TecC 27 - Keeping the Word: The Inscrutable Invocations of Innovation
What happens when attempting to dominate the forces of nature - the inconvenient truth of unintended consequences
In the previous episode we discussed how in continuation of notions going right back to the Proto-Indo-Europeans, one of their descendent groups, the Indo-Aryans, extended the idea and need to preserve the ‘spoken word’ to extraordinary lengths, anticipating some of our modern data-integrity technologies by around three thousand years.
But one might add, this ain’t nothing compared to what came next!
But let’s go back to our fictional friends first.
We saw last time that while trying to get Bruno the dog to eat its food, our friends Bryan, Brenda, et al started to wonder if the way you say something in entreaty matters. And whether it matters not just while trying to please the dog but possibly also when pleading to the gods. This matters, because, well, they want to be sure it rains properly so they can have a good crop and so on - the usual bread-and-butter issues.
So yes, their curiosity has been kindled after that experience, observation and let’s say, realization. They suddenly become sensitive to the whole question of how they chant their prayers to the gods. Furthermore, it’s not just what the gods think of how they do their chanting, but it’s also the others in the wider community. They start to notice that between how Steve does the chanting and how Bryan does it, there seem to be differences. And correspondingly there are differences in how the people react to each of these guys - the levels of enthusiasm and respect seem to vary.
So Brenda and Irene are sat there when Steve and Bryan are doing the chanting - you see, they have somewhat respectable positions in the village, and it seems to have something to do with how they do the public worshiping. But what is it? Is it the sounds? Is it the words they use? Maybe it’s the words, because people respect when you use certain ‘high-status words’ rather than the more mundane ones?1
So there’s something to it. They resolve that they must understand this further. But where do you start? Well, they say, let’s observe the chanting, the use of speech, more closely. Very very closely.
Utter Fascination
This in essence is the bug that bit the ancients in the tradition of the preservation of the sacred word we discussed in the last episode. We saw that this in some sense can be traced back to the Proto-Indo-European idea of ḱléwos, for being famous and having glory meant being, literally, for theirs was on oral society, heard far and wide. And of course for the right reasons.
The right reasons, in the calculation of the Indo-Aryans, necessarily implied the right utterance of the words that carry your fame. And as we’ve also been seeing, this mattered both godward and manward - it was bound up with notions and necessities of prestige in the here and now.
So alongside the effort in preserving every syllable unchanged in a body of text longer than the Iliad and Odyssey combined, the RigVeda, there emerged a pressing need also to understand how language itself worked.
One other impetus for this was the fact that language is always changing. Language change, as I’ve touched upon in TecC15 where I first introduced the Proto-Indo-Europeans, is a fact of life - human speech evolves, today we don’t go around speaking like people in Shakespeare’s time did, heck, even the speech of the actors in those black-n-white movies from the 1930s sound rather strange to us, and your friends would burst out laughing if you spoke like that!
But this – language change – bothered the Vedic preservers. And just as most people today, perhaps much more so in their time, there was little notion of this scientific fact of linguistic mutability and evolution: we realize this perhaps only because we can peer into the past of our own language, whether a recorded voice from about hundred years ago or some written material from around 500 years ago, as in the above examples.
The ancient Vedic people, being part of a nonliterate society, and especially without a notion of historiography as we expect, didn’t see it that way. Language was not supposed to change, the changing speech of the common folk was but a corruption, the doing of the devil, so to speak.
This is vital for us to appreciate the nature of innovation that emerged from this. For, in order to arrest language change, they got to the very heart of the matter - starting with developing a keen understanding, in the first place, of the very thing that is speech: vāk. (This Indo-Aryan term is the same PIE word as we see in Latin voc-, as in ‘vocal’, ‘invoke’, ‘voice’ and so many more!)
Let’s Vāk the Talk!
We saw in the previous episode that the RigVeda, their body of prayer and praise poetry, was formally compiled and canonized by around 1000 BCE, and our story here is roughly in the 500 years from that point. At some point in those centuries arose schools of meticulous investigation of the phenomenon of language. These developed as ‘limbs’ of the Veda (vedānga), 6 branches in all. We have looked at one of them in the previous episode: metre/prosody. Of interest to us here are 3 others: phonetics, etymology, and one more, the big one!
It’s easy to imagine the importance they gave to phonetics, given their obsession with the ‘correctness of utterance’. And etymology, what we now understand as the science of the origin of words, was back in the day focused a bit more on elucidation of meanings of words, especially archaic ones, and will have to be considered with the historiographical caveat I’ve already mentioned above. But ultimately what mattered was the formation of words and thus sentences: grammar. Or broadly, how it all fitted together - don’t worry, I’m not going to get into the grammatical specifics of the ancient language. I promise!
The Lean and Mean Word-Machine
Instead, let’s look at one very simple scenario in English relating to how we form words and sentences - if you’re reading this, this example is child’s play!
Let’s take the word ‘undanceable’. Yes yes, you’re right, that’s not in the dictionary, and probably no-one has ever used that form before me right now here. But, honestly now, you understood what it might mean - and those amidst us with two left feet would certainly relate to it eh? (Besides I get to copyright my example?!)
But given such a word, we can recognize the components, and composition, of it. We can see that the ‘root’ or ‘base’ form here is ‘dance’, the ‘-able’ is, well, an indicator of ability, and the ‘un-’ is a negator - the latter two being affixes (prefixes and suffixes) which are, again, fixed to base forms, rather than used on their own.2
Now here’s the beauty of what the ancient Vedic grammarians did. They not only analysed and broke down their entire speech patterns down to all such basic components, but on top of that built a sophisticated algebraic system to definitively allow for the generation of new words, on the basis of a finite set of rules they developed.
Yes, algebraically. Which means, in essence you have equations, with input, rules and output. Throw in the component forms plus your intent into this concrete mixer of rules and out you get a shiny new word - it perhaps never existed before, but now there it is - all singing, all dancing. (And the irony of my choice of example!)
Anyways, let’s play with our newfound linguistic-algebraic skills a bit more. For it might come in handy the next time you’re near the dance floor and need excuses to fend off those inevitable “come onnn, let’s dance!!!” types in your group! Excuses with gravitas!
So we all probably recognize that ‘undanceable’ is technically an adjective - describing the state of someone, in this case in their two-left-footed glory. What about the adverb, for in case your friend did venture out to the dance floor in spite of! Just throw in the adverb-maker into the concrete mixer and voila you get ‘undanceably’.
And if you need to reflect on the phenomenon the next day, while nursing your hangover, you can do further concrete-mixing so you can mull over your ‘undanceability’.
Words of Honor. Once. Forever!
In essence however this is what they developed, which culminated in a single authoritative work around 500 BCE: The what’s-called Aṣṭādhyāyī, also written ‘Ashtadhyayi’, attributed to the grammarian named Pāṇini.
This work, the title meaning ‘eight-chaptered’, is a compilation of about 4000 precise rules that define ALL correct and approved use of the formal language. Yes, it defined the full bounds of the formal language. Once. And for all!
It had rules yes, but also metarules about how the rules functioned. It had global rules and then in specific circumstances local qualifications or exceptions.3 And I mentioned it was algebraic: it had axioms, theorems, conventions; it had abstract elements, variables, equations. You would, very much like I depicted above, apply a combination of rules to a collection of data elements, and get the results: generate output in the form of correct, legitimized language text.
The resulting defined and formalized language was eventually referred to as the ‘well-done’, the ‘complete’, the ‘perfected’, their word for it being ‘saṃskṛta-’: Sanskrit.
Letter Perfect
What does such a system of rules and all the elements I outlined above look like, to us in modern times, in the information age?
This of course is the algorithm. Pāṇini’s work lays out a comprehensive algorithm that implements all the core features of what we expect of a modern algorithm: it’s step-by-step, deterministic, has a defined process and executability - with both linear and recursive symbolic manipulation, it has a finite and well-defined set of instructions, flags, parameters and metadata. This of course is at the very heart of modern computer science, software engineering and all programming.
But here’s the thing. All of this was compressed in those 4000 sūtras, or aphorisms, I mentioned, which were, wait for it, entirely orally applied!
Yes, there was no ‘concrete’ version of that concrete mixer: no computing machine, no mechanical device, not even any writing. The whole algorithm and its components were stored and executed in the brains of the memorizers of the Aṣṭādhyāyī. It was all in their biological wetware!
In other words, they could perform live grammatical analysis, applying complex rule-chains spanning the entire breadth of their language, and generate or parse words in real-time conversation. No external notation, no scratch paper, no scribbling!
And all of this 2,500 years ago.
In technical terms, it was the first generative grammar of all time. (I’ll come back to that word another time with more juicy stuff!)
But watch this in action in its full algebraic glory:
(Full detail in footnotes)4
Generations Ahead
Interest in the nature of one’s own language is of course not unique to the Vedic people, the Mesopotamians in the millennium before our 1000-500 BCE timeframe here were already doing advanced astronomical stuff, as we explored in TecC 24. Around 600-500 BCE, the Greeks were starting, as we explored in TecC 23, to systematically explore the nature of reality. Early on the Chinese developed advanced forms of analysing their intricate character forms. But we don’t know of any attempt of algorithmic linguistic generation anywhere, of having invented formal grammatical computation.
We see similar developments next only in the mid-1900s in the new fields of computational linguistics, formal language theory, machine learning/NLP, cognitive science and computer science proper.5
Workaround Words
But also, it was perhaps all futile. Yes, this comprehensive, all-encompassing, top-down imposition held for a while, in the rigorous way intended - for a few centuries, perhaps give-or-take even a thousand years, but it began to wither.
Epistemologically speaking – and this matters directly in our study of innovation and progress, there’s only so far you can go with a top-down approach to solving problems. It’s true they provide a certain guarantee of determinism, but slowly and surely actors will inevitably find loopholes and workarounds. This happened in the current case and similarly we can see myriad examples in our modern top-down experiments as well.
Think about an example of the modern age: a written, national constitution, say penned in an atmosphere of optimism, even noble idealism, 250 years ago, very well crafted, but nevertheless by a small number of people, from the elite, in an attempt to manage the affairs of a nation-state across time and space. How much of the spirit of such a great body of law has still held? Not just the letter.
Something similar happened.. Sanskrit, in line with other archaic languages descended from Proto-Indo-European had a huge number of changes in the word forms - especially endings but also elsewhere in the word - in English, which also descends from PIE, we have retained vestigially but just a small fraction of this phenomenon (sing -> sang, sung; talk -> talked, talking…; ox -> oxen; mouse -> mice…)
And of course the said system laid out in full detail every rule and every exception as well.
Imagine you’re writing a poem where a whale is pursued by more than one octopus. And for the life of it, you’re not sure what the plural of octopus is (there are three contenders: the straightforward logical English octopuses, the Latin-derived octopi and the Greek-derived octopodes!)
The loophole here (both in English and in Sanskrit, they are sister languages!) is that when forming compounds you don’t need to use modified endings. So for example, you can say “a three-foot plank” without having to know or bother that the plural form of ‘foot’ is ‘feet’ and not ‘foots’.
In later Classical Sanskrit (say in the mid-first millennium of our era), they exploited this to the hilt. So to say “3 octop-?… are chasing the whale”, you’d say “the three-octopus-chased whale”. This of course is a contrived example but such usage sprang up like mushrooms all over the place, with the result that the language of literature deviated quite a bit from normal speech, while, adhering to most of the rules of Pāṇini all along. A more real example from the literature is in the case where a father is grateful that his sons are alive. Instead of what you’d expect he’d say, he says the literal equivalent of “Thank god, I’m alive-childed”!
It was the aim of the ancient Indian grammarians of Pāṇini’s ilk to arrest language change once and for all. To that extent, the language that arose out of his definitive work, Sanskrit, by its very description the ‘perfected’ language, was also by definition, an ‘artificial’ language - in the sense that it was or was soon to be the language which had no native monolingual speakers: what we normally call a ‘dead’ language. (Latin being a familiar example.)
But in this ‘stillborn’ definition of this language lay its immortality. For, from that time on, around 500 BCE, the clarity of what is (and isn’t) Sanskrit (along with other factors) afforded it a form of stability, constancy, reliability that stood the test of time for at least two full millennia from that point, in its use as an international language of erudite communication and literary expression, across not just the Indian Subcontinent, but especially all over South East Asia,6 but also Tibet, China and indeed as far as Japan!
Article written by Ash Stuart
Images and voice narration generated by AI
Further Reading & Reference
Cardona, George. (1997). Pāṇini. His Work and Its Traditions. MLBD. ISBN 8120804198. (Online Preview)
Renou, Louis. (1966). La Grammaire de Pāṇini. 4 Vols. École Française d’Extrême Orient.
Watkins, Calvert. (1995). How to Kill a Dragon: Aspects of Indo-European Poetics. Oxford University Press. ISBN 0195085957.
Benveniste, Émile. (2016). Dictionary of Indo-European Concepts. Hau Books. ISBN 978-0986132599.
Parpola, Asko. (2015). The Roots of Hinduism: The Early Aryans and the Indus Civilization. Oxford University Press. ISBN 978-019022690-9.
Reich, David. (2018). Who We Are and How We Got Here. Oxford University Press. ISBN 978-0198821250.
Jamison, Stephanie W & Brereton, Joel P. (2014). The Rigveda: the earliest religious poetry of India. Oxford University Press. ISBN 978-0199370184.
Monier-Williams, Monier, Sir. (1899). A Sanskrit-English Dictionary: Etymologically and Philologically Arranged with special reference to Cognate Indo-European Languages. Oxford University Press.
Böhtlingk, Otto. (1887). Pāṇinis Grammatik. Leipzig.
Pāṇini. (±499 BCE). Aṣṭādhyāyī. ISBN 0-0.
Compare “Dinner is served milord”, to “Grubs up, guv”
Be sure not to spell it as ‘undancable’ as that’s pronounced not unlike ‘undankable’ and your German friends wouldn’t be happy!
This hierarchical rule architecture appears throughout complex systems: federal law provides overarching frameworks while state/local jurisdictions add specific implementations; software configuration files use global defaults with local overrides; even modern CSS cascades from general stylesheets to page-specific rules. The principle Pāṇini articulated - global rules with contextual qualifications - remains fundamental to managing systematic complexity.
Base Elements
\(\begin{align} R &= \text{dance} \quad [\text{+V, -N, -Adj}] \\ S_1 &= \text{-able} \quad [\text{+Adj-forming, +V-selecting}] \\ P_1 &= \text{un-} \quad [\text{+negation, +Adj-selecting}] \\ S_2 &= \text{-ly} \quad [\text{+Adv-forming, +Adj-selecting}] \\ S_3 &= \text{-ity} \quad [\text{+N-forming, +Adj-selecting}] \end{align}\)Transformational Rules
\(\begin{align} T_1: \quad R[\text{+V}] + S_1 &\rightarrow R \circ S_1 \quad [\text{+Adj, -V}] \\ T_2: \quad P_1 + X[\text{+Adj}] &\rightarrow P_1 \circ X \quad [\text{+Adj, +neg}] \\ T_3: \quad X[\text{+Adj}] + S_2 &\rightarrow X \circ S_2 \quad [\text{+Adv, -Adj}] \\ T_4: \quad X[\text{+Adj}] + S_3 &\rightarrow X \circ S_3 \quad [\text{+N, -Adj}] \end{align}\)Derivational Sequence
\(\begin{align} \text{Step 1:} \quad &\text{dance} + \text{-able} \xrightarrow{T_1} \text{danceable} \\ &[\text{+V}] \quad [\text{+Adj-form}] \quad \rightarrow \quad [\text{+Adj}] \\ \\ \text{Step 2:} \quad &\text{un-} + \text{danceable} \xrightarrow{T_2} \text{undanceable} \\ &[\text{+neg}] \quad [\text{+Adj}] \quad \rightarrow \quad [\text{+Adj, +neg}] \\ \\ \text{Step 3a:} \quad &\text{undanceable} + \text{-ly} \xrightarrow{T_3} \text{undanceably} \\ &[\text{+Adj, +neg}] \quad [\text{+Adv-form}] \quad \rightarrow \quad [\text{+Adv, +neg}] \\ \\ \text{Step 3b:} \quad &\text{undanceable} + \text{-ity} \xrightarrow{T_4} \text{undanceability} \\ &[\text{+Adj, +neg}] \quad [\text{+N-form}] \quad \rightarrow \quad [\text{+N, +neg}] \end{align}\)General Form
\(\begin{align} \forall R[\text{+V}]: \quad ((P_1 \circ (R \circ S_1)) \circ S_{2|3}) &\rightarrow \text{Output}[\text{+Adv|+N, +neg}] \\ \\ \text{where } \circ \text{ denotes morphological concatenation} &\text{ with feature unification} \end{align}\)
Key developments include: Noam Chomsky's transformational-generative grammar (1957), which explicitly built rule-based systems for language generation; Claude Shannon's information theory (1948) establishing mathematical foundations for symbolic processing; early formal language theory defining context-free grammars and parsing algorithms; machine translation projects like Georgetown-IBM (1954) attempting systematic linguistic computation; and the emergence of computational parsing techniques that, like Pāṇini's system, used finite rule sets to generate infinite possibilities. Chomsky is said to have specifically acknowledged that his 'generative' approach echoed ancient grammatical traditions, though he appears unaware of the full algorithmic sophistication of Pāṇini's work.
The full and ceremonial name of Bangkok is in fact a big sentence, where, with the exception of the first word ‘krung’ which is of Khmer origin, all the rest of the words are derived from Sanskrit. Check it out:
[krung] deva mahā·nagara, amara·ratna·kosindra, mahindra·ayudʰya mahā·tilaka bʰava, nava·ratna·rāja·dʰāni purī·ramya, uttama·rāja nivedana, mahā·stʰāna amarābʰimana, avatāra·stʰita śakradattiya viṣṇukarma·prasiddhi
The city of angels, great city of immortals, magnificent city of the Nine Gems, seat of the King, city of royal palaces, home of gods incarnate, erected by Vishvakarman at Indra's behest.
Indra on the seal of Bangkok, with the first 4 words of the above in the Sanskrit script of Thai:
This was fascinating! Can I ask what the mechanism for transmitting the 4000 aphorisms to the memorizers was? Was there a formal school? I haven't listened to the previous episode, so I'll go back and do it -- maybe that's in there. It's ironic because I am just publishing today an article about how our modern obsession with grammar and style gets in the way of writing things that humans connect with. But it's really cool to see what an innovation grammar rules were when they were developed. Thank you for this window into language. I loved it!