Stephantics

So, as the fast-paced story comes to its close, the characters discover that the sound-hunting monsters’ weakness is, fittingly, the very device that Simmonds’ character Regan uses to help her hear. But the detail that filmgoers seem to have missed is exactly what’s happening in those final few minutes before the credits. Many reviewers seem to have taken Regan to be wearing a conventional hearing aid, and have gone on to assume that the evil invaders induce a sound in it that both Regan and them found too painful to bear, giving our heroes some hope for the future. But you might have noticed that part of Regan’s equipment attaches to just behind her ear, as if by magnet. In fact, this is because she’s not using a standard hearing aid at all, but a cochlear implant, which doesn’t actually have the capacity to produce sound, seemingly opening up an enormous hole in the middle of the film’s plot!

Specifically, hearing aids contain both a microphone and a speaker, amplifying the sounds around a wearer and broadcasting them down the ear canal. But for those who have suffered severe damage to the cells inside the inner ear, which are responsible for processing that sound, amplification isn’t enough; the receiver simply isn’t operating, and no level of volume will do the trick. With a cochlear implant, electrical signals are transmitted directly into the cochlea instead, which contains a special membrane. And as I mentioned in another post about the inner workings of the ear, this membrane acts a tonotopic map, which means that different locations correspond to different frequencies. And so it follows that a direct electrical signal can cause a wearer to experience different sounds — at least, enough to perceive and understand speech.

But if cochlear implants don’t produce any audible sound, doesn’t the whole movie fall apart? Not quite, thanks to the filmmakers being clever enough to provide all we need to work out what’s really going on.

If you pay close enough attention, you might have noticed that every time the creatures come around, the lights flicker off and on. This strongly suggests they produce some kind of electrical field, maybe as a result of having electroreception, like a shark. In other words, they use their own electrical field, and its tendency to interact with the surroundings, to locate their prey in space (this would be on top of their remarkable hearing). This is made most obvious when one of them angrily attacks an old CRT monitor displaying static; as this handy explainer makes clear, the static we see is just the ambient electromagnetic radiation being translated into an image of a snowy mess. Which means the creature is reacting to its own body’s signal fed back to it. And if that same electrical field can induce a reaction in a cochlear implant, which goes on to generate its own electrical signal in response, the implant might be able to irritate the monster as much as the TV did. In other words, it isn’t really sound at all that ends up making the difference.

(One potential problem with this theory is that Regan seems to triumphantly dispatch the monster’s protective armour by holding the external part of her implant up to a microphone, apparently amplifying its sound — a sound that should’t even exist! But this becomes less of an issue when we remember that any device that’s capable of producing an electrical signal has the potential to interfere with audio equipment. For instance, just think about the interference you hear when your cell phone goes off near a set of speakers. And, so, it isn’t so much the sound that matters, but the creature’s own electrical signal being amplified back towards it.)

So, the short answer is pattern recognition. Just like how a physicist might toss a ball up into the air a bunch of times to figure out which rules it follows coming down (or toss some baryons at each other to see what comes out the other side), a linguist will observe how the sounds and symbols of a language regularly connect up to meaning. With even a small amount of data, a linguist can begin the process of uncovering the relationships that hold between speech and the world it’s used to represent.

What does this look like? Well, we don’t have any real alien languages to work with (for now), but we’ve got plenty of human languages to choose from -- between six and seven thousand. So, let’s pick one, look at a few examples, and see what we find!

To get as far away from English as we can, let’s go to Kashaya -- an endangered Pomoan language spoken in California. Unlike English, which is analytic and tends to keep its morpheme-to-word ratio low (relying heavily on word order to keep track of who’s doing what to whom), Kashaya is polysynthetic; it tends to pack lots of information into each word, by way of affixation, and has relatively free word order.

Now, about half the world’s languages don’t have any developed writing system associated with them; with those that do, many lack native representation, relying instead on some version of the Latin alphabet. Kashaya is one such language, with its written form having been developed by Dr. Robert Oswalt in the 1960s. In the examples that follow, you can think of each word as being written in a kind of quasi-IPA; the words on the left represent the actual sounds of the language, as understood and transcribed by a linguist, while the meanings hang in the right-hand column.

bahcúw ‘to jump’ bahcubedu ‘always jumps’ coqocedu ‘always shoots’ coqów ‘to shoot’ kelci 'peek!’ kelcíw 'to peek’ coqo ‘shoot!’

(At this point, I should admit I grabbed this data from a linguistics assignment. In the context of the film, the Heptapod language doesn’t come with an instruction manual, so Louise has to employ a monolingual demonstration --  a technique used to elicit linguistic data when no common tongue exists between the parties involved. If you want to see this in action [you do], I highly recommend having a look here. In under 40 minutes, and without any assistance, linguist Dan Everett works with a speaker of a language he has no knowledge of and manages to uncover basic facts about its word order, sound inventory, and morphology. It’s really an impressive thing to behold, even for other linguists.)

Let’s start by choosing two words which share some meaning, to see if we can figure out what else they have in common. Beginning with the words for ‘to jump’ and ‘always jumps,’ we might tentatively conclude that what they share -- “bahc” -- means ‘jump,’ and that “úw” and “ubedu” mark the verb as being tenseless and habitual, respectively. But throwing a third word into the mix -- “coqocedu” for ‘always shoots’ -- forces us to either reject or revise our hypothesis; “ubedu” can’t mean ‘always,’ since the only common sounds between the second and third words are “edu.” What can we do?

Here, it helps to know a bit about what languages sometimes have up their sleeves. In this case, it’s important to notice that the first consonants -- “b” and “c” -- are repeated in “bahcubedu” and “coqocedu.” This looks like reduplication, which involves the repetition of some or all of a root word, sometimes with some modification. It’s not as exotic as it sounds, since English speakers use this strategy when treating something dismissively (e.g., “alien-shmalien”). So, Kashaya might instead express ‘always’ using partial reduplication, followed up with “edu.” We’d need more data to know for sure, but it’s a good start.

As for those shifting vowels -- “bahcubedu” vs. “coqocedu”? Maybe they’re not part of the the affixes, after all; maybe they’re part of the verb! In fact, the fourth word -- “coqów” -- seems to confirm this, since they each tend to stick to one verb root pretty consistently. Which means we need to say something more about that tenseless marker: it’s really just a “w,” with some stress added onto the last syllable. And, again, it’s not so unusual for stress to play a role in grammar; “complex” can mean either “complicated” or “a bunch of buildings,” depending on where you put the main emphasis.

Looking at the rest of the list, we can add that the absence of an affix altogether apparently marks the verb as imperative, meaning it can be used to issue a command. So, all in all, just 7 words have told us something about the sounds used in Kashaya, whether or not it has stress and reduplication, how to mark verbs as tenseless, habitual, or imperative, and how to say “jump,” “shoot,” and “peek.” Not bad, eh?

And this is just how Louise approaches Heptapod B, the film’s alien writing system. You can see for yourself, in this behind-the-scenes snapshot, that each logogram can be divided up into 12 segments. (Having an upper limit to sentences isn’t something you really see in human language, but of course this isn’t a human language!)

Looking closer, we can see the kinds of patterns we’d need to pay attention to, to begin cracking open the language. In the image below, the lowest righthand portion is shared across both logograms. (Actually, the bottom twelfth is the same across both of them, too. Maybe, like in Kashaya, this absence communicates something meaningful!)

And armed with this understanding, Louise devises a program to automatically analyze the logograms’ parts.

She even reverses the process by taking that database of segments and using it to construct her own logograms, in order to pose the all-important questions at the heart of the story.

Too cool!

It’s encouraging that the movie managed to capture the process so well (if only briefly), and it’s impressive that it’s resonated with audiences so much. I know I speak for more that just myself when I dare to hope this means we’ll be seeing more of a willingness on moviemakers’ parts to represent real science on the big screen. It might’ve sounded farfetched to say so just a few years ago, but a little linguistics in Hollywood seems not to be such an out-of-this-wolrd idea, after all!

If you enjoyed this story, and want to learn more about the linguists who worked on Arrival, definitely have a look at this!

Like we’ve talked about before, the theory that linguists developed back in the 1970s had it that every kind of phrase fit into the same basic shape — the X-bar schema. This meant that even simple nouns like “tree” would have a structure that looks like this:

That might seem like we’ve needlessly put a very small word into a very big house, but that extra room is crucial when determiners, adjectives, and prepositional phrases come over to visit:

With English being pretty easy-going when it comes to more than one noun living under the same roof, we’ve got our Christmas tree:

And with more recent developments in syntactic theory, like the DP Hypothesis discussed back in Episode 79, we have some good reasons to think that even bare noun phrases are actually cozily packaged up inside determiner phrases, like a present waiting to be unwrapped (think about how, in morphologically richer languages like French, even a simple phrase like “trees” needs an article before it, as in ”les arbres”):

But what about that “O”? What is it, even? Some version of “oh”? And where should it go?

At first, I just assumed it was an interjection (like “wow”) — one of the eight parts of speech that make up traditional grammar. Not that I’d thought about it much up until now, mind you. But, that seemed to be a reasonable guess. Problem is, the expression “oh” is typically associated with fear, surprise, or delight. In this case, it looks more like it’s pointing out the addressee, which is the person (or thing) being, well, addressed. After all, the accompanying lyrics mostly go on to describe the eponymous tree. That “O” looks to be doing the same job here that it’s doing in Canada’s national anthem ‘O Canada,’ or Walt Whitman’s poem ‘O Captain! My Captain!’

The good news is, linguists have a name for this! That “O” is a vocative particle. The bad news? That doesn’t mean much. The word “vocative” is there to point out that we’re dealing with vocative phrases, which are typically used to explicitly address the person being spoken to. In the lyric below, from the 1953 song ‘Santa Baby,’ that first bit functions as a vocative phrase, addressing the whole song to Santa Claus.

(1) Santa baby, slip a sable under the tree for me

English doesn’t oblige us to use “O,” but many languages regularly make use of such words. Catalan employs the particles “ei” and “eh.” Irish uses “a.” And, in fact, the “O” we’re currently interested in is just a carry-over from the original German ‘O Tannenbaum.’

But “particle”? That’s just what linguists call anything that they don’t know what it is. The infinitival marker “to” in English is sometimes called a particle, along with many other words in many other languages. By itself, it doesn’t tell us much. So what does the literature on this stuff have to say?

Unfortunately, as it turns out, there doesn’t really seem to be much research to speak of. And what little there is doesn’t seem to point to much of a consensus on the matter. So, how did we settle on the structure for our shirt? Well, to start, there seems to be a fundamental disagreement about the nature of vocative phrases: are they constituents?

Informally, a constituent is just a string of words that behave as a group. And on the surface, it might seem strange to question whether or not such phrases are bound together. But consider the following Italian data; while language generally allows any two like phrases to connect together with an “and” sandwiched between them, vocatives resist the trend.

(2a) O Maria e Pietro, Gianni è arrivato.        o Maria and Pietro, Gianni is arrived

(2b) *O Maria e o Pietro, Gianni è arrivato.        o Maria and o Pietro, Gianni is arrived

Italian lets you double up on what’s inside the phrase (2a), but you can’t do the same for the whole thing over (2b). For some, this suggests that vocative particles aren’t really part of the noun phrases they’re coupled with, and instead form part of the structure of the overall clause. That way, only one ever rears its head, in the same way questions only ever have enough room for one set of question-y words at a time (3); “what will” and “when will” can’t coordinate with each other, since neither pair forms a constituent, with each word instead occupying a pre-determined position inside the sentence.

(3a) What will you be giving them Christmas Day?

(3b) When will you be giving them gifts?

(3c) *What will and when will you be giving them?

The general blueprint of sentences, then, would limit the number of particle-y vocatives to exactly one, at the very top, above everything else.

Still, others insist our dear “O” is closer to its noun phrase than to the clause that follows. And since it’s hard to make a shirt out of a non-constituent, we took the path of least resistance. That still leaves us with an open question, though: if vocative phrases are just bulked up noun phrases, why put the “O” up at the top, instead of putting it inside that empty Voc position a bit lower down? If anything’s going to form the core part of a vocative phrase, shouldn’t it be our particle?

Enter Latin. Many languages mark their vocatives with affixes instead of particles. William Shakespeare’s famous line “Et tu, Brute?” in the play ‘Julius Caesar’ has “Brutus” in his vocative form, with that “-e” replacing the nominative suffix “-us.” Other languages, like Romanian, can both directly mark their nouns (e.g., “grandpa”) as vocative, while also making use of a particle:

(4) bre tataie      PRT grand’pa.VOC

The takeaway is that the information that a phrase is being used vocatively spreads our across two separate positions — one that can connect up directly to the noun, and one that shows up beside it. (French negation works similarly, as in “Ils n'aiment pas Noël.”) Any given language can make use of one of these positions (Latin), the other (archaic English), neither (Modern English), or both (Romanian).

In cases where the noun bears an affix, we suppose it’s moved up out of the NP/DP combo and into that Voc position, so that (4)’s structure looks something like this:

We can tell this is probably the right way to think about what’s going on, since names in Romanian (e.g., “Ion”) contain information related not only to their being marked as vocative, but to their being definite (i.e., as though they’ve picked up a hitchhiking article along their way up the tree):

(5) Ionelule      Ionel.the.VOC

So, at last, we have our structure for “O Christmas tree.” Without any vocative affixes to speak of in English, the nouns stay put, and our lovely little particle “O” perches itself up on the highest branch. Mix in some atelierMUSE magic, et voilà!