Keeping the Sound

A reciter in a village somewhere in India draws a breath and begins a hymn. The man who taught it to him learned it the same way, by ear, and the man before that, and the chain runs back along the spine of three thousand years to a forest where no one had yet thought to write anything down. The pitch on which he lands the third syllable is not his choice.

It was fixed before the alphabet that could have recorded it existed.

This chapter is about how that is possible. Two of the six Vedangas, the limbs of the Veda introduced in the previous chapter, exist to guard one thing: the sound. Shiksha, the science of pronunciation, and chandas, the science of metre. The tradition placed them first among the limbs, and the order is not an accident. Before you can interpret a sentence of the Veda you have to keep it intact, and the Veda is sound before it is anything else. The question that organises everything below is a practical one. How do you hold a sound exactly, syllable for syllable and pitch for pitch, with no text to check it against, and what kind of science does that take?

The problem of the exact sound

Start with what was at stake. The Vedas were held to be effective as sound. A mantra mispronounced was, in the tradition’s own framing, a mantra that did not work, and there is a famous cautionary tale, recorded in later grammatical literature, of a sacrificer whose misplaced accent turned a blessing into its opposite. Whether or not the story is history, it tells you the stakes the culture assigned to the problem. Getting the sound wrong was not a small error.

So the sound had to be reproduced without drift. Not approximately. Exactly.

Shiksha (literally “instruction,” used here for the discipline of phonetics) is the answer to that requirement. It is the science of correct utterance, and it breaks the spoken Veda into parameters that can be taught and checked. The first is varna, the individual phoneme, the irreducible sound. The second is svara, the pitch accent, of which the tradition recognises three: udatta (raised), anudatta (not raised, lower), and svarita (a falling glide between them). The third is matra, the duration of a sound, measured in beats so that a long vowel and a short one are never confused. To these the manuals add the place and manner of articulation, the volume, the tempo. Every dimension along which a spoken syllable can vary was named, so that every dimension could be controlled.

Nothing was left to the ear’s good faith.

The manuals that measured it

The discipline did not float free. It was written into texts, and the most specialised of these are the Pratishakhyas. The name tells you their function: a Pratishakhya is the phonetic and sandhi (sound-combination) manual tied to one shakha, one branch or recension of the Veda. Each major branch had its own. The Rigveda-Pratishakhya governs the recitation of the Rigveda; the Taittiriya-Pratishakhya and the Vajasaneyi-Pratishakhya serve two branches of the Yajurveda; the Atharvaveda-Pratishakhya serves the Atharvaveda. These are not general guides. They are exacting rule-books for how the words of a particular text behave when they meet, which sounds change, which lengthen, which fuse, and how the accents fall across the joins.

Alongside the branch-specific manuals stand the general ones. The best known is the Paniniya Shiksha, a compact treatise on phonetics associated with the school of the grammarian whose work the next chapter takes up. Where the Pratishakhyas drill down into one recension, the Paniniya Shiksha states the principles of correct speech for the reciter as such.

What the texts contain, when you read them, is the part that has surprised modern linguists.

A sound-map built by anatomy

The ancient Indian phoneticians described speech by looking at the mouth. They classified every sound by its sthana, the place of articulation, the point in the vocal tract where the sound is formed, and by its prayatna, the manner of articulation, the kind of effort or constriction involved. They mapped the throat, the soft palate, the hard palate, the ridge behind the teeth, the teeth, the lips, and assigned each sound to its origin. This is articulatory phonetics, and it was being done with precision long before the field had that name in the West.

The proof is sitting in plain view in the varnamala, the “garland of sounds,” the ordered inventory that every Indian schoolchild still learns. It is not arranged by accident or by tradition’s whim. It is arranged by the anatomy of the mouth.

The vowels come first. Then the consonants, ordered by sthana, marching back to front through the mouth: velar sounds made at the soft palate, then palatal, then retroflex, then dental, then labial sounds made at the lips. And within each of those five groups the sounds are ordered again by prayatna, by manner: unvoiced before voiced, unaspirated before aspirated, with the nasal of that group placed last. The alphabet is a grid. Its rows are places in the mouth and its columns are manners of production.

The scholar W. Sidney Allen, in Phonetics in Ancient India (1953), set this tradition out for modern linguists and judged its articulatory description to anticipate the analysis that European phonetics arrived at only in the nineteenth century. That is a careful claim, and worth keeping careful. It is a statement about the description of articulation, made by a specialist, not a sweeping verdict that one tradition beat another at everything. Held to that scope, it holds.

A sound-map this precise gives you something to teach against. It does not, by itself, stop the text from drifting. For that the tradition built a machine.

The machine that corrected itself

The recitation disciplines are the engineering that turned exact phonetics into exact transmission. A reciter does not learn the Veda only one way. He learns it several ways at once, and the ways check each other.

The base form is the samhita-patha, the continuous recitation, in which the words run together with all their sound-changes applied, the way the verse actually sounds when spoken whole. Against it stands the pada-patha, the word-by-word recitation, in which every word is isolated and pronounced in its bare, uncombined form, so that you can hear exactly which word is which before the joins blur them. Already you have two views of the same line. If a sound-change in the samhita-patha has corrupted a word, the pada-patha exposes it, because it shows the word standing alone.

Then come the permutation recitations, and here the redundancy becomes formidable.

In the krama-patha, the step recitation, words are recited in overlapping pairs: first and second, then second and third, then third and fourth, each word spoken twice, bound to its neighbour on either side. In the jata-patha, the braided recitation, each pair is woven forward and backward and forward again. In the ghana-patha, the dense recitation, the pattern thickens further into a fixed scheme of forwards and backwards permutations that drills every word through its neighbours in a set order. A reciter trained in ghana-patha carries the most exalted title, the ghanapathi, and the discipline takes years to mount.

The point of all this braiding is not display. It is error-correction.

When every word is recited in fixed combination with the words around it, and in forward and reverse order, a single dropped or altered or misaccented syllable breaks the pattern audibly. The mistake has nowhere to hide, because it now contradicts several other recitations of the same passage. The text becomes self-checking. The scholar Frits Staal, among others, has argued that this is precisely how the Rigveda was carried, essentially unchanged, by mouth alone across thousands of years, a feat treated in the oral-preservation chapter of Volume I. Shiksha fixed the value of every syllable. The recitation patterns made sure no syllable could quietly change.

That is half the answer. The sound was held in place. The other half is the shape it was held in, and the shape is measured by number.

The counting of syllables

Chandas is the Vedanga of metre, the science of prosody, and at its root it is an act of counting. A Vedic verse is built from padas, metrical quarters, the lines of the verse, and each pada has a fixed number of syllables. Count the syllables and you have named the metre, and the metre is part of how the verse is remembered, because a line that runs short or long announces its own error.

The principal Vedic metres are defined by those counts. The gayatri has twenty-four syllables, three padas of eight, and it is the most sacred of the set; close to a quarter of the Rigveda is composed in it. The ushnih has twenty-eight. The anushtubh has thirty-two, four padas of eight, and it is the metre that later hardened into the shloka, the standard verse of the epics. The brihati has thirty-six, the pankti forty. The trishtubh has forty-four, four padas of eleven, and it is the single most frequent metre in the Rigveda. The jagati has forty-eight, four padas of twelve.

Take the gayatri to see the count made visible. The opening verse of Rigveda 1.7.1 runs as three lines, and in Ralph Griffith’s public-domain rendering of 1896 the first opens, “INDRA the singers with high praise.” The English is incidental here. The metrical fact is the thing: three lines, eight syllables each, twenty-four in all, the signature of the gayatri. Counting is the test. If a line does not come to eight, something has gone wrong, and the metre has caught it.

So metre is a second check running underneath the phonetic one. And it was about to become something stranger than a check.

Pingala and the light and the heavy

The treatise that turned the counting of metre into a formal system is the Chandahshastra of Pingala, also called the Chandahsutra, composed in the terse aphoristic style of the sutras, dated by historians to roughly the third or second century BCE. Pingala’s subject was metre. His method was mathematics, and the mathematics is what makes the work astonishing.

Pingala observed that every syllable in a metre is one of two things. It is laghu, light or short, or it is guru, heavy or long. Two values, and only two. A metre of a given number of syllables is therefore a string of lights and heavies, and Pingala gave a recursive procedure, the prastara or “spreading out,” that generates every possible such string in order. For a metre of n syllables there are exactly two-to-the-power-n patterns, and the prastara lays them all out, none missing and none repeated.

This is a binary enumeration. Two symbols, strung in every combination, counted by powers of two. Kim Plofker, in Mathematics in India (2009), treats Pingala’s procedure as the first known systematic enumeration of this kind, and notes a further detail: in carrying out his algorithm Pingala uses the word shunya, “empty” or “zero,” as a marker within the procedure. It is one of the earliest explicit uses of shunya as a symbol in an operation. It is worth saying plainly what it is not. It is not yet the full place-value zero of later Indian mathematics, the zero of Brahmagupta that functions as a number in arithmetic. It is a marker in an algorithm, an early step on a long road, and Plofker is careful to keep the two apart. So will we.

What grew out of Pingala’s spreading-out is where attribution has to be handled with care, because the popular accounts get it wrong.

Who did what

From the enumeration of metres come two famous mathematical objects, and neither is straightforwardly Pingala’s.

The first is the triangular array of numbers that the West would later call Pascal’s triangle, in which each number is the sum of the two above it. In the Indian tradition it is the Meru-prastara, the “spreading out of Mount Meru,” and it answers a natural metrical question: how many metres of n syllables have exactly so many heavy syllables. The array is implicit in Pingala’s procedure. The explicit triangular construction, and the name, come centuries later, set out in the tenth-century commentary on Pingala by Halayudha, the Mritasanjivani. The honest statement is that the structure is latent in Pingala and made explicit by Halayudha. Pingala should not be handed the finished named triangle.

The second object is the one most often mangled.

If instead of counting syllables you count beats, matras, and ask how many rhythmic patterns of a given total duration exist, the answer runs 1, 2, 3, 5, 8, 13, and onward, each number the sum of the two before it. This is the additive sequence the West knows under the name of Leonardo of Pisa, Fibonacci, who set it out in 1202 CE. In the Indian tradition it is older by centuries. The sequence is implicit in Pingala’s work on moraic metres, but it was first explicitly enumerated by Virahanka, writing somewhere around 600 to 800 CE, in his Vrittajatisamuccaya, and then again by Gopala and by Hemachandra in the twelfth century, all of them before Fibonacci. The accurate chain runs Pingala, then Virahanka who made it explicit, then Gopala and Hemachandra, and then, much later and independently, Fibonacci in Europe. Historians of mathematics who have traced this, including Parmanand Singh in his 1985 study in Historia Mathematica and B. van Nooten in his 1993 paper on binary numbers in Indian antiquity, use the historically honest label: these are the Virahanka numbers.

One more boundary, because it is easy to blur. Pingala is the man of binary and prosody. The grammar that the next chapter takes up belongs to Panini, a different figure entirely, and nothing here should be read as saying Panini invented the binary system. The two disciplines are neighbours among the Vedangas. They are not the same hand.

Step back and the whole apparatus resolves into one design. The sound of the Veda was fixed by shiksha down to the pitch of each phoneme, and held in place by recitation patterns that could not tolerate a single changed syllable. The shape of the sound was governed by chandas, which counted the syllables and so caught any line that ran short or long, and which, in Pingala’s hands, became a piece of formal mathematics that Europe would not reach by other routes for more than a thousand years. Sound and number, locked together, guarding a text with no pages.

What this preserved was the body of the Veda. What gives that body sense is the meaning of its words, and the limb that guards meaning is grammar. That is the most ambitious of the six, and it produced the work that some linguists still call the most rigorous description of any language ever made. We turn to it in the next chapter.

The reciter finishes the hymn. The last pitch lands exactly where it did three thousand years ago.