Pop culture

Constituency and linearity in Heptapod

The constructed language created for Arrival (2016) may be just a tiny bit more human-like than you would think.

I know what you’re thinking: “But Arrival was released in cinemas two years ago!” I didn’t have a linguistics blog two years ago, so here’s my contribution to one aspect of the linguistic details featured in the movie.

And if you’re one of those people thinking, “But Heptapod isn’t even a real language!” I know, but let’s imagine for a second that it is.

For those of you who haven’t yet watched Arrival, the plot can be summarized as follows: the linguist and polyglot Louise Banks is asked to help develop an interpretive method for communicating with the so-called Heptapods, a species of aliens who mysteriously came to Earth and don’t seem to be using any human-like means of communication.

The movie is great because it’s one of the only high-budget, award-winning movie whose protagonist is a linguistics professor. What this does is make the study of linguistics more accessible to the general public and inspire more young people to embark on linguistics studies. While the movie definitely has its flaws, especially by assuming the Sapir-Whorf hypothesis, is a topic for a different post.

But let’s assume for a second that Sapir and Whorf are right, and that the language you speak entirely shapes your perception of the world. Speaking Heptapod will make you see all of time simultaneously. Your entire timeline is put out in front of you, and this is because you master a language in which all separate elements occur at the same time. How does this compare to natural languages?

This is where constituents come in. If you’re new to syntax, let me give you a general, minimalist intro à la Adger (2012). You can also listen to the Lingthusiasm podcast if you want more explanation here.

Each word makes up a constituent of its own. In a tree representation, this is what that would look like:


Constituency is marked in the syntax tree through phrases. If we add a determiner, we have a constituent at a higher level.

(2)the car

This type of representation of constituents accounts for the fact that some combinations of words may be moved around, while others can’t. This can be exemplified through wh-questions. Consider a context in which A asks B what they did yesterday, and what they did was to eat pizza. (4-a) shows an acceptable response, while (4-b) sounds off.

(3) A: What did you do yesterday?

(4) a) B: Eat pizza.

b) B: #Pizza.

The reason why (4-b) sounds weirder than (4-a) can be explained by what it is that what replaces, namely a verb phrase. The role of B is then to fill in the gap left by the wh-word. This makes it sound weird when B replaces the VP gap with an NP.

Now, back to Arrival. The idea of constituents is important to consider because it puts emphasis on the fact that languages are not linear. Even though, when we talk, one word has to come before the other, that’s not necessarily how the structure is organized in our head. The way we move constituents tells us that syntactic structures are hierarchical rather than linear. So my question is: is Heptapod hierarchical? That is, does Heptapod have constituents?

Actually testing movement in the language is impossible, since it’s not actually a language. Some clues appear in the movie, though.


The written language of Heptapod was created by software designer Stephen Wolfram and his son, Christopher Wolfram. The program used by Louise in the movie to decode and create new sentences in the language divided each circle into twelve sections. Here’s some more information about the logograms, if you’re interested. It appears that each section may contain several subsections, too, at least in the code, but the way Louise uses the software in the movie makes it look like each section represents a word or concept.


While Louise is using the software, she picks out three symbols, reading out ‘give’, ‘technology’ and ‘now’. I think it’s fair to assume that each of the twelve sections of the circle represents a constituent. Whether ‘give’ in Heptapod is just one word, or whether that is just the English translation of a constituent including the verb to give with the imperative interpretation, does not make a difference in terms of whether it is a single constituent. The main difference between Heptapod and natural languages then seems to be that instead of having a hierarchical structure, Heptapod has the entire sentence produced or interpreted at once.

As far as I can tell, there is no hierarchy within the constituents, either: only loose words or concepts attached to it. Here is one example of the constituent structure within the software programmed by the Wolframs:


If anyone has any idea what sort of hierarchy or internal structure these constituents could have, I would very much be interested in hearing your thoughts. But for now I think I’ve established that Heptapod does have constituents, though not hierarchical like in natural languages, which is what makes it so weird and (supposedly) changes speakers’ view of space and time. But what do I know.



If you enjoyed this post, you may be interested in reading my article about interesting syntactic and sociolinguistic aspects of Klingon, Dothraki and Sindarin which was published in the Norwegian students’ journal Riss this year. You can find out how to order a copy of the journal here. (The article is in Norwegian, though.)