Context: https://twitter.com/svat/status/1142644432174342144
People often say that if you want to learn about some topic, the best way is to teach it to somebody else. Actually, it’s even better if you can teach it to a computer — then you really have to understand it.
— Donald Knuth, paraphrased (from e.g. here or 25:14 to 26:46 here)
This is my attempt to understand a few things, by writing them out in enough detail that it would be hard to fool myself into an illusion of understanding. Please point out mistakes.
On layānvita metres
In his work Chandogati, Sediyapu Krishna Bhat introduced1 the notion of laya, and defined as layānvita metres those metres that are (perceptibly) divisible into “chunks” of equal duration (mātrā length2). As a consequence, verses in them can each be naturally3 recited to a (simple) tāla.
This definition covers not only metres of the mātrābandha4 type (in which not every syllable has fixed weight), for example:
…but also many of those of the vṛtta5 type (in which the weight of every syllable is fixed), for example:
Further, Sediyapu observed that most metres in most languages (including even Sanskrit: all mātrābandha-s and most vṛtta-s listed in large catalogues) are of this layānvita type (often, especially in other languages, an even more restricted type consisting simply of repetition of a fixed pattern: what he calls satāna – like the second example above which is simply a repetition of the pattern LLG), and that Sanskrit is unique in having several popular (though relatively few in number) metres like Śārdūlavikrīḍita, Mandākrāntā, Vasantatilaka etc, that are not of the layānvita type.
On being “incomplete”
Something that perfectly fills a pattern can seem to go nonstop, and with no clear ending.6 This is often avoided with an “incomplete” group at the end (ūna-gaṇa), for example this:
or even:
In these cases, the “missing” syllables at the end actually provide a way for the singer/reciter either to stretch the last (non-missing) syllable or to add a pause (or some combination of both), in order to keep the rhythm while still keeping the end distinctive. Thus they provide some “breathing room” in the verse. Not only do they help the listener perceive the ends of lines, they also give the singer some flexibility or room to improvise, which I imagine they love.
This sort of “incomplete at the end” is very common (perhaps even occurring more often than not) across many languages, and is known as catalexis; such verses as catalectic.
For example, in English (trochaic tetrameter):
Lives of great men all remind us
We can make our lives sublime,—
And, departing, leave behind us
Footprints on the sands of time;—
…
Let us, then, be up and doing,
With a heart for any fate;—
Still achieving, still pursuing,
Learn to labor and to wait.—
(If the bold font doesn’t display properly: the odd lines are GLGLGLGL and even lines are GLGLGLG- though of course I’m using G and L here in the English prosody sense of stressed and unstressed syllables.)
Another example (dactylic tetrameter; compare with the “aigiri” nandini metre):
Just for a handful of silver he left us —
Just for a ribbon to stick in his coat — —
Found the one gift of which fortune bereft us, —
Lost all the others she lets us devote; — —
These “incomplete” chunks are mentioned on pages 302–5 of Sediyapu’s “Kannada Chandassu” book (which, disclaimer, I haven’t yet read except for these pages), below:
The obvious and the subtle
A digression (again, just my incomplete understanding of ideas from Sediyapu).
Something is said to be written in verse rather than prose (padya rather than gadya) if it is chandobaddha, i.e. if its scansion—the sequence of laghu-s and guru-s in it—follows some constraint. Not only that, but this constraint must be something perceptible to the ear (perhaps to a refined ear). In other words, the prosody of the verse must form a gati that is aurally striking. In yet other words, apart from whatever charm comes from the meaning and the choice of words (alliteration etc), the gati itself—shared with every other possible verse (or even nonsensical stream of syllables) in the same metre—is charming in some way.
Now there’s a trade-off here. The more charming/attractive the gati is, the more attention it draws to itself, and away from the specific verse, its meaning, etc. Additionally, this can also lead to monotony after listening to several of these verses in sequence.
A few factors that contribute to the metre (chandogati) being more prominent / more obvious to the ear:
- When the same sequence of laghu-s and guru-s repeats multiple times (e.g. LLG-LLG-LLG…)
- When the word boundaries fall at the same place as the gaṇa-boundaries in the metre
- When the metre is more constrained, i.e. less variation is possible
- Etc.
At the other extreme there is prose.
The way this tradeoff plays out in practice is that when the author wishes to draw more attention to the meaning, it would be wiser to choose a metre where the gati is less prominent. So for example we see narrative poetry and technical works written in metres like Anuṣṭubh and Āryā, which allow a lot of variation and can be listened to for hours without monotony, while for stotra-s and the like we do see very rhythmic metres being used, and they are popular with everyone whether or not they understand the meaning.
(To extend this: in English, rarely is something “non-poetic” written in metrical verse, and even there whatever “serious poetry” there is in metre is for instance more often in iambic pentameter than in iambic tetrameter—and only nursery rhymes tend to have perfect metre. Exceptions, like Vikram Seth’s novel The Golden Gate, are few. So the observation that things are more often written in verse in Sanskrit than in English is not just because Sanskrit has more word possibilities or Sanskrit poets are more skilled, but also because Sanskrit’s well-defined prosody allows a richer set of metres: if the only possibilities in Sanskrit were to write in a small handful of repetitive metres like Sragviṇī, we’d see less use of verse in Sanskrit too! Not all metres are so obvious that even a child can hear them.)
End of digression.
The Āryā family of metres
Āryā and related metres are what I’d call “subtle” — I find myself having barely any ear for them, further handicapped by not having any musical knowledge. :-) So trying to be very slow and detailed here.
(Further, Sediyapu mentions the great diversity of apparent gati in kandapadya verses…)
Metres in the Āryā family are made up of combinations of three kinds of “lines” (“halves”), which (besides other constraints to be discussed shortly) happen to contain, in descending order of length, 32 (=12+20) mātrās, 30 (=12+18) mātrās, or 27 (=12+15) mātras. Let’s call these A, B, C, respectively (suggestion for better names?). Then we have:
- Āryā: BC (i.e. 30=12+18, followed by 27=12+15)
- Gīti: BB
- Upagīti: CC
- Udgīti: CB
- Āryāgīti (kanda): AA
So we might understand the metres if we understand each of these lines individually. Some authors split each of these halves into pādas, with a break after 12 mātrās, but not all do, so I’ve not split further.
Before we try to understand them we must first define them: the more specific constraints on each of these types of lines A, B, C, are (TODO: Check whether this is correct, i.e. whether this matches what is seen from texts):
Type A (32)
This is each line of āryāgīti=skandhaka=kanda. (Note: I’ve extrapolated from definitions I found for āryā, so even less confident of this section.)
The line is divided into groups (gaṇa-s) of 4 mātrās. This is equivalent to saying that each gaṇa must be one of GG, GLL, LGL, LLG, LLLL.
The odd gaṇas (1, 3, 5, 7 of the line) cannot be LGL (ja-gaṇa). So each of them can only be one of GG, GLL, LLG, LLLL. This is equivalent to saying that it must be divisible as 2+2 (G or LL, followed by G or LL).
The sixth gaṇa must be either LGL, or L—LLL with a word-break after the first L, i.e. with a word starting on the second syllable.
If seventh gaṇa is LLLL, then there’s a yati at the end of the sixth gaṇa.
Schema, with L standing for Laghu, G for Guru, and – for yati:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Options | GG | GG | GG | GG | GG | GG | GG | |
| ↓ | GLL | GLL | GLL | GLL | GLL | GLL | GLL | |
| LLG | LLG | LLG | LLG | LLG | LLG | LLG | ||
| LLLL | LLLL | LLLL | LLLL | LLLL | L–LLL | –LLLL | LLLL | |
| LGL | LGL | LGL | LGL |
Simplified, with “2” standing for “either G or LL”, and “4” standing for “any of GG, GLL, LGL, LLG, LLLL”:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| 22 | 4 | 22 | 4 | 22 | LGL | GG | 4 | |
| L–LLL | GLL | |||||||
| LLG | ||||||||
| –LLLL |
Type B (30)
This is the first half of āryā and both halves of gīti. The difference from Type A above is that the last gaṇa has a single syllable (which on account of being at the end of the line, counts as guru, i.e. 2 mātrās):
Following the same notation:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Options | GG | GG | GG | GG | GG | GG | G | |
| ↓ | GLL | GLL | GLL | GLL | GLL | GLL | ||
| LLG | LLG | LLG | LLG | LLG | LLG | |||
| LLLL | LLLL | LLLL | LLLL | LLLL | L–LLL | –LLLL | ||
| LGL | LGL | LGL |
or equivalently:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| 22 | 4 | 22 | 4 | 22 | LGL | GG | G | |
| L–LLL | GLL | |||||||
| LLG | ||||||||
| –LLLL |
Type C (27)
This is the second half of āryā. The difference from Type B above is that the sixth gaṇa has a single laghu, and that if the fifth gaṇa is LLLL, then the fourth must end with a yati, i.e. a word must start with the fifth.
Following the same notation:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Options | GG | GG | GG | GG | GG | L | GG | G |
| ↓ | GLL | GLL | GLL | GLL | GLL | GLL | ||
| LLG | LLG | LLG | LLG | LLG | LLG | |||
| LLLL | LLLL | LLLL | LLLL | –LLLL | –LLLL | |||
| LGL | LGL |
or equivalently:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| 22 | 4 | 22 | 4 | GG | L | GG | G | |
| GLL | L | GLL | ||||||
| LLG | LLG | |||||||
| –LLLL | –LLLL |
Summary
To simplify even further, let ′2 denote “either G or –LL”, and let ″22 denote “either GG or GLL or LLG or –LLLL”. (The ′ or ″ are intended to suggest “something” at the start of the block, that is not quite stress or emphasis but I don’t know what: something capable of something like “descending intonation” but not that.) Then we can write the three types of lines as:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type A | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | 4 |
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| Type C | 22 | 4 | 22 | 4 | ″22 | 1 | ″22 | G |
Understanding
Assuming these definitions are correct, we can turn to attempting understanding them. The only difference between Type B and Type A is the ūna-gaṇa at the end, and this can be understood in terms of each other: for Kannada and kanda poets, for whom a non-stop recitation was strongly preferred, we have Type A a full gaṇa at the end (which was also aided by lack of yati and by the ādi-prāsa), and Type B for everyone else.
So we only have two kinds of lines to understand: (1) Type A/B, and (2) Type C.
Anyway, the question arises: why? Why all these weird-looking constraints, such as on where words can start? Why disallow LGL in the odd gaṇas — or equivalently (starting with say the “bhaja govindam” example as canonical), why allow it in the even gaṇas?
Obviously, being a metre, these constraints must be aesthetic in origin, and reflect something perceptible to the ear, i.e. some actual gati inherent in the metre.
The rest of this webpage is me trying to understand these constraints.
Arranged in roughly the order in which I encountered them:
Ganesh (2013)
Back in 2013, Shatavadhani Ganesh briefly mentioned the following (in this audio recording). If you consider the following āryā verse:
nītijñā niyatijñā
vādajñā api bhavanti vedajñāḥ
brahmajñā api labhyā
svājñāna-jñānino viralāḥ
(with breaks after 12), and match its two halves against Type B and Type C above, the first half is:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| Actual | GG | GLL | GG | GG | GLL | LGL | GG | G |
| Verse | nītij | ñā niya | tijñā | vādaj | ñā api | bhavanti | vedaj | ñāḥ |
and the second half is:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type C | 22 | 4 | 22 | 4 | ″22 | 1 | ″22 | G |
| Actual | GG | GLL | GG | GG | GG | L | GLL | G |
| Verse | brahmaj | ñā api | labhyā | svājñā | na-jñā | ni | no vira | lāḥ |
This he mentioned can be sung in either eka-tāla or rūpaka-tāla. I don’t know much about tālas, but what I gather is that a tāla is a way of measuring time, and:
(caturaśra) eka-tāla means a repeating sequence of 4 time units, of the form
-...where “–” denotes the beat and “.” denotes the other time units (cf. Hindustani tīn tāl which seems to be something like “4 4 (4) 4”),(caturaśra) rūpaka-tāla means a repeating sequence of 6 time units, of the form
-.-...where again “–” and “.” denote the beat and non-beat time instants. (Cf. Hindustani ek tal which seems to be a doubled version?)
And as this is a periodic sequence, we only care about it up to cyclic shifts, i.e. something sung to a tāla may in general start on any element of the sequence (anacrusis/pickup/eduppu) (i.e. the phase offset does not matter).
So, in the recording he recites as follows (again, I don’t have an ear for tāla so this is just my best guess from counting, and I have no idea about the phase offset — actually this is almost certainly wrong because in the audio I hear pauses not accounted for here, so leaving this up here just so that it can be corrected by someone reading this — if I had more time I’d look at the audio waveform to compensate for my ear not detecting time properly):
at 1:30, in rūpaka:
- . / - . . . [2] / [ 4 ] [G] / [ G G ] // nī / tijñā // [LL]/ [ G G ] // niya/ tijñā // [G] / [ G G ] // vā / dajñā // [LL]/ [L G L] // api / bhavanti// [G] / [ G G ] // ve / dajñāḥ // [G] / [ G G ] // brah/ majñā // [LL]/ [ G G ] // api / labhyā // [G] / [ G G ] // svā / jñānaj // [LL]/ [L G -] // ñā / nino - // [LL]/ [ G - ] // vira/ lāḥ - //at 1:46, in eka-tāla:
- . . . // [ 4 ] // [ G G ] // nītij // [G L L] // nāniya // [ G G ] // tijñā // vādaj // ñā api // bhavanti// vedaj // ñāḥ - // brahmaj // ñā api // labhyā // svājñā // najñā // ninovi // ralāḥ //
Note that above, the alliteration naturally supports/encourages the rūpaka / split-by-6s rendering. (Well, assuming the above splitting by 6s is correct, which I doubt, especially towards the end.)
Similarly, from the gīti verse
nindatu vā nandatu vā
mandamanīṣā niśamya kṛtim etām
harṣaṃ vā marṣaṃ vā
sarṣapamātram api naiva vindema
whose halves are:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| Actual | GLL | GG | LLG | GLL | GG | LGL | LLG | G |
| Verse | nindatu | vā nan | datu vā | mandama | nīṣā | niśamya | kṛtim e | tām |
and
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| Actual | GG | GG | GG | GLL | GLL | LGL | GG | G |
| Verse | harṣaṃ | vā mar | ṣaṃ vā | sarṣapa | mātram a | pi naiva | vinde | ma |
one can split by 6s or by 4s:
At 2:22, in rūpaka:
- . / - . . . [2] / [ 4 ] [G] / [L L G] // nin / datu vā // [G] / [L L G] // nan / datu vā // [G] / [L L G] // man / damanī // [G] / [L G L] // ṣā / niśamya // [LL]/ [ G G ] // kṛti/ metām // [G] / [ G G ] // har / ṣaṃ vā // [G] / [ G G ] // mar / ṣaṃ vā // [G] / [L L G] // sar / ṣapamāt // [LL]/ [L G L] // rama/ pi naiva// [G] / [G - ] // vin / de - // [G] / [- - ] // ma / - - //At 2:38 in eka-tāla:
- . . . // [ 4 ] // [G L L] // nindatu // [ G G ] // vā nan // [L L G] // datu vā // [G L L] // mandama // [ G G ] // nīṣā // [L G L] // niśamya // [L L G] // kṛtime // [ G - ] // tām - // [ G G ] // harṣaṃ // [ G G ] // vā mar // [ G G ] // ṣaṃ vā // [G L L] // sarṣapa // [G L L] // mātrama // [L G L] // pi naiva// [ G G ] // vinde // [ G - ] // ma - //
Again, note how the alliteration/rhyme, and also the repetition of the GLLG pattern, naturally supports/encourages the former split-by-6s.
This splitting by 6s actually seems to explain the mid-line yati (why some authors break the ārya half into two pādas, breaking after 12 mātrās).
The confusing thing to me about both these examples is that neither of them makes any use at all of the option to use LGL in the even gaṇas, except in the 6th gaṇa where it’s required. And that in fact seems to be a requirement to be breakable as units of (2+4)s, in addition to groups of 4: for every group of 12 (=lcm(4, 6)) mātras to be breakable as both 4+4+4 and as 2+4+2+4, we basically need the first 8 of them to be 2+2+2+2. For example, for Type B lines, the constraints superimposed:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| 2+4s | 2+2 | 2 + 2 | 4 | 2+2 | 2 + 2 | 4 | 2 + 2 | 2 |
| Result | 22 | 22 | 22 | 22 | 22 | 1′21 | ″22 | G |
So isn’t it “cheating” to pick such verses as examples, instead of ones that use LGL in gaṇas 2 and 4? Or is the statement about only some subset of āryā/gīti verses, those that don’t use LGL in gaṇas 2 and 4? In that case, what does this say about the āryā metre in general, and why all the other constraints (those in the 6th and 7th gaṇas) came to be?
(TODO: Look at the corpus of āryā verses and see how many actually have LGL in gaṇas 2 and 4.)
In the same recording, he also gives another gīti verse:
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| Actual | LLG | LLG | LLLL | LLG | LLLL | LGL | GG | G |
| Verse | kalamañ | jula-vā | ganumita | - galapañ | jara-gata | śukagra | hautkaṇ | ṭhyāt |
as first half, followed by
| Gaṇa | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|
| Type B | 22 | 4 | 22 | 4 | 22 | 1′21 | ″22 | G |
| Actual | GLL | LGL | GG | GLL | GG | LGL | GG | G |
| Verse | amba ra | danāmba | raṃ te | bimbapha | laṃ śam | barāri | ṇā nyas | tam |
This one happens to have LGL in second gaṇa in second half, but wasn’t sung in rūpaka so I’m not sure what conclusion can be drawn.
Chandogati (2018)
(The work is from ~1985; I read it in 2018. 😛)
Sediyapu Krishna Bhat in Chapter 6 of Chandogati has a section on the gati of kanda (type A above). The book is structured like a long thesis and it is hard to jump to a section and make sense of it without having understood everything written before, and my memory and notes are incomplete, but the following is what I understand (based on that section and earlier remarks in the chapter).
Firstly, in mātrābandha-s, the akṣaragati (specific combination of Ls and Gs) tends to recede except that all-Ls or all-Gs or especially ja-gaṇa LGL has a special prominence. My take-away is that LGL has a distinctive “sound” to it. (page 138ff and footnote 52).
Relevant? See his discussion of the 6 gati (sankalita-drutāvartagati), i.e. when a verse is divisible into chunks of 6 mātrās each of which is not divisible as 3+3. This non-division is achieved with either a G after 2 mātras that therefore straddles positions 3 and 4, or with a yati after 2 or 4 mātrās, or with the pattern GLLG. (page 148ff) So basically, of the 13 possibilities, we have:
GGG - not 3+3 GGLL - not 3+3 GLGL - 3+3 GLLG - not 3+3 (subtler reasons) GLLLL - 3+3 LGGL - 3+3 LGLG - 3+3 LGLLL - 3+3 LLGG - not 3+3 LLGLL - not 3+3 LLLGL - 3+3 LLLLG - not 3+3 (even subtler reasons) LLLLLL - 3+3
so made of these:
GGG
GGLL
GLLG
LLGG
LLLLG
LLGLL
plus the others with appropriate yati.
He claims that when a padya (in a layānvita metre, i.e. made of āvartas of equal mātrā duration) is read according to its chandogati, there is a barely perceptible pause between one chunk and the next, also resulting in the beginning of each chunk being more “sphuṭa”. (Elsewhere he is at pains to distinguish this “sphuṭate” from stress or accent or anything like that, but I’m not sure whether this is a distinction without a difference, so let me use “emphasis” for now — though perhaps “distinctness” may be better.)
Anyway, his main points about kandapadya are (last 3 pages of chapter 6):
It is a madhyāvartagati, i.e. made of catur-mātrā-gaṇa-s.
Its gati is very variable: “ī mātrābandhavu guru-laghu-vinyāsa prakāra-bhedagaḷiṃda hāgū yati-niveśana-sthānabhedagaḷiṃdāgi, vidhavidhada gati-vaicitryagaḷannu prakaṭisaballa adbhutavādoṃdu padyabhedavāgide” and “nānāvidha gativaicitryagaḷannu līlājālavāgi hommisaballa śaktiyalli kaṃdapadyakke saridoreyāda bēroṃdu chaṃdobaṃdhavu kannaḍadalli ēvarege kāṇisikoṃḍilla”. (See example c which is the “standard”, and the others give a different gati.)
If one ignores all the words and meanings and reads just the sequence of syllables, then the 4-gati becomes apparent, else the words can give it a different gati.
So, recalling the table:
Gaṇa 1 2 3 4 5 6 7 8 Type A 22 4 22 4 22 1′21 ″22 4
with yati after 3 gaṇas, if we look at the even pādas, i.e. “4 | 22 | 1′21 | ″22|4”, then this can sound like 6 + 3 + 3; thus breaking monotony.
There’s a bit more in Chapter 7, with a useful example of a change made by Kittel that destroyed the metre. Also something that seems to be about yati in āryā etc.
Others
In his page here, Andrew Ollett gives the names
syncopated for LGL or L–LLL, and
unsyncopated for the rest, i.e. GG, GLL, LLG, and the-other-kind-of-LLLL.
The terminology is suggestive (see syncopation), and in fact the page says:
A syncopated gaṇa begins a prosodic foot (i.e., usually a heavy syllable, or the first of two light syllables) on the second mora; an unsycopated gaṇa begins a prosodic foot on the first mora.
i.e. the odd gaṇas (breakable as 2+2) can have their “emphasis”(?) on the first mātrā, while the LGL and L–LLL gaṇas have it on the second.
Not yet read
GSS Murthy’s question on the mailing list: https://groups.google.com/forum/#!msg/bvparishat/Jb1NuvtYq_w/metCjcNTAAAJ;context-place=forum/bvparishat
Another thread with replies by Suhas etc: https://groups.google.com/forum/#!msg/bvparishat/4CMZmWiTYOU/L4axzaZZBwAJ;context-place=forum/bvparishat
Deo’s 2007 paper: http://www.safarmer.com/Indo-Eurasian/skt-meter.pdf
Ollett’s 2012 “The gaṇacchandas in the Indian metrical tradition”
Ollett’s 2012 “Moraic Feet in Prakrit Metrics: A Constraint-Based Approach”: https://docs.google.com/viewer?a=v&pid=forums&srcid=MTQyNTE0MTcxODQxNzMxMDM3MDQBMDc5NTIwMDU3NDgxMTUwMzk5NzEBTGViS2Fyb3NCd0FKATAuMQEBdjI&authuser=0
https://groups.google.com/forum/#!searchin/bvparishat/Arya$20singing%7Csort:date/bvparishat/8GR8eILlFno/3Y4Al5pX0hwJ (the https://archive.org/details/Vishnushatpadi-AryaMetre is good)
Norman: https://www.ancient-buddhist-texts.net/Textual-Studies/Prosody-Articles/Norman-Arya.htm
Footnotes:
- Like several other ideas of genius, this has the property of seeming very obvious in retrospect. Obviously laya is extensively discussed in music; did really no one set down this specific idea in the context of prosody? It appears not — at any rate, not as clearly as Sediyapu has done. [return]
- 3, 4, 5, or some combinations thereof: 3 to 8 mātras in length. [return]
- Obviously, when singing, a verse in any metre (or even prose for that matter!) can be sung to practically any rhythm or tāla. These are not of interest to us: we care about the gati of the metre itself, such that when simply recited clearly, the syllables themselves naturally fit into a rhythm. Related: see footnote 48 that opens Chapter 6. [return]
- Also sometimes called mātrāvṛtta, which Sediyapu argues is not ideal terminology. A subclass of jāti. [return]
- Also sometimes called varṇavṛtta, which Sediyapu argues is less preferable to akṣaravṛtta, which in turn is redundant. [return]
- I’ve recently got to experience this in practice, as my ~2-year-old often goes into an infinite loop singing certain nursery rhymes :-) [return]



