Reading Pāṇini · Part 1

Fourteen Drumbeats

About this series

It is an old custom to begin any study by honouring the teachers who carried it to us. The first verse salutes the whole lineage of teachers. The second, traditionally recited before the study of grammar, salutes Pāṇini himself. Notice what it credits him with: receiving the अक्षरसमाम्नाय (akṣara-samāmnāya), the “collection of sounds”, from Maheśvara, and building all of grammar on it. That collection is the fourteen Māheśvara Sūtras, and by the end of this post you will see why an entire grammar could rest on them.

Reading Pāṇini is a guided walk through the Aṣṭādhyāyī, India’s great grammar of Sanskrit, for readers who know no Sanskrit at all. Before we start, a word on why grammar deserves this attention, and on what to expect along the way.

Grammar: the mouth of the Veda

Traditional learning surrounded the Veda with six auxiliary sciences, the वेदाङ्ग (vedāṅga), “limbs of the Veda”. A verse from the पाणिनीय शिक्षा (Pāṇinīya Śikṣā) pictures the Veda as a person and gives each science a part of the body:

Why is grammar the mouth? Because the mouth is where meaning becomes speech. A text you cannot parse is a text that cannot speak to you. Every word of a Sanskrit verse can be split, joined and inflected in more than one way, and only grammar tells you which reading is correct.

Patanjali, Pāṇini’s great commentator, lists the purposes of studying grammar in five words: रक्षोहागमलघ्वसन्देहाः (rakṣo-ūha-āgama-laghv-asandehāḥ). They are: to protect the Veda from corruption, to adapt its formulas correctly to new contexts, because tradition enjoins it, for economy (learning rules instead of endless word lists), and to remove doubt.

His example of doubt is memorable. A ritual text calls for a स्थूलपृषती (sthūlapṛṣatī) cow. Does that mean a cow that is “large and spotted”, or one “with large spots”? The written letters are identical. Only the grammarian, who knows how the word’s accent changes with its meaning, can settle the question. Grammar, in other words, is the discipline that makes an ancient text say one thing, unequivocally.

What to expect

• No prior knowledge needed. We start from sounds and build up, one idea per post.
• Two scripts, always. Sanskrit appears in Devanagari with a romanised form (IAST) in brackets, so readers who don’t read Devanagari can follow every step.
• Real sūtras, decoded. Each post ends with you reading actual rules of Pāṇini yourself.
• Things to play with. Interactive tools, recordings of chanted texts, and small worked examples.
• Asides for the technically minded. Occasional notes link Pāṇini to chemistry, computer science and linguistics. You can skip them without losing the thread.
• Tradition and scholarship side by side. Where the traditional account and modern research differ, you’ll hear both.
• Read in order. Each part builds on the one before. The series box at the top of every post shows where you are.

Half a syllable

There is an old saying among Sanskrit grammarians:

It’s a joke, of course. But like most good jokes it hides something true. These were people who would rewrite a rule to make it half a syllable shorter, and then celebrate.

Why would anyone care that much? The answer leads us to one of the most remarkable intellectual works ever produced: पाणिनि (Pāṇini)’s अष्टाध्यायी (Aṣṭādhyāyī), a complete grammar of Sanskrit in about 4,000 rules, composed some 2,400 years ago and memorised, not written down.

This first post shows the scale of what Pāṇini achieved, and then gives you a first, hands-on taste of how his system works.

The scale of the achievement

Pāṇini did something no one else would do for another two thousand years: he wrote a complete, rule-based machine that generates every correct word of a language, and only those words.

What it meant for Sanskrit

Pāṇini was not the first Indian grammarian. He names about ten predecessors, among them आपिशलि (Āpiśali), शाकटायन (Śākaṭāyana) and शाकल्य (Śākalya). Their works are almost entirely lost, and the usual explanation is that Pāṇini made them unnecessary.

After him, his grammar became the yardstick of “correct” Sanskrit. Poets like Kalidasa, philosophers like Adi Shankara, and scientists like Aryabhata all wrote in a language whose correctness was measured against Pāṇini. The very word संस्कृत (saṃskṛta) means “refined, well-formed”. For more than two millennia, Pāṇini was the definition of well-formed.

What it meant for formal languages

Here is where a modern reader should sit up. To make 4,000 rules fit in human memory, Pāṇini had to invent tools that look startlingly like those of modern linguistics and computer science:

The third row is the most striking. In Pāṇini’s code, the grammatical case of a word tells you its role in the rule. The genitive marks what gets replaced. The nominative marks what replaces it. The locative marks what follows. So the three-word sūtra

is a complete rewrite rule: ik → yaṇ / _ ac. In plain words: “replace a sound of class ik by the matching sound of class yaṇ when a sound of class ac follows.” By the end of this post you’ll know exactly what ik, yaṇ and ac mean, and you’ll be able to run this rule yourself.

Western linguistics reached formal rewrite rules in the 1950s. Leonard Bloomfield, a founder of American linguistics, called Pāṇini’s grammar “one of the greatest monuments of human intelligence”. In 1967, the computer scientist Peter Ingerman, noticing how closely Pāṇini’s notation resembles the Backus–Naur Form used to define programming languages, proposed in a letter to the Communications of the ACM that it be renamed “Pāṇini–Backus Form”.

What the Aṣṭādhyāyī is, and isn’t

The Aṣṭādhyāyī (“eight chapters”) is a generative grammar of Sanskrit: a set of rules that, starting from roots and suffixes, builds every correct word form.

Who and when. Pāṇini came from शलातुर (Śalātura) in ancient Gandhāra, in the north-west of the subcontinent. Most scholars place him around the 5th–4th century BCE.

Why. Grammar, व्याकरण (vyākaraṇa), literally “taking apart”, is the “mouth” of the Veda in the verse above. Pāṇini describes both the Vedic language and the everyday educated speech of his own time, भाषा (bhāṣā).

How it’s built.

• 8 chapters (अध्याय, adhyāya) × 4 quarters (पाद, pāda) = 32 sections.
• About 4,000 sūtras (the exact count varies by edition, from about 3,959 to 3,996).
• Rules are cited by address: 6.1.77 means chapter 6, quarter 1, sūtra 77.
• Companion lists hold the raw material: the 14 Māheśvara Sūtras (the sounds), the धातुपाठ (Dhātupāṭha, about 2,000 verb roots) and the गणपाठ (Gaṇapāṭha, lists of words that behave alike).

Who carried it forward. कात्यायन (Kātyāyana) added critical notes (वार्तिक, vārttika). पतञ्जलि (Patanjali) wrote the great commentary, the महाभाष्य (Mahābhāṣya), around the 2nd century BCE. Together they are honoured as the त्रिमुनि (trimuni), the “three sages” of grammar.

What it is not

• Not a dictionary. It lists roots and word classes, not meanings of words.
• Not a phrasebook or textbook. It is the engine under the hood, not a guide to conversation.
• Not a scripture. It is deeply respected in tradition, but it is a technical text: no stories, no gods, no kings.
• Not the invention of Sanskrit. The language was already alive and old. Pāṇini described it so precisely that it has stayed stable ever since.

The sūtra: a theorem in a few syllables

A सूत्र (sūtra, literally “thread”) is the shortest possible statement of a rule that still says exactly one thing. A well-known verse lists what a sūtra must be:

“Those who know sūtras know a sūtra to be this.”

Why be so stingy?

Think about how knowledge travelled 2,500 years ago. There were no printed books. Writing materials like palm leaves were scarce and perishable. Knowledge lived in human memory and moved from teacher to student by recitation.

In that world, every syllable has a cost: it must be memorised, recited and passed on without error, generation after generation. Shorter rules are cheaper to store and harder to corrupt. Hence the joy over half a syllable.

Here’s a modern way to picture it. A sūtra text is like a compressed file. On its own, it looks cryptic, almost unreadable. But a trained mind carries the “decompression software”: the conventions for reading it. Unzip the Aṣṭādhyāyī with those conventions and out comes the entire grammar of a language.

The rest of this post is about the first and cleverest of those conventions.

Fourteen drumbeats: the Māheśvara Sūtras

Before the first rule, Pāṇini’s system begins with fourteen short lines that list the sounds of Sanskrit. They are called the माहेश्वर सूत्र (Māheśvara Sūtras), the “sūtras of the Great Lord”, or शिवसूत्र (Śiva Sūtras). (They are not to be confused with the Śiva Sūtras of Vasugupta, a text of Kashmir Śaivism.)

The story

Tradition gives them a beautiful origin, preserved in the नन्दिकेश्वरकाशिका (Nandikeśvara-Kāśikā):

As Naṭarāja (Śiva as cosmic dancer) finished his dance, he struck his डमरु (ḍamaru) fourteen times, and each beat rang out as one of these lines. Tradition holds that Pāṇini heard them and built his grammar on them.

Many modern scholars think Pāṇini (or a predecessor) composed the list himself. As you’ll see, either way, it is an extraordinary piece of design.

The fourteen lines

Read them slowly and one thing stands out. Every line ends in a consonant without a vowel, like ण् (ṇ) or क् (k). These are not sounds being listed. They are markers, Pāṇini’s bookmarks, and they are the key to everything that follows. (Consonants in the list are pronounced with an “a” only so they can be recited: ह is just h.)

A periodic table of sounds

At first glance the order looks scrambled. Why is ह (h) listed twice? Why are क (k) and प (p) cut off from the other voiceless stops? The answer: Pāṇini ordered sounds by how they behave in his rules, not simply by how they sound.

Roughly, the list runs like this:

Words like “voiced”, “aspirated” and “nasal” describe how and where the mouth makes each sound. Indian phoneticians mapped this with great precision long before Pāṇini, and their map is what gives the familiar alphabet, the varṇamālā, its neat rows and columns. That map is the subject of Part 2.

If you studied chemistry at school, this should feel familiar. When Mendeleev built the periodic table in 1869, he ordered the elements by atomic weight, except where that broke a chemical family. He put tellurium before iodine, against the weights, because iodine clearly belonged with chlorine and bromine. Behaviour beat the simple sort.

Pāṇini made the same kind of choice, about 2,300 years earlier. His goal: every group of sounds that a rule needs should sit in one unbroken stretch of the list. Some examples:

• k and p are split off (line 12) so that one rule can target exactly ch ṭh th c ṭ t, without k or p.
• h appears twice. In line 5 it falls inside the voiced sounds, where it behaves like them. In line 14 it joins the sibilants, the other “breathy” sounds.
• The same marker ṇ is used twice (lines 1 and 6). That’s a small cost for a shorter list, and the tradition resolves the resulting ambiguity from context.

Modern researchers have treated this as a mathematical puzzle: find the shortest list in which every needed group is a contiguous stretch. Linguist Wiebke Petersen showed that no such list can avoid repeating at least one sound, and that Pāṇini’s solution is about as compact as possible.

Pratyāhāras: the compression trick

Here is the payoff for all that careful ordering. With one rule, Pāṇini can name any stretch of the list using just two sounds.

A name made this way is a प्रत्याहार (pratyāhāra), a “drawing together”. The recipe:

1. Pick the sound where your group starts.
2. Walk forward through the list until your group ends.
3. Take the marker that comes right after that point.
4. Glue the start sound to the marker. Skip any markers you passed along the way; they are bookmarks, not sounds.

Try it. Start at a, walk to the end of line 4 (ai, au), and take the marker c. The name is अच् (ac), and it means all the vowels. Start at h in line 5 and run to the very last marker l: हल् (hal), all the consonants. Start at a and run to the end: अल् (al), every sound.

One more detail: a vowel in these groups also stands for its long form, so ac includes ā, ī, ū and so on (another rule, 1.1.69, takes care of that).

Try it yourself

Click a starting sound and then a dashed marker, or pick one of the famous groups. Watch the same group light up in the familiar alphabet underneath.

Why this is clever

Imagine writing a rule that applies “before any vowel”. Without pratyāhāras you’d have to list them all, every time. With them, you write ac: two syllables, reusable in hundreds of rules. The grammar uses a little over forty of these names, and they appear on almost every page.

If you write code, here’s the cleanest way to think about it. The fourteen lines are an array. The markers are bookmarks inside it. A pratyāhāra is a slice: “from this element up to that bookmark”. Because slices can overlap, one sound can belong to many groups. The vowel i sits inside ac, ik, aṇ and al at the same time, with no extra storage.

For programmers, here is Pāṇini’s rule 1.1.71 in about fifteen lines of Python. Markers are in capitals:

# The 14 Maheshvara Sutras: each line's last letter is a marker (anubandha)
SUTRAS = ["a i u Ṇ", "ṛ ḷ K", "e o Ṅ", "ai au C", "h y v r Ṭ", "l Ṇ",
          "ñ m ṅ ṇ n M", "jh bh Ñ", "gh ḍh dh Ṣ", "j b g ḍ d Ś",
          "kh ph ch ṭh th c ṭ t V", "k p Y", "ś ṣ s R", "h L"]

def pratyahara(start, marker, nth=1):
    """Rule 1.1.71: the start sound + a marker = everything in between."""
    group, collecting = [], False
    for line in SUTRAS:
        *sounds, mark = line.split()
        for s in sounds:
            if s == start and not collecting:
                collecting = True
            if collecting:
                group.append(s)
        if collecting and mark == marker:
            nth -= 1
            if nth == 0:
                return group

print(pratyahara("a", "C"))   # ac  -> ['a', 'i', 'u', 'ṛ', 'ḷ', 'e', 'o', 'ai', 'au']
print(pratyahara("i", "K"))   # ik  -> ['i', 'u', 'ṛ', 'ḷ']
print(pratyahara("y", "Ṇ"))   # yaṇ -> ['y', 'v', 'r', 'l']

A small puzzle the old commentators enjoyed: the rule that says “a final consonant in Pāṇini’s lists is a marker” is 1.3.3 हलन्त्यम् (halantyam). It uses the name hal to say so, but hal only works if you already know the marker l is a marker. The system defines its markers using one of its own abbreviations. The tradition has a careful answer, but it’s a wonderful example of a system describing itself.

Now read Pāṇini yourself

You now hold enough of the decompression software to read real sūtras. Let’s start where Pāṇini starts.

Rule 1.1.1

Split it: vṛddhiḥ + āt + aic. aic is a pratyāhāra you just met: ai, au. The t after ā is another marker, meaning “exactly this sound, ā”. So the whole rule says: ā, ai and au are called vṛddhi.

Four syllables, and we have a technical term. That’s a definition, a संज्ञा (saṃjñā) sūtra. Tradition also notes that Pāṇini chose to open his grammar with the word vṛddhi, “growth”, an auspicious start.

Rule 1.1.2

at = exactly a; eṅ = e, o. So: a, e and o are called guṇa.

Two rules, both decoded with pratyāhāras. These two labels, guṇa and vṛddhi, will do an enormous amount of work later in the grammar.

Put the labels to work

Here is where it gets fun. When two words meet, their sounds often fuse. This is सन्धि (sandhi), “joining”. Let’s run three real rules.

6.1.87 आद्गुणः (ād guṇaḥ): when a or ā meets a following vowel like i or u, both are replaced by one guṇa vowel. a + i → e, a + u → o.

6.1.88 वृद्धिरेचि (vṛddhir eci): when a or ā meets e, o, ai or au, both are replaced by one vṛddhi vowel: ai or au.

6.1.77 इको यणचि (iko yaṇ aci): the rule from the start of this post. An ik vowel (i u ṛ ḷ) becomes its matching yaṇ semivowel (y v r l) before any vowel ac. i → y, u → v.

Notice what just happened. You didn’t memorise five words. You learned three rules, and those rules will produce thousands of words you’ve never seen. That’s the generative idea in miniature, and it’s why Pāṇini’s grammar works more like a program than a book.

You may also have noticed that 6.1.87 is only two words long, yet my explanation of it was a full sentence. Where did “when a vowel follows” and “both are replaced by one” come from? They are silently inherited from the rules just before it. That inheritance, called anuvṛtti, is the subject of Part 3.

Why it still matters

Pāṇini’s grammar is not a museum piece. It is still taught and used in traditional Sanskrit schools, and it remains a working tool for modern researchers.

• Linguistics. Ideas we now credit to the 20th century, such as rewrite rules, zero elements and rule ordering, were already at work in the Aṣṭādhyāyī. Modern linguists still publish on how his system is built.
• Computing. Researchers have built computer programs that run Pāṇini’s rules to generate and analyse Sanskrit words. A grammar written for human memory turns out to be remarkably close to something a machine can execute.
• An open text. Pāṇini scholarship is still alive and still argued over. In 2022 a Cambridge PhD thesis by Rishi Rajpopat proposed a new reading of the rule that settles conflicts between rules (1.4.2). It made world headlines, and traditional scholars pushed back hard. We’ll look at that debate in Part 6.

The deeper lesson is about compression and design. Faced with a hard constraint (everything must fit in human memory), Pāṇini didn’t just shorten his text. He invented a notation, a way of encoding knowledge so that the structure of the language was visible in the structure of the rules.

Coming up next

In Part 2: Where Sounds Are Born, we’ll step back from Pāṇini’s list to the mouth itself. A short traditional formula, अकुहविसर्जनीयानां कण्ठः (akuhavisarjanīyānāṃ kaṇṭhaḥ), “a, the k-row, h and visarga are made in the throat”, begins a complete map of where every sound is produced. Once you see that map, the familiar order of the varṇamālā (ka kha ga gha ṅa, ca cha ja jha ña …) stops looking arbitrary, and Pāṇini’s own definition of “similar sounds” falls into place.

Notes and further reading