It turns out that English has a lot of syllables. While English has somewhere around 40 phonemes, it has enough syllables that I needed to find some way to cut the scope of the project.

I settled on creating a consonant-vowel syllabary, and to only create characters for the syllables needed to encode a few sample texts.

I was inspired by katakana, which is a secondary component of the Japanese writing system often used to transcribe foreign-language words. Katakana contain just 50 glyphs representing 50 syllables—all combinations of five vowels and nine consonants (plus the vowels on their own).

I decided to take this approach for ninescript. If I were to create glyphs for all possible English consonant-vowel combinations, I would have to create 110 characters. That’s really not that bad, but I managed to winnow that down to 100 by omitting C, X, and Q (though I added CH).

I then went through the first lines of Lilacs, pronouncing each syllable and noting which syllables from my syllabary I used to spell it so I would know how many of the hundred I needed to design.

While I planned to take advantage of diacritics to add semantic and phonetic meaning, the grid-based system I used for representing characters made this difficult. The two that I did manage to implement are using || to represent plurals and a sign for possessives.

This process of pronouncing words and comparing them against a limited set of syllables gave me a deeper understanding of why foreign accents sound like they do. Often, I would catch myself off-guard by pronouncing an English word as a native Japanese, Chinese, or Arabic speaker might. (Japanese was most common, possibly because I based my syllabary on katakana, so my syllables had no codas.) For example, pronunciations by Japanese speakers ending in “-u” made more sense once I realized they were using the consonant onset from a consonant-vowel syllable as it was the closest pronunciation available. (Prof. Houston broke down a hieroglyphics text in class that did the same thing with a rebus.) There are many cases in ninescript when I do the same; ideally, ninescript would have a large enough syllabary to encode all English syllables, as if it had developed organically.

I was impressed by how cuneiform, despite being a logosyllabic writing system with thousands of signs, had the visual simplicity of an alphabet. Each character is formed by a careful arrangement of just a handful of basic shapes.

I wanted it to be conceivable that a practiced scribe of ninescript could write it by hand at a useful pace, so the design of my characters would need to be simple and pragmatic, rather than intricate and ceremonial. I think there’s room for improvement on this front. The grid I chose was a useful creative constraint, but it meant that my symbols were more complicated than they needed to be—at most 9 strokes, which is a lot. Curved lines would allow me to create visually distinct signs with fewer strokes.

I knew that if I constrained the way my characters were constructed, it would be easier to maintain visual consistency. (Like how the wedge-shaped instrument used to write cuneiform dictated its visual language.) To this end, I was inspired by two very random things: garment care labels and the pattern lock available on some smartphones.

I coded an editor that would let me quickly create signs by connecting points on a three-by-three grid (or, nine points, hence “ninescript”) and save them as a series of coordinates. After spending a few hours playing around, I had a bit over a hundred glyphs.

You can try out this editor below for yourself. Click “reset” and draw your own signs. Click once to start a line, and again to finish it. Clicking on the same point twice will create an outlined circle—I found that outlined circles are easier to hand-write even than smaller dots. There are 2²⁹ = 536,870,912 possible visually-distinct signs.

Having a set of fairly arbitrary signs was fun because it allowed me to both use rebuses and semantically compose signs. For example, in Lilacs, what should be read as “false” is written as “not true.” (If a writer wanted to encode “not true,” they would have to use a space.)

While most of my syllabic glyphs are abstract, all of my pictograpgic glyphs have at least tenuous pictorial grounding. One could imagine them evolving from more detailed representations, like the transition from hieroglyphics to hieratic.

This was an area I had some fun with—like representing cardinal directions as combinations of ordinal directions, and colors with additive color mixing—even though these decisions might not be realistic for a writing system that evolved alongside English.