in the medieval ages, Germany was fragmented into many more or less independent smaller territories, there was much less travel for the masses, and of course no national tv or radio, so that the language could easily develop regional characteristics. It was much more a continuum than today. As someone from northern germany, i have a soft spot for Lower German, which has a common ancestor with english. I was very amused as i once heard a lower german version of Harry Potter, and it sounded much more like the english orginal as the standard german version. But enough about Germany!

    I wonder if the effectivity of the Duolingo method comes from its similarity to the language learning process of small childs or even babies. I mean, they too are exposed to large quantities of correct pronounced, but (for them) basically senseless speech. But maybe this ›subconscious‹ exposure gives one
    a basic unterstanding of the ›feeling‹ of a foreign language and hence a better base for the following ›real‹ lerning of it.

    1. >The dialect-neutral version of German is indeed called Standard German, and it’s clearly dominant in national-wide communications, so that’s probably your dialect.

      Ah, the article answers my question. I would hear this natively spoken in Hanover. And, I’m guessing, elsewhere near it, on the margins of the North German plain.

      >I was very amused as i once heard a lower german version of Harry Potter, and it sounded much more like the english orginal as the standard german version.

      Indeed. I once discovered by accident that I can read Old Low German pretty well, when I stumbled across a collection of folktales in it. Thus I can confirm that it seems much more like English to an English-speaker, too.

      >I wonder if the effectivity of the Duolingo method comes from its similarity to the language learning process of small childs or even babies.

      I am pretty certain your speculation is correct. It matches my experience of learning other languages.

  7. >What is the German equivalent of “standard American”?

    I would compare it to the British Received Pronounciation – it is not as much based on the evolved common dialect of a region but to a large extent constructed by academics or emerged from the usage of highly educated people. The most visible aspect of the constructedness is the Germanization of Latin terms, like translating television to far-seer (Fernseher).

    Another aspect is that basically the major, maybe one can say only virtue of German is being really precise, because otherwise it is not particularly an aesthetical or musical language. For example “average” is called cut-through, Durchschnitt, because the average of a function is the line cutting through in the middle when drawn as a graph. This sort of thing just must be constructed.

    You know, the usual top-down approach of Continental Europe, not a bottom-up distributed evolution. Of course in the past you could point aspects of distributed evolution: Luther’s Bible, Goethe’s literature, Prussian military command and governmental administration (Beamter) language…

    BTW I think that the legend that that “High German” (i.e. standard) comes from being spoken in the Alps highlands is bullshit – actually the higher you go in the mountains, the weirder it gets, even Bavarian is very un-standard (Peter could be pronounced as Beda) but Vorarlberger and Swiss, so the Allemanish dialects are practically a different language. Of course the plains near the northern coast speak a very different dialect too. I guess the place where the natural common dialect is closest to the standard is somewhere in the middle, say, Frankfurt.

  8. I agree with Dan C; the same way as some people have perfect pitch, and some don’t, some people can pick up differences others simply ignore. If you have a developed musical ability, it is likely you are just better at listening to different sounds.

  9. >>I’m a native german; If you can link to a speech examle somewere, i can tell you the dialect.

    > There’s a problem. What you’d have to do to identify the dialect I’ve locked on to is get me to speak German. Which I don’t.

    So find some standard passages, read them, and record the audio. Or can you not pronounce written German either? That isn’t too hard to learn, but hm, the process of learning might well muck with whatever your current German accent is (assuming you do in fact somehow have one).

    Are you sure you had no significant contact with native German speakers as a child?

    As an adult, my father found Italian very easy to pick up (“you can learn it from restaurant menus!”) when he was transferred there for work. Years later he realized that as a toddler, the nice old woman down the street who babysat him while his mother worked, fed him spaghetti, etc., had been Italian. He remembered *communicating* with her, he’d just never thought about what *language* it had been in.

    1. >Or can you not pronounce written German either?

      I don’t actually know. All I know is that (a) I can hear what sounds correct and incorrect in speech, and (b) I can often reproduce correctness. I can associate individual written words that contain sounds not present in English with correctness, like “schön”, but I am not confident of my ability to pronounce continuous written German. I think I might stumble at that.

      One time when I was flying Lufthansa I tried asking a stewardess for apfelschorle in German and thanking her in the same language (three whole words!). There was no indication in her response or body language that she thought I was anything but a native speaker using the language within norms. So there is at least a small bit of evidence that I can produce phonology correctly in short speech interactions.

      (I often do this sort of thing for fun when I travel.)

      >hm, the process of learning might well muck with whatever your current German accent is (assuming you do in fact somehow have one).

      Yes, I think that is rather likely.

      >Are you sure you had no significant contact with native German speakers as a child?

      I don’t recall any, and the circumstances of my childhood make it unlikely.

  10. I’ve learned something by listening to her – which is that somehow, somewhere, I have internalized a very precise understanding of German phonology and phonotactics. As in, I not only know right pronunciation from wrong, I give her detailed advice on how to match Duolingo’s model speaker that we can both tell is correct.

    I suspect that there is a class of people who have much better conscious access to the contents of their brains’ audio buffers than most. I think that when most people hear an utterance their brains decode what was said and then discard the “raw audio” whereas people in this other class can hold onto the “raw audio” and derive all sorts of information from it — the precise inflections and movements of the vocal tract, tongue, and lips to recreate the utterance for example. A synaesthetic association between hearing a sound and feeling how that sound was produced.

    This hypothesis would predict secondary effects, for example that skilled musicians (particularly singers), people who can do vocal impersonations, and people who can readily acquire foreign languages significantly overlap. I was once the first and am the last two; and I suspect that I am in this class of people; it’s entirely possible that you may be as well. Though I have problems of my own; oftentimes I do not immediately hear what was said by a person and so have to access the audio in my head and consciously decode it. Very frustrating and can sometimes make conversations difficult, especially in noisy environments.

    1. >This hypothesis would predict secondary effects, for example that skilled musicians (particularly singers), people who can do vocal impersonations, and people who can readily acquire foreign languages significantly overlap.

      That’s me, all right. My ability to do vocal impersonations is weak relative to the other two, but I think that is mainly due to lack of practice and attention; I definitely have the knack.

      It has recently become known that crib bilinguals like me often have a slightly unusual organization of the organ of Broca that shows up in brain scans. I think it is rather likely that you just explained what that does.

  11. This is the way Pimsleur works. Aside from their infomercial sales tactics, you can pick up a language in a month, and be reasonable in three. Without an accent. But they say you need to listen and imitate the sounds and not read (since english letters represent different sounds – I once put my tablet into a mode that did the reverse giving driving directions in a thick spanish accent). They call it the auditory loop. Your brain has a parrot mode. You do it with music, remembering and imitating the inflections. Pimsleur has a few youtubes. I didn’t think I could learn a language, but I can now speak and understand basic spanish.

  12. Some people can hear better than others. Many people can’t even hear the differences in regional variations of English. For example, some people say “caught” like “cot”. Many of them don’t even hear the difference when people pronounce them differently.

    Do you know any Germans? I find hearing people’s accents in English helps me get a better accent in their native language.

    1. >Do you know any Germans?

      There are no Germans in my face-to-face social network. A boatload in my on-line one, but that wouldn’t help.

  13. “I guess the place where the natural common dialect is closest to the standard is somewhere in the middle, say, Frankfurt.”

    This is common. It happened in English. People in the north and south of England were furthest away from each other, and their dialects differed the most. People in the middle were able to communicate with both groups, as the differences were smaller, and so the midlands dialect grew at the expense of the others, and became ‘standard’.

  14. I have a similar ability with music. I don’t know how to play any instruments, I can’t read music — but when my kids are practicing, even when they’re playing a piece I’ve never heard before, I can tell when they’ve hit the wrong note. I think I’ve just heard enough music over my life that I can recognize the internal logic of a musical phrase.

    Eric, you speak English, which is a Germanic-descended language. Maybe you’ve managed to tease out the Germanic roots of English somehow. Do you know any Romance languages well, or classical Latin? Perhaps you’re mentally “subtracting” those elements from the Germanic parts of English.

    1. >Do you know any Romance languages well, or classical Latin?

      I was a crib bilingual in Spanish. I have spoken French and Italian passably, though not with native fluency.

      >Perhaps you’re mentally “subtracting” those elements from the Germanic parts of English.

      That might be a plausible explanation for vocabulary. I don’t think it is for phonology.

      No, what is most likely going on is that my brain readily builds phonetic models from speech samples, and I am unaware this is going on until circumstances require me to use one of the models. Children do this sort of thing all the time; people who are natural polyglots, the kind who learn languages for fun, retain the knack into adulthood. I think I’m a sort of very low-grade natural polyglot.

    2. >I can tell when they’ve hit the wrong note.

      You should test yourself, if you haven’t, for mathematical ability. Chances are good that you have some. I’m serious.

  15. As a semi-relevant aside, there are different measurable abilities this is related to:
    * Memory for music
    * Tonal differentiation
    * Memory for words

    Somebody who is good at the first two I would expect to be a good musician. Somebody who’s good at the last two I would expect to be good at picking up foreign languages. Somebody good at all three I would expect to be great at impersonations.

    Personally, I score ~90th percentile for tonal differentiation but 10th percentile for memory for music and words. The result is that when I sing I am acutely aware that I’m off-key. I still enjoy it, though I try to spare others the agony.

    1. >Somebody good at all three I would expect to be great at impersonations.

      Counterexample: I’m good at all three, but my ability to do impersonations is not great. I mean, it’s sort of there, but it’s crude and not that effective. Maybe it would get better if I put effort into it, maybe not.

  16. @Shenpen: “BTW I think that the legend that that “High German” (i.e. standard) comes from being spoken in the Alps highlands is bullshit – actually the higher you go in the mountains, the weirder it gets”

    I thought this referred to German as opposed to what eventually evolved into other languages such as Dutch, Flemish, and English, rather than dialects of modern German.

  17. I think there is a linguistic thing, an actual physiological capability, akin to musical ability that means you hear nuances of sound better than others. You can train to be a good musician, but you are born a great musician, I think the same is true of linguists. I have a similar thing. I can speak some French and some Spanish, though I am far from fluent in either, however, I have an extremely good accent in both languages. It is funny to hear me speak with a perfect accent while fumbling my verb tenses like a first year learner.

    My brain notices subtitles of the sound that other people don’t. I think it is just the way your brain is made. Sounds like you are the same, just a built in brain capacity. On the flip side, I have a horrible sense of direction, and it frustrates the hell out of me when people just seem to know the way to go when I literally don’t have a clue which way is north. I think I am missing that particular brain capacity.

    (Oh and FWIW, since this is a blog frequented by programmers, I think that it is exactly the same when it comes to programming. Some people just seem to be missing a brain capability to really grok what is going on.)

    1. >It is funny to hear me speak with a perfect accent while fumbling my verb tenses like a first year learner.

      Heh. Yeah, I do this too and I also have no doubt it sounds funny. Like you, that fails to bother me.

      >I think I am missing that particular [wayfinding/sense-of-direction] brain capacity.

      I have it, big-time. My wife doesn’t. Most women don’t, most men do. It is what it is.

  18. Ya, that is a surprising observation, having knowledge of a language with such limited exposure. There seem to be two sorts of explanations for that. One is unknown exposure to the language. The other is that extrapolation from limited exposure is more powerful than expected.

    See the Aslin, Saffran and Newport articles on statistical language learning from the mid 1990’s. In a hand-wavy nutshell, the story is that a human brain runs a subconscious background process which acquires language by abstracting statistical properties from the corpus of speech to which it is exposed. The mind learns, for example, to identify word boundaries from conditional probabilities of phoneme sequences alone. The learning requires no attention and no semantics. There are probably better references on that subject published since the Aslin et. al. articles, but those were the first, dramatic published results in that area. Seems related to ESR’s experience, for example, both have the quality that the learner would not necessarily have meta-knowlege of his own language knowledge.

    Language acquisition also begins early in life. So one possibility is that if ESR had been exposed to German language in early childhood, for example by a German-speaking nanny or housekeeper. He would not recall that exposure but would have acquired and retained into adulthood some German language skills.

    Another line of explanation is that, to the degree that languages are constrained (i.e. regular or low-entropy), generalizations are more powerful. So the surprising phenomenology could also by explained by efficient generalization from limited exposure in adult life alone. It seems necessary to assume isomorphism between constraints on natural language and biological constrains on human language acquisition for that explanation to work.

    1. >In a hand-wavy nutshell, the story is that a human brain runs a subconscious background process which acquires language by abstracting statistical properties from the corpus of speech to which it is exposed. The mind learns, for example, to identify word boundaries from conditional probabilities of phoneme sequences alone. The learning requires no attention and no semantics.

      Ah, right. By coincidence, I recently stumbled over a thumbnail sketch of this work when reading about challenges to the Chomskian “Universal Grammar” theory of language acquisition. The argument is that this theory evades the “poverty of the stimulus” argument that the UG theorists have leaned on so heavily, because a running Bayesian analysis can acquire information from communication features outside of speech samples. Such as what people don’t say!

      I found this idea very appealing for two reasons:

      #1: It matches my own experience of language acquisition. I mean, I seem to be really good at extracting “conditional probabilities of phoneme sequences” from speech samples, and this appears to underly several neat tricks I can do, like correcting Cathy’s German pronunciation even though I don’t know German and generating plausible-sounding names in fantasy and SF languages on demand.

      #2: Ah. So maybe Noam Chomsky is full of shit even in his own field? Cool!

      I hope I’m not too influenced by the second point…

  19. Perhaps you watched lots of ‘Hogan’s Heroes” as a kid!

    Thanks for the interesting discussion. I learned my German in Texas public schools. Years later, while studying in Austria, I was surprised to learn that I speak German with an English accent.

    Of course it makes sense, we can hear the difference between a German and a Frenchman speaking English, but I never considered the possibility that I might not have an American accent when speaking German.

    1. >Perhaps you watched lots of ‘Hogan’s Heroes” as a kid!

      Urk. In fact, I did. But the likelihood that I could learn actual German phonology by exposure to exaggerated stage-German accents seems low.

  20. #2: Ah. So maybe Noam Chomsky is full of shit even in his own field? Cool!

    I hope I’m not too influenced by the second point…

    But the Aslin et al. paper doesn’t disprove UG in the light of your experience, Eric. It only means that your experience can’t be considered dispositive evidence for UG, as Chomsky and friends would claim.

    Here’s the generative phonology story on what’s going on in your head:

    You were born knowing a set of parameters, yes/no predicates whose settings between them determine the phonotactics of whatever language you end up speaking. For example, in English you can’t begin a syllable with the velar nasal [ŋ], so maybe there’s a universal parameter that says {[ŋ] tolerated in syllable onset}, and English has that parameter set to False (pardon my Python).

    The idea is that you get around poverty of the stimulus by having these parameters innate: all a language learner has to do is listen for evidence that sets each parameter one way or the other. Once they’re all set, you’ve got yourself a phonology. So, the Chomskyan story would be that the reason you’ve managed to absorb a knowledge of what is and is not proper-sounding German is that the Kolmogorov complexity of that knowledge is actually pretty low: it’s just {True, False} * n parameters.

    All Aslin et al. add to this is to demonstrate that the complexity of the message you have to receive is low enough in absolute terms; even without any parameters, apparently there isn’t that much there there. Even within generative phonology, this kind of reasoning gets some play: I myself am skeptical of linguistic nativism because it seems to be a diagnosis of exclusion, so I follow a school of thought that says parameters get bootstrapped up by children monitoring their own production as they acquire language: roughly speaking, if they find a particular sound hard to make, they infer that there’s a parameter that militates against it, then listen to see whether they need to make the effort, as it were, to switch the new parameter off. This means that the parameters are still innate and universal, but they come from us all being born with more or less the same anatomical hardware.

    This is just the tip of an iceberg of findings of that sort, by the way. Debating whether the theoretical constructs employed by linguists are necessary to explain language, or whether it’s all just frequency and other statistical effects, is by now almost a form of ritualized warfare between departments of experimental psychology and linguistics, to the point where theoretical phonologists like me wish, in our weaker moments, that the big x-psy meanies would just let us play with our toys in peace.

    Also, I have an anecdote that is vaguely relevant to this discussion. A few years ago I had a realization similar to the one you report here, when my mother told me that, when learning French at school, she could never hear the difference between un and une. Unfortunately for her, she made this revelation too late: I already had enough training in phonetics that not only could I say “how can you not hear that?” I was also able to follow it up with a precise and more than a little condescending description of exactly what the vocal apparatus does in each word (sorry, Mom!)

    Anyway, the vague relevance comes in in that I’m genuinely not sure what makes some adult learners more baby-like in their sensitivity to phonetic distinctions their own language doesn’t employ. It might be that our milk tongues are fixed in place (so to speak) by a “hardening” of parameter settings, so that the brain optimizes itself not to listen for the evidence for the other settings any more, or it may be that lifelong monoglots are less sensitive to all kinds of acoustic detail in the first place. There’s evidence for both propositions, and more likely than not both are true to varying extents. All interesting behaviour, as someone once said, is overdetermined, after all.

    1. >Here’s the generative phonology story on what’s going on in your head:

      Oh, good. I was hoping you’d show up for this thread!

      >This means that the parameters are still innate and universal, but they come from us all being born with more or less the same anatomical hardware.

      This seems like a weaker (hence, better) hypothesis than supposing that we have a bunch of linguistic rules for which there is no obvious a priori wired into our brains. Really. Do Chomsky et al. have any theory about how UG got there?

      >almost a form of ritualized warfare between departments of experimental psychology and linguistics,

      I’m sorry to tell you that I’m inclined to side with the Bayesians on this one (see “how did it get there”, above). Not that you guys don’t have some very fun toys.

      >There’s evidence for both propositions, and more likely than not both are true to varying extents.

      Agreed. But I think my experience tilts me towards “monoglots are less sensitive to all kinds of acoustic detail”. It would be an astonishing coincidence if my acute ear for musical patterns, pitch sensitivity, and unusually precise auditory memory were irrelevant to my stupid language tricks.

  21. Do Chomsky et al. have any theory about how UG got there?

    Natural selection. Seriously.

    OK, bear in mind that I’m summarising a position I don’t believe and only vaguely remember, but it goes like this:

    It doesn’t take much doing to imagine how being able to speak might net you more Darwinian fitness than not, but of course UG in all its (however fuzzily-defined) complexity is probably too much to arise out of a single mutation. Plus, language is for communication, so it wouldn’t be much use if that mutation only affected a single hominid.

    The idea is that there are individual components of UG that are adaptive all on their own, and so can evolve one at a time, much like to get an eye out of natural selection you start with a single light-sensitive cell, then recess it a bit so it can focus more, and so on and so on.

    The argument is usually made from syntax, rather than phonology: you start with the notion of having maps as well as territory, mapping single arbitrary mouth-noises onto real-world experiences. Clearly adaptive: if you can get the other monkey-things to grok that ug means “big scary predator” and ook means “food”, then they’ll scatter and hide when you yell ug! and gather round (and potentially have kids with you) when you yell ook (and actually have food. Even at this stage, General Semantics is important.)

    The next big step is having a principled way of combining two words together. If you have a rule that says compounds are right-headed (a housecat is a kind of cat, not a kind of house), then you can say a bunch more things without ambiguity. If you yell ug-ook because you’ve got some nice ug-steaks over the fire, then the guy who hasn’t figured out endocentric compounds yet, and thought you meant that you were being eaten by an ug, is likely to run away screaming and get eaten by an angry, recently widowed she-ug.

    Eventually more benefits get conferred by cool tricks like recursion and whatnot, but that’s the basic idea. Of course, it does mean that everything in UG, from headed compounds to center-embedding, is there in our DNA, and I don’t think the Human Genome Project has found it yet, but if you believe this stuff is innate, it’s a position you’re kind of committed to, short of claiming it’s absorbed from the amniotic fluid.

    Entertainingly, Chomsky has taken a more hard-line position recently, arguing that syntax in basically its modern form emerges more or less fully formed in early homo sap. I went to an interesting talk last week about how his arguments for this are mainly flat assertions that have little basis in the archaeological, or even historical linguistic record.

    On another note entirely, and speaking of syntax, I guess I should give you a heads-up that your posts on how you’ve changed the syntax of reposurgeon got me thinking about applying Lexical-Functional Grammar (the theory of syntax I learned as an undergraduate) to the parsing of command lines. If my copious free time permits, you may get a fork of the project for your consideration that is backwards-compatible between v2 and v3 syntax (with the user asked to resolve any ambiguities the program can’t), and has tab-completion and some DWIM-ish natural language features besides. Also, I’m making it my python3 learning project, so you might get the conversion to python3 as a freebie. Or then again I might be too busy. Hard to say.

    1. >Of course, it does mean that everything in UG, from headed compounds to center-embedding, is there in our DNA, and I don’t think the Human Genome Project has found it yet, but if you believe this stuff is innate, it’s a position you’re kind of committed to, short of claiming it’s absorbed from the amniotic fluid.

      I find this highly implausible. But this is an area where I am very aware of the limits of my knowledge, so I’m not going to lean on that feeling too hard.

      Um. I’m guessing you have a pretty good idea of the bits and pieces of linguistics I’ve managed to pick up by now – mostly historical/taxonomical linguistics and descriptive phonology, some very very basic stuff about language morphology, a dab of psycholinguistics. One of my minor goals is to be able to talk with real linguists like you without feeling ignorant as shit. Can you point me at good books (preferably but not necessarily online) to learn my way a little deeper in?

      >Also, I’m making it my python3 learning project, so you might get the conversion to python3 as a freebie.

      I will examine any of the patches you are contemplating with great interest.

  22. >#2: Ah. So maybe Noam Chomsky is full of shit even in his own field? Cool!

    UG seems to be the right level of generality that if wrong we still learn a lot from finding so.

    I would say that even if false UG is nonetheless a good theory, that theories which prove true are no less valuable than theories which prove false. That instead, the worth of a theory is its efficiency, which depends upon how evenly it partitions the set of unknowns. The worst case would be a set of theories which enumerates every unknown to be tested individually. The best case would be a set of theories which recursively divide the remaining unknowns into two testable sets. The former resolves all unknowns on average in 1/2 n time, the latter in 1/2 log base 2 (n) time. Strong inference is another criterion; experiments testing a theory should yield definite conclusions and therefore good theories are those subjectable to definitive testing.

    Debugging is only sequential hypothesis testing and so the standard applies there. Anecdotally, I have seen novices in a domain, who are good theorists, beat the pants off of experts. This is especially common for novel bugs, where the expert tests for specific bugs encountered in his past experience or presented in training, whereas a good theorist will quickly converge on any bug.

    As valuable a theory is UG by the proposed standard of worth, Chomsky’s rantings outside of linguistics are worthless.

  23. >#2: Ah. So maybe Noam Chomsky is full of shit even in his own field? Cool!

    FOS might be overly harsh. The conditional element of UG seems incontestable: To the extent that innate biological constraints exist the work of language acquisition could be reduced.

  24. Just a guess, but did you at some point study any Chaucer or Ancient English? English, at it’s heart, is Germanic after all. I know we spent a couple weeks on it in College English II, and you’re education is undoubtedly a lot better than mine in that area, so that, along with your early-developed ear for foreign languages and broad exposure, would mean you’d very quickly pick up on basic phonetics, pronunciation and rhythm I would think.

    Hearing Cathy’s lessons fired some grey cells you haven’t used in several decades :^).

    As I said, just a guess.

    1. >Just a guess, but did you at some point study any Chaucer or Ancient English?

      Unless you count having a good time puzzling out Chaucer on my own and reading parallel-translated Old English epic poetry for fun, no.

  25. @esr:
    >However I should also note that dialect variation in American English is much narrower than in German; only a very few extreme Southern dialects vary from “neutral” as much as any random two German dialects do from each other.

    Being a fair bit younger than you, and having been raised quite a bit further west (Denver) than I gather you were, there are a few northeastern dialects that I perceive to be significantly off-neutral, in addition to the extreme Southern dialects.

    >>I can tell when they’ve hit the wrong note.
    >You should test yourself, if you haven’t, for mathematical ability. Chances are good that you have some. I’m serious.

    What sort of mathematical ability? I have a fairly good ear for relative pitch (not so much for absolute pitch), and an annoyingly good memory for music (pieces I haven’t heard in months or years will bubble up without any trigger I’m conscious of and remain stuck in my head for days), as well as very solid linguistic ability.

    My mathematical abilities are mixed: I have good mathematical intuition (tied into a knack for visualization), but my raw number crunching abilities, while not bad, aren’t great either. My dad can do multi-digit arithmetic in his head, whereas I need paper for that except in the case of a few memorized cases and a few simple patterns. On the other hand, he doesn’t seem to have my aforementioned ability for visualization.

    >>I think I am missing that particular [wayfinding/sense-of-direction] brain capacity.
    >I have it, big-time. My wife doesn’t. Most women don’t, most men do. It is what it is.

    A couple questions: Is your sense of direction ever wrong? And, when wrong, is it wrong consistently?

    I always have a solid sense of which way is north (solid in the sense of mental certainty, not necessarily accuracy), and it is generally accurate, but in some locations it can be off by a full 180 degrees. In any given location it is either always accurate or always off by a consistent angle, for example, by 180 degrees in the gym at the elementary school I attended, by 90 degrees (such that perceived north is true west) at my church, and so forth.

    I *think* there’s a tendency for linear features to line up to a perceived cardinal direction (such that a northwest road will be perceived to run north), but I’m not entirely sure of that.

    1. >there are a few northeastern dialects that I perceive to be significantly off-neutral, in addition to the extreme Southern dialects

      Yup, those exist. I’ve been to Maine. :-) Even the thickest of those isn’t as far off neutral as any of several deep Southern accents, though. My point is that you have to go to the most extreme variations in American English to find anywhere near the degree of variation that is typical in German dialects.

      >What sort of mathematical ability?

      Musical ability is a pretty good predictor of the ability to do creative mathematics. Most mathematicians are aware of this. Curiously, musicians aren’t. Your weakness at mental arithmetic is insignificant; many mathematicians aren’t very good at that either.

      >A couple questions: Is your sense of direction ever wrong? And, when wrong, is it wrong consistently?

      It works less well indoors, and there are specific outdoor locations where it packs in for no reason I understand.

      My errors aren’t as consistent as you report.

      I think I may be a sun-tracker via UV or something, patched with angular dead reckoning indoors. You sound more like you’re reading magnetic fields (yes, that is possible). These would produce different error patterns.

  26. Kerbal: Spassi Ishosh yi Aton (Kerbstomp Edition) Chapter 2: The Lone Angel

    Links:
    Game: http://www.kerbalspaceprogram.com
    Full Story: http://redd.it/1rgldc

    Gary tentatively reaches up with his five fingered hand and gently touches the tiny hand of the little alien standing over him, and utters the response he can’t understand. He’s astonished at how formal it sounds and how the ‘r’ rolls as he slowly speaks: “Kerbal, spassi ishosh yi aton.”

    He then nervously looks past as five more of the creatures [kerbals] approach, two of which are obviously larger. The one that greeted him flees back, but the sounds he makes are joyful.

    “Kerdman,” the alien emphasizes the difference in their names, “Gary Kerdman.”

    “Definitely a ‘d’ there, I’d say,” Jebbers remarks. He turns back to his patient and says, more formally than the first time, “Gary Kerdman, welcome to Kerbin.”

    “Kerbal,” the giant alien says slowly, but fluently, “Space welcomes you.”

    —-

    The first one smiles as the other four gather around and stare at his face with their enormous, unblinking, dilated eyes. Jebediah turns to the largest one and smiles, “Chii? Yi gianto.”

    “Aak sput!” the big one gasps, “Mebs ni Obewann aak yi lunatik.”

    I know exactly what this language is supposed to sound like, but I don’t do it particularly good justice in the title theme: https://www.youtube.com/watch?v=KKT1ZSePZqI

    Unfortunately, I can’t really tell you how… but it does work.

  27. @esr:
    >I think I may be a sun-tracker via UV or something, patched with angular dead reckoning indoors. You sound more like you’re reading magnetic fields (yes, that is possible). These would produce different error patterns.

    I don’t think I can be reading magnetic fields, for two reasons:

    1) Otherwise I’d expect my directional sense to pack in near artificial magnets, but it doesn’t.

    2) My directional sense is still quite strong in computer games.

    I’m betting on memory and angular dead reckoning.

  28. Addendum:

    In fact, I’d say that if one of us is perceiving magnetic fields, it’s you. And I have a test case for your sun-tracking theory: does your sense of direction pack in at night?

    Another bit of evidence against me being a magnetic tracker is that it is possible for me to become briefly directionally confused (though it takes some doing) when blindfolded. In middle school I sometimes would close my eyes and pull my hood down over my face on the bus to see if I could generate a directional shock when we got to school and I opened my eyes. Every once in a while I’d be 90 or 180 off when we arrived, but my frame of reference would immediately snap back (with a second or two of confusion) as soon as I saw anything outside the bus.

    1. >And I have a test case for your sun-tracking theory: does your sense of direction pack in at night?

      No, but that’s not dispositive. My dead-reckoning sense of distance and direction is so good that I seldom need to orient by asking where North is, and it’s distinctly possible I’ve never had to do that at night.

  29. Eric, I’ll quibble just a little with your description of the reach of Standard American: lower-class Philadelphia and Pittsburgh are definitely outside it, and Main Line Philadelphia sounds a bit too stilted to someone expecting it, while Minneapolis and Wisconsin, at least the western part, sounds too short and clipped. (Not sure whether to draw that line east or west of the People’s Republic of Madison.)

    Houston is right out.

    1. >Eric, I’ll quibble just a little with your description of the reach of Standard American:

      These are fair qualifications, but remember that I was addressing someone who is not a native speaker of English. It is highly doubtful he would be able to hear these or any of the other not-quite-Standard dialects in the Middle American region as being different from it.

      Depending on experience, he might not even be able to tell that you and I have different accents – as impossible as that seems to your American ear. Trust me on this one. Our difference (e.g between a Texas accent and East-Coastal Standard American) is tough to process even for a lot of Brits who watch American TV.

  30. According to Wikipedia you once were in Germany in Würzburg on May 22, 1997 to speak at the 4th international Linux Kongress ;)

    1. >According to Wikipedia you once were in Germany in Würzburg on May 22, 1997 to speak at the 4th international Linux Kongress ;)

      That’s right, that’s the one visit to Germany I referred to. My other venture into German-speaking territory was to Zurich to give a talk at ETH.

  31. There are a couple of reasons it seems like I have a model of German, though. The biggest one is that I am aware that I would not pronounce German exactly the way the model speaker does, but that Cathy should. The model speaker is female…

    Now that is interesting! I don’t have a handy explanation for that one.

    (Some years ago I wrote about hearing Liv Tyler pronounce Elvish in the first Rings movie and realizing that (a) she was doing it slightly wrong, and (b) I knew why her speech coach wasn’t fixing it. She was female-izing Sindarin a bit too much. Could you hear this?)

    Not that I remember… *finds some YouTube clips* Ah. No, what stands out for me is national origin: Tyler has an obvious American accent (and yes, now that you mention it, it’s a very feminized accent at that; and she does the same thing in English), Orlando Bloom’s sounds somewhere between American and British, and Viggo Mortensen’s is… is… wonderfully unplaceable, actually! Is that Gaelic??? I didn’t notice that one the first time around.

    The Internet tells me these actors are an American, an Englishman, and a globetrotting polyglot, so that maps out well. Sindarin enthusiasts peg Mortensen as having the best pronunciation, so the Gaelic-sounding bits I’m hearing might even be a legit Sindarin accent per Tolkien.

    Even though I can hear their Anglicization clearly, I wouldn’t be able to give Tyler or Bloom any specific advice about how to fix it. My explanation-powers only kick in with bigger issues than they have.

    1. >Sindarin enthusiasts peg Mortensen as having the best pronunciation, so the Gaelic-sounding bits I’m hearing might even be a legit Sindarin accent per Tolkien.

      I agree Mortensen had the best accent. But I think Sindarin (and Mortensen speaking it) sounds more Welsh than Irish Gaelic. It’s always seemed obvious to me that the phonology/phonotactics of both Elvish languages owed the most to Welsh and Finnish; the Welsh a little stronger in Sindarin, the Finnish a little stronger in Quenya.

      I didn’t notice the American vs. British accent thing at the time, but I did notice that Mortensen seemed to be getting inside the phonology of Sindarin whereas the other actors were speaking with accents. And Tyler’s feminization jumped out at me hard. I thought it was a forgivable error on the speech coach’s part not to fix it; the character design is supposed to be ultrafemale/sexy, after all.

      (One way you can tell if you’re a “Sindarin enthusiast” is if you recognized bits of vocabulary that weren’t subtitled. Why yes. Yes, I did.)

  32. Never considered the phoentics aspect before – Eric you’ve given me something new to think about, so thanks.

    I’ve always picked up languages easily and relatively naturally. I always assumed it’s because in my formative years I developed English primarily but my great-grandfather spoke French as his primary language and my grandfather and father spoke Portuguese, so I spent my formative years hearing and speaking all three.

    But, like you, I have that music passion and naturalness. Very interesting to things to consider here now.

    Funny thing is, folks can understand me in their language, but my own countrymen often ask me what I said or what what I said meant. I speak with a bastardized accent of mainly blue collar Boston with some southern coastal Maine thrown in there (amusingly mentioned above).

  33. Another question:

    What got German to do with the ability to understand iron age-texts? Or are these text just written in German?

    Time passed by, even in Europe ;)

    1. >What got German to do with the ability to understand iron age-texts?

      A lot of sources on the archeological record pertaining to Iron Age and Viking costume are in German.

      My wife does museum-quality reproductions of womens’ costumes from these eras for fun, and maintains a substantial research library on the topic. She has a blog about it at Lohttp://cathyscostumeblog.blogspot.com/ose Threads.

  34. >My dead-reckoning sense of distance and direction is so good that I seldom need to orient by asking where North is, and it’s distinctly possible I’ve never had to do that at night.

    That’s different from my case: I never need to ask to get a sense of where north is, I always have that sense wherever I am. In the cases where that sense is wrong, I have to remember that it is wrong in that location, and then ask for true north and remember the angle between true and perceived north. Perceived north never changes after I ask, but I can correct for it when I know it’s wrong.

  35. Our difference (e.g between a Texas accent and East-Coastal Standard American) is tough to process even for a lot of Brits who watch American TV.
    Even post-Dallas? Whatever else you could say about that series – and one could say one hell of a lot – they got the accents right. Helped immensely that Larry Hagman was a native speaker.

    You and I have two of the most distinctive voices I know, and quite distinctively different. I would be very, very surprised if someone could not easily tell us apart by voice, even if they do ask if we’re brothers when they see us together (a fact that amuses me probably more than it should).

    1. >Even post-Dallas?

      Yes, even post-Dallas. When I say the dialect differences in American English are slight compared to those in other languages, and even compared to other dialect subgroups in English, I’m not kidding. You hear them as large divergences only because you’re a native speaker of American English yourself.

      I’m sure almost anyone could tell you and me apart by voice, but that doesn’t necessarily mean they could tell we have different accents.

  36. To me, from far away, most American voices sound alike at least what I’ve heard on TV. Except when the accents are played up for deliberate comic effect like the so-called “redneck” accent.

    I can make out the differences between American, British, Australian, South African English accents though. Distinct accents and dialects like Cockney are more easily recognizable.

  37. @esr
    “I once discovered by accident that I can read Old Low German pretty well, when I stumbled across a collection of folktales in it. ”

    Did you ever look where the Anglo-Saxons came from?

    1. >Did you ever look where the Anglo-Saxons came from?

      Of course I know where the Anglo-Saxons came from. I never said that I found being able to read Old Low German surprising, did I? :-)

      Actually I did, a bit. But mainly because construing Anglo-Saxon seems slightly more difficult.

  38. Interesting. I am a Czech, studying German for three years now, and I visit German-speaking countries a few times each year. The differences among the dialects can be heard very well. Low German is not that far from Dutch. Saxons speak quite hard to my ear, Bavarian villagers are really hard to understand etc.

    Some words will betray the origin of the German speaker immediately, such as many expressions from the cuisine. There is a specific regional expression for “gingerbread” in almost every Land, etc.

    Once the natives notice that you speak as a foreigner, they immediately switch to Hochdeutsch, though (or English).

    As far as Duolingo goes, I am level 15 in German now and I am unhappy about the limited size of the vocabulary. 1700 words is just not enough. I am thinking about buying 3 months at babbel. They promise about 4000 words and, being Germans, quite a lot of culture-specific courses.

  39. >dialect differences in American English are slight compared to those in other languages, and even compared to other dialect subgroups in English, I’m not kidding. You hear them as large divergences only because you’re a native speaker of American English yourself.

    There are southern and rural dialects in the United States which are completely incomprehensible to native SAE speakers. Those dialects are not heard on radio or television and native speakers only use them when conversing with each. So one must travel to were those dialects are spoken and eavesdrop on the locals to discover this.

    While waiting for a mechanic to replace a radiator hose I sat in a rural Kentucky garage for hours listing to the good ol’ boys converse. It was the local hangout. Could not understand one single word. Though when I asked a question or paid the bill the cashier and mechanic could switch into accented SAE from the local backwoods dialect.

    African-American dialects in the Carolinas are also utterly incomprehensible to native SAE speakers. At the extremes north-american creoles such as Gullah constitute separate languages.

  40. > As someone from northern germany, i have a soft spot for Lower German, which has a common ancestor with english. I was very amused as i once heard a lower german version of Harry Potter, and it sounded much more like the english orginal as the standard german version.

    Not only Lower German. All western european languages have common ancestors (and you realize that as soon as you try to learn three or more of them). The reason why lower german has retained more ties with English is probably because it’s spoken in the seafarer region of Germany.

  41. >Depending on experience, he might not even be able to tell that you and I have different accents – as impossible as that seems to your American ear. Trust me on this one. Our difference (e.g between a Texas accent and East-Coastal Standard American) is tough to process even for a lot of Brits who watch American TV.

    And there’s an East Texas accent that’s different from the Texas accent.

    On the other hand, Americans have trouble telling the different Brit accents apart. Silly example- my kids love to watch ‘Peppa Pig’. I can tell that the different characters have different accents, and that there is a whole world of subtext packed into those accents and who speaks them, but I don’t know enough to know what it *is*, just that it’s there.

    1. >I can tell that the different characters have different accents, and that there is a whole world of subtext packed into those accents and who speaks them, but I don’t know enough to know what it *is*, just that it’s there.

      Don’t feel bad, you pretty much have to have lived in England to decode this. Unlike regional American accents, British regional accents don’t get enough media exposure that a foreigner with a good ear has any chance of developing a key just on what’s exported.

      I have lived in England, so I get a lot of it, but I’m sure I don’ t get all of it.

      I found an episode on YouTube and watched a few minutes. One piece of subtext is the way Grandpa Pig’s accent differs from Papa Pig’s. Both are speaking BRP (British Received Pronunciation) which is the prestige dialect of Southern England and the London region, but Grandpa’s version is a generation older and significantly less Americanized than present BRP. This is expected of older people and has connotations of authority and authenticity.

  42. Pardon the digression…. class enters in to it as well. The Miss Rabbit character who seems to do all the work in the series, has some kind of working class accent, I think. And the postman and his daughter sound (to me) vaguely Irish. The veterinarian lady’s accent is just wacky.

    Divided by a common language, indeed. I understand the situation is even worse w.r.t. Spanish, and the different (mutually loathing) countries that speak it.

  43. Actually, Greg, there are at least five different Texas accents, subtly different, but on a continuum. Mine is quite easily identifiable as Houston, to a Texan. I can simulate a deep East Texas accent pretty well (say, Jacksonville or Jasper). The rest, maybe not so much, with the possible exception of west Texas (Midland/Odessa).

  44. >Actually, Greg, there are at least five different Texas accents, subtly different, but on a continuum. Mine is quite easily identifiable as Houston, to a Texan. I can simulate a deep East Texas accent pretty well (say, Jacksonville or Jasper). The rest, maybe not so much, with the possible exception of west Texas (Midland/Odessa).

    Interesting, thanks. I work with a lot of people in a Dallas office, and one of the people I’m friendliest with there grew up in East Texas. There are accent and cultural incompatibilities. Another guy I was friendly with, from Oklahoma, was much better able to blend in with the Dallas accent.

    I know there are also places in Texas where German was spoken as a first language (entire towns) until very recently, but I don’t know how that affects things.

  45. Going a little off, as a non-native English speaker (I’m from Brazil”, I have always been curious to know why English orthography is much more irregular than that of romance languages, and even German. Then I stumbled upon the following Wikipedia article, which appears to give a big part of the answer:
    http://en.wikipedia.org/wiki/Great_Vowel_Shift

    Does someone know of any other online study on such topic beyond the above article, especially the causes of such a sudden pronounciation change?

  46. @Cleverson
    You may also want to look into the (2nd) Germanic Consonant Shift, which also models/explains some of the differences between the Germanic languages.
    And while the vowel shift explains the sometimes odd discrepancies between English pronunciation and spelling, one could also just say they neglected to adapt their orthography to the changes in pronounciation.

    Also, those changes aren’t all that sudden ; they happen over the course of several hundred years – say, something in the order of 20 generations for the Great Vowel Shift ?.
    As to why they happen, I don’t know. But as sound shifts are apparently not unique to the Germanic languages (see Grimm’s Law, or the evolution from Latin into French/Italian/Spanish), I’d guess you have to look for linguistics or philology and see if they offer an explanation.

  47. OK, thank you for replying.

    The Wikipedia article says that the GVS occurred between 1350 and 1500, so it may then be wrong, as 20 generations comprise about 500 years.

    And yes, languages from other families also change. Being a Brazilian Portuguese speaker, I have identified by personal listening observation what I believe to be at least one true consonant change (not just accent variation), that tends to propagate over time here in Brazil.

    Thanks for the studies suggestions as well.

  48. In the obituaries for Sid Caesar, there is some talk about how he learned his characteristic “doubletalk” — sounding as if he was speaking a langage while making no sense at all. Ben Zimmer has a post at the linguistics blog Language Log, linking to Caesar’s own memoir about how he learned to do that:
    http://languagelog.ldc.upenn.edu/nll/?p=10429#more-10429

  49. Eric: There is no major part of the United States in which the “neutral accent” (in the sense of the one that broadcasters are taught to use) is spoken. (There was, fifty to seventy years ago, which is where the pronunciations given in American dictionaries come from. Taken literally, they’re nonsense; taken as “use the sounds in your accent that match these”, they can still be quite useful.) The closest thing to it is the area marked “North Central” on Roger Aschmann’s North American English map.

    Shenpen: The grammar and vocabulary of Standard German does come from the High German area, though not the Highest. The pronunciation has been described as a Low German trying to speak High German.

    1. >The closest thing to [neutral accent] is the area marked “North Central” on Roger Aschmann’s North American English map.

      I am aware of this map. I think that to a significant extent it represents a pursuit of phonological distinctions that (a) vary as much or more across idiolects within some of the supposed regions than they do across the regions, and (b) are not recognition features for actual speakers of any dialect.

      In particular, I question from experience the separations made among the North, North Central and Midlands dialect regions. I realize it’s fun for sociolinguists to write papers about this sort of thing and I suppose they have to climb the tenure track somehow (poor things), but when even a layman with a dialect ear as sensitive as mine can’t hear their categories I have to wonder if there’s much real point to the exercise.

Leave a comment

Your email address will not be published. Required fields are marked *