Short: Infinite Fractal Complexity

Podcast – Conlangery Podcast

This episode, George gives a short discussion of the idea of language as having infinite fractal complexity, and what this means for conlangers building fictional worlds.

Special Mention: Resources on the Line 3 protest: Stop Line 3, Center for Protest Law and Litigation, Sierra Club Fact Sheet, Line 3 Legal Defense Fund

Original Script

Welcome to Conlangery, the podcast abou t constructed languages and the people who create them. I’m George Corley.

Today, I’m going to talk a little about the realities of what naturalistic conlangers are trying to simulate. What does it mean for a language to look natural or realistic, and can a conlanger actually create something as complex as a natural language? I’m going to suggest that you ultimately can’t, but I also think that you don’t have to. Most people’s goals in conlanging will not really approach that, and I’m going to talk a little bit about how to decide what you really need out of your conlang.

Instead of doing my normal Patreon pitch, I wanted to draw attention to something I think is important to point to. You may have heard of Line 3, the pipeline that is being built in Minnesota to bring tar sands from Alberta into Wisconsin. This pipeline is going through Anishnaabe land. It has the potential to pollute waters through much of the United States, and it’s going to contribute greatly to climate change. I would encourage you guys to go to stopline3.org. I’m also going to link a couple of other resources in the shownotes, and I have decided to make a small donation to the Center for Protest Law and Litigation, which is providing some legal defense funds to people who are protesting the pipeline. I would encourage people to learn about what’s going on here.

Of course, there are many important reasons to oppose Line 3. It’s going to have huge ecological impacts. It’s going to impact water in a huge area. It’s going to contribute to climate change. And it is going through treaty land. I, personally, feel the need to highlight it in this podcast specifically because we, as conlangers, often draw inspiration from indigenous languages, and I, myself, have drawn inspiration from the Nishnaabemowin language, also known as Ojibwe, so it seems kind of wrong to take that inspiration and not care about the issues of the actual people who speak those languages. But this is up to you guys as individuals, what you want to do to support this cause. I just want to raise some awareness and let you guys know that I’ll make that small donation. Thank you.

Now on to natural languages. Many of us conlangers have a goal of creating a language that at least looks like a natural language, and people do succeed at that in varying degrees. In some ways, it’s not so difficult. There is a reason that ANADEW, A Natlang Already Did it Except Worse, is such a common term in the community. You have to almost deliberately go out to the edge to come up with some grammar or phonology that really looks impossible for humans to come up with naturally. The lexicon can be a bit harder, but with some work, you can avoid relexing and come up with realistic senses for words.

What is difficult, and likely impossible, is to come up with the massive amount of variation in language. You can work on dialects and registers all you want, but you won’t really get to the complexity we see in the real world. The reason for that, I’m going to propose, is because in the real world, natlangs have infinite fractal complexity.

What do I mean by infinite fractal complexity? Let’s start with your language. For the sake of this exercise, let’s assume a language that is relatively unified and not part of a dialect continuum. That language can naturally be divided into a number of dialects, based either on geography or on social divisions, though most likely both. But those dialect divisions are not hard lines, and there is variation within each dialect. You can subdivide and subdivide until you get to the idiolects of individual people.

You might assume that idiolects are atomic, but they’re not. Even there, you’re going to run into variation. Most people code switch between several different varieties, even if they are monolingual. Not only are there different registers for different situations, but you subtly change the way you speak in individual interaction. Sociolinguists often model this by simply taking statistics on how often one variant of a word or construction is used as opposed to others and reporting the percentage. Those percentages change depending on who you’re talking to, and it may even change over the course of a conversation.

Add to this the fact that every speaker knows thousands, even tens of thousands of words, and each speaker may have slightly different understandings of their meanings. There are also thousands of collocations, idioms, and combinations involved. And of course, infinitely many unique sentences that can be constructed.

All of this extends back into history as well. Historical linguistics attempts to classify languages into neat family trees, but this is still an abstraction. In reality, every reconstructed proto-language is really just an approximation of a messy collection of different dialects. There were many branches off of our languages that we will never know, and others that were reabsorbed into another, surviving branch, possibly leaving traces behind. Words take their own individual journeys, branching out through derivation or hopping across languages in unique patterns reflecting trade routes or migrations. We even have mysterious words that may be from languages that we otherwise know nothing about.

So far, I think I’ve impressed on people the fact that it is truly impossible for anyone to construct a language that truly approaches the complexity of what happens in natural languages. The question, I guess, is “What do we do about it?”

The seeds of this short are in my interview with Lauren Gawne. We talked a little bit there about determining how much of a language you need for worldbuilding. Lauren told me that Aramteskan mainly just needed enough fleshed out for the needs of the book, though she did go beyond that a bit. My own current project mostly requires a number of naming languages, with some of them related to others. That led me to do some significant historical work on sound changes, in order to create families, but only to get me to phonology and basic morphology. Depending on how the story goes, I might actually do some significant grammar work on two of the languages, but that really depends on whether I decide to include any dialogue in those languages.

To understand what my goal should be, I had to think about what these languages were for. This is a story set in what is, for lack of a better term, magic grad school. I have two protagonists and three other viewpoint characters, all from different cultures, and probably from three different language families. There are also other characters: students and professors from a variety of cultures, plus locals to the area the university is in, which I plan to be speaking another language. Being in something resembling a modern academic environment, I intend for characters to be citing names from various cultures, too, potentially including names from significantly earlier times.

All of these factors led me to think that I needed the basic structure of at least two language families with enough sound changes and morphology to make names, and possibly a few extra isolates with room to expand on. In addition, at most two languages might need to generate sentences for this story, but it remains to be seen whether I’ll actually have to go that far. My historical work, for now, is limited to sound changes. I may need to add a few grammaticalized forms for the more synthetic languages, but I’m not going to go all out on verb forms or syntax evolution until I decide I need to make sentences.

There isn’t really a standard shape for what you need here. The complexity is still something to think about, but precisely what parts of that complexity you need to simulate is something you need to consider when you’re building. For one world, you may decide that you will only ever need names for one culture, and so you just make one naming language. For another, you might decide that you want a few phrases in an archaic form of a language, so build one language out to build those sentences, and maybe do a little bit of historical work to get modern names or a bit of a modern version of a language.

But what if you want two characters from the same linguistic culture communicating in the language, well, then you want to have that one language built out, but you also want to think about not just broad strokes diversity, but the particular idiolects of these two characters. You want to ask if they come from different regions. You want to ask if their social class, gender, or other social identifications affect how they speak. Perhaps most importantly for this one dialogue, you want to work out just how these two particular characters would speak to each other. That involves some thinking about register, politeness, and relative social standing, but also about the emotional state of these particular characters and their own tendencies to adhere to norms or not. That sounds like a lot of detailed conlanging work, but it’s specific to the two characters in the dialogue. Some of it has implications that could affect other things down the road, but you only need enough framework there that you can expand later if you need to. You can focus on the differences between these characters, and along the way, work out some basic ideas of how to expand things in the future.

Now, there are plenty of conlangers out there who enjoy going deep into a language without some particular application in worldbuilding, and there are always people who don’t mind building more than you need. If you are building a language for a work of fiction, though, I hope these ideas are helpful to you. The bottom line is, language is fantastically complex in myriad ways, but you don’t need to deal with all of that complexity at once. Consider what you need right now, and where you need room to expand in the future.