Language trees and script trees

« previous post | next post »

[The following is a guest post by Jim Unger (J. Marshall Unger), who wrote it in response to my invitation to him to draw up a Stammbaum to show the relationships of the world's scripts.]

The rationale for tree structures in language history is that languages never completely converge. When speakers of two languages come into contact, there are always clues in the resulting language(s) that reveal the identity of the input languages: apart from the effects of contact, languages diverge over time.

In the case of writing, one must first of all distinguish graphic methods per se from writing systems. The adaptation of existing graphic methods that originated in one time and place to a different time, place, and, usually, language, does not, in my opinion, show the spread of a writing system, just the diffusion of a technology.  It only makes sense to speak of a writing system with respect to a particular language at a particular time and place.  This is a corollary of the fact that every practical and learnable writing system co-extensive with the potential output of a natural language necessarily utilizes a certain amount of phonography with respect to that language (unless the system is contrived expressly for cryptographic purposes).  Logography arises because of sound changes that obscure the motivations for some previously phonographic inscriptions, the purposeful suppression of certain phonographic information for the sake of brevity, or, as in the case of Chinese, historical accidents that militated against the adoption of an abjad, abugida, or alphabet.

Furthermore, the diffusion of knowledge of graphic methods has become less and less tightly tied to the diffusion of the graphic methods themselves.  I think DeFrancis was right:  full writing originated just three times in human history.  (There's a little evidence of very early Egyptian writing prior to contact with Mesopotamia, so perhaps there was a fourth kind of ab ovo writing.)  Not counting the Cree and Cherokee syllabaries, for instance, as "ab ovo" is legitimate because they are very late historically and were obviously motivated by (though not merely copied from) European writing.

For these reasons, I think the best one can do is a historical survey that describes how graphic methods originated, been borrowed, been imitated, and, in some cases, died.  Here is a short overview* I put together for an exhibit at Ohio State in 2009.

[*See the first two comments below.]


  1. Allan from Iowa said,

    December 27, 2021 @ 8:57 am

    Link is broken.

  2. Brian said,

    December 27, 2021 @ 9:07 am

    Navigating Unger's home page reveals that the correct link for the overview is:

  3. Terpomo said,

    December 27, 2021 @ 11:00 am

    "The Tamil alphabet is thought to have evolved from the Brāhmī script, though some scholars believe it may have roots in the so far undeciphered Indus script" – Maybe Unger knows more than I do but this sounds suspiciously like an insufficiently-critical repetition of nationalist posturing.
    "Amharic, the second most widely spoken language of the Semitic family after Arabic, is written with letters that function much like modern Japanese kana" – Not really, they're an abugida.

  4. Philip Taylor said,

    December 27, 2021 @ 12:26 pm

    Is there a number missing between "first" and "centuries" in the following sentence (in re Manchu

    For roughly the first centuries, Manchu was the main language of government in China and served as a lingua franca.

  5. Cervantes said,

    December 27, 2021 @ 2:29 pm

    Was John really right about that? What about Nahuatl?

  6. J.W. Brewer said,

    December 27, 2021 @ 2:55 pm

    In terms of "function" an abugida is really just a subgenre of syllabary, where there are meaningful visual patterns and combination that help you learn/remember which glyph denotes which syllable. The glyphs of kana lack that sort of visual organization, but once you know the system you know it and can read with presumably equal fluency.

    The origin of kana is maybe a good example of how the "tree" model of descent via branching from a common ancestor doesn't necessarily always work well for writing systems,* if you accept the view that kana was at a conceptual level a "hybrid" approach, where certain Japanese already familiar with the Chinese-derived practice of using certain kanji as purely phonetic symbols for representing semantically-unrelated words phonetically became aware via the diffusion of Buddhism of the Brahmic abugida approach to writing and then somehow came up with the clever idea that you could use glyphs derived from the former approach as the component pieces of a comprehensive system like the latter approach.

    *Of course there are already lots of arguments out there that the "tree" model doesn't work well or at least not universally well for languages either …

  7. Jim Unger said,

    December 27, 2021 @ 6:15 pm

    Thanks for the comments. I will fix the typo when time allows. I didn't take the photos—I just supplied the captions for the exhibition, which had to be brief, not too technical, and encyclopedic in style. The mention of the suggestion of a link between Harappa and Tamil was not meant as an endorsement!

    DeFrancis's "three origins" were Mesopotamia, China, and Meso-America. As for Nahuatl, Mayan was a precedent in the area, so I think his claim stands.

  8. Terpomo said,

    December 28, 2021 @ 7:00 am

    J. W. Brewer, not really- the difference is that a syllabary has separate, unrelated glyphs for syllables, whereas an abugida has separate graphemes for vowel and consonant, it just doesn't write the vowels as separate letters. At least in the terminology I've heard, abugidas are a separate thing from syllabaries, not a subset.
    Jim Unger, I have to admit that "some scholars believe" without further comment seems to me like it would create the impression, to a layman, that it's far more plausible than it actually is.

  9. Philip Taylor said,

    December 28, 2021 @ 7:36 am

    Terpomo, to me "some scholars believe" means just what it says (neither more not less), whereas "most scholars believe" would create an impression in my mind that the hypothesis is very definitely plausible.

  10. Terpomo said,

    December 28, 2021 @ 10:05 am

    I feel like that phrasing is somewhat irresponsible when it's such a fringe and unlikely theory. Would you say "some scholars believe anthropogenic climate change is not a real phenomenon"? After all, there are some people with scientific degrees who believe that. There's such a thing as 'undue weight' as Wikipedia calls it.

  11. Mehmet Oguz Derin said,

    December 28, 2021 @ 10:18 am

    In case anybody else gets access denied with the URL in the comments, too, here is the link, which works fine for me at this time:

  12. Chris Button said,

    December 28, 2021 @ 12:23 pm

    Of course there are already lots of arguments out there that the "tree" model doesn't work well or at least not universally well for languages either

    Yes, I was thinking the same thing. Horizontal transmission as opposed to vertical “trees”.

    Separately, an interesting note to add to the Burmese section would be how the writing became more circular (an acquaintance of mine once called it hole-punch script) as it shifted from inscriptions to softer materials that couldn’t tolerate sharp angles.

    Also, might I recommend correcting the sentence in the introduction that says 勿 “began as a pictogram for ‘creature’”?

  13. Jim Unger said,

    December 30, 2021 @ 12:06 pm

    To access the PDF, go to and click on the link.

RSS feed for comments on this post