Introduction
The Atharvaveda (AV) is undoubtedly one of the most fascinating Vedic texts. On the one hand, it belongs to the oldest layers of the Vedic heritage, being, in its oldest parts, essentially contemporaneous with the younger parts of the earliest Vedic text, the 
gveda (RV). Furthermore, this collection of hymns and spells contains extremely archaic fragments of common Indo-Iranian and Indo-European magic and ritual knowledge as well as the earliest specimens of Old Indian philosophical and cosmogonic thought (see, for instance, Bloomfield Reference Bloomfield1899, which to this day remains a useful survey of the field). On the other hand, since it was canonized much later than the RV and even some younger Vedic texts of the Yajurveda school, the AV reveals numerous traces of relatively recent editorial work. This also explains its lesser degree of preservation as compared with the RV and some other Vedic texts. Accordingly, the study of the textual history of the AV remains one of the most vital issues of Vedic philology. In particular, the relationship between the two recensions of the AV, Śaunakīya and Paippalāda, is one of the most challenging topics for a Vedicist. The Śaunakīya, which is much better studied and has many translations, represents a more recent development within the Atharvavedic tradition, while ample evidence from the more authentic Paippalāda (which is therefore presumably closer to the hypothetical Atharvavedic archetype), originally only scarcely known from the Kashmir manuscript, has only recently become available to Indologists, thanks to the discovery of the much better-preserved Orissa manuscript tradition (see, in particular, Witzel Reference Witzel1985; for a general survey of research in the field, see also introductory chapters in Zehnder Reference Zehnder1999 and Griffiths Reference Griffiths2009).
Furthermore, although it is contemporaneous with the late gveda, the AV was probably not quite identical to the RV as regards its linguistic and dialectal basis: it exhibits several linguistic features quite different from both the language of the RV and the varieties attested in younger (middle Vedic) texts of the Yajurveda.
In this regard, book 19 of the Śaunakīya recension is of particular interest. This part of the Śaunakīya is borrowed almost entirely from the Paippalāda and, in many cases, helps us to trace editorial changes applied to the original Atharvavedic text. It also contains several important hymns testifying to new trends in the Vedic pantheon and religious system at the end of the early Vedic period.
This is also the case for a group of four hymns (AVŚ 19.49–52) dedicated to the goddess of night, Rātrī (only once becoming the object of devotion in the gveda, in RV 10.127).
The present paper offers an analysis of an Atharvanic hymn addressed to the goddess of night, Rātrī. The hymn is found in both recensions of the AV, in the Śaunakīya (AVŚ 19.50), and in the Paippalāda (AVP 14.9). In the Śaunakīya-Saṃhitā it concludes the group of the four Atharvavedic hymns (AVŚ 19.47–50) dedicated to Rātrī.Footnote 1 According to AV-Pariśiṣṭa 4.3–5, hymns 47–8 and 49–50 are thematically grouped into two “sense hymns” (arthasūkta; see Griffiths Reference Griffiths2003: 5 f. and, most recently Rotaru (forthcoming)), applied in the ritual of worshipping night. The metre is anuṣṭubh: that is, stanzas normally consist of four octosyllabic verses or pādas (the number of syllables is indicated in brackets after each pāda).
In what follows I offer a new translation of the hymn on the basis of textual evidence available from both recensions, accompanied by a philological and linguistic commentary. In spite of its relatively short length (seven stanzas), this hymn offers valuable evidence for a study of the peculiarities of the language of the Atharvaveda, which exhibits both archaisms typical of the early Vedic period and some innovations peculiar to the end of this period and marking a transition to the later, Middle Vedic, stage of Old Indo-Aryan.
I will use the standard edition by R. Roth and W. D. Whitney (hereafter abbreviated as RW), the edition by Viśva Bandhu (for the Śaunakīya recension) and the edition by Bhattacharya (for the Paippalāda). Instead of providing a full critical apparatus (which can be found in Pandit (ed.) and Viśva Bandhu (VB) (ed.) for the Śaunakīya recension; while for the Paippalāda recension, Bhattacharya's edition provides the necessary references to the evidence from both the Orissa (Or.) and Kashmirian (Kashm.) manuscripts). I will only note the most important unclear passages and discrepancies between the two recensions.
Alongside the standard English translation by Whitney and Lanman (Reference Whitney, Lanman, Whitney and Whitney1905), there is also a German translation by Ludwig (Reference Ludwig1878: 465–6), an Italian translation by Sani (Orlandi and Sani Reference Orlandi, Sani, Orlandi and Sani1992: 194–5), and a Russian translation by Elizarenkova (Reference Elizarenkova1976: 329–30, comm. on p. 396). Some verses are also translated by Insler (Reference Insler1970: 143–7).
Śaunakīya 19.50 = Paippalāda 14.9: translation and commentary
AVŚ 19.50.1 = AVP 14.9.1 abd: AVŚ 19.47.8abd = AVP 6.20.9abd
+ándhaṃ rātri t
ṣṭádhūmam (+tíṣṭhad-dhūmam ?) (8)
aśīrṣ
ṇam áhiṃ kṇu (8)
+akṣyàu v
kasya nír jahiy (8)
++sténaṃ drupadé jahi (8)
O Night, make the tṣṭadhūma-snake (?) blind, headless [or: make the snake blind, breathless (?), headless]! Strike out the wolf's (robber's) eyes! Strike at the thief [bound] to the post.
a. Śaun. mss. áṃdha, ed. VB ádha, Paipp. andhaṃ
Śaun. mss. t 
ṣṭádhūmam, Paipp. Or. mss. tiṣṭhadhūmam, Kashm. tiṣṭadhūmam
c. Śaun. mss. akṣáu, akṣyáu, Paipp. akṣau, ed. RW +akṣyàu
cd. Śaun. mss. nír jahyās téna táṃ, Paipp. Or. nir jahyās tenaṃ
d. Śaun. drupadé, Paipp. Or. mss. dupade, dupade, Kashm. n pate
ab: t ṣṭadhūma-snake (?) [or: +tíṣṭhad-dhūmam ?] – an unclear word, attested also in AVŚ 19.47.8 = AVP 6.20.9. Śaunakīya reads t ṣṭádhūmam, Paipp. has tiṣṭhadhūmam (tiṣṭha dhūmam?). All interpreters understand this word as the epithet or name of a snake (thus rendered by Ludwig and Whitney; cf. also Griffiths Reference Griffiths2009: 221: “the tiṣṭhadhūma snake”). The variant of this bahuvrīhi compound preserved in the Śaunakīya recension is translated as “mit dem gift[i]gen hauch” (Kuhn Reference Kuhn1864: 132), “von beissendem Hauch” (Zimmer Reference Zimmer1879: 180), “von widerlichem geruche” (Ludwig), “harsh-smoked (?)” (Whitney), “of pungent/poisonous smoke” (Griffiths).
The reading attested in Paipp. mss. (tiṣṭhadhūmam) appears semantically more attractive: this might suggest the emendation +tíṣṭhad-dhūmam. Assuming the meaning ‘breath’ for the word dhūma- ‘smoke, vapour’ (cf. one of the glosses of the indigenous commentary ni[ś]śvāsa-dhūma (viśvāsa-dhūma ?) ‘vapour of exhalation’, which connects this epithet with the stinky breath of a snake), one might tentatively translate the compound in question as ‘(the one, whose) breath stopped’ – i.e.: +tíṣṭhaddhūmam … k ṇu ‘make (the snake) breathless’.
d: thief [bound] at the post – this translation is convincingly substantiated by Griffiths (Reference Griffiths2009: 222), instead of Whitney's rendering ‘cast the thief into the snare’. As Griffiths explains, drupadá- is “not a snare into which a robber can be cast, but a post to which or a fetter in which he can be bound […], and then beaten”.
AVŚ 19.50.2 = AVP 14.9.2
yé te rātriy anaḍvhas (8)
+tīkṣṇáśṅgāḥ suvāśávaḥ (8)
tébhir no adyá pāraya- (8)
-áti durgṇi viśváhā (8)
b. Śaun. mss. t 
kṣṇa°
Those draft-oxen of yours, O night, which are sharp-horned, very swift – with them make us today pass over [regions which are] difficult to traverse, [and please do so] always.
cd. The juxtaposition of adyá ‘today’ and viśváhā ‘always, every day’ is apparently nonsensical. Some translators tried to find a reasonable interpretation by separating the time of causing the event and the time of caused event from each other (cf. for instance, Whitney: “do thou today pass us always […] over difficulties”; Sani: “facci oggi oltrepassare sempre i pericoli”), but this interpretation is hardly possible for a morphological causative (pāraya), which strongly implies the condition of the unity of time of the causing and caused events (see Wierzbicka Reference Wierzbicka1975: 497–9).Footnote 2 Ludwig suggested a different (but hardly more acceptable) solution, rendering viśváhā as ‘everywhere’: “mit denen sollst du uns heute retten über alle gefar und überall”. The problem can probably be solved by analysing the last request of this verse as containing a co-ordinating structure, i.e.: ‘today [and] always’ – as suggested by Elizarenkova (Reference Elizarenkova1976: 329): ‘… With them pass us today, Every day (pass) over dangers!’ [‘Na nix ty segodnja perevezi nas, / Každyj den' (perevozi) čerez opasnosti’].
AVŚ 19.50.3 = AVP 14.9.3
*rtrīṃ-*rātrīm áriṣyantas (8)
tárema tanuv
vayám (8)
gambhīrám áplavā iva (8)
ná tareyur árātayaḥ (8)
May we ourselves (tanv 
) cross night after night unharmed, [while our] enemies may not cross [it], like floats that do not (properly) float [will not cross] a deep [water].
a: *r 
trīṃ-*rātrīm: According to Wackernagel (Reference Wackernagel1930: 185 f.) and Bloomfield and Edgerton (Reference Bloomfield and Edgerton1934: 79 f.), after the gveda, r trī- ‘night’ also occurs with the short vowel stem (r tri-). Note, however, that of eighteen occurrences of the accusative form in the Atharvaveda (AV), consistently transcribed in all editions as r trim, in accordance with mss. readings and the Padapāṭha, the only three occurrences that appear in metrically distinctive contextsFootnote 3 (AVŚ 19.49.5a; 19.50.3a; and 19.55.1a) are attested in book 19. All these occur in contexts where we typically expect long syllables (all – in the fourth syllable, shown in bold type), in accordance with the triṣṭubh schemes ( ∪ – ∪ ─, ∪ ∪ – | – ∪ – ∪ and ∪ – ∪ ─ ∪ , ∪ ∪ | – ∪ – ∪ ) and the two possible schemes for anuṣṭubh stanzas: ∪ – ∪ ─ | ∪ – ∪ ∪ (the standard pattern prevailing in the early gveda) and ∪ – ∪ ─ | ∪ – – ∪ || ∪ – ∪ – | ∪ – ∪ ∪ (the innovative pattern, which becomes common in the late gveda and Atharvaveda; see e.g. Macdonell Reference Macdonell1916: 438 and 439, with n. 5). The relevant passages run as follows: śiv ṃ r trim +ahuvi s 
riyaṃ ca (AVŚ 19.49.5a); r triṃ rātr i m áriṣyantas (AVŚ 19.50.3a); and r triṃ rātr i m áprayātaṃ bháranto (AVŚ 19.55.1a). This means that we have good reason to read *r trīm in these three occurrences (and, accordingly, in all other cases); see also Kulikov Reference Kulikov and Nikolaeva2010: 174, n. 1.
ab: tanv is the instrumental singular form of the emphatic reflexive pronoun, or intensifier, tan -, employed in adnominal usage (see Kulikov Reference Kulikov2007b: 1416 f.) (cf. Russian ‘sam’, German ‘selbst’). The emphatic reflexive usage of this lexeme often poses problems in translation, and we should recall here the main conclusions of a series of studies on reflexive constructions and the expression of reflexivity in Vedic published within the last decade (Kulikov Reference Kulikov, Ofitsch and Zinko2000, Reference Kulikov2007b, Pinault Reference Pinault2001, Hock Reference Hock, Tikkanen and Hettrich2006, and, most recently, with some criticisms contra the first four, Hettrich Reference Hettrich and Fincke2010). tan -, alongside its original meaning (‘body’), has two grammatical functions. First, it can be used as a reflexive pronoun (although, according to Hettrich, not completely grammaticalized), i.e. for the expression of co-reference with the subject (constructions of the type “Peter cured himself” or “Peter heilte sich”: the Agent and the Patient are referentially identical). Second, tan - can be employed as an emphatic reflexive, or intensifier.Footnote 4 One of the main functions of intensifiers is to signal that the referent “is to some degree unexpected in the discourse role or clausal role where it occurs” (Kemmer Reference Kemmer, Stein and Wright1995: 57). This type can be illustrated by such usages as “Peter cured John himself / Peter heilte Hans selbst” (that is, without the help of a professional physician or someone else). Another subtype, called “adnominal”, singles out the referent from a set of items somehow related to it (cf. Mary ~ Mary's children, Mary's brother etc.; London ~ centre of London, London's suburbs), as in “I prefer the surroundings of London to London itself” (note the ungrammaticality of “*I prefer Paris to London itself”; example from König and Gast Reference König and Gast2006: 228 ff.). The two functions, i.e. reflexive proper and emphatic reflexive, or intensifier, can be expressed by two different forms in some languages – for instance, in German (sich vs. selbst) or Russian (sebja vs. sam). By contrast, some other languages syncretically express them by means of the same form – as is the case with English (-self). Vedic exemplifies the latter type of language, using the same word, tan - (and, in the later language, ātmán-), for both functions.
Although the reflexive function of tan - is by no means a novelty, having been noted as early as Grassmann's (Reference Grassmann1873) dictionary and, later, in Wackernagel's grammar (1930)Footnote 5 (though without due attention to emphatic and heavy reflexives of the type ‘sich selbst’), we still, and not infrequently, find inexact or confusing renderings of this word. For instance, RV 10.128.5c = AVŚ 5.3.7c m hāsmahi prajáyā m tan bhir “let us not suffer loss in progeny, not in ourselves” is translated by Geldner (Reference Geldner1951: III, 358) as “[n]icht möchten wir der Kinder noch des eigenen Leibes verlustig gehen”.
Returning to our Atharvavedic passage. tanv must belong with the finite verb (tárema), as Whitney suggested in his comments, rather than with the participle in pāda a (as it is rendered in Whitney's translation: “receiving no harm with ourselves”). The sense must be as follows: “May we ourselves cross [or: as for us, may we cross] …, (while our) enemies may not cross…”.Footnote 6 Both the personal pronoun vayám (which is not necessarily overt in Vedic, typically being dropped) and the intensifier tanv are used to emphasize the opposition “we” ~ “our enemies”.
AVŚ 19.50.4 = AVP 14.9.4
yáthā +śyāmkaḥ (śāmykaḥ?) prapátan (8)
+preyivn nnuvidyáte (8)
ev rātri prá pātaya (8)
yó asm
abhyaghāyáti (8)
As a [tiny] millet-seed (or: particle of śamī-wood?) that is flying away cannot be found (anymore), once it has gone away,– so, O night, make [him] fly away who is trying to harm us.
a. Śaun. mss. śāmy kaḥ, ed. RW +śyām kaḥ, Paipp. śāmyākaḥ
b. Śaun. mss. apav n, ed. RW +aparáṃ, Orissa mss. predivān (ed. Bhattacharya predivān, crit. app. pre(ti → )divān), Kashmir ms. tedivāṃ
a: [tiny] millet-seed (or: particle of śamī-wood (?)) +śyām kaḥ (śāmy kaḥ?) – Ed. Roth/Whitney emends +śyām kaḥ ‘millet-seed’ (against the mss.), which perfectly fits the context. The reading śāmy kaḥ, found in the mss. of both recensions, Śaunakīya and Paippalāda, is, however, not entirely impossible. The word śāmy kaḥ ‘made from śamī-wood’ (Prosopis cineraria), attested in the Kauśika-Sūtra, might refer to a tiny particle of śamī-wood, which is used for producing fire (A. Lubotsky, personal communication).
b: … is gone away … (+preyiv n) – Śaunakīya mss. read an unclear word (apav n), for which ed. Roth/Whitney emends +aparáṃ ‘far, further’ (and this emendation is adopted in Ludwig's translation). In his translation Whitney returns to the original reading, taking apav n as pres. participle of the verb ápa-vā and translating it as ‘blowing away’. Note, however, that the subject of the verb (ápa-)vā ‘blow (away)’, both in transitive and intransitive usages, is typically wind, air flow, etc., not the object being blown or driven (away) by a wind. Accordingly, the participle of the transitive verb ápa-vā ‘blow away; drive away by blowing’ can hardly be an epithet of the subject of this sentence (be it a millet-seed or a particle of śamī-wood). An entirely different reading (which, obviously, cannot be reconciled with the Śaunakīya version) is attested in the Paipp.: Orissa mss. read predivān, while in Kashmir ms. we have tedivāṃ. A possible conjecture might be +preyiv n (= pra-īyiv n) ‘gone away’ (with the d/y confusion, quite common in Orissa mss.), i.e. a perfect active participle of prá-i (Werner Knobl, personal communication).
AVŚ 19.50.5 = AVP 14.9.5
ápa stenáṃ +vāsomatháṃ (*8)
+goajám utá táskaram (8)
átho yó árvataḥ śíro (8)
’abhidhya nínīṣati (8)
[Keep] away the thief who steals clothes(?), and the robber running off cattle (cattle-lifter), but also [the one] who, covering [their = coursers’] head[s], tries to lead [away] coursers.
a: Śaun. mss. v so, Paipp. Or. vāsomathaṃ, vāsamathaṃ Kashm.
b: Śaun. mss. gór ájam, Pp. mss. go'ajám, gorájam, Paipp. Or. goajam, Kashm. gotham (ed. Barret gomatham)
a: … who steals clothes (?) (+vāsomatháṃ) – Pāda a is lacking two syllables in Śaun. mss. The heavy emendation suggested in ed. Roth/Whitney, ápa… +avāsayas (in Whitney's translation: “Thou didst make [the thief] stay away”) is unlikely. More plausible is the emendation suggested by Insler (Reference Insler1970: 147 f.) on the basis of the reading attested in the Paipp. (rejected by Whitney), +vāsomatháṃ: ‘clothing-stealing (thief)’ (with the final accentuation, in accordance with the agentive meaning of this compound); a similar reading (vāsamathaṃ) was noticed for Paipp. already by Barret (Reference Barret1927: 248).Footnote 7 A parallel emendation in pāda b, +gomátham (‘cow-stealing’), suggested by Insler on the basis of the Kashm. ms. of the Paipp. and in analogy with +vāsomáthaṃ, is unnecessary. The reading gór ajám (ajá- is a derivative of the root aj ‘drive’) is quite possible in the context; and the compound go-ajam (the emendation suggested already in Whitney/Lanman's commentary ad loc.) attested in the Orissa mss. of the Paipp. is even more appropriate in this passage (W. Knobl, personal communication), being perfectly parallel with the compound vāsomathá- and indirectly corroborated by the compound go-ájana- attested in the gveda.
cd: …covering the head of a courser … (árvataḥ śíro ' ’abhidh ya) – i.e. slipping on a halter on the muzzle of a horse; thus in accordance with Whitney's interpretation, against that of Insler (Reference Insler1970: 147–8), who believed that this phrase must refer to a thief hiding his face (i.e.: ‘covering [his] head’), rather than to covering the head of a horse with a halter.
AVŚ 19.50.6 = AVP 14.9.6
yád +adyá (ady?) rātri subhage (8)
vibhájantiy áyo vásu (8)
+tád etád asmn +bhojaya (8)
+áthéd anyn +upyasi (8)
When, you, O fortunate night, will be sharing (or: will come sharing?) out wealth today, then make this [wealth] benefit us, and only then you will go to the others!
a. Śaun. mss. ady , Pp. adyá, Paipp. adya
c. Śaun. mss. yád etád asm n bhójaya [ed. RW +tád etád …]; Paipp. yadehy asmān bhrājaya-
d. Śaun. mss. yáthedáṃ n nup yási or n nup yáti [ed. Vishva Bandhu yáthéd any n up yasi]; Paipp. -athéd anyān upāyasi
This stanza is discussed at length by Zehnder (Reference Zehnder2011: 59 f.), who offers the following translation: “Wenn Du heute, glückbringende Nacht, Gut verteilen gehst, so (*tád) lass dieses (Gut) uns Nutzen bringen; und dann erst (*áthéd) magst du zu anderen hingehen”. I essentially follow Zehnder's interpretation.
a: ady – Whitney explains the final length as sandhi with the preverb (i.e. adyá- … áyo ‘you will come’), contra the Padapāṭha (Pp.) and ed. Pandit, which saw here a common metrical lengthening that is not very common in this metrical position, however (W. Knobl, personal communication). Accordingly, there are good reasons to adopt the reading attested in the Paippalāda (adya). áyo, taken by all translators as the subjunctive form of the verb i ‘go’ (or + i ‘come’), could be interpreted as a (semi-)auxiliary verb connected with the participle vibhájantī, to form a particular periphrastic formation that can be tentatively rendered as future continuous. Such periphrastic formations are quite common in the language of the AV.
c: Roth and Whitney's conjecture +tád (for yád, which must also be a secondary replacement, obviously responsible for the accent on bhójaya) is very likely. The Paipp. version must be corrupt, also in Orissa mss. The causative bhrājaya- hardly makes any sense in the context (‘make shine’?), while bhojaya- is attested from the AV onwards and perfectly fits the context.
d: and only then you will go to the others – A difficult pāda; the emendation of ed. Roth and Whitney: yáthedáṃ +n p yati (‘that it [i.e. goods, riches] go not away’) is unlikely. The Paippalāda version seems to preserve the original reading. The negative particle ná must result from misinterpretation of +áthéd +any n … (→ yáthédaṃ ná…). In contrast to pāda a, the subjunctive ayasi should be taken as a full verb (‘you will go (to the others)’), rather than as an auxiliary (‘you will be [making this wealth benefit (?)] (the others)’), since in that case the preverb úpa will be unexplained.
AVŚ 19.50.7 = AVP 14.9.7
uṣáse naḥ pári dehi (8)
sárvān rātriy anāgásaḥ (8)
uṣ no áhne bhajād (8)
áhas túbhyaṃ vibhāvari (8)
O Night, commit us undamaged, free from sins, to the Dawn; may the Dawn deliver us to the Day, the Day (back) to you, O multiply shining one.
a: Śaun. mss. pári dehi, Paipp. pari dhehi
b: sárva- is rendered in accordance with the interpretation of this word as ‘whole, undamaged’ (cf. Lat. salvus; see Gonda Reference Gonda1955), rather than as ‘all’ (Whitney: ‘commit us all’, etc.).
d: The epithet vibh van- (vocative singular feminine form vibhāvari) ‘multiply shining, wide-shining’ (derived from the verb ví-bhā ‘shine out, shine widely, multiply shine’, where the preverb ví should probably be understood in the distributive meaning; see, for instance, Kulikov Reference Kulikov and Nedjalkov2007a: 723) is common in the context of hymns to night (cf. AVŚ 19.49.6ab = AVP 14.8.6ab stómasya no vibhāvari ' r tri r jeva joṣase ‘you will enjoy our prayer, O multiply-shining Night, like a king’, AVŚ 19.49.4d = AVP 14.8.4d purú rūp ṇi k ṇuṣe vibhāt ‘You make for yourself [i.e. you take] many forms, multiply shining’; see Kulikov Reference Kulikov2009: 10 ff.) and may refer to the starry night-sky: multiply shining = shining with many stars.
