Constructed Languages

There's something almost cosmically arrogant about inventing a language from scratch — the presumption that you, a single human, can improve on what billions of speakers and thousands of years of drift and collision have produced. And yet people keep doing it. More than nine hundred constructed languages exist on record, from Hildegard von Bingen's twelfth-century Lingua Ignota (a thousand words, listed hierarchically from God down to cricket) to Tolkien's Elvish to the language spoken by pseudo-Mongol nomads on HBO. Almost all of them have failed. The ones worth thinking about are the ones that failed in interesting ways.

The Two Impulses

Conlangs tend to spring from one of two impulses: the desire to make language more perfect or the desire to make it more universal. These pull in opposite directions, and the tension between them is the story of the field.¹

The perfection impulse traces back to Leibniz, Bacon, and Descartes, who saw natural languages as corrupted instruments that clouded thought. If you could build a language from rational principles — a taxonomic tree where each syllable encoded a branch of meaning — you could think more clearly, maybe even resolve philosophical disputes by calculation. John Wilkins actually tried this in 1668, producing a sprawling classification scheme in which every concept had a precise address. Robert Hooke was so impressed he published a paper on pocket watches in it. Nobody else bothered. The language was too burdensome, though Wilkins's organizational scheme survives — it was a precursor to the modern thesaurus.

The universality impulse came later, in the nineteenth century. Forget perfection; just get everyone talking. Solresol, built from the seven notes of the musical scale, let you sing your sentences or weave them into textiles as colored thread. Esperanto, invented in the 1880s by a Jewish doctor from Bialystok who watched Poles, Russians, Germans, and Jews fail to communicate in his hometown, came closest to success — two million speakers at peak, its own literature, and at least a thousand native speakers including George Soros. Two world wars and the rise of global English killed the dream, but Esperanto still has more speakers than most of the world's six thousand natural languages.

Ithkuil: The Perfection Limit

John Quijada spent thirty-four years building Ithkuil while working at the California DMV. It is possibly the most ambitious conlang ever attempted: a language designed to be simultaneously maximally precise and maximally concise. Ideas that require clunky circumlocution in English collapse into single words. "On the contrary, I think it may turn out that this rugged mountain range trails off at some point" becomes Tram-mlöi hhâsmařpţuktôx.¹

Ithkuil has twenty-two grammatical categories for verbs (English has six), eighteen hundred suffixes, and fifty-eight phonemes including a voiceless uvular ejective affricate found in only a handful of other languages. To translate "gawk" requires fifteen minutes of introspection — selecting the degree of surprise (more than unpreparedness, less than shock), the degree of impropriety (less than scandalous, more than eyebrow-raising), the physiological response format — and produces apq'uxasiu. Every possible human thought gets a unique set of coordinates.

The problem is exactly what you'd expect: nobody can speak it. When cognitive linguist George Lakoff finally examined Ithkuil, he called it "an impossible language" but also "fascinating as conceptual art." His deeper critique was that Quijada had misunderstood the nature of metaphor — you can't grammar your way out of figurative thought, because metaphors aren't bugs in language, they're how cognition works. Life is a journey in some deep structural sense that no suffix can disambiguate.

What happened next is stranger than the language itself. A group of Ukrainian psychoneticists adopted Ithkuil as a consciousness-expansion tool, invited Quijada to conferences in Kalmykia (the only majority-Buddhist region west of the Urals), and made it mandatory curriculum at their university. Their leader talked about creating "intellectual special forces" for a reborn Slavic superstate. One conference attendee turned out to be Ukraine's second-most-wanted domestic terrorist. Quijada, horrified, cut all ties. The history of constructed languages is full of this kind of thing — creators losing control of their creations, communities warping the original intent. Charles Bliss, a Buchenwald survivor, became unhinged when teachers modified his pictographic language to help children with cerebral palsy. Volapük died because its creator refused to let anyone else coin words.

Toki Pona: The Simplicity Limit

Where Ithkuil pushes toward infinite precision, Toki Pona pushes toward radical reduction. It has 123 words. Five colors. Numbers up to about twenty (and even that's contested — purists think you only need wan, tu, and mute: one, two, several). You can learn it in thirty hours.²

The creator, Sonja Lang, was inspired by Taoist minimalism and hunter-gatherer life. What is a car? It depends: tomo tawa (space for movement) if you're driving it, kiwen utala (hard striking object) if it just hit you. There's no word for "please" or "thank you." If you want to be polite, you do a little Japanese-style nod. The word for "simple" and the word for "good" are the same: pona.

This isn't just whimsy. Toki Pona makes a genuine philosophical claim: that the precision we think we need is mostly noise, and that stripping it away forces a kind of mindful attention to what you actually mean. Speakers report feeling more honest. "I'm more comfortable now with the things I don't know," says one polyglot. The language's dependence on context and subjectivity becomes an exercise in perspective-taking — you have to model your listener's understanding of the world to be understood at all.

The comparison between Ithkuil and Toki Pona is itself illuminating. As Quijada put it with uncharacteristic concision: "It's the difference between John Cage's 4'33" and a Beethoven symphony." But there's a deeper lesson. Both languages assume you can separate the structure of thought from the habits of culture. Both discover, in different ways, that you can't. Toki Pona speakers immediately reinvent politeness norms through gesture. Ithkuil speakers can describe gawking with surgical precision but can't actually hold a conversation.

The Sapir-Whorf Shadow

Behind every conlang lurks the Sapir-Whorf hypothesis — the idea that language shapes thought, or even constrains the thoughts you can have. Named after Edward Sapir and Benjamin Whorf (an insurance inspector who studied linguistics in his spare time), the strong version has "sunk into disrepute among respectable linguists." But weaker versions keep getting empirical support: speakers of gendered languages really do think about objects differently depending on grammatical gender, and those conceptual biases persist even after learning a non-gendered language.

Conlangers are some of the last true believers in a strong Sapir-Whorf. If language constrains thought, then building a better language is building a better mind. The psychoneticists in Kiev took this to its lunatic conclusion, but the impulse itself isn't crazy — it's the same impulse behind mathematical notation, programming languages, and the periodic table. All are constructed symbol systems that make certain thoughts easier to have.

The real question, the one Riau Indonesian answers from the opposite direction, is how little language you can get away with. Riau Indonesian has no tense, no articles, no fixed word order, no distinction between "he," "she," "it," and "they." The phrase ayam makan — chicken eat — can mean the chicken is eating, the chicken was eaten, someone is eating for the chicken, or where the chicken is eating. Context does all the work.³ This didn't happen by design; it happened because generations of second-language learners stripped away complexity until only the communicative essentials remained. It's a natural-born Toki Pona — proof that humans will always find ways to understand each other, even with almost nothing to work with.

The Conlang Community

The interesting thing about constructed languages isn't just the languages — it's the communities that form around them. Toki Pona demonstrates this best. When someone translated Andy Weir's short story "The Egg" into Toki Pona — sike mama, literally "mother sphere" — the exercise revealed how the language forces a particular kind of thinking.⁴ The concept of reincarnation becomes mi sin e lon sina — "I renew your existence." Heaven and hell become ma moli pona and ma moli ike — "good death-place" and "bad death-place." The vocabulary constraints don't impoverish the story; they strip it to its philosophical skeleton. Every translation choice is an argument about what the original sentence actually means.

Conlang relays — games where a text is translated from one constructed language to another, telephone-game style — show how different design philosophies interact. The Nineteenth Conlang Relay passed a story about two stones and a pterodactyl through UNLWS (a visual language without inherent reading order), requiring translators to make dozens of interpretive decisions the original text left open.⁵ The UNLWS version lost explicit temporal sequence entirely because the language "isn't big on stating temporal sequence," but gained something unexpected: the translators rewrote the snake's observations to favor smell and vibration over vision, because real snakes can't see well. The language's constraints forced the translators to think harder about the story's internal logic than the original author did.

The relay tradition also shows how conlangs develop in unexpected directions once communities adopt them. Scott Alexander's Kadhamic — a fictional philosophical language for his utopian society of Raikoth — was designed so that "cheap shots, ad hominem attacks, rhetorical tricks, appeals to emotion" would be grammatically impossible.⁶ The thorniest problem was defining "good": the logician-monks who built Kadhamic spent millennia refining a utilitarian superstructure until it described exactly how to aggregate preferences into a utility function. Then the government banned political discussion in any language except Kadhamic, which had the dual effect of excluding anyone too dim to learn it and making specious arguments structurally inexpressible. It's the ultimate perfect language thought experiment — and Alexander, characteristically, follows it to its unsettling conclusions about what happens when a society optimizes its language for truth-telling and then hands governance to the AI that reads the language best.

Giuseppe Peano, the mathematician who gave us the Peano axioms, tried something less radical: take Latin and simply remove the inflections. Latino sine Flexione kept the entire Latin vocabulary but stripped case endings, conjugations, and grammatical gender.⁷ The 1903 article introducing it was written in classical Latin, gradually dropping inflections until there were none — a live demonstration of the language's thesis. Peano published his famous Formulario Mathematico in it. Most mathematicians were put off by the artificial appearance and didn't bother reading it, which tells you something about the gap between logical elegance and social adoption. A language that makes perfect rational sense but looks weird loses to a language that makes less sense but feels natural.

The most conceptually ambitious conlang project may be "Europan," a fictional alien language designed to have sentence structures that are genuinely incompatible with tree representation — the universal structural property shared by every known human language. Where all human grammars organize words into hierarchical trees, Europan allows cyclic graphs where dependencies loop back on themselves, producing non-tree sentence structures that no human language has ever exhibited.⁸ It's the strongest possible version of the conlang-as-experiment: not just tweaking parameters within the human design space, but asking what lies outside it entirely.

What Conlangs Reveal

The deepest value of constructed languages isn't the languages themselves — it's what they reveal about the natural ones. Each conlang is a controlled experiment: what happens when you remove ambiguity? When you eliminate verbs? When you reduce vocabulary to 123 words? When you forbid rhetorical tricks? When you abandon tree structure? The answers are never what the creators expect. Ambiguity turns out to be load-bearing. Verbs turn out to be optional. A hundred words turn out to be enough for love but not for math. Rhetorical tricks turn out to be structurally identical to sincere persuasion. And tree structure turns out to be so deeply embedded in human cognition that abandoning it requires positing alien neurology.

Maybe the most honest assessment came from Quijada himself, at the end of thirty-four years: "If linguistics is the best window into the mind that we have, why wouldn't you want to manipulate it for artistic purposes?" That reframing — from engineering project to conceptual art — is probably the healthiest relationship anyone has ever had with a constructed language.