उपसर्गाः — twenty-two prefixes and what they do

उपसर्ग is a job, not a spelling

1.4.58 – 1.4.59 · 1.4.83

The same twenty-two words are called three different things depending on what they are doing in the sentence. Only one of those three names is "उपसर्ग", and it is earned, not given by default.

प्रादयःThe list itself

1.4.58 — "the प्र-etc. [words]" — names the closed list of twenty-two by pointing at its first member and saying "and those like it." The Kāśikā spells the list out in full:

That is C.krt()['upasarga'] from the corpus's own core/vidyut.json — entries, the same twenty-two the vidyut-prakriya engine accepts as its prefix parameter. One spelling difference is worth naming rather than stepping around: the Kāśikā cites the sixteenth member as उत्, the corpus's own table as उद्. They are the same morpheme — उत् is what its final consonant surfaces as before an unvoiced sound (8.4.55 खरि च), उद् its citation form — and this page, like the corpus, uses उद्.

उपसर्गाः क्रियायोगेThe condition that earns the name

1.4.59 — the twenty-two are called उपसर्ग only when joined to a verbal action. Joined to a verb, प्र and परि turn नी ("lead") into two distinct professions:

प्रणयति "leads forth", परिणयति "leads around" — hence प्रणायकः "a leader", परिणायकः idiomatically "one who leads [a bride] around [the fire]", a husband. Same root, two prefixes, two words with nothing in common but नी.

क्रियायोग इति किम् — and what if it isn't joined to a verb? The Kāśikā's own answer, on this very sūtra:

"[Someone] departed [प्रगतः] as leader from this country — a प्र-leader country." प्र is written directly against नायक, exactly as it would be against a verb, and it still is not उपसर्ग — because नायक is a noun, not a verbal action, and 1.4.59's condition simply is not met. Spelling together is not the test. क्रियायोग is.

कर्मप्रवचनीयाःThe same words, a different job

1.4.83 opens a second adhikāra running to 1.4.98, under which several of these same twenty-two words — not joined to a verb, but governing a noun the way an English preposition does — get a third name, कर्मप्रवचनीय, and a case-assignment rule of their own (2.3.8 onward). Three of the Kāśikā's own examples:

अभि1.4.91, "toward, in the sense of no [claimed] share" सु1.4.94, "in praise" अधि1.4.97, "over, of an owner"

"Lightning flashes toward (अभि) the tree." "Devadatta [behaves] well toward (अभि) his mother." — अभि governing a noun directly, no verb of "flashing toward" or "behaving toward" grammatically underneath it as the thing अभि attaches to.

"Well (सु) sprinkled by you" — praise of the sprinkling, सु doing the praising rather than a directional job.

Why this matters for reading, not just for taxonomy. A word like सु or अधि sitting in front of a noun is not automatically the "prefix" half of a compound verb. Whether it is उपसर्ग, कर्मप्रवचनीय, or (1.4.56 onward) plain निपात depends on what it is doing in that sentence — and that is also why §3 below can say, honestly, that सु and अपि do not get a fixed directional gloss on this page: in Classical Sanskrit they most often are not functioning as verbal prefixes at all.

गतिसंज्ञा — a second name for the same twenty-two, not covered further here

1.4.60 assigns the identical प्रादयः a fourth technical name, गति, active under the same क्रियायोगे condition. गति governs accent and compounding rules (6.2.49 and others) that this page does not open; उपसर्ग and गति overlap in membership but are different संज्ञा for different rule-sets, and the two should not be conflated.

The prefix drags the meaning elsewhere

हृ · six ways

A line quoted constantly in the grammatical tradition — and worth testing, not just repeating.

उपसर्गेण धात्वर्थो बलादन्यत्र नीयते"The prefix drags the root's meaning elsewhere, by force"

Where this line is from. It is not a sūtra and not in this corpus — none of the nineteen commentary layers on 1.4 contain it, checked directly. It is a maxim widely quoted in the teaching tradition (often traced to Vākyapadīya-adjacent grammatical literature), presented here as received wisdom, not as something this page can cite chapter and verse for. What can be checked is whether it is true.

हृ () with six of the twenty-two, each an established word, each already carrying an English verb that has nothing to do with the other five:

आहरति, प्रहरति, संहरति, विहरति, उपहरति, प्रतिहरति — six words a dictionary lists as six unrelated verbs. Read as हृ plus a prefix, they are one root, moved six different directions by six different prefixes. The line is not hyperbole.

All twenty-two, on हृ

The systematic senses where the tradition states one, हृ's actual derived form (both पद — हृ is उभयपदी), and whether the corresponding -हार noun (उपसर्ग + हृ + घञ्) is one of this corpus's 9,007 paradigm stems. Most cells in the last column are honest gaps, not denials — प्रहार, संहार, प्रतिहार are entirely real, common words that this particular corpus build simply does not carry as a nominal stem.

उपसर्गः	सामान्यार्थः	हृ (परस्मैपदे)	हृ (आत्मनेपदे)	-हार अत्र कोशे?

Where every form in this table comes from. Every हृ+उपसर्ग form was read once from the live vidyut-prakriya engine —

AD.engine.call
  ('deriveTinantas', {dhatu: AD.engine.dhatu(hṛ, {prefix: u.slp}), lakara:'Lat',
  prayoga:'Kartari', purusha:'Prathama', vacana:'Eka'})

for each of the twenty-two core/vidyut.json prefixes — driven over CDP against dhaturupa.html, exactly as pages 12 and 18 read their forms. Every -हार attestation is C.has() against the corpus's own shabda/index.json, checked live at build time, not assumed. One honest gap: for उद् the engine returns two candidate forms, उद्धरति and उद्हरति (the द्+ह् sandhi applied and not applied); this table shows only उद्धरति, corroborated by उद्धार being one of the attested corpus stems — उद्हरति is not asserted to be wrong, just not shown.

Two of the twenty-two are the same word, twice

1.4.58 · 8.2.66 · 8.3.34

Pāṇini's own list cites निस् and निर् — and दुस् and दुर् — as four separate entries. Run all four through the engine against the same root and see what actually comes out.

निस् । निर् । दुस् । दुर्Four citations, how many outputs?

Against हृ (voiced-initial — ह् is a voiced sound) and against कृ (voiceless-initial — क् is not), all four selections, run through the same derivation:

चयनम्	+ हृ (लट्, प्रथम, एक)	+ कृ (लट्, प्रथम, एक)

The two selections in each pair produce the identical form, every time. निस्+हृ and निर्+हृ both give निर्हरति; निस्+कृ and निर्+कृ both give निष्करोति. Choosing "निस्" from the twenty-two makes no difference to the output — the engine's own sandhi resolves the citation form to whatever the following sound actually calls for: र् before a vowel or voiced sound (हृ), ष् before क् (8.3.59's chain, below). निस्/निर् and दुस्/दुर् are not four prefixes; they are two, each cited under both of its own historical spellings because Pāṇini's text needs to name both — the "two that behave differently" among the twenty-two are exactly these, each showing synchronic allomorphy that most of the other twenty simply do not have.

What carrying an उपसर्ग actually changes

7.1.37 · 8.4.14 · 8.3.59

The senses above are a minority case — most prefixed roots are simply lexicalised, one word to be learnt at a time, not computed from a table. What a reader can rely on are three concrete grammatical facts that follow whenever a root carries an उपसर्ग at all, regardless of which one.

क्त्वा → ल्यप्The gerund changes shape

7.1.37 — with any उपसर्ग in front of the root, क्त्वा (the bare gerund, "having done") is replaced outright by ल्यप्. Page 18 derives this rule by rule for नम् and कृ; the same two roots, checked again here directly against the engine:

धातुः	bare क्त्वा	+ उपसर्ग → ल्यप्

Both गम् and नम् confirm the same fact page 18 already established: the engine derives two spellings after the prefix (6.1.71's तुक्, or 6.4.38's optional nasal loss), and both are grammatical. This page follows the spelling used across the literature — आगत्य, प्रणम्य — exactly as page 18 does.

णत्वA root's own न् turns to ण्

नम् is cited in the Dhātupāṭha as णमँ — a णोपदेश root, whose न् is really ण् underlyingly, surfacing as न् by a general rule except where something restores it. 8.4.1 and 8.4.2 turn a following न् to ण् after an उपसर्ग ending in र् or ष्, reaching past an intervening vowel; 8.4.14 applies that specifically to a णोपदेश root's own initial letter, and — unlike 7.1.37 above — it is not limited to the gerund. It fires in the plain present too:

नमतिnot attested bare — see below प्र (ending in र्) + नम्, लट्

षत्वA following स् turns to ष्

8.3.57 — स् becomes ष् after इण् (an इ, उ, ऋ or ऌ) or a क-वर्ग sound, reaching through the adhikāra 8.3.59 and — for roots not directly adjacent — 8.3.65 , which names सिच् and सिध् directly among the roots it covers. अभि and नि both end in इ — a member of इण् — so a स्-initial root right after either one triggers 8.3.57 without needing any further extension:

धातुः	bare	+ उपसर्ग

अभिषिञ्चति "anoints, sprinkles upon" — the same rule the Kāśikā cites its own example under, अभिषिक्तम् (1.4.60's commentary); निषेधति "forbids, restrains" — नि, ending in the same इ, does the same thing to सिध्.

प्राक् प्रयोगे, छन्दसि परेऽपिWord order: fixed in Classical, free in Vedic

1.4.80 — a गति or उपसर्ग must be used immediately before the verb it belongs to. That is Classical Sanskrit's rule, full stop — प्रगच्छति, never *गच्छति प्र. But two rules immediately after it open a Vedic exception:

छन्दसि परेऽपि1.4.81 — in the Veda, even after [the verb] व्यवहिताश्च1.4.82 — and [even] separated [from it]

"He goes down with the elephant" — नि याति (Vedic word order, उपसर्ग after the verb) means exactly what नियाति (Classical order, compounded) means; likewise नि हन्ति / निहन्ति, "strikes down". Classical Sanskrit allows only the second form of each pair; the Kāśikā gives both as equally licensed in the Veda.

व्यवहित — not just reordered, but separated by other words. The Kāśikā's own citation, quoting the Ṛgveda directly:

आ...याहि — इन्द्र is invoked with आ and याहि (from या, "come") pulled apart by an entire verse's worth of vocative and instrumental phrases between them (Ṛgveda 3.45.1), where the ordinary compounded आयाहि (Ṛgveda 6.16.10, quoted right after it) shows the same two morphemes fused as they always are in Classical prose. Only the Veda permits the gap.

Walk the twenty-two

Interactive

हृ, one prefix at a time. The root never changes; watch the form and the meaning both move.

Check yourself

अभ्यासः

Six questions.