Can you really learn a language in ten days?

I’m not the only linguist in my family. My father has worked as a professional linguist his whole life… but with a slightly different definition of “linguist”. His job is to use his specialist knowledge of a language (specifically Mandarin Chinese, Mongolian or one of the handful of other languages he speaks relatively well) to solve a problem. And one problem that he’s worked on a lot is language learning.

There’s no doubt that knowing more than one language is very, very useful. It opens up job opportunities, makes it easier to travel and can even improve brain function. But unless you were lucky enough to be raised bilingual you’re going to have to do it the hard way. And, if you live in America, like I do, you’re not very likely to do that: Only about 26% of the American population speaks another language well enough to hold a basic converstaion in it, and only 9% are fluent in another language. Compare that to Europe, where around 50% of the population is bilingual.

Japanese language class in Zhenjiang02
“Now that you’ve learned these characters, you only need to learn and retain one a day for the next five years to be considered literate.”
Which makes the lure of easily learning a language on your own all the more compelling. I recently saw an ad that I found particularly enticing; learn a language in just ten days. Why, that’s less time than it takes to hand knit a pair of socks. The product in this = case was the oh-so-famous (at least in linguistic circles) Pimsleur Method (or approach, or any of a number of other flavors of delivery). I’ve heard some very good things about the program, and thought I’d dig a little deeper into the method itself and evaluate its claims from a scientific linguistics perspective.

I should mention that Dr. Pimsleur was an academic working in second language acquisition from an applied linguistics stand point. That is, his work (published mainly in the 1960’s)  tended to look at how older people learn a second language in an educational setting. I’m not saying this makes him unimpeachable–if a scientific argument can’t stand up to scrutiny it shouldn’t stand at all–but it does tend to lend a certain patina of credibility to his work. Is it justified? Let’s find out.

First things first: it is not possible to become fluent in a language in just ten days. There are lots of reasons why this is true. The most obvious is that being a fluent speaker is more than just knowing the grammar and vocabulary; you have to understand the cultural background of the language you’re studying. Even if your accent is flawless (unlikely, but I’ll deal with that later), if you unwittingly talk to your mother-in-law  and become a social pariah that’s just not going to do you much. Then there are just lots of little linguistic things that it’s so very easy to get wrong. Idioms, for example, particularly choosing which preposition to use. Do you get “in the bus” or “on the bus”? And then there’s even more subtle things like producing a list of adjectives in the right order. “Big red apple” sounds fine, but “red big apple”? Not so much. A fluent speaker knows all this, and it’s just too much information to acquire in ten days.

That said, if you were plopped down in a new country without any prior knowledge of the language, I’d bet within ten days you’d be carrying on at least basic conversations. And that’s pretty much what the Pimsleur method is promising. I’m not really concerend with whether it works or not… I’m more concerned with how it works (or doesn’t). There are four basic principals that the Pimsleur technique is based on.

  1.  Anticipation. Basically, this boils down to posing questions that the learner is expected answer. These can be recall tasks, asking you to remember something you heard before, or tasks where the learner needs to extrapolate based on the knowledge they currently have of the language.
  2. Graduated-interval recall. Instead of repeating a word or word list three or four time right after each other, they’re repeated at specific intervals. This is based on the phonological loop part of a model of working memory that was really popular when Pimsleur was doing his academic work.
  3. Core Vocabulary. The learner is just exposed to basic vocabulary, so the total number of words learned is less. They’re chosen (as far as I can tell, it seems to vary based on method) based on frequency.
  4. “Organic learning”. Basically, you learn by listening and there’s a paucity of reading and writing. (Sorry about that; paucity was my word of the day today 😛 ).

So let’s evaluate these claims.

  1. Anticipation. So the main benefit of knowing that you’ll be tested on something is that you actually pay attention. In fact, if you ask someone to listen to pure tones, their brain consumes more oxygen (which you can tell because circulation to that area increases) if you tell them they’ll be tested.  Does this help with language learning? Well. Maybe. I don’t really have as much of a background in psycholinguistics, but I do know that language learning tends to entail the creation of new neural networks and connections, which requires oxygen. On the other hand, a classroom experience uses the same technique. Assessment: Reasonable, but occurs in pretty much every language-learning method. 
  2. Graduated-interval recall: So this is based on the model I mentioned above. You’ve got short term and long term memory, and the Pimsleur technique is designed to pretty much seed your short term memory, then wait for a bit, then grab at the thing you heard and pull it to the forefront again, ideally transferring it to long-term memory. Which is peachy-keen… if the model’s right. And there’s been quite a bit of change and development in our understanding of how memory works since the 1970’s. Within linguistics, there’s been the rise of Exemplar Theory, which posits that it’s the number of times you hear things, and the similarity of the sound tokens, that make them easier to remember. (Kinda. It’s complicated.) So… it could be helpful, assuming the theory’s right. Assessment: Theoretical underpinnings outdated, but still potentially helpful. 
  3. Core Vocabulary. So this one is pretty much just cheating. Yes, it’s true, you only need about 2000 words to get around most days, and, yes, those are probably the words you  should be learning first in a language course. But at some point, to achieve full fluency, you’ll have to learn more words, and that just takes time. Nothing you can do about it. Assesment:  Legitimate, but cheating. 
  4. “Organic learning”: So this is in quotation marks mainly because it sounds like it’s opposed to “inorganic learning”, and no one learns language from  rocks. Basically, there are two claims here. One is the auditory learning is preferable, and the other is that it’s preferable because it’s how children learn. I have fundamental problems with claims that adults and children can learn using the same processes. That said, if your main goal is to learn how to speak and hear a given language, learning writing will absolutely slow you down. I can tell you from experience: once you learn the tones, speaking Mandarin is pretty straightforward. Writing Mandarin remains one of the most frustrating things I’ve ever attempted to do. Assessment: Reasonable, but claims that you can learn “like a baby” should be examined closely. 
  5. Bonus: I do agree that using native speakers of the target language as models is preferable. They can make all the sounds correctly, something that even trained linguists can sometimes have problems with–and if you never hear the sounds produced correctly, you’ll never be able to produce them correctly.

So, it does look pretty legitimate. My biggest concern is actually not with the technique itself, but with the delivery method. Language is inherently about communicating, and speaking to yourself in isolation is a great way to get stuck with some very bad habits. Being able to interact with a native speaker, getting guidance and correction, is something that I’d feel very uncomfortable recommending you do without.


How Many Languages Are There?

This Christmas, my dad gave me a very special gift: a hardcopy of Ethnolouge. It was a great gift for two reasons. The first is that Ethnolouge is probably the single best resource for looking up basic information about the languages of the world, and the second is that I had a door that just wouldn’t stay open. (Seriously, the hard copy is about the size of a layer cake and weighs three times as much).

But the point is that you would think that if a body of linguists is going around printing big annotated lists of all the world’s languages, they’d have a hard-and-fast number ready to throw at you when you ask them how many there are. Instead you get answers like, “Around 6,500” or “Somewhere between 6000 and 7000” or “It really depends what you mean by language” or “Fewer all the time” or “6500! Who told you that? Their figure is clearly wrong!” Why all the hedging?

Language families! Well, one person’s idea of language families, ‘cuase there’s disagreement. And the consensus is constantly changing, so it’s one person’s idea of language families at one point in time… Long story short, linguists love to argue.

Well, there’s a couple reasons, and they all seem to boil down to the same thing: language is complicated. Seriously, I spend most of my spare time thinking about language and have for the past several years… and I keep coming back to “language is complicated”.

But let’s say you’re not satisfied with our sort of fluffyish numbers of how many languages there are. So you put on your linguist outfit (which, honestly, is probably just jeans and a t-shirt) and pack your rucksack and head out to count all the world’s languages.

But right away you start to run into problems. Let’s say you’re in India. You’ve been counting the Dravidian languages and getting some pretty good data, and you decide to really quickly check off Hindi. TotalNumberOfLanguages = TotalNumberOfLanguages + 1. So next you look at Urdu. Well, huh. The people that you’re interviewing say that they’re clearly different languages… but they’re mutually intelligible, so if one person is speaking Hindi and one person is speaking Urdu they can carry on a normal conversation. And you notice that they sound alike. A lot alike. In fact, their rules for how sounds change seem to be the same. And they have overlapping speaker populations. The best way you can find to tell them apart is that one (Hindi) is written using Devanagari script, and the other (Urdu) is written using Arabic script. Soo…. are they one language or two? Two different linguists might give you two different answers and then, BAM, you’ve got a different count.

And you’ll encounter similar problems with other languages and language families all across the globe. Do it three or four hundred times and you’ve got a radically different number from the next linguist counting.

But, despite these problems, you finally manage to count all the languages. You stumble home, travel-worn and weary, and settle down to check a truly massive back log of e-mails. And it looks like one of  them is very sad news. One of your friends you met along the way, the last living speaker of an Amazonian language, has died. And with her death, her language is gone. It’s a double tragedy, to be sure, and when you think back on, a lot of languages had only one or two very old speakers. Who knows how many of them have died in the years you’ve been travelling? So it looks like, due to language death, your count is even more off.

So you write up your findings and try to get them published, but people keep pointing out the hard and fast number number that you were searching for and collecting data for all along isn’t really reliable. Even if you end up giving a range of numbers, languages are dying so quickly that they won’t be accurate for long.

Man, I’m glad that this situation is a hypothetical one, and you didn’t actually spend years chasing after a mythic number that it turns out is unobtainable. Aren’t you?

Can Animals Talk?

There’s a long-running (and I really mean long-running: Plato chimed in on this one) debate about what language is. Now, as a linguist, you’d think I’d have the inside scoop on the subject. I mean, I pretty much sit around and think about language all day, so I should have this one down cold, right?

Well, as much as I hate to admit it, not really. And it’s not just me. Ask three different linguists what language is and you’ll probably get six to twelve answers. There is one point that I’m firm on, though: language is a human phenomena. Outside of the internet and fantasy worlds (which tend to overlap a lot, now that I think about it), animals don’t talk.

You talking to me
“…and then Mildred told Randolf that she thought his new haircut made him look like a basilisk. Well, you can imagine how he took *that*.”
I’m not claiming that animals don’t make communicative noises. Far from it! As someone who has bottle-fed more than one lamb, I can tell you that there’s a definite difference between cries that indicate genuine hunger and cries that are transparent ploys to get cookies.  But there are a lot of differences between communicative behavior and language.

  1. Lying is part of language.  By lying here, I mean a wide variety of linguistic expressions that express information that is counter to the truth, including joking. Language is separate from the things it describes (there’s nothing inherently tree-ish about the word tree, for example, ditto arbor, boom and träd, though there are inherent respect points in correctly identifying all three languages) and because of this can communicate abstract thought. Abstract thought, as evinced through lying, is in inherent part of language. There’s some evidence that Koko the gorilla is capable of lying, but one isolated incident really isn’t a sound basis for scientific argument.
  2. Language is generative. I’ve written an entire post about generativity, but it’s worth repeating. Language has to have underlying structures that can be used to produce new and novel utterances. Otherwise, you’re just saying random words.
  3. Langauge is communicative. This is part of the reason why music isn’t language, though it’s completely abstract and (at least in the Western tradition) generative. Abstraction is required, but so is a connection to thoughts and ideas. Tied to this is the fact that you have to have a community to speak in, even if it’s a community of two.
  4.  Language can communicate events at a temporal distance. This is a a biggie, and one of the main reasons that I really think that Koko and other talking animals are really using language. (Quick aside: Did you know that the Nazis attempted to train talking dogs as part of the war effort? True story.) It’s pretty easy to teach a dog to bark for a treat, but try teaching it to bark because you gave it a treat two days ago. You may think that a specific bark means “treat”, but without temporal distance and repeatability, it’s pretty much just pigeons in boxes.

Now, other linguists will take other positions (or, you know, the same position ;), but this is how I see it. So what do you think: can animals talk?

Why are tongue twisters tricky? (Part 2)

Oh, so in my last post I talked about the tongue part of tongue twisters. But, as I mentioned, the really interesting part of tongue twisters comes from the brain, not the tongue. It all has to do with lexical access.

Lexical access: The process through which a speaker or listener accesses their mental lexicon (i.e. the not-so-tiny brain dictionaries we all have and are all constantly changing).

If you were a computer, your mental entries on various words would be like files you needed to access. Like, if I write “kumquat”, you probably have some sort of mental entry for it. Even if you’ve never eaten a kumquat, you’ve probably seen them, so you have that mental image associated with the words–like a .txt file with a picture in it. So, once you hear or read “kumquat”, you need to rummage around until you find that file, then open it and access the information inside. And you do! In fact, you do it very, very quickly. You do it for every single word you ever read or hearand you do it in reverse for every word you ever write or say.

Brain Surface Gyri

This is your brain on language. Well, some of the bits that deal with language, at any rate.

So lexical access is a very important process. You need it for every single aspect of language use. The good news is, there’s a lot we know about lexical access! Remember when I talked about priming? That’s an aspect of lexical access. The bad news is, there’s a whole bunch we don’t know about lexical access.

(Is lexical access starting to sound like a fake word yet? That process, by the way, is known as semantic saturation.)

This is where tongue twisters come into play. (No, I hadn’t forgotten them.) Why? Well, it turns out that tongue twisters are a really good way to get at what happens during the lexical access process. Like many things that happen in the human brain, it can be difficult to study lexical access. Unlike physicists, linguists can’t break apart the mind to see what happens and figure out what’s going on. First, it would be deeply unethical. Second, when you break a brain open, it stops working. Sometimes, however, the brain does something weird. Like with tongue twisters. If you read my previous post, you know that the tongue itself can cause problems… but not enough to explain the most common errors, like saying “How can a clam cram in a clean cream can?” as “How clan a cam cram in a clean cleam can?”

In a 1999 study, Carolyn E. Wilshire found that there were two main contributors to making tongue twisters tricky. The first factor that made it easier to confuse sounds was whether the confuseable sounds were similar. There’s lots of technical reasons this is, but basically sounds can be grouped together and some sounds are more like other sounds. “t” and “d” are really similar, for example, whereas “k” and “m” are not really that alike. Unsurprisingly, sounds that are alike are easier to confuse. Basically, you reach for something that sounds similar, then realize that you made a mistake and try to correct your error.

The second factor was that it was easier to confuse sounds that were repeated. This is because you’re more likely to “reach for” something you’ve already gotten out once, even if it’s the wrong thing. Together, these factors make for some really awesome tongue twisters. Awesome for two reasons: the first is that they’re really, really hard to say. (Try moss knife noose muff!). The second is that we can use tongue twisters like these to help increase our understanding of the human mind. And that’s what it’s really all about.

Why are tongue twisters tricky? (Part 1)

For me, the best part of the 2009 Star Trek movie was the scene where Kirk tries to pick up Uhara in a bar. After she  says he probably doesn’t know what Xenolinguistics is, he replies:

The study of alien languages. Morphology, phonology, syntax. Means you’ve got a talented tongue.

Well. Four out of five isn’t bad. Linguists know about language, not the languages themselves–so tongue talents are a skill that linguists only develop tangentially. (Although, of course, a lot of linguists do end up learning the languages they work on.) But knowing about tongues is still pretty useful. For example, it helps explain why tongue twisters are so hard.

Rolled tongue flikr
Tongue rolling is actually probably not controlled genetically. That’s right, your introductory biology textbook lied to you.
Basically, your tongue is a muscle like any other muscle, and it has certain limits. For example, there’s just a certain upper limit to how fast you can type, knit or eat, you can only produce recognizable words so fast. In addition to speed, however, there are certain motions that are difficult to make. Linguists often refer to correctly producing a given sound as “hitting an articulatory target”, and that’s a useful metaphor. For each sound in your repertoire, your tongue (and other parts of your articulatory system) have to be in certain positions.

Exercise time! Try saying “s sh s sh s sh” and “t k t k t k”. In the first, the tip of your tongue should move from that little ridge in your mouth (just behind your front teeth) to behind that ridge. In the second, the “t” sound should be made with the tip of your tongue against the roof of your mouth, and the “k” with the very back of your tongue. (I’m assuming that you have the “sh” sound in your native language, otherwise this exercise might have been a little fruitless for you. Sorry.)

You might have noticed that it was a little easier to make the “t k” sound than it was the “s sh” sound. That’s because you’re using two different parts of your tongue to make the “t k” pair, so while one is making a sound, the other is preparing to and vice versa. On the other hand, if your only using the very tip of your tongue, you have to finish one task before you can move onto the next, so your rate of making sounds is much lower. It’s the same reason that assembly lines are so much faster–you don’t have the additional time it takes to switch tasks. (BTW, that’s why, if you play a wind instrument, you can tongue faster with “t k t k” than “t t t t”.)

Ok, so we have two pressures working against your ability to produce tongue twisters. The first is that your tongue can only move so fast. You can train it to move faster, but eventually you will (barring cybernetic implants) reach the limits of the human body. Secondly, you have limited resources to make sounds, and when sounds that draw on the same resources are produced too close together, they both become more difficult to produce, and the speed at which you can produce them is reduced even further.

And that wraps it up for tongues, because the dirty secret of tongue twisters is that they’re mainly actually brain twisters. But I’ll cover that in part two. That’s right: brraaaaaaains.

Do subliminal messages (join the Illumanati) work?

Since I recently joined the Illumanati, I’ve been thinking a lot about subliminal messages.  A “subliminal stimulus” is literally something that’s below–“sub”–your perceptual limit– “liminal”. So, linguistically speaking, it would be a word that was presented to you to be read but removed before you could actually read it. Or a sound that’s too soft for you to hear.

Dollar Illuminatus
Holy crap, look at that subliminal owl! Oh wait, you see it? Well shoot. Guess it’s just a regular old liminal Illuminati owl.
So if you can’t really perceive subliminal messages, why are they even a thing?

Well, if there’s one thing I’ve learned by studying linguistics, it’s that language is complex and that there are huge gaps between what we know and what we think we know about language–at least on an individual level. (I’ve also learned that I can’t count, becuase that’s definitely two things.) We all do things that we have no idea we’re doing, and so quickly and easily that they just slip below our notice. Linguistics is all about figuring out what those things are.

One of those things is priming. Basically, when you hear or read a word it gets “warm”, like how you leave a heat signature on your sheets after you get out of bed. And if, later, you’re looking for a similar word, you’re more  likely to go back to your warm bed than another word you haven’t used lately. Of course, the effect fades over time, but it does fade very slowly. And priming effects are where you really see an effect of subliminal messages. (And, just to be clear, not really anywhere else… at least linguistically speaking. 🙂

For example, this study by Abrams, Klinger and Greewald found that, if participants had been exposed to a word earlier in their study, they were able to recognize it later when it was presented subliminally–but it only works really  well when participants had not only read the word before, but had had to think about it some by assigning it to a category. So the effect isn’t really strong enough to, say, help you lose weight or stop smoking.

The fact that it exists at all, though, does tell us something interesting about the human brain and how it uses language. For example, is our ability to interpret stimuli that are degraded tied to the pressure to understand conversational speech, even in noisy environments? What does it mean that the effect is also present with visual stimuli? Like all sciences, linguistics is all about asking the right questions, and research on subliminal stimuli opens up a whole barrel-full of questions.

Seeing noise?

Some of you may be familiar with synesthesia, a neurological condition where you perceive sensory input from one sense as if it were another sense–with synesthesia the color yellow might taste like root beer, or the sound of a bassoon may feel like bread dough. Even without synesthesia, however, linguists (particularly phoneticians and phonologists) see sound all the time. What does it look like? Something like this:

Auuuugh what is this? It looks so boring and spiky! My eyes!

These, boys and girls and others, are what your speech sounds look like. Spectrograms are one of the most useful tools in the speech scientist’s tool shed. Heck, they’re pretty much a Swiss army shovel. You can spend your entire career basically only looking at data in this one form.

Why? Well, there’s a lot of data in a spectrogram. Big things, like whether a sound’s a ‘b’ or a ‘p’ (there’s a big black bar on the bottom if it’s a ‘b’, but not if it’s a ‘p’), but also really small things that we as humans have have a really hard time hearing. Like, remember what I said earlier about your ears lying to you? Turns out it’s a lot easier to sort out the truth if you can see what you’re hearing. Plus, by looking at spectrogram we can quantify things like average vowel frequencies really quickly and easily. (Turns out, by the way, that you can [maybe, kinda, if you squint just right and have just the right voice sample] judge how tall someone is based on their vowel frequencies.)

But spectrograms aren’t just a serious scientific tool; they’re also pretty fun. Aphex Twin, an ambient musician (I mean, he makes music in the ambient genre, not that he provide background music at canape parties. Sheesh.) uses spectrograms as an art form. This song, for example, has a picture of his face encoded in it’s spectrogram. Give it a listen and see if you can find it!

On a more general note, the study of images made with sound is known as cymatics. I’m just going to leave this video here for the more physics-minded among you: