Sanskrit phonology

Classical Sanskrit distinguishes about 36 phonemes. There is, however, some allophony and the writing systems used for Sanskrit generally indicate this, thus distinguishing 48 sounds. The sounds are traditionally listed in the order vowels (Ac), diphthongs (Hal), anusvara and visarga, plosives (Sparśa) and nasals (starting in the back of the mouth and moving forward), and finally the liquids and fricatives, written in IAST as follows:



An alternate traditional ordering is that of the Shiva Sutra of.

Vowels
The vowels of Classical Sanskrit written in Devanagari, as a syllable-initial letter and as a diacritic mark on the consonant,  pronunciation transcribed in IPA, IAST, and approximate equivalent in English:

The long vowels are pronounced twice as long as their short counterparts. Also, there exists a third, extra-long length for most vowels. This lengthening is called pluti; the lengthened vowels, called pluta, are used in various cases, but particularly in the vocative. The pluti is not accepted by all grammarians. The vowels and  continue as allophonic variants of Proto-Indo-Iranian,  and are categorised as diphthongs by Sanskrit grammarians even though they are realised phonetically as simple long vowels.

Additional points:
 * There are some additional signs traditionally listed in tables of the Devanagari script:
 * The diacritic called anusvāra, (IAST: ). It is used both to indicate the nasalisation of the vowel in the syllable  and to represent the sound of a syllabic  or ; e.g..
 * The diacritic called visarga, represents  (IAST: ); e.g.  . While pronounced as a fricative, it's assimilable into a succeeding stop,
 * The diacritic called chandrabindu, not traditionally included in Devanagari charts for Sanskrit, is used interchangeably with the anusvāra to indicate nasalisation of the vowel, primarily in Vedic notation; e.g..
 * If a lone consonant needs to be written without any following vowel, it is given a halanta/virāma diacritic below.
 * The vowel in Sanskrit is realised as being more central and less back than the closest English approximation, which is . But the grammarians have classified it as a back vowel.
 * The ancient Sanskrit grammarians classified the vowel system as velars, retroflexes, palatals and plosives rather than as back, central and front vowels. Hence and  are classified respectively as palato-velar (a+i) and labio-velar (a+u) vowels respectively. But the grammarians have classified them as diphthongs and in prosody, each is given two mātrās. This does not necessarily mean that they are proper diphthongs, but neither excludes the possibility that they could have been proper diphthongs at a very ancient stage. These vowels are pronounced as long  and  respectively by learned Sanskrit Brahmans and priests of today. Other than the "four" diphthongs, Sanskrit usually disallows any other diphthong—vowels in succession, where they occur, are converted to semivowels according to sandhi rules.
 * The pronunciation of vocalic r is as above: much the same as in English "butter" and the Hindi "patr", meaning letter.   This cause a great deal of confusion for those who start from the writing systems of West European and modern Indian languages, which do not recognise the vocalic r as an allophone of the consonant r, and which therefore require that a vowel be included in the syllable - "e" in "butter",  "i" in "bird", etc. -.  Other languages have writing systems that include vocalic r.   In Croatian, for example, "corner" is written "vrj", the "r" doing double duty as a vowel in addition to a consonant. Indian descendant languages of Sanskrit lost the vocalic r, and Indian pandits came to pronounce it as variously "ri" in the North - as in the name of the god known there as "krishn" -, and "ru" in the South - where the same god is commonly referred to as "krushna".  When vocalic r reappeared in modern Indian languages due to the loss of final "a" in Sanskrit loan words such as "patra" > "patr" in Hindi, (which replaced prakrit and apabramsha "patta", the natural derivative of the Sanskrit) it was not recognised by modern speakers that this corresponded exactly to the vocalic r of ancient Sanskrit.   The pronunciation remains a bone of contention even now amongst those who are unfamiliar with the ancient Indian science of phonetics (discussed above) in which the sound system is perfectly described.

Consonants
IAST and Devanagari notations are given, with approximate IPA values in square brackets.

The table below shows the traditional listing of the Sanskrit consonants with the (nearest) equivalents in English (as pronounced in General American and Received Pronunciation or the Indian English pronunciation if specified), French and Spanish. Each consonant shown below is deemed to be followed by the neutral vowel schwa, and is named in the table as such.

Accent
Vedic Sanskrit had pitch accent; some syllables had a high tone, and the following syllable a falling tone, though through ellipsis a falling tone may occur elsewhere.

Phonology and sandhi
The Sanskrit vowels are as discussed in the section above. The long syllabic l is not attested, and is only discussed by grammarians for systematic reasons. Its short counterpart occurs in a single root only,  "to order, array". Long syllabic r is also quite marginal, occurring in the genitive plural of r-stems (e.g. ' "mother" and ' "father" have gen.pl. ' and '). are vocalic allophones of consonantal. There are thus only 5 invariably vocalic phonemes:.

Visarga  is an allophone of  and, and anusvara , Devanagari  of any nasal, both in pausa (i.e., the nasalised vowel). The exact pronunciation of the three sibilants may vary, but they are distinct phonemes. An aspirated voiced sibilant was inherited by Indo-Aryan from Proto-Indo-Iranian but lost shortly before the time of the Rigveda (aspirated fricatives are exceedingly rare in any language). The retroflex consonants are somewhat marginal phonemes, often being conditioned by their phonetic environment; they do not continue a PIE series and are often ascribed by some linguists to the substratal influence of Dravidian or other substrate languages. The nasal is a conditioned allophone of  ( and  are distinct phonemes— 'minute', 'atomic' [nom. sg. neutr. of an adjective] is distinctive from  'after', 'along'; phonologically independent  occurs only marginally, e.g. in  'directed forwards/towards' [nom. sg. masc. of an adjective]). There are thus 31 consonantal or semi-vocalic phonemes, consisting of four/five kinds of stops realised both with or without aspiration and both voiced and voiceless, three nasals, four semi-vowels or liquids, and four fricatives, written in IAST transliteration as follows: or a total of 36 unique Sanskrit phonemes altogether.

The phonological rules which are applied when combining morphemes to a word, and when combining words to a sentence, are collectively called sandhi "composition". Texts are written phonetically, with sandhi applied (except for the so-called ).