The
anusvāra in Sanskrit is an absolutely fascinating topic, but it is not a straightforward one. It is not straightforward at all.
There was a very brilliant American Sanskrit scholar called William Dwight Whitney whose
Sanskrit Grammar is almost unsurpassed in its combination of thoroughness and concision. It is a superb reference work. It is a little bit out of date, but you can only be so out of date when it comes to a language which is thousands of years old. He is, however, out of date in quite another, more unfortunate way. Whitney did not like later Sanskrit literature, or even the whole Classical language itself, really. He was also quite dismissive of the Indian grammatical tradition at times and extremely blunt in giving his personal opinions. To a slightly less unenlightened modern Western reader it is almost comical that he felt emboldened to write some of the things that he wrote, although at times he has a point. Nevertheless I will refer to his discussion (with due apologies for his attitude) which can be read in its original here:
Sanskrit Grammar (Whitney)/Chapter II - Wikisource, the free online library beginning number 70.
When certain nasal sounds appear before other sounds in Sanskrit, sandhi changes occur. The nasal sounds that concern us here are word-final
m, the penultimate nasal of a root and a nasal increment (you don't need to worry about what that is; in fact forget the last 2, word-final
m will do for the purposes of illustration). If a word that ends in
m is followed by a word that begins with a stop, the final
m is assimilated into the class nasal that corresponds to that stop. Thus, as in
Sanskrit/Hindi: Anusvaara अं:
रामम् करोति -> रामङ्करोति
रामम् च -> रामञ्च
रामम् टीका -> रामण्टीका
Etc.
What if the next letter is not a stop, though? What if it's a semivowel (like य) or a fricative? In this case there is no contact made. The articulators are brought close together, but do not touch. We are told that in this case the nasal element that the final
m is converted into is also without contact. "It is a nasal utterance with unclosed mouth-organs. The question is, now, whether this nasal utterance becomes merely a nasal inflection of the preceding vowel, turning it into a nasal vowel (as in French
on, en, un, etc., by reason of a similar loss of a nasal mute [stop]); or whether it is an element of more individual character, having place between the vowel and the consonant; or, once more, whether it is sometimes the one thing and sometimes the other."
Unfortunately, there is no single answer to this question. The Indian grammatical treatises do not agree with each other.
The
Atharva-Prātiśākhya says that this combination (of word-final
m + semivowel or fricative) always results in a nasal vowel (as in मैं हूँ (both vowels)) unless the next sound is a
l, in which case you get a nasalised
l. So रामम् वदति ->
rāmã vadati (with nasalised
a), but रामम् लोके -> r
āmal̃loke with a nasalised
l, which isn't an easy thing to type!
The other
Prātiśākhyas say that this also happens before
y and
v, but not
r. I.e.
m +
y -> nasalised
ỹ and the same also for
m + v. Before
r and the sibilants (
śa, ṣa, sa) and
ha the
Atharva-Prātiśākhya says you get a nasalised vowel, but the others say you get
anusvāra (literally 'after-tone').
However, there is no general agreement on what exactly the
anusvāra was.
Of the nature of this nasal afterpiece to the vowel no intelligibly clear account is given. It is said (RPr. [Rig-Veda-Prātiśākhya]) to be either vowel or consonant; it is declared (RPr., VPr. [Rig-Veda-Prātiśākhya, Vājasaneyi-Prātiśākhya]) to be made with the nose alone, or (TPr. [Tāittirīya-Prātiśākhya]) to be nasal like the nasal mutes [nasal stops]; it is held by some (RPr. [Rig-Veda-Prātiśākhya]) to be the sonant [voiced] tone of the nasal mutes; in its formation, as in that of vowel and spirant [fricative], there is (RPr. [Rig-Veda-Prātiśākhya]) no contact. As to its quantity, see further on.
I'm sure you will find this document absolutely fascinating
http://www.sanskritweb.net/sansdocs/anusvara.pdf
It focuses on the pronunciation of
anusvāra in the
Tāittirīya school of the
(Black) Yajur-Veda. Among other things it describes the
Śuddhānusvāra 'pure
anusvāra':
So what does 'pure nasal' mean then?
The answer is that though no sound can be strictly called a pure nasal by the above definition, the tradition still calls the anusvāra a 'pure' nasal because it does not involve any 'impurities' which here means articulations that are used in creating other (nasal) sounds such as [ŋ] [ɲ] [ɳ] [n̪] or [m]. This means that there is no stricture in the oral cavity.
This does not mean that there is an open stricture, since that would cause only a nasalized vowel. So what does this mean? The answer must come only from tradition, which tells us to simply close the mouth, without forming any particular stricture. This may seem a ridiculously simple answer, but it is the truth. When the mouth is simply closed, the air does not pass through the oral cavity, although it does, unavoidably, come into contact with the oral cavity, and then passes through the nasal cavity.
It also notices a peculiarity of certain traditions whereby
anusvāra is pronounced
gṃ in certain places.
It really is fascinating. It really isn't straightforward. Pāṇini takes a different approach. He generally prescribes
anusvāra everywhere, but a nasal semivowel is also allowed before a semivowel.
It is evidently a fair question whether this discordance and uncertainty of the Hindu phoneticists is owing to a real difference of utterance in different classes of cases and in different localities, or whether to a different scholastic analysis of what is really everywhere the same utterance. If anusvāra is a nasal element following the vowel, it cannot well be any thing but either a prolongation of the same vowel-sound with nasality added, or a nasalized bit of neutral-vowel sound (in the latter case, however, the altering influence of an i or u-vowel on a following s ought to be prevented, which is not the case: see 183).
If you followed that, frankly, well done!
This isn't adhered to in the modern classroom. Well it wasn't in mine anyway! What we were taught to do is pronounce an
anusvāra as the class nasal before the stops and everywhere else as [ŋ].
