Unified Arabic solved a machine problem

I recently read Fighting Illiteracy With Typography about Nasri Khattar's attempt to redesign printed Arabic. He called his system Unified Arabic. Since Arabic letters change shape depending on their position in a word, he proposed giving each letter a single shape.

I admire the engineering behind the idea. As a native Arabic reader, though, I find the result uncomfortable to read. Khattar was responding to a real problem. Arabic had to work with machinery designed around the Latin alphabet. A Latin typewriter could approximately map one key to one letter. However, Arabic's connected letters, contextual forms and ligatures made that arrangement much harder. Each additional form could require another piece of metal, key combination or mechanism.

Khattar reportedly arrived at his idea while teaching Arabic typing in 1932. After accidentally typing the "wrong" contextual form of a letter, he noticed that the word remained readable. He eventually reduced printed Arabic to about 30 principal forms, allowing typists to use one form for each letter. That was an elegant response to the limits of a typewriter. The trouble begins when the argument moves from printing to literacy. Two claims became intertwined: "Arabic is difficult for machines to print." and "Arabic is difficult for humans to read." The machinery of the period gave good reason to accept the first. It did not establish the second.

Much of the case for Unified Arabic rests on counting forms. A beginner learns the letter meem, then encounters its different shapes as its position and connections change. Include ligatures, and the number of printable forms grows well beyond the alphabet's 28 letters.

For a machine, those forms are separate things to accommodate. I don't experience them that way when I read. I don't consciously classify four versions of a letter before deciding which character they represent. I recognize words, and the connections between letters help give those words their familiar appearance.

Reducing 150 printable glyphs to 30 therefore tells us something useful about mechanical complexity. It tells us much less about the effort of fluent reading. For me, Unified Arabic takes more effort: I can read it, but I have to slow down and reconstruct words I would ordinarily recognize almost immediately. The simpler set of forms has made the text less familiar.

This doesn't settle how a beginner would learn it. One of the striking stories in Khattar's history concerns literacy educator Frank Laubach. After teaching Unified Arabic to an illiterate girl, he became convinced that it could help people learn Arabic dramatically faster.

I can believe that a single form per letter makes the alphabet easier to teach. There are fewer forms to memorize. But I would want to distinguish "How quickly can someone learn to identify the letters?" from "How efficiently can someone become a fluent reader?"

A writing system could make the first task easier while making the second harder. As readers gain experience, they recognize familiar combinations and word forms without decoding every character independently. Arabic's connections, positional forms, proportions, dots and overall geometry contribute to those patterns. Unified Arabic removes some of that information to make the letters more regular.

Perhaps that is a useful trade in the first few hours of instruction. I'm less sure it remains useful after ten years of reading. The historical anecdotes suggest that beginners could learn Unified Arabic's letters quickly, but they don't, by themselves, demonstrate better reading speed, comprehension or long-term literacy.

Adoption would also have introduced a problem that the letter count leaves out. Arabic already had an enormous written culture. A child educated exclusively in Unified Arabic would eventually encounter conventional Arabic in books, handwriting, signs, historical documents and, most importantly, the Quran.

Even a complete transition to Unified Arabic would leave that literary inheritance in a script new readers still had to learn. A partial transition would require students to learn both systems. Either way, the educational benefit becomes harder to assess. Learning contextual variants of familiar letters might be less work than learning Unified Arabic and then learning conventional Arabic as well.

These objections don't diminish Khattar's achievement as an engineer. Traditional movable type could require enormous collections of Arabic forms. Typewriters had limited keys, and mechanical typesetting favored scripts that could be broken into independent, reusable units. Khattar understood those constraints and designed a way around them.

His proposal received substantial attention. IBM supported the project, and the Ford Foundation funded it. Academy of the Arabic Language in Cairo selected it as one of three finalists for debate from 266 submissions. Unified Arabic addressed a technological limitation that institutions were taking seriously.

Computers eventually gave us another way to handle it. When I type the Arabic letter beh, I don't normally choose between "initial beh," "medial beh," and "final beh." I enter the character, and the text system determines its appearance from the surrounding letters. The distinction is between a character and a glyph. The stored text says, in effect, "this is the letter beh." The shaping system says, "given the surrounding letters, draw it like this."

That separation preserves a simple underlying representation while allowing readers to keep the forms they know. Unicode provides an abstract representation of Arabic characters, shaping engines and fonts handle their presentation. Contextual shaping is routine, and a font containing hundreds of glyphs presents little of the difficulty that the same inventory posed for mechanical equipment.

Khattar worked within the limits of his machinery by simplifying the printed forms. Modern computing allows the machinery to handle the contextual variation. As computation became cheap, the original engineering case for Unified Arabic largely disappeared.

This is what interests me most about its history. Early machines can put pressure on people to simplify established practices. Once those machines improve, the compromise deserves another look. Arabic's cursive structure was once expensive to reproduce mechanically. Computers can handle it through shaping, without asking readers to give it up.

OCR and machine learning offer a related example. Making writing more regular was an understandable way to help machines recognize it. Modern recognition systems can learn variation instead, shifting more of the work to the software.

There is still one use for Unified Arabic that I find plausible: helping people whose native language uses the Latin alphabet begin learning Arabic. Someone accustomed to English, Spanish, or Italian generally expects letters to remain discrete and relatively stable. Arabic introduces a new alphabet and unfamiliar sounds, along with right-to-left writing, contextual forms, connections and dots that distinguish otherwise similar letters. Stable, disconnected forms might make those first encounters easier to manage.

My response as a native reader could work in reverse for such a learner. I miss the connected structure because it is familiar to me. Someone accustomed to Latin typography might initially find the disconnected letters more familiar and easier to decode. I can imagine a textbook using Unified Arabic temporarily alongside conventional Arabic: first helping learners recognize the letters, then introducing their contextual forms and connections. That would need to be judged by how well students make the transition. I wouldn't want Arabic publications written this way, but I can see a possible educational role for it.

It's a narrower claim than the original promise of reforming literacy, and one I find more defensible. Khattar correctly saw that Arabic was disadvantaged by machines designed for another writing system. His solution made sense within those constraints. What I find unconvincing is the further assumption that reducing the number of printable forms would necessarily make people better readers.

For native readers like me, the familiar visual structure is worth keeping. For beginners, any benefit from fewer forms would have to be weighed against fluency and the need to read conventional Arabic. Unified Arabic remains an interesting typographic and educational experiment, with a possible use for foreign learners. Its history is especially revealing because we eventually addressed the mechanical problem while retaining the forms Khattar proposed removing.