Hindustani phonology

Consonants

Hindustani has a core set of 28 consonants inherited from earlier Indo-Aryan ancestors. Supplementing these are two consonants that are internal developments in specific word-medial contexts, and seven consonants originally found in loan words, whose expression is dependent on factors such as status (class, education, etc.) and cultural register (Hindi vs Urdu).

Most native consonants may occur geminate (doubled in length; exceptions are ). Geminate consonants are always medial and preceded by one of the short vowels (that is, , or ). They all occur monomorphemically except , which occurs only in a few Sanskrit loans where a morpheme boundary could be posited in between, e.g. for niśśīl ('without shame').

For the English speaker, a notable feature of the Hindustani consonants is that there is a four-way distinction of phonation among plosives, rather than the two-way distinction found in English. The phonations are:

1. tenuis, as , which is like ⟨p⟩ in English spin
2. voiced, as , which is like ⟨b⟩ in English bin
3. aspirated, as , which is like ⟨p⟩ in English pin, and
4. murmured, as .

The last is commonly called "voiced aspirate", though Shapiro (2003:260) notes that,

Evidence from experimental phonetics, however, has demonstrated that the two types of sounds involve two distinct types of voicing and release mechanisms. The series of so-called voice aspirates should now properly be considered to involve the voicing mechanism of murmur, in which the air flow passes through an aperture between the arytenoid cartilages, as opposed to passing between the ligamental vocal bands.

The murmured consonants are believed to be a reflex of murmured consonants in Proto-Indo-European, a phonation that is absent in all branches of the Indo-European family except Indo-Aryan and Armenian.

Notes

• ¹Only present in Sanskrit loanwords, predominant in Hindi (and marginally in Urdu as well).
• ²Only present in Arabic and Persian loanwords, predominant in Urdu.
• ³ Post-alveolar is sometimes regarded as a substitution for retroflex in colloquial pronunciation of Hindi. In Urdu (as well as in Hindi), on the other hand, is a phoneme coming from Arabic and Persian loanwords; though this phoneme also does come from Sanskrit tatsamas as well.
• ⁴A non-standard phoneme only found in a few words, or in a few dialects. Otherwise an allophone of before palatals.
• Marginal and non-universal phonemes are in parentheses.
• is lateral for some speakers. ⟨The sound is lateral in Dravidian languages, see the voiced retroflex lateral flap⟩
• is mostly (but not always) tends to get de-aspirated to in colloquial western Hindi, though explicitly retained as a distinct phoneme in eastern Hindi.
• , , and are usually (though not always) perceived post-velar.
• , , , and are mostly replaced by , , , and respectively in Hindi, except in the careful speech of educated speakers. is found in Urdu and is rarer in Hindi, often being replaced with (or further by , or even in rarer cases) in the latter; an example of a word containing this sound is aźdahā (अझ़दहा – اژدہا 'dragon').
• /ŋ/ mostly only occurs in clusters before velars as in aṅkit, but there are also words like tinkā, ākramaṇkārī, mumkin, tanqīd making it phonemic. Sanskritic loans with ṅ occurring elsewhere are made ṅg as in Sanskrit vāṅmaya being pronounced /ʋaːŋ(ɡ)mɛː/.

Stops in final position are not released, although they continue to maintain the four-way phonation distinction in final position. varies freely with , and can also be pronounced . is usually flapped or trilled. In intervocalic position, it may have a single contact and be described as a flap , but it may also be a clear trill, especially in word-initial and syllable-final positions, and geminate is always a trill in Arabic and Persian loanwords, e.g. zarā (ज़रा – ذرا 'little') versus well-trilled zarrā (ज़र्रा – ذرہ 'particle'). The palatal and velar nasals occur only in consonant clusters, where each nasal is followed by a homorganic stop, as an allophone of a nasal vowel followed by a stop, and in Sanskrit loanwords. However /n/ + velar clusters also occur, eg. /ʊn.kaː/ making /ŋ/ phonemic. There are native murmured sonorants, , eg. nanhā, lhēsnā, kulhāṛī, tumhārā (< ślakṣṇa, ślēṣayati, kuṭhāra, yuṣmē-) there are cases where mbh becomes mh as in sam(b)hālna, also from tatsamas e.g. arhat, but these are considered to be consonant clusters with in the analysis adopted by Ohala (1999).

The fricative in Hindustani is typically voiced (as ), especially when surrounded by vowels, but there is no phonemic difference between this voiced fricative and its voiceless counterpart .

Hindustani also has a phonemic difference between the dental plosives and the so-called retroflex plosives. The dental plosives in Hindustani are laminal denti-alveolar (as in Spanish, Portuguese, etc.), and the tongue-tip must be well in contact with the back of the upper front teeth. The retroflex series is not purely retroflex; it actually has an apico-postalveolar (also described as apico-pre-palatal) articulation, and sometimes in words such as ṭūṭā (टूटा – ٹوٹا 'broken') it even becomes alveolar (especially in Urdu).

In many Indo-Aryan languages, the plosives and the flaps are allophones in complementary distribution, with the former occurring in initial, geminate and postnasal positions and the latter occurring in intervocalic and final positions. Standard Hindi also does this, as in nīṛaj (नीड़ज – نیڑج 'bird'), phonemically . However this is moreso morpheme initial than word initial in Standard Hindi, thus some instances of or even seemingly occur in medial positions as well. A classic example is niḍar (निडर – نڈر 'fearless'), from ni- (नि- - ن) 'negative prefix') + ḍar (डर - ڈر) 'fear'). .

Allophony of [v] and [w]

Hindustani does not distinguish between and , specifically Hindi. These are distinct phonemes in English, but conditional allophones of the phoneme in Hindustani (written ⟨व⟩ in Hindi or ⟨و⟩ in Urdu), meaning that contextual rules determine when it is pronounced as and when it is pronounced as . is pronounced in onglide position, i.e. between an onset consonant and a following vowel, as in dvīp (द्वीप دویپ, 'island'), and elsewhere (though it is not universal), as in vrat (व्रत ورت, 'vow'). Native Hindi-Urdu speakers are usually unaware of the allophonic distinctions, though these are apparent to native English speakers (as well as usually to the speakers of the Eastern Indo-Aryan languages, such as Odia, Bengali, Assamese, etc.).

When is preceded by a consonant which itself is preceded by a vowel (i.e. in the environment VC_), the allophony is non-conditional, i.e. the speaker can choose , , or an intermediate sound based on personal habit and preference, and still be perfectly intelligible. This is due to ambiguity in the syllabification of such words. For example, advait (अद्वैत ادویت) which is underlyingly , may be syllabified as , , or . Accordingly, the word can be pronounced equally correctly as , , or .

In final clusters, is when preceded by a more sonorous consonant and otherwise. For example, dvandv (द्वंद्व دوندو, 'pair') is pronounced , but garv (गर्व گرو, 'pride') is usually (though not always) pronounced .

External borrowing

Sanskrit borrowing has reintroduced and into formal Modern Standard Hindi. They occur primarily in Sanskrit loanwords and proper nouns. In casual speech (as well as in Urdu), they are usually replaced with and . does not occur word-initially and has a nasalized flap as a common allophone.

Loanwords from Persian (including some words which Persian itself borrowed from Arabic or Turkic languages), or even directly from Arabic, introduced six consonants, . Being Perso-Arabic in origin, these are seen as a defining feature of Urdu, although these sounds officially exist in Hindi and modified Devanagari characters are available to represent them. Among these, , also found in English, French and Portuguese loanwords, are now considered well-established in Hindi (though not in the rural parts of India, as well as among older generation-speakers); indeed, appears to be encroaching upon and replacing even in native (non-Persian, non-English, non-Portuguese) Hindi words as well as many other Indian languages such as Punjabi, Bengali, Gujarati, Marathi and Malayalam as happened in Greek with phi. This to shift also occasionally occurs in Urdu (especially in India). While [z] is a foreign sound, it is also natively found as an allophone of /s/ beside voiced consonants, eg, rasgullā, nasbandī. Similarly, v can get devoiced before voiceless consonants, eg., (when pronounced as "bēvqūf").

The other three Persian loans, , are still considered to fall under the domain of Urdu, and are also used by some Hindi speakers; however, other Hindi speakers may assimilate these sounds to respectively. The sibilant is found in loanwords from all sources (Arabic, English, Portuguese, Persian, Sanskrit) and is well-established. Some Hindi speakers (especially those from rural areas, as well as among older generation-speakers) pronounce the sounds as , though these same speakers, having a Sanskritic education, do hyperformally uphold and . In contrast, for native speakers of Urdu, the maintenance of is not commensurate with education and sophistication, but is characteristic of all social levels. The sibilant , found in loanwords from Persian, Portuguese, French, and English, is very rare and is considered to fall under the domain of Urdu; although it is officially present in Hindi, many speakers of Hindi assimilate it to , , or even .

Being the main sources from which Hindustani draws its higher, learned terms, English, Sanskrit, Arabic, and to a lesser extent Persian provide loanwords with a rich array of consonant clusters. The introduction of these clusters into the language contravenes a historical tendency within its native core vocabulary to eliminate clusters through processes such as cluster reduction and epenthesis. Schmidt (2003:293) lists distinctively Sanskrit/Hindi biconsonantal clusters , and distinctively Perso-Arabic/Urdu biconsonantal clusters .

Suprasegmental features

Hindustani has a stress accent, but it is not as important as in English. To predict stress placement, the concept of syllable weight is needed:

• A light syllable (one mora) ends in a short vowel : V
• A heavy syllable (two moras) ends in a long vowel or in a short vowel and a consonant: VV, VC
• An extra-heavy syllable (three moras) ends in a long vowel and a consonant, or a short vowel and two consonants: VVC, VCC

Stress is on the heaviest syllable of the word, and in the event of a tie, on the last such syllable. However, the final mora of the word is ignored when making this assignment, so if all syllables are light, the penult is stressed (Hussein 1997) [or, equivalently, the final syllable is stressed either if it is extra-heavy and there is no other extra-heavy syllable in the word or if it is heavy, and there is no other heavy or extra-heavy syllable in the word]. For example, with the ignored mora in parentheses:

Content words in Hindustani normally begin on a low pitch, followed by a rise in pitch. Strictly speaking, Hindustani, like most other Indian languages, is rather a syllable-timed language. The schwa has a strong tendency to vanish into nothing (syncopated) if its syllable is unaccented.