धातुरूपाणि — the ten classes and their stems

Why a vikaraṇa shows up at all

3.1.67

Before any of the ten can be described, one switch has to be understood — because it is the switch that turns them on.

सार्वधातुकेThe word every vikaraṇa rule is standing under

3.1.67 opens this stretch of the Aṣṭādhyāyī, and its word सार्वधातुके — "when a सार्वधातुक [affix] follows" — carries on by अनुवृत्ति through every rule after it: 3.1.68 , through 3.1.81 , and the rest. Every one of the ten vikaraṇa is conditioned on exactly one thing — a सार्वधातुक ending is coming. Take that ending away and none of them are licensed to appear.

What counts as सार्वधातुक

3.4.113 — तिङ् endings (broadly), and शित् affixes: anything carrying a श् as its own इत्, which is most of the ten vikaraṇa themselves (शप्, श्यन्, श्नु, श, श्नम् all literally contain श्). 3.4.114 sweeps up everything left over — क्त्वा, तुमुन्, क्त, the future's स्य, the desiderative's सन् — and calls it आर्धधातुक.

The one deliberate exception. 3.4.115 — "and लिट् [too]". The perfect's endings are तिङ्, so 3.4.113 would sweep them into सार्वधातुक by default; this rule reaches in and pulls them back out, into आर्धधातुक. That single word is the entire reason the perfect (बभूव) has no vikaraṇa and reduplicates instead, while the present (भवति) does neither — it takes a vikaraṇa and does not reduplicate. The present, aorist, imperative and optative look like one family; the perfect, and everything built with a kṛt affix, look like another. This is the line between them.

What a gaṇa actually is

3.1.68 – 3.1.81

Ten rules, each naming one root-list and one affix. The affix is what the gaṇa is; the root-list is only which roots happen to answer that way.

इत्Reading a vikaraṇa's own name

Every vikaraṇa is cited with its full इत्-marked spelling — शप्, not अ — and the same three rules that closed page one's प्रत्याहार system strip it down to what actually attaches: 1.3.3 removes a final consonant, 1.3.7 and 1.3.8 remove certain initial consonants, and 1.3.9 elides whatever got marked. That is why शप् is pronounced with a bare अ in every form and the प् is never heard — and why णिच् (चुरादि) loses both ends, ण् by 1.3.7 and च् by 1.3.3, and surfaces as a bare इ.

All ten, with what each name is worth once its इत् letters are gone, whether 7.3.84 guṇa gets through, and how many roots the corpus actually files under it:

गण	विकरणम्	अवशिष्टम्	गुणविषये	उदाहरणम्	धातवः

One sentence per gaṇa

The same ten roots, in a sentence each — every noun form pulled from the corpus's own declension tables, every verb form the one this page's table just showed.

Guṇa, and what blocks it

7.3.84 · 1.1.5

Two roots can take what looks like the identical अ, and come out differently — because the rule that names that अ decides more than the sound does.

भू vs तुद्Same shape, two different sūtras, two different outcomes

7.3.84 — before a सार्वधातुक or आर्धधातुक affix, the अङ्ग takes गुण. भू's vikaraṇa is शप् (3.1.68); once its इत् letters vanish, nothing is left of it but the bare अ that follows — and 7.3.84 fires cleanly, turning भू into भो: भो + अ + ति, which 6.1.78 then turns into भवति.

तुद्'s vikaraṇa (3.1.77 ) is named श, not शप् — and after its own इत् letters vanish, what is left is also a bare अ. The two look identical on the page. But 1.2.4 — "a सार्वधातुक affix that is not पित् counts as ङित्" — applies to श (it has no प् among its letters at all) and not to शप् (whose very इत् letter is प्). And 1.1.5 blocks गुण and वृद्धि whenever a क्-or-ङ्-marked affix follows. So तुद् never reaches 7.3.84 at all: तुद् + अ + ति stays तुद्, giving तुदति — not the *तोदति a surface reading of "root + अ + ति" would predict.

The lesson, stated plainly. गुण is not a property of a vowel or a sound-shape. It is a property of which affix follows, traced back through what that affix's own इत् letters were before they were elided. Two identical-looking अ can carry opposite outcomes.

The same mechanism outside the present system. Compare भवति to भूत (the past participle, क्त — निष्ठा, 3.2.102 ). क्त is आर्धधातुक (3.4.114), and its own इत् letter is क् — so 1.1.5 blocks गुण directly, no detour through 1.2.4 needed: भू stays भू, giving भूत. भवति has भो- because its affix is not marked to block; भूत has भू- because its affix is. Same rule, same reason, two different routes to it.

परस्मैपदम् · आत्मनेपदम् · उभयपदम्

1.3.12 · 1.3.72 · 1.3.78

The class decides what attaches. A separate, independent choice decides which set of endings — and that choice belongs to the root, not to the speaker.

पदम्A lexical fact, not a free choice

Three rules divide the whole Dhātupāṭha. 1.3.12 — a root cited with an अनुदात्त (grave) accent, or marked ङित्, is आत्मनेपदी. 1.3.72 — a root cited स्वरित (circumflex) or ञित्, when the fruit of the action accrues to the agent, is उभयपदी. 1.3.78 — everything else, the शेष, is परस्मैपदी. The accent is written into the root's own citation form in the Dhātupāṭha; it is not something a speaker decides sentence by sentence.

तुद् and चुर् — two of the four roots this page follows — are both उभयपदी, which is exactly why the table above could show two forms for each: तुदति / तुदते, चोरयति / चोरयते. भू and अद् are परस्मैपदी only; the corpus's tables never had a second column to show for them.

Where this page's पद labels come from. C.dhatus() carries gaṇa, pada and artha per the brief, but the specific file it reads — dhatu/vidyut.json — does not carry a per-root पद field in this build of the corpus. The label shown here is read from its sibling dhatu/index.json instead, joined by the same Dhātupāṭha code; see the source note below the browsable table.

Watch a stem get built

Interactive

One root across all ten classes cannot happen — a root belongs to exactly one gaṇa (occasionally two, with two senses). But one root's stem being built, step by step, can. Four roots, four classes, four different things a vikaraṇa can do: arrive cleanly, arrive and get blocked, arrive twice, or never arrive at all.

Every step below is the actual rule the vidyut-prakriya engine cites when it derives these four forms — read off it once, not composed by hand. For any other root, any lakāra, any person or number, derive it live and see every rule the corpus's engine actually fires, not a curated four steps.

All 2,259, browsable

Coverage

धातुः	गणः	पदम्	अर्थः	meaning

Where this comes from. The 2,259 roots, their gaṇa, their aupadeśika citation and their artha are C.dhatus() — a straight read of the Dhātupāṭha as vidyut-prakriya parses it. पद is joined in from the corpus's own dhatu/index.json, as noted above. Every "प्रक्रिया →" link opens that exact root in the live engine, with nothing pre-computed or cached — the same engine this page's four-step walkthrough was read from.

Check yourself

अभ्यासः

Six questions.