¹ Visarga is apparently a short form for visarjanīya. It does not figure in the Prātiśīkhyas (cf. Wackernagel, Aind. Gr. 259, for the history of the use of the term). It means ‘ender, terminator, terminal sound’ (< sṛj ‘let go‘). There is also a visarga of r and one of n (cf. AVPr. 2.28).

² Bollensen, ZDMG 22.632, 47.583; Benfey, Gr. 6, ZDMG 11.348; GGA 1711 (1859).

³ The following editions of the Prātiśākhyas were used: Vājasaneyi Pr. (ed. Weber), Ind. Stud. 4.65–171, 177–331; Atharva Veda Pr. (ed. Whitney), JAOS 7 (1862); Tāittirīya Pr. (ed. Whitney), JAOS 9 (1871); Rig Veda Pr. (ed. M. D. Shastri), Lahore, 1937; Rig Veda Pr. (ed. Max Müller), Leipzig, 1869; Rig Veda Pr. (ed. Regnier), Journal Asi astique, V* série, tomes 7–12 (Paris, 1856–8).

⁴ However, the natives are not in complete agreement as to which sounds constitute the ūṣman category. The number is four in the VPr. (8.22), six in the TPr. (1.9), seven in the AVPr. (1.31: not specifically mentioned but implied throughout the text and commentary), eight in the RVPr. (1.10). Visarga is not included in the ūṣman category in the TPr. (1.9) although it tells us (1.12) that ūṣman and visarga are voiceless (aghoṣa). The Pāṇinīya Śikṣā 15 (Yaj. recension, ed. Weber, Indische Studien 4.345–60) includes visarga in the ūṣman category. The articulation of the spirants is described in detail at RVPr. 13.3, RVVyāk. intro. 2, Pāṇinīya Śkṣā 29 (Yaj. rec.), AVPr. 1.30–1, VPr. 1.72, TPr. 3.44–5. The commentator to RVPr. 1.11 glosses ūṣman by vāyoḥ ‘wind’.

⁵ Cf. fn. 4 for the Prātiśākhya references to these spirants.

⁶ Whitney (note to AVPr. 2.40), ‘the division of this indistinct and indefinite sound [i.e. visarga] into three kinds of indefiniteness savors strongly of over refinement of analysis.’ Whitney, Skt. Gr. §69, ‘It may be fairly questioned, perhaps, whether these two sounds [i.e. and ḫ) are not pure grammatical abstractions devised (like the long ḷ-vowel) in order to round out the alphabet to greater symmetry.’ Beames, Comp. Gr. of the Mod. Aryan Lang. of India 1, §70 (London, 1872), ‘It was at best a mere grammarian's fiction to be classed with ṛ and ḷṛ and such like inventions.’ Benfey, Gr. 6, ‘Die Hauchzeichen sind kaum, vielleicht gar nicht hörbar gewesen.’ More favorably disposed are the following: Ebel, KZ 13.277, ‘Der Visarga ist jedenfalls etwas mehr als ein blosser Hauch. . . . Es fragt sich ob Gutturale und Labiale gar keine harte Spirans aufzuweisen haben. Ich glaube ja, und zwar beide im Visarga, der vor ihnen die Stelle der Zischlaute vertritt.’ Kirste, Ausprache des Visarga, Sitzungsber. d. phil.-hist. Klasse d. Kais. Akad d. Wiss. 121.11.18, has his own theories. Wackernagel, Aind. Gr. 1, §225, in general accepts the sounds. S. K. Chatterji, The Origin and Development of the Bengali Language 242–4, likewise accepts them as phonetic entities.

⁷ By ‘takes place’ is meant that the Sanskrit word stock, as collected in the dictionaries from MSS and inscriptions, exhibits both . These two forms have, according to their usage, two different functions. This in itself is no guarantee, however, that these two forms were phonetically distinct.

⁸ Phonetic existence is made probable on the basis of the many references to it as an actual sound.

⁹ The passage reads na plākṣi plākṣāyaṇoḥ.

¹⁰ Note that in every case a spirant of some sort is retained before an initial simple consonant. The pattern suggests school, class, or regional variants.

¹¹ Kaṇṭhya ‘throat’ is used to describe the k-series also (cf. Pāṇ. 1.1.9). Physiologically, it refers to the entire area from the velum to the bronchial tubes (Jolly, Medicin, in Grundr. d. indo-ar. Philologie u. Altertumskunde 3.10 s.v.). It is not necessary to interpret kaṇṭhya as Techmer's ‘strictura spiritus asperi’, as did Kirste (Wiener Sitzungsber. 121.11.7), for this is simply a spirant with closed glottis and arytenoid friction (Arabic ḥa). This last may be ruled out since a close spirant (i.e. one with friction) cannot stand in final position.

¹² Visarga is also described as abhiniṣṭāna (Aśv. Gṛ. 1.15.4; Pār. Gṛ. 1.17.2; Gobh. Gṛ. 2.8.4; Āpas. Gṛ. 15.9; Hir. Gṛ. 2.4.10; AVPr. 1.42). The form abhiniṣṭhāna is also found frequently (cf. PW s.v.). Whitney (note on AVPr. 1.42) and Kirste (Wiener Sitzungsber. 121. 11.16 and MSL 5.92) prefer the form in -ṣṭhāna (< sthā-) and interpret it as ‘Anlehnung, Hinzusetzung, Anhängsel’. PW takes it as ‘ein verklingender Laut’ (< abhi + niḥ + stan, Pāṇ. 8.3.86). And Wackernagel (Aind. Gr. 1.259) takes it as ‘Kehlkopfgeräusch’. A scholiast's note to Pāṇ. 8.3.86 states: nāma cāsmāi dadyurghoṣavadādyantarantastham abhiniṣṭānāntam ‘they give the name abhiniṣṭāna to this when a voiced sound comes to stand between the beginning (of one word) and the end (of another)’. Literally, abhiniṣṭāna means ‘sounding out (from one word) towards (another)’. It is, then, a technical term referring to the close phonetic link between a terminal sound and a following initial. The scholiast's use of ghoṣavat ‘voiced’ is somewhat puzzling. He undoubtedly intends a special class of voiced sounds here since visarga is clearly voiceless. With the interpretation of abhiniṣṭāna as a term referring to final-initial relations, the AVPr. expression visarjanīo ‘bhiniṣṭāna becomes clear: ‘visarga can also be considered as a sound which occurs at the end of a word when it (the word) is immediately followed by another word’.

¹³ These transcriptions are simply working approximations based on the evidence at our disposal. They are not intended as absolutely conclusive.

¹⁴ Kirste (Wiener Sitzungsber. 121.11.7) believed many points spoke against Hoffory's suggestion (KZ 23.557) that ḥ was a whispered vowel. Yet he himself suggested (among others) a pronunciation of ḥ ‘als h mit der Mundstellung des a, z.B. ah^a, ih^a, uh^a. Still another suggestion was ‘als h mit dem sthāna des vorhergehenden Vocals, z.B. ah^a, ihⁱ, uh^u’. New Bengali voiceless h [h], according to S. K. Chatterji, The Origin and Development of the Bengali Language 558 (Calcutta, 1926), ‘is optionally changed to the voiceless velar, palatal or bilabial spirant according to the nature of the preceding vowel, e.g. [αh - αx], [ih: iç], [eh: eç], [oh: of], [uh: uf]’.

¹⁵ Passage cited by Kirste.

¹⁶ A long list of these has been collected by Bollensen (ZDMG 22.632) with the comment that ‘wir nicht zweifeln können, in der lebendigen Sprache sei wirklich s vor den genannten Consonanten gesprochen worden’.

¹⁷ After terminal s > ḥ the new allophones of ḥ were transferred from pausa position to preconsonantal position.

¹⁸ For their occurrence in inscriptions cf. Fleet, Corpus Inscr. Ind. 3, Ind. Ant. 5.699; Kielhorn, Epigr. Ind. 3.307, Ind. Ant. 15.34. The haphazard use of these symbols and the lack of regularity in the conversion of visarga to a sibilant (symbol) before a sibilant (cf. Whitney on AVPr. 2.40) may be explained by the treatment of visarga as a subclass of the s phoneme.