Linguistic evidence for Kuṣāṇa trade routes Bactrian *λιρτο ‘load, cargo’ and Sanskrit lardayati ‘to load’

: Late Sanskrit lardayati ‘to load’ is probably not inherited from a PIE root * lerd- , as has recently been argued by Kaczyńska (2020), but can be explained as a denominative of * larda - ‘load, cargo’. This noun * larda - could be a borrowing from Bactrian *λιρτο /lirtə/ ‘load, cargo’ < Old Iranian * dr̥šta -. This etymology fits well with the fact that lardayati is phrased together with sthora - ‘pack-animal’, likely another instantiation of the Iranian collocation of * staura - ‘animal’ and *√ darz ‘to load’, for which I discuss evidence from Niya Prakrit, Parthian and Khotanese. In addition, further support is drawn from the independent historical evidence for the domination of the main trade routes of Central and South Asia by the Kuṣāṇa dynasty in the first centuries of our era.

In this short paper, I will critically evaluate an Indo-European etymology of lardayati which has recently been proposed by Kaczyńska (2020) (see §2), and I will, as an alternative, argue in favour of a Bactrian etymology (see § §3-4). Before doing so, I first want to make a few notes on the actual attestations of lardayati and its connection to the word sthora-'pack-animal'.
The relevant passages from the Divyāvadāna are cited below (exx. 1-3), accompanied by my own translations. 1 (1) yāvat paśyati sthorāṃl lardayantaṃ sārtham so 'pi sthorāṃl lardayitum ārabdhaḥ When he saw that the caravan was loading the pack-animals, he started to load his pack-animals as well.
(2) sa sārthaḥ sarātrim eva sthorāṃl lardayitvā saṃprasthitaḥ The very night they had loaded the pack-animals, the caravan departed.
In contrast to the general practice in the recent English translations of (parts of) the Divyāvadāna by Tatelman (2005) and Rotman (2008Rotman ( -2017, I have translated lardayati as 'to load' and not as 'to unload', because the translation 'to unload' is not consistent with the meaning of the Middle Indo-Aryan and New Indo-Aryan forms and because it lacks good support from the context. Moreover, both Tatelman and Rotman follow the glossary of Cowell & Neil (1886: 695) and translate sthora-, the default object of lardayati (cf. infra), with 'cargo', but this word more likely means 'pack-animal', as I will discuss now.
No further instances of sthora-are attested in the Divyāvadāna (Edgerton 1953: 2, 611), but some related forms are known from Sanskrit lexica. An entry sthaurī is glossed as pr̥ ṣṭyaḥ 'pack-animal' in two of the most famous dictionaries of Sanskrit synonyms, i.e. in the Amarakośa (2.8.2.14 ed. Colebrooke) and in Hemacandra's 1 Here and below, Cowell & Neil's orthography has been adapted to modern standards and I have adopted Cowell & Neil's (1886: 703) minor emendation sthorāṃl instead of the sthorāṃ of the manuscripts, because I agree with them that this word is acc.m.pl. rather than acc.f.sg. (see below) and because the language of the Divyāvadāna is, apart from its vocabulary, close to standard Classical Sanskrit. The acceptance or non-acceptance of the emendation, however, does not affect my argument in any way. Abhidhānacintāmaṇi (verse 1263). Later lexica only add variants such as sthorī and sthūrī to this. In addition, Wilson (1819Wilson ( : 1027 quotes an entry sthaura-'a sufficient load for a horse or an ass' from a late lexicon, the Śabdārthakalpataru. 2 Still more important for our purposes is the testimony provided by Niya Prakrit stora-'large animal ' and Khowar istṑr 'horse' and istōri 'mounted, horseman' (cf. Turner 1966: 796). For a discussion of the Khowar words, the reader is referred to Morgenstierne (1936: 659), but the Niya Prakrit evidence deserves to be examined here in somewhat more detail. The first person to relate Niya Prakrit stora-to the sthora-from the Divyāvadāna was Burrow (1934: 514), who at the same time suggested that these words are borrowed from a Middle Iranian reflex of *staura-'large animal', which because of geographical and historical reasons one may wish to identify as Bactrian (α)στωρο /(ə)stōrə/ 'idem' (Sims-Williams 2007: 266). 3 Niya Prakrit stora-can refer specifically to 1) a 'horse', so for instance in CKD: 13, where the plural form storaṃca is paired together with vaḍ ̱ avi 'mares' and taken up in the following sentence by a[śpa] 'horse', or 2) a 'camel', as in CKD: 367, where uṭa 'camel' is taken up in the sentence after it by stora. Yet a more general meaning 'animal' should also be posited, e.g. in CKD: 435 manuśa atha vā stora 'a man or a beast'. These animals were used, for instance, for military aims (cf. e.g. CKD: 292 seni storasa 'cavalry'), but more relevant to our concerns are references to their use as pack-animals. I cite two examples below (exx. 4, 5)  Both of these examples bear witness to a collocation stora-+ √darṣ 'to load' < Iranian *√darz-(cf. infra), which finds a close parallel in a passage in Manichaean Parthian, where 'stwr'n 'pack-animals' and drznd/'bdrzynd 'they load'/'they unload' are used together, cf. ex. 6.
(6) cw'γwn kd wd'nm'n'n ky 'd wxybyh wd [']n 'stwr'n 'wt gr'mg 6 'c wy'g 'w wy'g drznd 'wt 'bdrzynd 7 …, just like nomads who, (moving) with their own tents, pack-animals and wealth from place to place load and unload (their stuff). (tr. mine). 8 ed. Andreas & Henning 1934: 850 From Khotanese, one can furthermore cite the phrase drąysi-barā stūra (nom. pl.) 'animals who carry a load' (Pelliot chinois 5538a; ed. KT II 127.34) and one may also note that example (4) combines Niya Prakrit stora-with nadha-< Skt. naddha-'bound', the word that is usually used in the Niya documents to refer to a burden bound around a transport animal and which can alternatively be denoted as darṣa-'load, cargo'. It is because of the etymological identity between Niya Prakrit stora-'large animal' and Sanskrit sthora-and in view of the collocation *staura-'animal' + *√darz 'to load' that I agree with Burrow that Sanskrit sthora-+ √lard should be translated as 'to load the pack-animals'. The fact that sthora-is likely borrowed from Bactrian will also become an important argument in section 3, where I will argue that √lard is borrowed from a Bactrian cognate of the same Iranian root *√darz-'to load' that underlies Niya Prakrit √darṣ. Before coming back to this, I will first evaluate Kaczyńska's (2020) recent Indo-European etymology of lardayati on its own terms in section 2.
Kaczyńska's etymology would add an additional example to the list of Sanskrit words where a PIE *l has possibly been preserved unchanged and could thus be quite important, but how likely is Kaczyńska's proposal, in fact? First of all, one does not really expect the preservation of Indo-European material in a late source 9 The non-exhaustive presentation of the evidence which follows is mainly based on a comparison between Pokorny (IEW: 679), Beekes (EDG: 871f.), and Kaczyńska (2020: 418) and is not identical in every detail to Kaczyńska's overview. For instance, I use *-sk-and not *-sḱ-in my reconstructions, following Lubotsky (2001) in this respect. 10 I leave Latin luscus 'blind in one eye, one-eyed' aside, because, as Kaczyńska rightly notes, it is not generally accepted that this form also belongs here. 11 Because of their feminine gender, Kashmiri lad 'heaped-up load' and Hindi lād 'load, burden' are derived by Kaczyńska from *lardā-and so from a PIE nomen rei actae *lord-éh₂ (the notation with a laryngeal is mine). As the meaning of these nouns is so close to their masculine counterparts in other New Indo-Aryan languages, it seems better to me to assume that the feminine gender is secondary and to simply put them together with the other words from *larda-. In this, I follow Turner (1966: 636). Masculine forms derived from *larda-can also have the meaning of a nomen rei actae; cf. e.g. Punjabi ladd 'load'. like the Divyāvadāna and, precisely for this reason, Mayrhofer's etymological dictionary makes a clear break between material from Vedic and Classical Sanskrit, as loanwords are much more frequent than inherited material in the latter, with some notable exceptions like Classical Sanskrit parut 'last year'~Greek πέρυσι 'idem' < PIE *perut(i). 12 Kaczyńska furthermore assumes that lardayati would be the only vestige of the original verbal use of the root *lerd-, but as τόμος-type nouns (for which see now Nussbaum 2017) are generally derived from verbal bases, this is not too much of a problem and it would also be possible to explain lardayati as a denominative formation from *larda-, 13 which would mean that only *lórd-o-'loading' would have to be reconstructed for PIE.
The main argument against Kaczyńska's etymology is rather of a semantic nature. If one compares the meanings of the outcomes of PIE *lerd in the various branches, it is clear that they have two things in common. On the one hand, these cognates seem to point to an underlying meaning 'to curve'. On the other hand, words referring to various bodily defects are particularly common among them. Armenian lorcʿ-kʿ (pl.), for instance, translates Ancient Greek ὀπισθότονοι in the Armenian translation of Plato, Timaeus 84e (cf. Lidén 1906: 46f.), which in this case refers to a disease that is also called ὀπισθοτονία and which is "a disease in which the body is drawn back and stiffens, tetanic recurvation" (LSJ⁹: s.v. emphasis mine). 14 Likewise, Greek λορδός and the derived noun λόρδωσις refer to "a curvature of the spine which is convex in front" (LSJ⁹: s.v.), the antonym of which is κυφός/κύφωσις, referring to hunchbacks. These types of curvatures are explicitly described as bodily defects in the Greek medical tradition, given that the Graeco-Roman physician Galen (Commentary on Hippocrates' de articulis 18a.493.17; 18a.553.5) defines λόρδωσις as "διαστροφὴ τῆς ῥάχεως εἰς τὸ πρόσω", i.e. 'a forward distortion of the spine'. 15 In the same vein, the Germanic and Celtic evidence generally has to do with the notion 'crooked' and contains forms specifically referring to distorted feet. Semantically, Welsh lurc m. 'crooked foot', 12 The above should not be taken to imply that later strata of Indo-Iranian can simply be neglected for the reconstruction of Proto-Indo-Iranian (cf. Kümmel 2017), and an up-to-date presentation of what Middle and New Indo-Aryan has to contribute to Indo-Iranian and Indo-European linguistics is a desideratum. 13 I thank the anonymous reviewer for this suggestion and cf. other denominative forms such as Greek λορδόομαι. My own proposal, to be discussed in section 3, also starts from the idea that lardayati is denominative to *larda-. 14 The aspect of curvature is explicitly mentioned in the relevant passage from Plato (εἰς τὸ ἐξόπισθεν κατατείνῃ 'it strains backwards'). 15 Similarly, Hippocrates (De fracturis 16.10) pairs the pass. inf. perf. διεστράφθαι 'to be distorted' with λορδόν, a passage that is also cited by Galen (Commentary on Hippocrates' de fracturis 18b.498), who has a lectio facilior pas. inf. pres. διαστρέφεσθαι.
Gaelic loirc f. 'deformed foot' and Old High German lerz fuoz 'clubfoot', a gloss of Latin scaurus 'idem' (Köbler 1993: 717), form a particularly close match and the Germanic denominative verb *lurtjan 'to deceive' (cf. e.g. Old English belyrtan and Middle High German lürzen) fits a basic meaning 'crooked' as well, in that 'crooked' can be a synonym of 'deceitful' in English. Sanskrit lardayati, on the other hand, has no direct relation to either a meaning 'to curve' or to the category of 'bodily defects', as a result of which Kaczyńska's suggestion to reconstruct an earlier meaning 'to bend backwards due to a heavy burden' is but an ad hoc postulate.

A new explanation of lardayati
In view of the problems with Kaczyńska's etymology, I want to propose in this section a different and, in my opinion, more straightforward etymology of Sanskrit lardayati and its derivatives. lardayati looks like a productive denominative tenth-class verb based on an underlying noun *larda-'load, cargo', which, even though not attested as such in Sanskrit, can be safely reconstructed on the basis of various New Indo-Aryan cognates such as Punjabi ladd 'load' or Assamese lād 'an elephant's load' (cf. Turner 1966: 636). As noted above, the late attestation of lardayati/*larda-speaks against them being directly inherited from PIE, because of which we may rather be dealing with a loanword.
Speaking more generally, several parallels can be adduced for borrowings in this semantic field. First of all, Khowar drazēik 'to load up' is, just like Niya Prakrit √darṣ, borrowed from a derivative of the Iranian root *√darz (Morgenstierne 1936: 667). In the same vein, Tocharian B perpente* 'burden, load' has been convincingly compared to Sogdian prβnty 'burden' < *paribandaka-and is thus also Iranian in origin, whatever its exact source (cf. Adams 2013: 426f.). From Armenian, one can cite beṙnawor 'burdened', which is calqued on Parthian (or, less likely, Middle Persian) b'rwr 'loadened, burdened' < OIr. *bāra-bara-'he who carries a load' (cf. Olsen 1999: 364), which is also indirectly preserved by way of the Elamite title bara-bar-ráš (cf. Tavernier 2007: 417). In addition, Armenian grast 'beast of burden' has been identified as a loanword from Parthian grst*, indirectly attested in grstpty < *grasta-pati-, an official in charge of transporting provisions and supplies (Olsen 1999: 873f.), while, somewhat more speculatively, Armenian patat 'camel's load' has been argued to come from an Iranian *patāta- (Olsen 1999: 901). Outside the Iranian domain and closer to our times, another example is furnished by English 'cargo', which was borrowed in early modern times from Spanish cargo 'load, burden'.
In view of these considerations, it seems plausible that Sanskrit *larda-'load, cargo' is borrowed from a so-far unattested Bactrian *λιρτο /lirtə/. This *λιρτο would be the regular outcome of the past participle *dr̥ šta-'that which is loaded > load, cargo' from this same Iranian root *√darz 'to load'. *d > l is, of course, a typical feature of Bactrian (e.g. λαδο /lādə/ 'law' < *dāta-), and for the second part of the word, one can compare γιρτο /γirtə/, past stem of 'to complain' < *gr̥ šta-(√garz) and υιρτο /hirtə/, past stem of 'to set free, to permit etc.' < *hr̥ šta-(√harz) (cf. Sims-Williams 2007: 207, 272f.). Once borrowed as *larda-, a verb lardayati can easily be made within Indo-Aryan (cf. supra). 16 Two notes on the phonetic correspondences are still worth making. First, it may initially seem unexpected that Bactrian ι would have been borrowed with Sanskrit a. This is less of a problem, however, when we take the ancient descriptions of the phonetic character of a in Sanskrit into account. Pāṇini and other sources make it clear that a was a more closed vowel (saṃvr̥ ta-) than ā, which they describe as more open (vivr̥ ta-), from which we may infer that a stood for some kind of schwa (cf. Allen 1953: 57-61 and especially 58 fn. 4). Bactrian ι may, at least in some phonetic environments, represent a central vowel as well (perhaps [i̵ ]?), as is suggested by a couple of other loanwords where Bactrian ι is substituted by a schwa. 17 For instance, Bactrian φρομιγγο /frəmiŋgə/ 'hope' is borrowed into Tocharian A/B as pärmaṅk 'hope ' (e.g. Sims-Williams 2007: 276), where according to the orthographic rules of classical Tocharian B, a stands for /ə́/. In addition, if Niya Prakrit lastuga 'some type of textile product' comes from a Bactrian *λιστογο 16 The same is true about Prakrit laddaṇa-'load' < *lardaṇa-, which can easily have been made on the basis of *larda-: cf. e.g. doublets such as kara-and karaṇa-, inter alia meaning 'doing, acting'. A similar explanation can account for the fact that some of the New Indo-Aryan derivatives of *larda-have the meaning of a nomen actionis, i.e. 'loading', seen e.g. in Odia ladā/nadā, whereas I assume *larda-to have been in origin a nomen rei actae, i.e. 'that which is loaded'. Just as karafrom √kar 'to do' can mean 'doing', *larda-from a theoretical root *√lard 'to load' can have been understood by native speakers to mean 'loading' next to 'load, cargo'. 17 The vowel should in origin still have been palatal, because of the palatalisation of dental sibilants adjacent to -i-which is suggested by Bactrian in Manichaean script (cf. Sims-Williams 2011). Bactrian ι can also be rendered with -i-in loanwords, e.g. Niya Prakrit -vita/-vida 'lord' < -βιδο /vidə/ < *-pati. /listugə/, i.e. a derivative of λιστο /listə/ 'hand' < *dasta-, 18 this would be nicely parallel to Sanskrit *larda-corresponding to Bactrian *λιρτο /lirtə/.
Second, the discrepancy between Bactrian -ρτ-/rt/ and Sanskrit -rd-should be briefly addressed because both Bactrian and Sanskrit normally keep -rt-and -rddistinct. Admittedly, no cogent explanation for this peculiarity has presented itself so far, and I will only make a tentative suggestion. Given that Bactrian *λιρτο /lirtə/ may have entered Sanskrit and other Indo-Aryan languages through the intermediary of Gāndhārī, 19 an explanation involving Gāndhārī phonetics may not be out of the question. Old Indo-Aryan consonant clusters of the type -rC-were regularly affected in Gāndhārī, as in other Middle Indo-Aryan languages, by assimilation of the preconsonantal -r-to the following consonant, resulting in a long version of the second consonant. As a result, clusters with preconsonantal -r-preserved in writing are best seen as historical spellings (see Baums 2009: 162f.). This probably means that /rt/ and /rd/ were no longer part of native Gāndhārī phonology, which in turn makes it conceivable that speakers of Gāndhārī unconsciously applied voice assimilation in this type of cluster when pronouncing e.g. Bactrian or Sanskrit. In other words, /rt/ may have become /rd/ in the mouths of Gāndhārī speakers, even though the evidence for this is unfortunately meagre. 20 In theory, an alternative solution could be to assume a Bartholomae variant *dr̥ žda-'loaded'. 21 Although the outcome of *-r̥ žd-does not appear to be attested in Bactrian, *dr̥ žda-would probably, through an intermediary *liržda-, eventually yield *λιρδο. 22 However, a Bartholomae variant *dr̥ žda-would be without close parallel in Bactrian and forms such as γιρτο /γirtə/ < *giršta-< *gr̥ šta-(√garz) and υιρτο /hirtə/ < *hiršta-< *hr̥ šta-(√harz), already cited above, make it clear that the voiceless variant is the default in Bactrian for these verbal adjectives. 23 This is in accordance with the general tendency of Middle Iranian languages to eliminate these voiced Bartholomae variants (cf. Harmatta 1964: 406-408). So, while *λιρδο < *dr̥ žda-is not to be excluded as such, it still seems safer to assume that *lardacomes from *λιρτο, despite the uncertainty which in that case remains concerning the rendering of -rt-.

Some notes on the larger historical context
This new analysis of lardayati also fits well with the compositional history of the Divyāvadāna and the larger historical context. However, an important caveat is necessary here because it can neither be deduced with certainty when, where and by whom the Divyāvadāna was composed, nor can the stories contained within this collection be used as historical sources in any straightforward way. 24 According to the communis opinio, the Divyāvadāna was compiled in the early centuries of our era by Buddhists of the Mūlasarvāstivādin-school in the northwest of South Asia (cf. Rotman 2008Rotman -2017. A Bactrian loanword is most likely to have entered Indo-Aryan languages in more or less this time frame and at the northwestern borders of South Asia. It would thus not be surprising if a Bactrian borrowing is attested for the first time in precisely the Divyāvadāna. 25 The larger historical context in which such a borrowing fits has to do with the trade routes connecting Central and South Asia. The Divyāvadāna is an important witness to the connection between Buddhism and mercantilism, as nine of the stories contained in this compilation contain descriptions of caravans and maritime trade. While one cannot confidently say more than that these accounts of trade 23 If Niya Prakrit avaliḱa derives from a Bactrian *αβαλιþκο /əvališkə/ 'swaddle' < *upadr̥ štaka-, as I have recently argued (Schoubben 2021: 55), *αβαλιþκο would also show the expected form with *-r̥ št-. Alternatively, one could also reconstruct a Bactrian *αβολιþκο /avəliškə/, also meaning 'swaddle', < *abidr̥ štaka-, which would then be an exact cognate of Sogdian βδ'yštk 'swaddled'. Both /əvališkə/ and /avəliškə/ could have been rendered as avaliḱa in Niya Prakrit. 24 See Rotman 2008Rotman -2017 where the reader can easily find references to older literature. 25 It is also possible that the Divyāvadāna was composed slightly later than this period, which is, for instance, the opinion of Neelis (2011: 28 fn. 82), who suggests the third to fifth century. No scholar seems to assume that the Divyāvadāna, which is generally taken to be based on the Vinaya of the Mūlasarvastivādins, was composed before the start of the Christian era. In general, the text probably contains material from different layers which cannot be easily stratified. are in some way connected to historical reality, one option is that the Divyāvadāna mirrors the blooming trade of the Kuṣāṇa empire (cf. Rotman 2009: 12-14).
More significantly, there is secure historical evidence that the complex of trade routes in the northern part of South Asia, which was called the Uttarāpatha 'northern route' at the time, had been unified under Kuṣāṇa control. At the same time, the Uttarāpatha was connected to Bactria, which functioned as the major node for connecting the different trade routes of South and Central Asia, including those nowadays known under the name "Silk Road(s)". 26 Because of this, it would not be too remarkable if the Bactrian language had also left its traces in the trade vocabulary of neighbouring languages such as Gāndhārī, Sanskrit or Sogdian, and *λιρτο would be a good example of this. The same would be true of st(h)ora-'packanimal', if that is indeed borrowed from Bactrian (α)στωρο /(ə)stōrə/. Sogdian s'rtp'w 'caravaneer' and Chinese sàbǎo 薩保 'an official in charge of Iranian rituals' could be yet another example of the role Bactrian played in the international trade routes of Central Asia, as Sims-Williams (1996: 51 with fn. 37; cf. also 2010: 126) has convincingly derived these from a Bactrian compound *σαρτοπαο(ο) /sārtəpāwə/ 'protector of the caravan'. 27 These historical circumstances also explain why lardayati is, on the one hand, found in a Buddhist Sanskrit text as the Divyāvadāna and in later Middle and New Indo-Aryan languages, but, on the other, not in Classical Sanskrit. The Uttarāpatha stretched over much of northern India, splitting itself into at least three main branches connected to one another in Mathurā, a hallmark of Kuṣāṇa influence in South Asia (Neelis 2011: 197-200). As a result, a word like *λιρτο can easily have been adopted into the various local vernaculars in northern India that form the 26 Cf. Neelis 2011: 132: "Kuṣāṇa control of a network of routes between western Central Asia and the northern Indian subcontinent accelerated patterns of cross-cultural exchange, longdistance trade, and religious transmission from the first to third centuries CE," and Neelis 2011: 144: "From about the second half of the first century CE to the middle of the third century CE, the Kuṣāṇas maintained control over important nodes on a network of overland routes connecting Bactria in western Central Asia with the heartland of northern India. A chain of cities … linked the multicultural empire of the Kuṣāṇas to the Northern Route (uttarāpatha)." In general, see Neelis 2011: 132-144, 186-204. 27 Because Sogdian normally forms this type of compound with -p'k < *-pāka-and not with -p'w < *-pāwa(n)-, Sims-Williams infers that s'rtp'w cannot be genuine Sogdian. The first part of the compound is Indo-Aryan sārtha-, which has recently (Schwartz 2009) been connected to Semitic words like Arabic sayyāra(t)-and Aramaic šayyārtā-, both 'caravan'. The Sanskrit word could then go back to a South Arabic cognate of these etyma, folk-etymologically re-analysed as sa-artha-'having a goal'. This would mean that sārtha-reflects the maritime trade with Arabia and the Spice Road there, in a similar way as I argue *larda-to be a linguistic remnant of the trade with Central Asia and the Silk Road. basis of the New Indo-Aryan languages containing derivatives of the root √lard. In high-style Classical Sanskrit, a more technical and vernacular word of this type tended to be avoided, whereas the lexicon of Buddhist Sanskrit and Prakrit is more open to including such a word (cf. Burrow 1973: 61f.). One can compare the analogous case of Sanskrit moca(ka)-'shoe' and mocika-'shoemaker', also loanwords from Iranian (Bailey 1955: 21). These words and derivatives of them are attested in Buddhist Sanskrit and recorded in lexica, yet not found in Classical Sanskrit sources, but are still found in Prakrit (moca-'shoe') and are well attested in New Indo-Aryan languages (e.g. Hindi mocī 'shoemaker') (cf. Turner 1966: 597;Burrow 1973: 389). 28

Conclusion
To conclude, it seems best not to derive late Sanskrit lardayati 'to load' from a PIE root *lerd-. Rather, lardayati can be interpreted as an inner-Indo-Aryan denominative of *larda-'load, cargo', conceivably borrowed from a Bactrian *λιρτο /lirtə/ 'that which is loaded > load, cargo' < *dr̥ šta-. An important argument in favour of this etymology is the collocation of lardayati with sthora-'pack-animal' because of the evidence in favour of an Iranian collocation of *staura-and *√darz-. If this etymology is accepted, lardayati is another piece of evidence for the role Bactrian played as one of the main languages in the international trade routes which were under the control of the Kuṣāṇa dynasty.
Acknowledgment: This paper was financed by the European Research Council (ERC) Starting Grant project "The Tocharian Trek" (Grant agreement ID: 758855). For discussions of the idea presented in this paper, I am grateful to Sasha Lubotsky and Julian Kreidl. Michaël Peyrot and the anonymous reviewer are thanked for their valuable comments on earlier drafts, while I thank Elwira Kaczyńska for kindly sending me her paper in advance of publication. Elly Dutton, finally, offered welcome advice on English style and diction.
28 Similarly, the Greek loanword paristoma-'cushion' < περίστρωμα is found in more vernacular sources such as the epics and the Arthaśāstra, but not generally in Classical Sanskrit. Iranian loanwords also often only occur in Sanskrit works that are restricted to the more western regions, for instance, in works from Kashmir (e.g. divira-'scribe') or in works connected to Harṣa's court in 7 th -century Kannauj (e.g. khola-'helmet' < Bactrian χωλο* < *xauda-, which is indirectly attested in a personal name, see Sims-Williams 2010: 153).