Materia Medica Tocharo-Hvatanica

Abstract The two languages once spoken in the oases in the North of the Tarim basin, Tocharian A and B, have preserved many Iranian loanwords. These belong to different chronological layers and are of different dialectal origins. Whereas the oldest layers are now most likely seen as belonging to an unattested Old Iranian dialect, more recent layers have not yet been studied in detail. In this respect, the vocabulary of medical texts represents an important field of enquiry. Most terms come from Middle Indian, but a significant number are of Middle Iranian origin. This component, mostly ingredients and technical vocabulary, seems to be largely of Khotanese origin. The article introduces the material and examines possible scenarios for historical transmission and contact between the North and the South of the Tarim Basin.


Introduction
Tocharian vocabulary contains a large number of Iranian loanwords, belonging to different chronological layers and of different dialectal origins (Peyrot 2015). The oldest layers are now most likely seen as belonging to an otherwise unattested Old Iranian dialect (Peyrot 2018), rather than to a reconstructed "Old Sakan" (Tremblay 2005). More recent layers of borrowings, however, have so far received little or no attention. The medical vocabulary is in this respect emblematic. Whereas most technical terms are of Middle Indian origin, a significant number are Iranian (Carling 2007: 330). The scholarly literature on the subject tends to view the Iranian component as being overwhelmingly of Khotanese origin. If this is true, it will enable us to uncover scenarios of historical transmission and contact between the North and the South of the Tarim Basin.

Siddhasāra and Yogaśataka
The preface to the Khotanese Siddhasāra, 2 the great medical work preserved also in Sanskrit (Emmerick 1980) and Tibetan (Emmerick 1982), may offer us a rare glimpse into the reception of medical texts in Central Asia: yauga-mālyo jsa yuḍāṃdä śau-kṣīrä krra. apaysāṃdä āchai cvai nayä ni bvīra. *viysaṃ dūṣä' kālä u rve haṃdari pātcä. pijsanīra-ṃ aprrasama arve muḍa phari satva 5 "By means of collections of prescriptions they performed (medical) practice in the whole country. 3 Disease (was) unrecognised because they did not know the theory of it: the unequal humour, time and seasons, (their) intervals too. Inappropriate medicines struck them down: many beings died." (Emmerick 1983: 20-1) It was R.E. Emmerick's idea (1983: 22) that the "collections of prescriptions (yauga-mālyo jsa)" could refer to the Yogaśataka, "which was popular not only in India and Ceylon but also in Central Asia". 4 Such a reading of the passage is well worth considering, although I have not found any mention elsewhere of a rivalry between the Yogaśataka and Siddhasāra traditions. No Yogaśataka manuscript has been found in the South of the Tarim basin, but this is not sufficient to justify such an enmity. Moreover, Siddhasāra traditions are present in the North, although they are quite late. 5 It is possible, however, that the polemic passage of the Siddhasāra does not refer to a contrast between Southern and Northern oases. It could simply remind the reader of the contraposition existing between longer works that explained the medical theory and the popular collections of recipes such as the Jīvakapustaka, which were clearly made for practical use. 6 At any rate, if Emmerick's idea proves right, the preface of the Khotanese Siddhasāra might witness the late echoes of a contact scenario between the South and the North, which was already taking place at the time of the first Tocharian translation of the Yogaśataka in the North. 7 One could surmise that not only the Yogaśataka, but also other medical texts, were circulating widely between the South and the North. This could have been the reason why the Tocharian medical lexicon seems to be so composite, and the Iranian part appears to be overwhelmingly of Khotanese origin. 8 In such a contact scenario, one should obviously not underestimate the oral component, as pointed out by Carling (2007: 332).
In what follows, the Tocharian medical vocabulary of alleged Khotanese origin will be presented and analysed, in an attempt to verify whether such a contact scenario has to be assumed or not.

The Tocharian medical vocabulary of alleged Khotanese origin
Twelve medical lexical items have been selected. A distinction can be made between names of ingredients and technical vocabulary. Individual studies will attempt to verify whether the items have a clear Khotanese origin. Among the ingredients, we find: • TB aṅkwaṣ(ṭ) subst. 'Asa foetida' • TB eśpeṣṣe subst. 'spreading hogweed (Boerhavia diffusa)' • TB kuñi-mot subst. 'wine' • TB kuñcit ∼ kwäñcit A kuñcit subst. 'sesame' • TB kurkamäṣṣe ∼ kwärkamäṣṣi adj. 'pertaining to saffron' • TB tvāṅkaro subst. 'ginger' In the following, the items belonging to the technical vocabulary are listed: to rub in, rub on, anoint, embrocate (prior to washing)' It is important to note the presence of three verbs in this group, a feature that might suggest deeper linguistic contact (Thomason 2001: 70). 2. Names of ingredients 2.1. TB aṅkwaṣ(ṭ) subst. 'Asa foetida' Tocharian occurrences: • aṃkwaṣ PK AS 2A a5, aṅkwaṣ PK AS 2A b2. 9 Both forms appear in a list of ingredients belonging to the Tocharian bilingual (Sanskrit-Tocharian) fragments of the Yogaśataka. The Sanskrit equivalent is hiṅgu-'id.' 10 in both cases (Tib. śiṅ-kun). • aṅwaṣṭ PK AS 3B b5. 11 The word appears again in a list of ingredients, although the text has not yet been identified. It was classified as a medical/magical text. The title of the section to which the text should refer is given in line b4 as a generic bhūtatantra "Treatise against the demons".
In view of these problems with a derivation of aṃguṣḍa-from Proto-Iranian directly, it is preferable to see in LKh. aṃguṣḍa-a loanword from an Iranian language in which intervocalic *-ǰ-underwent fricativization (> *-ž-). This might be e.g. Sogdian, in which old *-ǰ-gives regularly -ž-(GMS: 42), or even Parthian, for which the same sound change is attested (Durkin-Meisterernst 2014: 96). Although highly speculative, a Sogdian or Parthian form might also be at the origin of the irregular -ž-found in New Persian angu-žad, which seems to alternate with a native form with -z-(angu-zad, Hassandoust 2015: I n°525).
The dating of the syncope is crucial to determining whether the Tocharian form was borrowed directly from the unattested Sogdian (or Parthian, or another, This is a graphic phenomenon associated with older stages, but without phonological relevance (Peyrot 2008: 178). 12 The edition of P 2893 is to be found in KT III: 82-93. 13 See DKS: 1, Bailey (1957: 50) and Rastorgueva and Èdel'man (2000: 166). 14 See Hassandoust (2015: I n°525). Compounds with another second member are also present, cf. angu-yān (Hasandoust 2015: I n°535) and angu-dān (Hasandoust 2015: I n°5 23), all meaning 'Asa foetida'. 15 For the Late Khotanese alternations u:ū and u:au cf. Dresden (1955: 406 [4], [5] unknown, Middle-Iranian language of the area) cognate that may be posited, or from Khotanese. It seems that the attribution of the syncope to Khotanese is not problematic: -a-was first weakened 17 to -ä-in unstressed syllable (*angùžata-> *angùžäta-) and then lost. Moreover, New Persian angu-žad, if borrowed from Sogdian or Parthian, may show that the unattested form had no syncope (although this is far less certain). In other words, the Tocharian form needs a source language in which syncope has already taken place. This may be identified with Khotanese, in which the loss of -a-can be accounted for without problems. More questionable would be the possibility that loss of -a-was already realized in the unattested Middle-Iranian antecedent. Therefore, the chance that the Tocharian form was borrowed directly from Khotanese may seem higher than the possibility that Tocharian borrowed from Sogdian or Parthian. Nevertheless, this second possibility cannot be excluded.
As far as Tocharian is concerned, Iranian *-u-was reinterpreted as w + ǝ and, more precisely, as k w + ǝ, so that the word takes the aspect /ank w ǝṣt/. This phenomenon is to be observed also for a series of other Tocharian medical terms (TB kuñcit ∼ kwäñcit, kurkamäṣṣe ∼ kwärkamäṣṣi and kwarm < Skt. gulma-). 18 Since the development of u to u ∼ wä ∼ wa is thus understandable within Tocharian, the form may be derived from Khotanese without any problem. 19 The form aṅwaṣṭ with final -ṭ is older than the form without -ṭ, as aṅkwaṣ can be derived from the form with final -ṭ by sound law (Peyrot 2008: 67).
The Khotanese occurrences are attested in a puzzling series of different orthographies. It is immediately clear that such a vowel alternation in the first syllable is unprecedented, and therefore difficult to assess: Five of fifteen total occurrences show a back vowel (au-, u-), whereas the rest point to a front vowel (i-, ai-, e-). H.W. Bailey's tentative explanation (DKS: 48) takes the forms with back vowel as original and posits a hypothetical *ā-vastyā-. 24 However, this leaves the forms with front vowel, i.e. the large majority, unexplained. The subscript hook, which occurs five times, might signal the earlier presence of a lost -l-, as in the case of OKh. balysa-and LKh. ba'ysa-, be'ysa-, bi'ysa-, bai'ysa-. Only a few occurrences of the word have a subscript hook, but in the case of ba'ysa-, too, the subscript hook is often omitted. 25 Indeed, the presence of both front and back vowels in the Late Khotanese notation might also point to a lost -l-, which is normally associated with fronting, as noted by an anonymous peer-reviewer. The case of hälsti-'spear', however, which occurs in Late Khotanese both with initial ha'°and hu'°(DKS: 486), apparently shows that loss of -l-could also be associated with a back vowel. For the Khotanese word for Boerhavia diffusa, a hypothetical Old Khotanese form *alśta or *älśta can be then reconstructed. *älśta could be further interpreted as an inflected form of a stem *älsti-, a variant of OKh. hälsti-(SGS: 288) without initial h-(< PIr. *Hr̥ šti-'spear', cf. Av. aršti-and OP r̥ šti-'id.'). 26 The use of terms for 'spear' to describe plants with reference to the oblong form of their leaves is documented in Latin, where the adjective lanceolātus 'lanceolate' is used as a botanical term. Since the leaves of the Boerhavia diffusa are not oblong or spear-shaped, the term may refer here to the form of its roots.
However, given the tentative nature of this explanation, there is always the possibility that the word could represent a borrowing from an unknown language. Adams (DoT: 104) compares the Khotanese word with Tocharian eśpeṣṣe. The meaning is secured by the Khotanese and Sanskrit parallel (Maue 1990: 163 fn. 20). If -ṣṣe is an adjectival suffix, then we are left with something that closely resembles the Khotanese word, although Tocharian -śp-for Khotanese -śt-is not paralleled elsewhere. A possibility to obtain the cluster -śp-would be to consider the Tocharian word as a compound from LKh. *aiśti-+ *bā(ga) > *aiśtäbā > *aiśtbā > TB eśpe. 27 However, this leaves the Tocharian vocalism of the final syllable unexplained, since it is very unlikely that LKh. <ā>, which probably had the value /ɔ/ (Emmerick 1979: 245), could have resulted in TB -e-. Overall, the comparison seems rather doubtful.
• gūräṇai mau 'grape wine' P 2895.29 (Paris Y). 29 Adams (DoT: 193) puts forward the hypothesis that the first part of the word may derive from LKh. gūräṇaa-(KS: 142), adjective to gūra-'grapes', with loss of the medial syllable. LKh. gūräṇaa-is an adjectival formation which was formed with the suffix -īnaa-(PIr. *-ainaka-). The long -ī-of the suffix was shortened to -i-or -ä-in unstressed position. This phenomenon may be part of a more general tendency of vowel weakening before the nasal -n-, which is already attested in Old Khotanese (KS: 136). For the adjective gūräṇaa-, therefore, a proto-form *gudrainaka-may be reconstructed. If TB kuñi is really derived from the adjective gūräṇaa-, we must reckon with a loan from Khotanese, after the shortening of the long -ī-of the suffix (already Old Khotanese) and the loss of intervocalic -k-: kuñi < gūni < gūrṇi < LKh. gūräṇai (< PIr. *gudrainakah). At first sight, Adams' suggestion might appear rather far-fetched. However, the occurrence of the adjective gūräṇaa-with mau 'wine' in the Late Khotanese lyrical poem contained in the manuscript P 2895 30 might support his hypothesis. Indeed, the parallel TB kuñi-mot ∼ LKh. gūräṇai mau seems rather striking. The Tocharian B form would then be a partial calque with TB kuñi < LKh. gūräṇai and TB mot for LKh. mau. As suggested by the reviewer, it might be worth noting here that TB mot cannot have been borrowed from Sogdian, as stated e.g. by X. Tremblay (2005: 438). The form mwδy quoted by Gershevitch (GMS: 408) from the Ancient Letter IV, l. 5, is now recognized to mean 'price' (LW < Skt. mūlya-). 31 2.4. TB kuñcit ∼ kwäñcit A kuñcit subst. 'sesame' Tocharian occurrences: • TB kuñcit PK AS 3A a1; a3 (medical), PK AS 8C a7 (medical) The most recent Tocharian lexicographical works consider the word a loan from Khotanese. 36 This communis opinio is probably to be traced back to a note by H.W. Bailey (1937: 913). However, he does not state directly that the form is Khotanese. He writes rather that the Tocharian B word represents "an older stage than Saka kuṃjsata-". He further derives the Khotanese form (DKS: 61) from a reconstructed *kuncita-, which is based on Skt. kuñcita-, even if this seems to be used for another type of plant, the Tabernaemontana coronaria. 37 In fact, the Tocharian and Khotanese occurrences both in the Yogaśataka and in the Siddhasāra translate Skt. tila-'Sesamum indicum', (KEWA I: 504), not kuñcita-. Tremblay (2005: 440) does not give any identification more precise than "Middle Iranian". If the form is really Iranian, it might not be easy to find out if the Tocharian word actually derives from the protoform *kunčita-, which seems to be at the origin of Sogdian kwyšt'yc, 38 Khotanese kuṃjsata-, Old Uyghur künčit 39 and Middle Persian kwnc(y)t (CPD: 52). For what concerns Pashto kunjǝ́la, an Indian origin is preferred by Morgenstierne. 40 He further extends his hypothesis to all Iranian forms, which he considers old loans from Indian. In general, the Pashto form seems to share with Khotanese the voiced affricate and a different vowel in the second 34 Not to be read kuṃjsąna, see Luzzietti 2018-19: 45-6. 35 The numbering follows Emmerick 1970: 43-7. 36 See Carling et al. 2009: 148 andDoT: 193. 37 Gharib 1995: 202. Henning (1946) proposes the following: "kwyšt-(if = sesame) = ku̯ išt < *ku̯ inšt < *ku̯ inčt < *kunčit". A graphic explanation is preferred by Benveniste (1940: 180) ("Est-ce une mauvaise graphie pour *kwnšt-?"). The anonymous peer-reviewer noted that a form kwync'[ is also attested in P 29.9 (Sims-Williams and Hamilton 1990: 33), which seems to be phonetically closer to the forms occurring in the neighbouring languages. 39 An old loan from Sogdian, according to Tremblay 2005: 440 (?). 40 See Morgenstierne 1927: 33 and EDP: 39 "certainly" old LW < Indo-Aryan (Skt. kuñcita-) in Pashto. Here is not the place to reconsider the whole history of the word, which does not seem to be specifically Iranian and can be traced back in time as far as Akkadian kurkanū and Greek κρόκος. 45 The basis for the Tocharian form must have been provided by an unattested *kurkuma-. As in the case of aṃkwaṣṭ and kuñcit ∼ kwäñcit (cf. 2.1 and 2.4), *ku was reinterpreted as k w + ǝ, so that we obtain the spelling /k w ǝrk w ǝm/, further dissimilated to /k w ǝrkǝm/. This (*kurkäm) might have been the original form from which the adjective was derived through accent shift (/k w ǝ́rkǝm/ > /k w ǝrkǝ́m°/). The tiny fragment THT 2676 is one of the earliest Tocharian manuscripts (Peyrot 2014: 139 andMalzahn 2007: 267)  undissimilated form /k w ǝrk w ǝm/. Since all Indian forms (CDIAL: 3214, cf. Skt. kuṅkuma-) have a nasal instead of the expected -r-, it is more probable that the Tocharian word derives from Iranian. Given the fact that saffron is known to grow in Persia (Laufer 1919: 320), a Middle Persian origin (Pahlavi kwlkwm (CPD: 52) and New Persian kurkum 46 ) is suggested by Tremblay (2005: 437). Otherwise, the Middle Persian form might have reached Tocharian through Khotanese *kurkuma-(DKS: 63). In fact, this is the form which might be reconstructed for Old Khotanese based on the Late Khotanese occurrences. 47 However, as noted by a referee, there is no special phonetic feature that might be attributed to Middle Persian proper. Tremblay's idea seems thus quite arbitrary and a Middle Persian origin remains highly doubtful. For the time being, it seems safer to consider the origin of the Tocharian word as coming from a general 'Middle-Iranian' context, without further specification. It might be noted further that Sogdian kwrkwnph, 48 because of the final labial plosive, remains a less probable candidate. An Iranian origin has been also suggested for Tib. kur-kum (Laufer 1916: 474).

M A T E R I A M E D I C A T O C H A R O -H V A T A N I C A
H.W. Bailey's initial idea (1937: 913) sought to explain TB -vā-against Khotanese -u-by comparing TB aṅkwaṣ(ṭ) and Khotanese aṃguṣḍa-, simply taking note of the same correspondence, without offering any further explanation. This is not possible because the Tocharian form contains here clearly /wá/ (<wā>) and not /wə́/ (<wa>) for /u/ as in aṅkwaṣṭ (see 2.1). Some time later, however, he developed a new etymological proposal. 51 He derived the Khotanese word from *tuwam-kara-with *tuwam°from the Proto-Iranian root *tauH-'to be strong, swell' (Cheung 2007: 386). In this case, the Tocharian form would have conserved the Pre-Khotanese state of affairs and should be considered as a very old loan (Tremblay 2005: 428 and DoT: 343). Bailey's derivation seems to imply a nominal form *t(u)v-a-from the verb *t(u)v-'to be strong ' (DKS: 144). This root is attested as verb with causative suffix -āñ-in LKh. tv-āñ-'to strengthen' (SGS: 41). Several nominal forms from the same root are also to be found as medical terms, e.g. LKh. tv-āñ-āka-'strengthener' (KS: 46) 52 and LKh. tv-āmā-(< *tv-āmatā-) 'strengthening' (KS: 94). 53 The case ending of the first member of the compound would have been preserved in the nasal *-m-before the second member *-kara-, as is the case in similar compounds, cf. e.g. dīraṃggāra-'evil-doing' (SVK I: 56, Degener 1987: 39). This derivation, however, seems semantically difficult. tv-a-must be a substantive (KS: 1) with the meaning 'strong one', 'strong thing' or 'fat'. The resulting compound could be then approximately translated as 'maker of strong (things or beings)'. Admittedly, such an attribute would be suitable for a person, not for a plant. It would be then desirable to have an adjective as first member of the compound. This is indeed possible if one starts with a form tv-āna-, an -āna-derivative (pres. part. mid. KS: 78) from the root tv-, which could produce a proto-form *tvāna-kara-'strong-maker'. This would yield OKh. *tvāṃgaraa-54 through syncope of internal unaccented -a-. Both Old Khotanese reconstructed forms, *tv-aṃ-garaa-and *tv-āṃ-garaa-, may have been antecedents of the attested LKh. ttūṃgara-, since both OKh. tvā°a nd tva°may result in LKh. ttū°. For tvā°> ttū°one may compare the possessive adj. OKh. tvānaa-'your' (KS: 85) which occurs in LKh. as ttūnā (IOL Khot S. 15.11) and for tva°> ttū°OKh. tvaṃdanu 'reverence' (SGS: 219) and its Late Khotanese counterpart ttūda (IOL Khot S. 6.27). Both Old Khotanese reconstructed forms may as well have been borrowed into Tocharian B. There is indeed no need to consider TB tvāṅkaro as a Pre-Khotanese loanword. The evidence suggests that the word may have been borrowed from the Old Khotanese antecedent of LKh. ttūṃgara-. 55 It might be worth noting that Tib. li doṅ-gra, which translates Skt. nāgara-'ginger' in the Siddhasāra (Emmerick 1985: 313 andBielmeier 2012: 21-
In the manuscript PK AS 3A it is used consistently in the gen. sg. with sāṃtke 'remedy'. The text describes four remedies against ampoño. All other occurrences refer to medical texts. Adams' second edition of his Tocharian B dictionary contains the following statement s.v. ampoño: "A nomen actionis from āmp-'rot,' q.v., from Khotanese hambu-, i.e., hambu-+ the Khotanese abstract-forming suffix -oña" (DoT: 21). In Old Khotanese there is indeed a word haṃbūta-occurring in Z 5.16 and 5.18, 56 two passages which present us with two literary similes involving medical terminology: Z 5.16 trāmu māñaṃdu kho hvą'ndä haṃbūtä haṃbaḍä ysūna "Similarly, in the case of a man's fester full of pus, when one puts ointments on it on the outside, there is only so much alleviation of it." (Emmerick 1968a: 99) cvī ye ālīva nitcana īndä samvī ttaṃdu hamārgya Z 5.18 samu kho haṃbūvu bei'ttä . harbiśśī āchai jīye . trāmu nairātma-hvanaina uysnori ysaṃtha jyāre "Just as when one cuts open a fester all disease is removed for one, so through the doctrine of selflessness (nairātmya) births are removed for a being." (Emmerick 1986: 73) This has the aspect of a past participle from the Proto-Iranian root *pauH-'to stink, smell, rot' (Cheung 2007: 302), to which a preverb *ham-has been added. 57 In the corresponding stanzas of the Mañjuśrīnairātmyāvatārasūtra, the word appears regularly as ha (ṃ) the past participle haṃbūta-is derived. However, the suffix -ūña-/-auña-can be added to past or present participles but there is no example with the suffix being added directly to a present stem (KS: 159). If one were to add it to haṃbūta-, one would expect *haṃbūttauña-, in line with the attested hämättauña-(< past part. hämäta-) (KS: 164). The resulting intervocalic -t-seems to undergo strengthening rather than be lost altogether. However, one cannot exclude the possibility that intervocalic -t-was lost in this case already in Khotanese. In fact, as suggested a referee, -tt-in the hapax hämättauña-might be an example of 'morphologische Verdeutlichung' (KS: 162), i.e. a way to stress the presence of a morpheme boundary before the suffix. If this is correct, one could see in ampoño the past part. LKh. haṃbva-to which the suffix -auña-has been added. This would confirm the hypothesis of a Late Khotanese origin of ampoño, as suggested by Adams.
From the Tocharian point of view, however, there is still the possibility that ampoño is a genuine Tocharian formation based on the verb TB ampa-(borrowed from LKh. haṃbva-, see 3.2). In fact, all attested forms point to a nom. sg. ampoño or ampoña*. Because of the palatalization, ampoña would be the expected form (M. Peyrot, personal communication). THT 510, the fragment containing the only occurrence for ampoño, is normally classified as late, so the form might be simply interpreted as secondary for earlier ampoña (Peyrot 2008: 99-101). This form would have the appearance of a derivative in -'eñña from a verbal root, 59 which in this case could be ampa-'to rot' (see 3.2). For the forms with single -ñ-for the expected -ññ-one might compare the obl. sg. of wṣeñña, which is attested four times with a single -ñ-(IOL Toch 117 b4, Km-034-ZS-R-01 a7, PK AS 16.7 a4, IOL Toch 62 a3).
D.Q. Adams (DoT: 48) regards it as a Middle Iranian loanword from the same root as Khotanese haṃbūta-, New Persian ambusidan, etc. Malzahn (2010: 525) seems to be of the same opinion and would rather take the word as a Khotanese loanword. If from Khotanese, one might envisage the possibility that the form has the aspect of a denominative formation from LKh. haṃbva (< Old Khotanese haṃbūta-, see 3.1), resulting in TB amp(w)a-. This verb can be thus traced back with a fair degree of certainty to Late Khotanese.  It seems that the first scholar to put forward this etymological proposal was Van Windekens (1949). Isebaert (1980: §180) finds the derivation unconvincing and suggests an Indo-European origin. His main criticism of Van Windekens' proposal is based on morphological reasons. According to him, Middle Iranian loanwords never receive the masculine -e. Whereas Bailey's Dictionary (DKS: 74) does not seem to take note of the possibility of a loanword, Tremblay (2005: 433) returns to Van Windekens' proposal and reports it without any further comment. The Khotanese word is formed from the Proto-Iranian root *xard-'to defecate' 61 to which the suffix -ka-has been attached (KS: 181), resulting in *xardaka-. In order to obtain the attested forms, one has to assume a series of metatheses which took place very early, at least earlier than the sound change -rd-> -l-in Khotanese: *xardaka-> *xadraka-> *xadarka-. This might have been the base for Yidgha xǝlarγo (from a feminine *xadarkā-, EVSh: 79) and Khotanese khārgga-, through loss of intervocalic -d-and voicing of -k-.
Given the specificity of the formation, if the word is a loanword, it cannot come but from Khotanese. After all, it seems that Khotanese 'mud' refers to the same semantic areas of Tocharian 'dirt ' and 'filth'. 62 In this case, the Khotanese form would have undergone in Tocharian a further metathesis to become krāke.
As reported by Adams (DoT: 229), the meaning 'to be dirty' was suggested by M. Peyrot (apud Malzahn 2010: 612) on the basis of the substantive TAB krāke, from which the verb is derived. The passage in question, which refers to poor, blurred eyesight, seems to justify such an interpretation.
The first scholar to make known the word was H.W. Bailey (1935: 137). The striking correspondence with the Tocharian word was again noted by him some years later (Bailey 1947: 149). A further clarification of the meaning and the etymology is offered by R.E. Emmerick (1981: 221). 63 There the meaning is established as 'suppository' against Bailey's 'pastil'. The etymology is given as < PIr. xšaudakā-, a formation from the root *xšaud-'to wash' (Cheung 2007: 455). Since the word is a very specialized medical term, one should assume that the borrowing took place quite late, when Indian medical texts were already circulating within the Tarim basin. As it is attested only in the Late Khotanese Siddhasāra, the word was possibly borrowed from Late Khotanese, although it is not to be excluded that Old Khotanese translations of medical texts existed, even if they are no more extant. In this case, a possible Old Khotanese form may have been *ṣṣūdakā-or *ṣṣūvakā-, as intervocalic -dmight have been lost already in Old Khotanese (see e.g. OKh. pāa-< PIr. *pāda-). The preservation of intervocalic -k-is noteworthy. The possibility that the Tocharian word was borrowed from Late Khotanese may seem more probable, as the nearest antecedent of the Tocharian initial cluster ṣpmay have been LKh. ṣvrather than OKh. *ṣṣūv-. Thus, TB ṣpakīye must be considered a Late Khotanese loanword in Tocharian.
In conclusion, Adams is probably correct in interpreting the word as a loan from Iranian. Further, it seems clear that sanapa-can only be derived from Pre-Khotanese, as this is the only Iranian language which has a -p-increment to the root PIr. *snaH- (Cheung 2007: 348), no word-initial palatal, 68 and an extra epenthetic vowel in the first syllable.

Conclusion
Of the twelve lexical items analysed, one (sanapa-) can be derived from Pre-Khotanese and nine (aṅkwaṣ(ṭ), eśpeṣṣe, kuñi-mot, tvāṅkaro, ampa-, ampoño, krāke, krāk-, and ṣpakīye) can be ascribed to Khotanese proper. Among these, tvāṅkaro is certainly an Old Khotanese borrowing. For eśpeṣṣe, kuñi-mot, ampa-, ampoño (if not directly from ampa-), and ṣpakīye a Late Khotanese origin can be posited, although for eśpesse this remains for now too uncertain. There is unfortunately no way to determine whether aṅkwaṣ(ṭ) and krāke (with its derivative krāk-) have been borrowed from Old or Late Khotanese. For the remaining two items (kuñcit and kurkamäṣṣe), an Iranian origin can be given as certain, although the dialect affiliation is still not completely clear.
I am aware of the fact that the dimensions of the analysed corpus are quite small. Nevertheless, from these results one may argue that contact between Khotanese and Tocharian took place uninterruptedly from prehistoric times until the epoch of the first written attestations. In particular, the medical lexicon may bear traces of contact both at an oral level in the Pre-Khotanese epoch and at a written level in the historical epoch. Five items that can be attributed to Khotanese proper, ampa-, ampoño, kuñi-mot, krāke and ṣpakīye, are in fact technical terms which show a high level of semantic specialization. They must be assigned to a period in which Indian medical knowledge was already circulating widely in written form in both the South and North of the Tarim basin.