Numbers that
decline, until they don't

— 1.1.23 — names बहु ("many"), गण-words, words ending in the affix वतुँ (यावत्, तावत् — "as much/that much") and words ending in डति (कति, यति — "how many/as many") as सङ्ख्या. Read that list again: एक, द्वि, त्रि, चतुर्, पञ्चन् … are not on it. They don't need to be — every commentary on this sūtra makes the same point Vāsu's English gloss makes explicitly: these four kinds of word are "certain words, besides the ordinary numerals," pulled into the same technical name because they behave like counting words without looking like one. एक through दशन्, and beyond, are सङ्ख्या from the outset; 1.1.23's job is only to extend the label to their look-alikes.

Why bother defining सङ्ख्या as a संज्ञा at all, rather than leaving it a plain English-style word class? Because, as with every technical name in this grammar, other rules key off it by name — and one of them is the reason this page exists at all.

— 2.1.52: a compound whose first member is सङ्ख्या is named द्विगु. Page 9 introduced द्विगु as "a numeral first member" and moved on, because it hadn't yet earned the right to ask which word can stand there and what it does grammatically once it does. That is exactly this page's content — and it is why 2.1.52 needs the technical name 1.1.23 built first: without a checkable संज्ञा सङ्ख्या, "compound whose first member is a numeral" would just be an intuition, not a rule a student could apply to a word never seen before. See the द्विगु entry again, now that you know what its first member is allowed to be: page 9 · द्विगु ↗.

एक is not just adjective-shaped, it is one of the forty-odd words page 7 named सर्वादि (1.1.27) — the corpus's own entry for एक carries the tag सर्वादीनि_सर्वनामानि, exactly the tag page 7 read off every सर्व/तद्/किम्-type stem. That is why the dative singular is एकस्मै, not *एकाय, and the genitive plural is rather than an ordinary अ-stem's -आनाम् — compare page 7's , the same substitution on सर्व. एक has a full singular, dual and plural, which looks odd for "one," until the gloss is read in full: केवलः "alone," अन्यः "other," भिन्नः "different." एके वदन्ति, "some say," is एक's ordinary nominative plural, doing exactly what "one" cannot do in English but "a certain" can. Its feminine, एका, is not a liṅga of एक in this corpus — it is its own headword, exactly the way page 7 found सर्वा beside सर्व.

द्वि is सर्वादि too — the corpus tags it सर्वादीनि_सर्वनामानि, same as एक — but its own entry carries two more tags no other numeral on this page does: एकवचनाभावः and बहुवचनाभावः, "absence of the singular" and "absence of the plural," stated outright rather than left for the reader to notice from empty table cells. — 1.4.22 — is the general rule behind that: it assigns द्विवचन to exactly two of anything and एकवचन to exactly one, so a word whose entire meaning is "two" can never appear in any other number. The irony is that being सर्वादि buys द्वि nothing visible: every सर्वादि-specific rule this series has met (7.1.14's -स्मै, 7.1.17's प्रथमा-बहुवचन शी) targets either the एकवचन or the बहुवचन — cells द्वि simply does not have. सर्वादि-hood is a real, checkable fact about द्वि; it is also, for this one word, grammatically inert.

त्रयः, त्रीणि — masculine and neuter look like an ऋ/इ-stem behaving normally. The feminine does not follow page 5's ordinary route (adding टाप् to make an आ-stem, the way रमा is made): — 7.2.99 — replaces the whole stem outright, तिसृ for त्रि and चतसृ for चतुर्, whenever a case-ending follows and the sense is feminine. तिस्रः, चतस्रः are not त्रि/चतुर् with an ending stuck on; they are exactly what 7.2.99 licenses in their place — the same kind of by-name substitution page 5's "अनियमाः" used for गो and पथिन्, just applied to a gender-split rather than a whole word.

— 1.1.24: सङ्ख्या words ending in ष् or न् are named षट्. That is exactly six words — पञ्चन्, षष्, सप्तन्, अष्टन्, नवन्, दशन् — and every one of them is attested in this corpus with exactly one paradigm: no म./स्त्री./नपुं. split at all. षट्-संज्ञा numerals count everything the same way, which is the grammar's own explanation for why पञ्च modifying पुरुषाः, नार्यः and फलानि never itself changes shape.

They still take all seven cases, though — पञ्चभिः, पञ्चभ्यः, पञ्चानाम्, पञ्चसु are as real as पञ्च itself, and every one is pulled from the same paradigm cell देव's तृतीया/चतुर्थी/षष्ठी/सप्तमी came from:

— 5.1.59 — is a निपातन sūtra: it simply lists पङ्क्ति, विंशति, त्रिंशत्, चत्वारिंशत्, पञ्चाशत्, षष्टि, सप्तति, अशीति, नवति and शतम् as irregularly-formed measure-nouns and licenses whatever derivation each one actually needs. Patañjali and the Kāśikā spell the etymology out for विंशति specifically: द्वौ दशतौ परिमाणमस्य संघस्य विंशतिः — "twenty is [the name of] a group whose measure is two tens." विंशति is grammatically a noun meaning a-twenty, built once, the way "dozen" is a noun in English and "twelve" is not.

Being an ordinary feminine noun, it declines in all three numbers — विंशतिः is just its nominative singular; विंशतयः, "several twenties," is a real plural:

Once सङ्ख्या is a noun in its own right, the word for what's being counted — Pāṇini's own term for it, used in , 2.2.25, is सङ्ख्येय, "the thing to be numbered" — has two ways to relate to it. It can simply stand beside विंशतिः in apposition, itself carrying the plural and whatever case the sentence needs:

Or it can go genitive, exactly the way any possessed noun does after a quantity-word (page 5's षष्ठी शेषे, "genitive marks whatever relation is left over"):

2.2.25 itself supplies a genuine attested example of the second pattern — अधिका विंशतिः , "more than twenty, of the cows" (गो's own षष्ठी-बहुवचन, pulled the same way every other form on this page is). Both विंशतिः पुरुषाः and विंशतिः पुरुषाणाम् are correct Sanskrit; which one a given sentence uses is style, not a grammar error either way.

त्रिभुवनम्, "the three worlds," is a द्विगु (2.1.52, above): त्रि is सङ्ख्या, भुवन is सङ्ख्येय, and the whole compound refers to the group — one neuter singular word standing for a collection, exactly the समाहार ("aggregate") sense page 9's द्विगु entry named but didn't need to unpack. That is the payoff of keeping the two terms distinct: सङ्ख्या never survives compounding as a separately-agreeing word the way it does in विंशतिः पुरुषाः — inside a द्विगु it fuses with its सङ्ख्येय into a single new word, and that new word takes its own gender and number (almost always neuter singular, समाहार), unrelated to either input's original gender. पञ्च itself has no gender to lose; त्रि and चतुर् do, and lose it the moment they enter a द्विगु — त्रिभुवनम् is नपुंसक, not the त्रि-masculine's त्रयः or the त्रि-feminine's तिस्रः. Full mechanics, and more worked compounds, are page 9's job: see them there ↗.

— 5.2.48 — is the general rule: the affix डट् (leaving just अ, इत्-markers dropped) means "completing [that count]," after a सङ्ख्या in the genitive. That is the default. Two of the words this page has already met override it by name:

• — 5.2.54 — after द्वि specifically, तीय replaces डट् outright: द्वि + तीय → द्वितीयः.
• — 5.2.55 — after त्रि, the same तीय applies, but with सम्प्रसारण (the semivowel र् of त्रि becomes the vowel ऋ first): त्रि + तीय → तृ+तीय → तृतीयः.

An ordinal's feminine follows exactly the two patterns page 5 already taught: तृतीया is an ordinary आ-stem (रमा's pattern), and चतुर्थी is an ordinary ई-stem of the नदी class. Both are genuinely attested in this corpus — and both are spelled identically to words you already met on page 5, where तृतीया and चतुर्थी are two of the seven case names. This corpus's own gloss disambiguates them: तृतीया here is glossed "" — completing three — and चतुर्थी "" — completing four, [also] a particular tithi. Same spelling, same declension pattern (both are सर्वादि- free, ordinary triliṅga adjectives), two different jobs depending on which noun they modify — a विभक्ति or a सङ्ख्येय.

These ten glyphs are not a grammatical form — no सङ्ख्या word underlies them any more than "7" is a form of "seven" — so they are the one thing on this page not pulled through C.paradigm. They are simply listed:

"१० of 24" in this series' own page numbers is exactly this system — positional, base ten, no declension involved at all. Compare the topbar above.

Sanskrit poets and colophon-writers had a third option, older than either of the first two in everyday use: pick a noun everyone associates with a specific count — "the moon" for one, because there is only one; "the eyes" for two — and use the noun itself as the digit. String several together and you have a number, read by a fixed traditional maxim: अङ्कानां वामतो गतिः, "the digits move leftward" — the first word named gives the ones place, the next the tens, the next the hundreds, and so on, the reverse of how the compound is normally read aloud. This convention is not a Pāṇinian संज्ञा — no rule in this corpus's sūtra index assigns a digit-value to a noun — so the value beside each word below is stated, traditional knowledge; only the word's citation form and gloss are pulled the way every other form on this page is. Four of the ten happen to be corroborated by the corpus's own gloss, which is flagged where it applies:

A few digits have more than one traditional word, chosen to fit a verse's metre: . And two words outside the single-digit decoder below are worth knowing for real colophons: .

Try decoding one

Click up to four words, in the order a colophon would name them. The strip below shows what gets written (left to right, as named) and, beside it, the number that convention actually means — read right to left.