Ten shapes
for one abstract ल

3.4.77 — “in place of ल.” Everything from here through 3.4.117 replaces that one syllable with something concrete: 3.4.78 lists the eighteen personal endings (तिप्, तस्, झि… nine पर्स्मैपद, nine आत्मनेपद) that fill the slot in the ordinary case. But ल itself is never typed bare — it always shows up already wearing one of ten names: लट्, लिट्, लुट्, लृट्, लेट्, लोट्, लङ्, लिङ्, लुङ्, लृङ्. Those names are what 3.2.123 and the other defining rules below actually license — then 3.4.78 replaces whichever one got licensed with an actual ending.

What the extra letters are doing

Every name is ल plus a vowel plus one final consonant — ट् or ङ् — and every one of those three added sounds is marked इत् and vanishes before anything is pronounced: the vowel by 1.3.2 (a nasalised vowel in the citation form is इत्), the consonant by 1.3.3 (हलन्त्यम् — a final consonant is इत्), both erased by 1.3.9 . None of these ten letters is ever heard. So why do they exist at all? Because 3.4.78 will delete the whole ल-name regardless — the letters are not there to be pronounced, they are there to be read by other rules before that deletion happens. That is the one idea this page keeps coming back to: an इत् letter is an index, filed for rules downstream to search on, not a sound anyone ever says.

Six vowels would give six टित् names and six ङित् names — twelve — but Pāṇini only used four of the six vowels for the ङित् half. That asymmetry is exactly why there are ten names, not twelve: the table has two empty cells, not because of an error, but because nothing was assigned there.

A classification that changed nothing would be exactly the decoration it looks like. It changes two things, each stated by its own rule, and भू’s own present and imperfect show both:

Every cell that differs, differs for one of exactly two reasons: 3.4.100 shortened the ending, or 6.4.71 — लुङ्लङ्लृङ्क्ष्वडुदात्तः, “in लुङ्, लङ्, लृङ्, [the अङ्ग takes] the अट् आगम, and it is उदात्त” — put an अ on the front. (Before a vowel-initial root the same rule’s neighbour, 6.4.72 , lengthens it to आट् instead — भू begins with a consonant, so this page never shows that half.) The augment carries the accent itself: it is the one loud syllable of every लङ्/लुङ्/लृङ् form, which is also why it survives every subsequent sandhi untouched. लिट् gets neither change — it is not ङित् at all, and its personal endings are not even the ones 3.4.78 lists, as the interaction below shows.

3.2.84 opens an अधिकार — “in the sense of the past” — that governs every rule to 3.2.123. Inside that domain the three past lakāras are not three parallel options; they are one rule and two apavāda narrowing it, in this exact order:

1. 3.2.110 — the general case, no condition attached. Any past, unqualified, is लुङ्.
2. 3.2.111 — अनद्यतने, “not of today”: carve out everything before today and give it लङ् instead.
3. 3.2.115 — परोक्षे, “beyond [the speaker’s] sight”: carve out, from what is left, whatever the speaker did not personally witness, and give that लिट्.

So the three sentences below are not three ways of saying the same thing. Each is the only correct choice for what it describes: