Was rongorongo an independent invention of writing?

« previous post | next post »

Below is a guest post by Kyle Gorman and Richard Sproat:

Ferrara et al. [1] report on the results of a study of several specimens of kohau rongorongo, the enigmatic, undeciphered texts of Easter Island (also known as Rapa Nui). These texts, inscribed on wood—mostly driftwood that washed ashore on the island—may have numbered in the hundreds during the mid 19th century, when the system is known to have been in use. Roughly two dozen inscribed artifacts survive today. Ferrara et al. claim, on the basis of carbon dating, that one of them was inscribed before European contact in the 18th century, and thus represent “one of the few independent inventions of writing in human history”.  

Naturally it is this latter point in particular that has attracted attention in the popular science press. See for example here, here, here and here.  So, while the actual results of the paper are quite modest in that they establish the dates of one piece of wood that ended up being carved with glyphs, the authors clearly intend a much more sweeping interpretation of these results. And true to form, the popular science press is happy to help spread a story that, in the words of one of the articles linked above, “could rewrite history as we know it”.

It has long been an open question whether rongorongo was first developed before or after European contact in the 18th century; if the system was developed after contact, then there is a possibility that its invention was a case of stimulus diffusion [2] rather than an independent invention. Ferrara et al. [1] (henceforth, F24) estimate that the wood used for one tablet, known as tablet D, or Échancrée, is from a tree felled around 1500. If the wood (an African species, Podocarpus latifolius) somehow made it to Easter Island in the period between 1500 to before the early 18th centuries, and if it was inscribed with glyphs during that period, then clearly rongorongo was an independent invention. As the authors admit, the dating of the wood merely provides a terminus post quem for this text's creation. Échancrée was not discovered in its archaeological context. We do not know how or when the wood actually reached the island nor when it was inscribed, and F24 provide no specific proposals regarding these matters.  Indeed this is one of the weaknesses of the paper: providing a plausible mechanism for how a piece of African wood could have made it to the Eastern Pacific between 1500 and the early 18th century would seem to be of some importance for the authors’ intended thesis.  After all, a plausible alternative is that the wood first came to Europe via some established trade route, and only later dropped off a ship near Easter Island once regular exploration of the Pacific began in the 18th century. In that case, we could no longer be sure it did not arrive there after contact.

Still, while one cannot draw many firm conclusions from F24’s results, they are at least consistent with their claim that rongorongo is an instance of a very rare phenomenon in human history: the “pristine” invention of writing by a culture not in contact with any other literate culture.

But is this conjecture warranted? In order to answer that one needs to be much clearer on what this rare event consisted of. Putting Easter Island to one side, writing is known to have been invented in four ancient cultures: Mesopotamia, Egypt, China and Meso-America. It has even been suggested that Egypt may have borrowed the idea (though not the details) of writing from Mesopotamia [3]. Some would add the Indus Valley as a possible fifth site of invention, but thus far nobody has convincingly demonstrated that the cryptically short Indus Valley texts were a true writing system.

But what does “true writing system” mean? Humans have invented hundreds if not thousands of symbol systems that convey some sort of meaning, but what is rare was the discovery that one could use symbols not for their meaning but for their sound. The first step of this process is the so-called rebus principle, whereby one can write “I can see you” as ️️ This principle ultimately led scribes to discover that a word can be decomposed into a sequence of semantically-meaningless units of sound, units that can be used to organize a writing system. This realization—in some sense the discovery of phonology itself—was made in every one of the pristine inventions mentioned above, and it is this discovery that has so rarely occurred in human history. All of these ancient systems were mixed systems in that they had symbols used for their meaning, or to represent individual words, but they also had symbols that were used for their sound(s). Indeed, as DeFrancis [4] argues, there is no way to construct a true writing system without being able to notate phonological information—if by writing one means the ability to notate in graphical form basically anything one might say out loud.

Now there are some who take a more inclusivist view of writing opposed to the exclusivist view that we  sketched in the previous paragraph. Powell [5], for example, defines writing as “a system of markings with a conventional reference that communicates information”, a definition that does not even mention the notion “language”. On that definition, writing could include mathematical or musical notation, road signs, or Ikea assembly instructions, and thus there have been hundreds if not thousands of “writing systems” that have been invented, some by non-literate cultures [6]. But if one adopts this broad view of what writing constitutes, then it is not clear that the pristine invention of “writing” was rare at all.

This brings us back to rongorongo and F24's central thesis. To date it has not been demonstrated that rongorongo was a writing system in the exclusivist sense discussed above. Many researchers have attempted to decipher rongorongo as a mixed semantic-phonetic system along the lines of Sumerian, Egyptian, Ancient Chinese, or Mayan. Yet no one has yet succeeded in proposing more than tentative suggestions about possible interpretations of a handful of rongorongo glyphs.

The most recent attempt is by Davletshin [7], who uses evidence from “cross-readings” (cases where different glyphs are inter-substitutable in identical environments, and where one finds multiple instances of these patterns) to suggest that the language underlying the system was “East Polynesian”. Yet the set of proposed readings is very small, and many of them seem equivocal at best. As Davletshin himself notes, rongorongo presents as ideal a situation as a would-be decipherer could hope for. There is a lot of text—several thousand glyphs spread over a few dozen tablets, all of it digitized; we know what language the islanders spoke, and we know a lot about its structure; and, a great deal is known from ethnographic studies about how the texts were used. If the system was a true writing system in the exclusivist sense, why has it been so resistant to decipherment? If on the other hand it was some sort of mnemonic system—like Naxi symbology, Dakota winter counts, Australian message sticks or Lukasa memory boards [6]—then any attempt to decipher it as a semantic-phonetic writing system is bound to fail.

So while F24 might be correct that rongorongo was invented prior to European contact and therefore could not have been inspired by outside influences, nothing in their demonstration supports the notion that its invention falls into the category of rare invention that characterized the known invention of writing in Mesopotamia, Egypt, China, or Meso-America. F24 see rongorongo as a parallel to these inventions, and while they may well believe that rongorongo does indeed fit that bill, the evidence provided does not justify this claim. If and when the system is successfully deciphered as a true writing system, then and only then will that claim be justified.

The authors submitted an earlier version of this piece to Scientific Reports as a reply to the Ferrara et al. paper and, as one might have anticipated, it was rejected. According to the editors: 

In the present case, while we appreciate the interest of your comments to the community, we do not feel that they advance or clarify understanding of the paper by Ferrara et al. to the extent required for publication in Scientific Reports. Namely, while we appreciate the discussion of whether rongorongo should conceptually be considered or not a true writing system, this point does not necessarily fall within the scope of the Ferrara et al. study, which focused on the origin and dating of rongorongo.

Which of course is nonsense, since while indeed the paper’s actual demonstrated result was rather narrow in scope, the claimed implications of the study were far reaching and were clearly the main point that the authors wanted to push. But of course we know the real reason the editors do not want to publish a response of this kind: the paper has attracted lots of attention, which in turn translates into eyeballs migrating to the journal’s site. A piece that detracts from the narrative that is pushed in all of this publicity does not serve their needs.

That said, we rather doubt that this kind of laxness would pass muster in the “hard” sciences. Suppose there were some group of organisms X that are of uncertain relation to another lineage Y, but where X was believed in any case to have evolved after Y. Let us assume both of these lineages are long extinct, so that there is no way to use gene sequencing to establish the relationship, and one must do it solely on the basis of morphology. Suppose then someone publishes a paper reporting on a possible case of a much earlier specimen of X and presents the conclusion that “this suggests that the X-Y lineage evolved much earlier than previously believed”.  Suppose further that such a submission made it past review with that claim intact—doubtful in the case of biology, but not uncommon in work related to topics such as writing systems. Then if someone wrote a reply pointing out that the relation of X and Y has by no means been established, we hardly think that would result in an editorial response of the form “while we appreciate that the discussion of whether X and Y are related is of general interest, the point of this paper was to establish a dating for X”.

On the other hand, such sloppiness seems to be commonplace in the top science journals when it comes to issues related to language, and especially when it comes to writing systems and symbol systems more generally. The second author has had to deal extensively with this kind of stuff before [8]. But to be fair, many who work on writing systems are also not clear and consistent about what they mean by “writing”, and sometimes seem happy to capitalize on the vagueness surrounding the term.

We want to end with one point that should perhaps go without saying, but is nonetheless important to make clear. Even if rongorongo is finally established as not being a writing system in the exclusivist sense, this would not detract from its significance as a cultural artifact. Clearly rongorongo was a central part of Rapa Nui culture, and understanding how it worked and how it was used by the culture is of importance to understanding one of the most enigmatic human societies in history. The Easter Islanders clearly invented something of great value. We only ask that researchers not muddy the waters by claiming an affinity to other systems, where that affinity has in fact never been demonstrated.


[1] Ferrara, S. et al. 2024. The invention of writing on Rapa Nui (Easter Island). New radiocarbon dates on the Rongorongo script. Scientific Reports 14, 2794.

[2] Langdon, R. and Fischer, S. R. 1996. Easter Island's 'deed of cession' of 1770 and the origin of its Rongorongo script. The Journal of the Polynesian Society 105, 109–124.

[3] Daniels, P. 2006. Three models of script transfer. Word 57, 371–378.

[4] DeFrancis, J. 1989. Visible Speech: The Diverse Oneness of Writing Systems. Honolulu: University of Hawaii Press.

[5] Powell, B. 2009. Writing: Theory and History of the Technology of Civilization. Chichester: Wiley-Blackwell.

[6] Sproat, R. 2023. Symbols: An Evolutionary History from the Stone Age to the Future. Cham, Switzerland: Springer Nature.

[7] Davletshin, A, 2022. The script of Rapa Nui (Easter Island) is logosyllabic, the language is East Polynesian: Evidence from cross-readings. The Journal of the Polynesian Society 131, 185–220. 

[8] Sproat, R. 2010. Ancient symbols, computational linguistics and the reviewing practices of the general science journals. Computational Linguistics 36, 585-594.

Above is a guest post by Kyle Gorman and Richard Sproat

The extant rongorongo tablets are inventoried and transcribed here.


  1. Chester Draws said,

    March 21, 2024 @ 10:45 pm

    People write when they have a need to write. They don't just write because it is possible.

    Even now there are places in the world where people know of the concept of writing, but where they don't bother themselves, because they have no need for such a thing. A stone age culture gains nothing by being able to write, because they already have ways to transmit their myths and genealogies and they don't have sufficient surplus to require book keeping. What would they write about?

    So why would a society with no need for writing develop it?

  2. Peter Grubtal said,

    March 22, 2024 @ 3:30 am

    Many thanks for this well-balanced and well-grounded assessment which impresses with its objectivity. And the strictures about the "popular science press" are all too justified, unfortunately. I often go to these sources myself and have become leery about the amount of sensationalism in them, It's the lure of free content, which I'm afraid is difficult to resist.

  3. Jerry Packard said,

    March 22, 2024 @ 9:40 am

    Thanks to Gorman and Sproat for a great post. I feel publication of their article would have benefited the discussion – especially the exclusivist/inclusivist dichotomy, which I have always pitched to my students as linguistic/non-linguistic.

  4. Sean M said,

    March 22, 2024 @ 1:29 pm

    Chester Draws: there are gardening peoples in island Southeast Asia who used writing to learn as many love songs as possible, the better to impress people they want to boink. See Ray McDermott, "Hanunoo Literacy: A Report for the Wenner-Gren Foundation for Anthropological Research," September 1993

  5. Richard Sproat said,

    March 22, 2024 @ 5:53 pm

    Thanks everyone for the comments.

    @Jerry Packard: (Hi Jerry). Yes their article certainly would have been better for making that distinction clear. However this is far from being the first time that authors have been vague about what they mean by the term "writing", and have benefited from the fact that most readers will not know enough to understand that there are important differences in how the term gets used.

  6. TonyG said,

    March 22, 2024 @ 7:05 pm

    The date that a tree was felled in Africa has literally no evidential weight when it comes to estimating when symbols might have been carved on its wood half a world away. How could anybody propose a connection, even tentatively? This is laughable.

  7. Richard Sproat said,

    March 22, 2024 @ 9:21 pm


    I agree totally.

    It would be nice to have a good model of how that wood got from Africa (the species is originally from South Africa but has spread widely to both West and East Africa) to Easter Island and then eventually to Bishop Jaussen in Tahiti. By odd coincidence this was the first tablet that drew attention to rongorongo.

  8. Jonathan Smith said,

    March 22, 2024 @ 9:59 pm

    Irrelevantly but FWIW, Chinese writing didn't/doesn't involve "[decomposition] into a sequence of semantically-meaningless units of sound"; no notion of a subsyllabic segment resembling the phoneme existed until the (Indian-influenced?) fanqie decompositions of the 3rd century-ish CE. Indeed, conversely, due to the nature of the system, independently meaningless syllables like those within disyllabic monomorphemes including e.g. products of morphological reduplication are often regarded as meaningful.

    IDK why DeFrancis (1989) came to be so singularly focused on regarding Chinese writing as "syllabic" — the fact that Chinese writing systems from early inscriptions to the present have always involved the repurposing of extant glyphs for new coinages based on the sounds of the words those extant glyphs represented does not a syllabic script make; i.e., this productive process is not relevant to a consideration of the nature of the relationship(s) pertaining between glyphs and units of language (monosyllabic words or monosyllabic portions of longer words, thus ≈logography.) A good read on this question is Geoffrey Sampson (2009), "Chinese script and the diversity of writing systems" in Linguistics.

  9. Richard Sproat said,

    March 23, 2024 @ 2:09 am

    @Jonathan Smith.

    Well the notion of "semantically-meaningless units of sound" does not imply that one has a notion of subsyllabic units. It only requires that the system allow for symbols to be used purely for their sound value without regard to whatever they may have originally meant. In that sense Chinese is no different from what happened in Egyptian, Sumerian or Mayan. The only difference is that in Chinese the phonetic elements were grouped together along with semantic elements into something that is regarded as a single unit.

    That is the only difference between Sumerian "u-ri-in BIRD" for "hurin", 'eagle', and say, 鴨 yā 'duck' in Chinese is that the semantic bit 鳥 and the phonetic bit 甲 are glommed together into what is considered a single unit. But the 甲 bit is every bit as "meaningless" in 'duck' as the "u-ri-in" bits are in the Sumerian example.

  10. Chris Button said,

    March 23, 2024 @ 10:16 am

    @ Richard Sproat

    To your point, it's also interesting how Egyptian writing (and to an extent also Mayan, albeit slightly differently) is based on a similar combination of semantic determiners and non-semantic phonetic elements as Chinese writing.

  11. Chris Button said,

    March 23, 2024 @ 10:21 am

    Sorry – I just noticed that you already mention Egyptian and Mayan in the first half of your post!

  12. Philip Taylor said,

    March 23, 2024 @ 3:04 pm

    "The date that a tree was felled in Africa has literally no evidential weight when it comes to estimating when symbols might have been carved on its wood half a world away" — I disagree. Would you not agree that the date that the tree was felled places a definite lower bound on the earliest possible date for writing to be found on the wood therefrom, at least if the sample of wood is sufficiently large for it to be certain that the wood came from the trunk and not from a branch ?

  13. Richard Sproat said,

    March 23, 2024 @ 5:31 pm

    @Philip Taylor

    Échancrée is 30 × 15cm, and Podocarpus latifolius can grow to be 35m high, so your guess is as good as mine of whether this could have come from a branch or not. But does it make any difference? Once the wood is cut it is dead and I assume would show the same date whether its a branch or the whole tree?

    In any event, the real earliest date of possible carving is not 1500, but 1500 + Y, where Y is the number of years it would take the wood to leave Africa and arrive on Easter Island. Missing in Ferrara et al's piece is a reasonable estimate of Y, which would seem to be rather crucial, since Y could be as large as 300, if the wood was part of some European heirloom that fell overboard in the late 18th century.

  14. Richard Sproat said,

    March 23, 2024 @ 5:45 pm

    @Chris Button

    Yes, all the early systems were the same: all of them involved some combination of purely morphographic elements (a symbol standing for a whole morpheme with no written clues to the phonology), combinations of semantic determiners and phonographic symbols, and sometimes purely phonographic spellings.

    For some reason some Chinese scholars seem to view Chinese as being different. I suspect that is partly because in Chinese, fairly early on, the semantic-plus-phonetic combinations got analyzed as composite units that we now call "characters". In Sumerian there is no sense that the "u-ri-in BIRD" example is a single unit.

    (At least so it seems: we don't of course know what the Sumerians and Akkadians thought. These words were often written out in extensive lexical lists, so maybe they thought of the list elements as being single units.)

    Whereas in Chinese 鴨 is considered a single, decomposable, unit. If Chinese had been like Sumerian, they would have written it 甲鳥. So that's an interesting further development (IIRC, Amalia Gnanadesikan has an analysis that analogizes the Chinese process to morphological affixation). But it doesn't change the basic point that all these early systems behaved the same way.

  15. Jonathan Smith said,

    March 23, 2024 @ 10:45 pm

    @Richard Sproat

    The devices at work in Chinese vs. e.g. Sumerian are analogous in some respects but not in others. Since scholars of early Chinese tend to ignore Sumerian (and everything), and Western generalists tend to hand-wave Chinese, serious close comparisons are hard to come by.

    Re: coinage of novel Chinese glyphs, 'rebus'-type redeployment is seen in the earliest materials and remains the most common means of generating novel characters for unwritten words down to the present (say, within vernacular scripts for regional languages.) So yeah, clearly a kind of abstraction comparable to that demonstrated by reuse of Sumerian logographs for syllables of sound is at work here.

    But "甲" within "鴨" and the rest are only "phonetic (or semantic) bits" with respect to the historical matter of coinage: this history is irrelevant and obscure from the POV of latter-day users of the script (until it is explicitly pointed out by, say, elementary school teachers.) That is, "甲" does not function synchronically as a 'phonetic' or a 'speller'; no one (generalizing here) cares why "甲" is in "鴨"; and "鴨" is in fact a single unit (not "considered" a single unit) in light of its function — viz., writing words that mean 'duck' in Mandarin, Taiwanese, Vietnamese, and so on, no matter what they sound like or what the word that means 'first of the 10 Heavenly Stems' sounds like.

    So no this is rather, rather different from "u-ri-in BIRD," where we see incipient spelling. (Note though that while the cultural center has held to date, Chinese characters really do become syllabograms within e.g. Japanese kana writing and to some extent within Chinese itself to write foreign-language proper nouns and stuff… where we can find an analogy to things that happened when Sumerian writing was adopted by Semitic speakers.)

  16. Richard Sproat said,

    March 24, 2024 @ 12:18 am

    @Jonathan Smith

    As you say, coinage of new characters (for vernacular languages, for new chemical elements, and indeed for 95% of the characters that have ever been invented), makes use of that semantic-phonetic composition mechanism, so clearly it is an important part of Chinese writing and remains so to the present day. Whether in any particular case people view the character as decomposable is neither here nor there: 鴨 in Modern Chinese may not be generally understood that way. Clearly in Japanese where it represents the morpheme "kamo" it *cannot* be understood that way. But the point is you can't build a whole writing system based on purely morphographic symbols that carry no phonological information.

    OK, I take that back: of course you *could* do that in the sense that one could certainly come up with thousands of glyphs with no phonological cues at all. Computers would of course have no problem with that — and indeed that's effectively what they do given that what large language models like ChatGPT are trained on is just sequences of UTF8-encoded characters, with no useful internal structure. But such systems would be rather hard for humans to memorize, which is presumably why no full writing system in existence has ever been designed that way. The fact that Chinese developed the semantic-phonetic mechanism, and that this was by far the most productive of all the mechanisms it had for inventing new characters, seems to me to be a rather significant point (which of course was also DeFrancis's point). So I am having trouble seeing this as different in kind from the situation in Sumerian. Obviously there are many differences of detail, especially when the subsequent history of the two systems is considered. But the basic mechanism at work seems to be the same.

    Anyway, however one resolves this dispute, the fact remains that phonetics was important in the development of every single ancient writing system, including Chinese. And it has not been yet demonstrated that rongorongo encodes phonetics.

  17. Jonathan Smith said,

    March 24, 2024 @ 11:05 am

    @Richard Sproat

    Yes, on that level the comparison is certainly valid, and I appreciate the argument re: rongorongo.

    IDK re: "you can't build a whole writing system based on purely morphographic symbols that carry no phonological information": we might divide your statement into a claim that such a thing can't be built (by humans) and that such a thing can't be used (by humans).

    In the case of Chinese, 'phonetic' elements allow ready expansion of coverage. That is, they seem to have been instrumental (and remain so) in *building* the systems. But to readers/decoders, these things come to constitute just substructure, not substantive phonetic cues — so unnecessary for *using* the systems. (One could of course shift here to arguing that it is the substructure per se that makes the system manageable — that seems plausible.)

    So it's much more accurate to say that modern Chinese writing consists of thousands of morphographic characters with no phonological cues than to say, following DeFrancis, that it's a kind of bloated syllabary. I suppose Tangut approached such an extreme still more closely. Both systems prove readily *usable* by humans. E.g. non-native students can learn 10 units of introductory Chinese — many hundreds of characters — with no notion of what a historical 'phonetic' is or was… few characters are even plausibly explicable in such terms.

  18. Chris Button said,

    March 24, 2024 @ 12:50 pm

    I think it was Geoffrey Sampson in his "Writing Systems" book (possibly someone else; don't have the book to hand) who talks about how readers of English and Chinese have more in common in how they process text that readers of say English and Spanish.

    Ken-ichi Takashima has discussed how already in the time of the oracle-bone script, the character 口 "mouth" seems to have been used as a de-semanticizer for characters.

    Personally, I think the Chinese developed a proto-alphabet of sorts as represented by the ganzhi (as recently discussed here https://languagelog.ldc.upenn.edu/nll/?p=63100#comment-1615142 ) in a similar way to say how Egyptian hieroglyphics evolved into a proto Semitic alphabet. But for political/social/cultural reasons, the ganzhi were not used in that manner.

  19. Richard Sproat said,

    March 24, 2024 @ 5:03 pm

    Yes I think @Jonathan Smith you have hit the nail on the head: it's a heck of a lot easier to expand a system on the basis of sound than meaning, once you learn the principle. Charles Bliss discovered this (though he would never have admitted it), when he started inventing arbitrary indices to distinguish horses from mules in his purely semantic system.

    But I would guess it also aids with memorization. While my duck example is obviously not a good example of a phonetic match, given that the phonetic component's reading as an independent character is rather different in Modern Chinese from the reading of "duck", still it must be of some utility that the character represents a bird, and it sort of sounds like jiǎ. Better than nothing, anyway. And surely you must have had the experience of not knowing how to write a given word, and then seeing it written, realizing that that must be how you write it because the semantics and phonetics seem right.

    @Chris Button: yes Sampson pretty explicitly proposed that English is somewhat logographic. So does DeFrancis. And if you take Zev Handel's definition that basically defines logography as a spelling system where two words that are pronounced the same, but mean different things have different spellings — well English has many such cases.

    That's exactly why I proposed (in my 2000 book) that logography be treated as a separate dimension from whatever phonological units (syllables, morae, consonants, all segments …) a system is basically used to write. But it has been a tough sell getting people in the writing systems biz to buy into this idea.

  20. Yves Rehbein said,

    March 25, 2024 @ 5:08 pm

    The problem this critique and the paper have in common is the concept of "invention". Once you accept this, you are left to attack the premise of "writing". There is no objective reason to make non-writing the null-hypothesis. Even if it were, the paper wouldn't confirm it, commentators agree. The title ("The invention of writing on Rapa Nui"), which is questionable, is clearly synecdoche.

    You cannot argue rigorously that you can't claim something unless the proof is solid, only for you to claim the opposite without any evidence either way. The content of the paper makes no such claim. It merely relies on a well known hypothesis backed up by appropriate citations. The problem is introduced in about one page to surmise:

    "The question is of crucial importance, as it implies the possibility of an independent invention of writing, …" (Ferrara et al. 2024: 2)

    It implicitly assumes that earlier diffusion is too unlikely to be noted. That's not in the argument simply because the premise relies on a standard of comparison:

    "…, similar to what happened in other parts of the world where writing was an original creation, e.g., in Mesopotamia, Egypt, China and Mesoamerica," (loc.cit.).

    G&S' commentary in fact supports the premise:

    >Putting Easter Island to one side, writing is known to have been invented in four ancient cultures: Mesopotamia, Egypt, China and Meso-America. It has even been suggested that Egypt may have borrowed the idea (though not the details) of writing from Mesopotamia [3].

    As a consequence it is not clear why they complain. If the inclusionist sense of writing does not admit "invention", the title in Scientific Reports would imply the exclusionist definition. And if the exclusionist definition has not proved to be applicable to Rongorongo, one should not speak of an invention of "writing" to begin with. This would also apply to Linear A, which remains undeciphered. But they do not show that Powell's inclusive definition does not apply and that it cannot support invention. Nor do they show that invention is required at all. To require that Ferrara et al. expound on this is too much to ask.

    There is no chance that writing was invented many times independently in the estimated 300,000 year existence of humans within a span of 1% of that time, while we know that symbolic expression had developed between 30,000 to 10,000 BP at least.

  21. Richard Sproat said,

    March 25, 2024 @ 6:16 pm

    @Yves Rehbein

    I'm not sure I understand all your points, but it seems to me in any case that the definition of what one means by writing is crucial if one is going to make claims about its invention in one or another place.

    I happen to know (or at least have been told) that Ferrara herself believes that rongorongo is writing in the exclusivist sense. But since that has not been demonstrated, despite nearly a century of effort, it remains merely a wish. In my opinion—but obviously the popular science press disagrees with me—one should not be using a solid result (carbon 14 dating of one piece of wood) as a vehicle for smuggling through one's main point, which seems to be a groundbreaking "discovery", but is really just speculation.

    And if rongorongo isn't writing in the exclusivist sense, then it could be a (sophisticated) version of one of the more general symbolic expression systems you allude to, and therefore not so rare after all.

    I think the case of Linear A is a red herring. We have good reason to believe that Linear A was true writing in the exclusivist sense: it was the precursor to Linear B, which we know was writing. And the issue with Linear A is not so much that it hasn't been deciphered but that, like Etruscan, we don't really know much about the language underlying it, so it just reads as gibberish.

    Regarding your last point: well I have some very specific ideas on the preconditions and conditions that led to the discovery of (true) writing that I lay out in my 2023 book that we cite in our critique. One of the upshots of that is that the invention/discovery/evolution (whatever one wants to call it) of true writing is almost inevitable — if the right set of conditions prevail. One of the reasons it only happened within the last 5000 years or so is that it wasn't until around then that the kinds of institutions necessary for its development were in place. And it was rare because only a few cultures had the institutions in place and an appropriate precursor symbol system that could have evolved into writing. Oddly I happen to think that the cultural context was present for rongorongo to have become true writing. Similarly for khipu, which some (e.g. Sabine Hyland) have argued to have included true writing. But so far the evidence does not support the conclusion that either of those systems made the transition.

  22. Jonathan Smith said,

    March 25, 2024 @ 8:43 pm

    @Richard Sproat

    re: logography as "a separate dimension from whatever phonological units (syllables, morae, consonants, all segments …) a system is basically used to write"… interesting — though IMO one might just grade scripts on (1) parsability with respect to and (2) unique identifiability of units across the above levels of analysis inclusive of morphemes/words; i.e., just regard all of the above, etc., as non-mutually-exclusive dimensions.

    re: "surely you must have had the experience of not knowing how to write a given word, and then seeing it written, realizing that that must be how you write it because the semantics and phonetics seem right"
    naturally, but what happens far more often is not knowing how to write a word, seeing it written, remembering it, and appreciating only later that it involves or historically involved some apt phonetic (or semantic) component at time of coinage ;)

    As far as how frequent real phonetic cues are in the modern script(s), rather than observe that X% of characters are historically 'phono-semantic' or collect bunches of characters that feature 'phonetic' component P a la Defrancis (1989), one should really study actual corpus: I suspect that in say an intro simplified script textbook, %75+ of characters have Literally No Phonetic Cue Whatsoever, not even a crappy/unnoticed one as in "鴨".

    Incidentally it is easy to find people asking "why is '鴨' made of '甲' + '鳥'" online… I see a jokester has answered that "'鴨' was supposed to be used to write 'phoenix' because the character means first (甲) of the birds (鳥), but it got mixed up with '鳳'." Not a joke, really, as this is just the kind of explanation that 2000 year old dictionaries like Shuowen jiezi are full of. Components — phonetic, semantic or other — melt and shape-shift over time since the script is morpho/logographic and these historical roles are irrelevant; the modern Simplified script is of course full of the same sorts of distortions for the same reason.

  23. Richard Sproat said,

    March 26, 2024 @ 12:18 am

    OK fair enough.

    But since you asked about an introductory text … just for amusement, let's take a classic:


    I count 5 in that set though admittedly one is a repeat. So that's 58% that have no clues. So maybe you are not far off: it's easily tested if you have a particular text in mind.

    Anyway, not to belabor the point, but I think from the previous discussion that we at least agree that phonetics has been an important part of the design of the Chinese writing system, no matter what the utility of the phonetic component may be in Modern Chinese. Otherwise they would have just stuck with 回憶 compounds.

    What may be less obvious is just how lousy a phonetic system Sumerian was. But still it couldn't get by without symbols for sound.

  24. Jonathan Smith said,

    March 26, 2024 @ 7:07 pm

    Hi again Richard Sproat, Absolutely re: design or shall we say elaborational mechanism.

    There are competing thoughts about what "the utility of the phonetic component" as you say may have been at various historical moments esp. Warring States period, but the present situation seems to have obtained more or less from Han forward…

    Interesting example text :) if Mandarin, then: "近" clearly counts… "生" in "性" could be interpreted as a phonetic cue despite tone/onset/vowel not matching (though its occurrence in the latter compound character was likely to begin with a function of a morphological relationship between the two words at issue; I've noted in some paper somewhere that such twinning due to morphology could have been a model for the phonetic mechanism)… and while it is odd to count "袁" in "遠" since the latter is many times more common, there is a clear relationship here which we could nonetheless regard as phonetic cuing / 'spelling'. So among these 12 tokens I will give you 3 / 10 types = 30% 'phonetic' vs. 70% 'non' :P

  25. Was rongorongo an independent invention of writing? – Wellformedness said,

    April 4, 2024 @ 4:16 pm

    […] My colleague Richard Sproat and I recently went through it with "general science" journal editors regarding an overhyped- and undercooked paper about the origins of writing on Easter Island. Language Log has the scoop. […]

  26. V said,

    April 6, 2024 @ 5:04 am

    SFReader on languagehat suggested that the idea of writing might have come with the introduction of chariots, and the charioteers who had an idea of writing but not writing itself, and that the arrival of chariots in China and oracle bone script are both dated to 13th century BC. So China might be another not original but inspired invention of writing.

  27. Richard Sproat said,

    April 7, 2024 @ 7:04 pm


    Yes, the idea that China did not invent writing but learned about it from the West (as in West Asia) has been proposed. And indeed there is some suggestive evidence that it might be true: for one thing, our very earliest evidence of writing in China looks like a system that is fairly advanced in development. Whereas in Sumerian, individual glyphs often have a variety of readings and interpretations that are broadly related to a particular semantic category, in Chinese the glyphs seem to be much more related to specific morphemes. I discuss this issue a bit in my 2017 paper in Written Language and Literacy: https://typeset.io/papers/a-computational-model-of-the-discovery-of-writing-psomse9228

    That said, there is of course no direct evidence for this.

RSS feed for comments on this post