D1 is not ā

Emmerick's treatment of the āa-declension in SGS divides forms into monosyllabic and polysyllabic stems (SGS: 305–8). While historical and comparative evidence may suggest that the monosyllabic stems should end in /ā/, synchronically they behave no differently from stems in the aa-declension. Both share NS -ei, -ai, AS -au, -o, GDS -ai, -ei, IAS -eina, -aina, NAP -ā, and GDP -ānu (SGS: 297–9, 302–8). Emmerick's monosyllabic āa-stems synchronically belong in the aa-declension. GDSm dai ‘fire’ Z 2.14 does not show *ā+i>ai but a+i>ai. There are three additional inflectional forms from this group that merit comment: LS, IAP, and LP.

The LS dāña ‘fire’ Z 2.43 shows daa- plus the LS ending -aña attested in a-declension nouns and adjectives (SGS: 262, 283) /δāña < δa- + -aña/. The long ā in dāña is not from the stem, but from the a+a>ā contraction.

The IAP päyau jsa ‘foot’ Z 22.148 shows two processes. As shown in Hitch 2016b: 14–5, the oblique plural suffixes, IAP, LP, VP and possibly GDP, delete a stem final /a/. For example, the IAP morpheme -yau deletes the stem vowel /a/ in the aa- and aā-declensions as in spätyau Z 3.60 < spätaa- ‘flower’ + -yau, baśdyau jsa Z 6.22 < baśdaā-‘sin’ + -yau. With päyau jsa, the /a/ is deleted in paa- /ba-/ giving underlying *pyau jsa (cf. IAP pyau SI M13.14v3) which then undergoes glide resolution /bĕyau̯ dᶻa < byau̯ dᶻa < ba- + -yau̯ dᶻa/ (see below and Hitch Reference Hitch2016b: 23, fn. 29).

The LP for foot is variously spelled päto’ Z 3.5, po’ Z 2.132, pau’ Z 5.105, pvo’ Z 5.40. The shape päto’ shows final /a/ deletion and a spelling variant of the archaic LP suffix -äṣo Footnote ¹² /bĕō < bĕẓō < ba- + -ĕẓō/ with -ä- instead of later usual -u- (cf. SGS: 270). The shape pvo’ reflects the usual LP -uvo’ /buō < buẓō < ba- + -uẓō/. It is not clear which of the two suffixes is reflected in the shapes po’ and pau’.

While Emmerick's monosyllabic āa-stems are synchronically aa-stems, his polysyllabic āa-stems are in fact synchronically āa-stems. Here we can look for the contraction behaviour of stem final /ā/. An -ā- /ā/ is retained in almost all endings: NS -ā, AS -ā, NAP -ā, IAS -āna, LS -āya, LP -āvo’ (SGS: 305). The exception is that some instances of the GDS of haḍāa- ‘day’ show -ai, but these are alongside uncontracted -āyä, -āyi, e.g. haḍāyi Z 11.17~haḍai Z 11.18.Footnote ¹³ It seems that -āyä is original and -ai a later development in this common word, perhaps showing transfer to the aa-declension. The alternate haḍai does not provide proof for *ā+i > ai.

The NS and IAS of the polysyllabic āa-stems show ā+ä>ā, not *ā+ä>ei > ai. E.g.:

Similarly, the AS forms show ā+u>ā, not *ā+u > au > o:Footnote ¹⁴

The D1 stem-final vowel cannot be -ā-. We thus have the synchronic D1 stems ttätsa- ‘cross (a river)’, naltsa- ‘go out’, nimandra- ‘invite’, pa- ‘protect’, paṭha- ‘burn (tr)’, buysa- ‘extinguish’, and vatsa- ‘go down’.

ai in D1

The synchronic derivations in the D1 verbs showing -ai- may now be described. There are two distinct contractions. The first, a+ī >ai is phonologically simple with a low and a high vowel combining to give a low-to-high diphthong, /ai̯ < a + ī/.

The second contraction giving -ai- is more complex; it involves two historical stages. As outlined above, in the older, or archaic, stage, the low vowel a combines with a mid vowel ä to give the low-to-mid diphthong ei /ae̯ < a + ĕ/. This is kept distinct from a+ī>ai. Some time within Old Khotanese the dipthongs ai and ei merged and the spellings interchanged with a preference for ai. Both archaic contracted diphthongs may be seen in NSm baśdainei ‘of sin’ Sgh[1] 36.3, 4 < baśdaā- ‘sin’ + -īnaa- + -ä. In a later copy of this same text passage the diphthongs are no longer distinct in this word: baśdainai Sgh[17] 36.3, 4.

With the D1 verbs there is one example of a+ä>ei in an archaic text, 3Sij.m vatseiya Neb 50.30Footnote ¹⁵ • / watˢae̯ʔa < watˢae̯da < watˢa- + -ĕda/. There are two examples from Z, 3Sp.m naltseiye Z 22.202, naltseiyä Z 24.405 • /naltˢʰae̯ʔē < naltˢʰae̯dē < naltˢʰa- + -ĕdē/. Elsewhere ai has replaced ei.

a+ä>ei>ai

pāmä: a+ā>ā

It is worth pointing out that 1Pp.a pāmä Suv[Or] 6.1.39, pāmī (+ -i) Suv[Or] 11.9 shows ordinary a+ā>ā • /bāmĕ < ba- + -āmĕ/. Compare this regular formation with Emmerick's derivation from pai-: “*payāmä > * päyāmä > *pyāmä (cf. däyāmä, dyāmä < *dayāmä < dai- p. 47) > pāmä” (SGS: 86).

D2 behaves partly like i

With the contraction V+ā>yā the D2 vowel behaves like i /i/. In Hitch Reference Hitch2016b: 29–31 I show how several verb stems ending in -i- show i+ā>yā. These are ji- ‘cease, be destroyed’, kṣi- ‘be troubled, disturbed’, hävi-~hävi- ‘appropriate’, āri- ‘mix, meet (intr)’, pari- ‘order’:

The same pattern is found with the nominals. When the GDP suffix -ānu attaches to an iā̆-stem, the result is -yānu, -iyānu:

With the contraction V+a>ya the D2 vowel again behaves like i /i/. The attested examples are restricted to the 2Siv.a, which also show the resolved forms.

This process has the same result as the contraction i+a>ya known from both verbs and nouns. Two examples with verbs in -i- are known:

More examples are found with the iā̆-nominals with several suffixes of the shape -a:

Orthographically this contraction is i+a>ya, but phonemically the result of this contraction is a diphthong /i̯a/, not a glide + vowel /ya/. This is clear from the meter of Z where orthographic ya arising from contraction is two moras long while orthographic ya not from contraction is a single mora. The last syllable of LSf brītya Z 2.169 < brītiā- ‘love’ + -ⁱa, and of NAP indrya Z 1.85 < indria- ‘sense’ + -a is two moras long, while the last syllable of NAPm banhya Z 2.45 < banhya- ‘tree’ + -a, and of NSf karya Z 2.15 < karyā- ‘effort’ + -a, is one mora long. This is because underlyingly these pairs are respectively /i+a>i̯a/ and /y+a>ya/ (Hitch Reference Hitch2015b: 304–6). Another difference is that Cya reflecting the two-mora diphthong /i̯a/ may resolve to the two syllables Ciya~Cäya, while Cya reflecting a stem final glide+a /ya/ appears never to resolve. Beside indrya there is indriya (19× in Z) and beside brītya there is brītiya IOL Khot 155/1v3, while banhya never resolves to *banhiya and karya never to *kariya (Hitch Reference Hitch2015b: 296).

There is one example of the D2 vowel in the contraction V+a>ya in Z. In the hemistich below, 2Siv.a dya < dV- + -a counts as two moras /δi̯a/ in the final five-mora segment:

The evidence of the two contractions V+ā>yā and V+a>ya suggests that i /i/ might be a candidate for the D2 vowel. However, other D2 contractions show that the vowel cannot be i /i/.

D2 is not i

When suffixes beginning with -ä- attach to a D2 stem, there is contraction resulting in -ai-, that is, V+ä>ai: dV- + -ätä > 3Sp.a daitä Z 24.416; ysV- + -äte > 3Sp.m ysaite Z 2.212; bāysdV- + -äta > 2Pp.a bāysdaiya Z 23.8. This contraction, V+ä>ai, is not attested where V=i. At the same time, the contraction i+ä is well attested with the iā̆-nominals where the result is ī, that is, i+ä>ī.Footnote ¹⁹

That the D2 stem vowel cannot be i is also evident with suffixes beginning with -ī-. 3Pp.a forms like uysdaindi Z 20.20 ‘look up (at)’, daindä Sgh 51[2] ‘see’, and bāysdaindä ‘observe’ Z 3.17 cannot derive from *-i- + -īndä, nor can 1Sp.a daimä Z 6.7 from *di- + -īmä, because elsewhere in the grammar i+ī results in ī, not ai. The contraction i+ī>ī is attested among the iā̆-nominals with denominal adjective suffixes beginning with -ī-(KhSuf: 136 §21.5.6, 143a):

i+ī>ī with derivational suffixes

D2 cannot be i, but should have a value enough like i to be able to become y when an ā or a is suffixed. Phonemically, /yā < D2+ā/ and /i̯a < D2+a/.

D2 is not ī

As argued recently there are no nominal or (known) verbal stems ending in /ī/ (Hitch Reference Hitch2015b: 297, fn.9; Hitch Reference Hitch2016b: 26–32). But there may be other places to look for the behaviour of ī as the first vowel in contraction.

The 3S enclitic pronoun has two forms. After phoneticallyFootnote ²⁰ short vowels it most often has the form -i (earlier given as -ī Footnote ²¹ ) and contracts with the vowel to a diphthong: a+i>ai, e+i>ai, ä+i>ī, u+i>vī~uī /u̯i/, o+i>vei~vai (cf. Glossar: 398b–400a). After phonetically long vowels and rarely after short it has the form yä (ye, yi) and there is no contraction: -ā yä, -e yä, -ī yä, -o yä, -ū ye, -ai yä, -au yä (cf. Glossar 488a–b). After long -ī the form is yä: uvī yä ‘one's senses’ Z 3.30; khvī yä pratäbibau yīndi ‘indeed, how does one make an image of him’ Z 23.39. Final -ī does not contract with the 3S enclitic -i.

Another place with potentially relevant information is the behaviour of long vowels and diphthongs at the end of nominal and verbal stems. These do not contract in inflection or declension. Instead, there is epenthesis of y Footnote ²² between the two vowels.

Presumably there would also be epenthesis after a stem-final long ī. The evidence of these two phenomena suggests that D2 cannot be ī.

D2 is not u, ū, o, au

The rounded vowels u, ū, o and au could probably be eliminated as D2 vowel candidates on general phonological grounds alone. None would likely semivocalize or diphthongize to y. There is some specific evidence. The ua-declension of nouns and the uā̆-declension of adjectives given by Emmerick show a rounded element, either ū, u, v, or uv, after contraction in all forms (SGS: 322–8). Similarly, all au-declension nouns show a rounded element, either au, u or v, after suffixation (SGS: 328–30). Verb stems in -au also retain a rounded element in all forms (skau- ‘touch’ SGS: 134, vanau- ‘become inactive’ SGS: 118; byau- ‘be’ SGS: 107) . There is similarly a rounded element regularlyFootnote ²⁴ after the contraction of the 3S enclitic pronoun -i with -u and -o:

o+i>vei, vai

D2 cannot be a rounded vowel.

D2 is not ei or ai

No noun stems end in either -ei or -ai. There is no synchronic evidence requiring any verb stem to end in -ai and no verb stem has been suggested as ending in -ei. Stems do end in -au, as mentioned above. One might expect, on the basis of symmetry, to find stems in -ai also. As shown above, when a vowel V is added to a stem ending in au, the result is -auyV, with both au and V staying distinct and with no loss of syllables. The same might be expected with an -ai stem, that is, *ai+V>aiyV, but there is no hint of such epentheses in the language. The D2 stems show V+ā>yā, with semivocalization of the first vowel, and a loss of a syllable. as in dV- + -āmä > 3Pp.a dyāmä ‘we see’. The alternate form, däyāmä, arises from the well-attested process of glide resolution (Hitch Reference Hitch2016b), not from a stem *dai-. It is unlikely that the D2 vowel could be a diphthong ei or ai. dyāmä > däyāmä ‘we see’, may subsequently restore a syllable and a mora. Synchronically the first stage is contraction with semivocalization. It is unlikely that the D2 vowel could be a diphthong ei or ai.

D2 is not ä

The D2 vowel contracts with ī giving ai (ei), e.g., 1Sp. daimä < dV- + -īmä. There are instances of ä+ī>ī elsewhere in the grammar (Hitch Reference Hitch2016a:4.4, 4.5.5). These involve type B stems in -ä-: dä- (SGS did-) ‘appear’, nä- (SGS näd-) ‘sit down’, and pajä- (SGS pajäd-) ‘ask for’.Footnote ²⁵ These stems show ä+ī>ī not ai, implying that D2 cannot be ä.Footnote ²⁶

The D2 vowel also contracts with ä giving ei>ai. I find no examples of a contraction ä+ä.Footnote ²⁷ However, we do have ä+i, which might be expected to have similar results. The 3S enclitic pronoun -i contracts with a final -ä producing -ī not ei > ai.

There are two other enclitics with the shape traditionally given as -ī. The deictic -ī appears to reinforce pronouns (Glossar 400). There is no extensive list of examples available so no analysis of the contractions is given here. The 2Sp.a of the copula and auxiliary verb ‘be’ is observed only in contracted forms. I find no evidence allowing us to determine a different shape,Footnote ²⁸ so I tentatively keep the traditional -ī. This enclitic also contracts with final -ä giving -ī:

The evidence of the type B stems in -ä- and of the enclitics -i or -ī show that D2 cannot be ä.

D2 is e

By a process of elimination, the last candidate for D2 is e. This would be an abstract or underlying shape. There is no e on the surface in any inflected or derived D2 present stem.Footnote ²⁹ But there is contraction evidence. The two enclitcs with the shape -i or -ī offer confirmation of e as the D2 vowel. They show a contraction e+i>ai or e+ī> (ei).

3S enclitic pronoun -i

2S enclitic verb -ī ‘be’

Three other enclitics lend support to the identification of -e- as the D2 stem vowel. The conjunction u ‘and’ has an enclitic form -u. The 2P pronoun uhu has an enclitic form which is traditionally given as -ū. The emphatic enclitic particle is similarly traditionally given as -ū. These latter two enclitics may have a short vowel but as there is no decisive evidence on that point -ū is retained here. The conjunctive -u is attested after -a, -u, -ä and -e.

Conjunctive enclitic -u ‘and’

The conjunction -u contracts with -a to -au, -o, and with both -u and -ä to -ū-:

u+u>ū

ä+u>ū, o

As far as I can discern, the enclitic conjunction -u is attested only one or two times with final -e:

2P enclitic pronoun -ū

The 2P enclitic pronoun -ū ‘you (pl)’ is similarly attested after -a, -u, -ä, and -e. The results of contraction are identical after -a, -u and -ä with those for the conjunctive enclitic -u:

u+ū>ū

ä+ū>ū

But with final -e we have two behaviours. Sometimes there is e+ū>au (o), which is parallel to e+ī>ai, but sometimes with the negative adverb ne ‘not’ there is e+ū>yau (yo) which may show a process similar to the semivocalization characteristic of the D2 stem vowel.

e+ū>au, o

e+ū>yau, yo

Emphatic enclitic particle -ū

Emmerick (St. III: 27) was agnostic about whether the shape is -ū or -u. He offered examples of this particle suffixed to ne:

He also noted, “when it contracts with ce the result is cau or co”, without references. I can only find the starred reading *kyau Sgh[10]53.2. The instances in Z of these spellings reflect the 2P enclitic. But, if ño = nyau, the patterns are consistent with the conjunctive -u and 2P -ū enclitics.

The contractions e+u>au>o and e+ū>au>o are parallel to the D2 contractions e+ä>ei > ai and e+ī>ai. In both sets there is e + high vowel > low-to-high diphthong.

Type A verbs in -e-

With the stem vowel of the D2 verbs identified, the attested forms of uysde- ‘look up (at)’, de- ‘see’, bāysde- ‘observe’ and yse- ‘be born’ may now be described. They exhibit several processes. There are contractions e+ī>ai and e+ä>ei > ai, and there are actually three environments producing orthographic y. There are many examples of semivocalization e+ā>yā, two of diphthongization e+a > ya, and one of diphthongization e+e>ye.

The contraction e+ī>ai phonemically involves dissimilation. A mid + high vowel sequence becomes a low-to-high diphthong /ē + ī > ai̯/. Phonetically, non-final e is [ǣ]Footnote ³¹ and the development [ǣ+ī>ai̯] involves less phonetic change of vowel quality than one might assume from the orthographic or phonemic transcription.

e+ä>ei>ai

The contraction e+ä>ai is less straightforward. Phonologically /ai̯/ results ultimately from /ē+ĕ/. In contraction the first vowel appears to lower, and the second appears to rise. However, as mentioned above, the first vowel was phonetically [ǣ]. Further, as regards the second vowel, there is evidence suggesting that in the oldest period the contraction was e+ä>ei /ē+ĕ>ae̯/, with the second vowel not changing height as it became an offglide. Manuscript 4 of the Saṅghāṭasūtra has 3Sp.m yseiye ‘he is born’ Sgh[4] 56.5.Footnote ³⁴ That manuscript is described by Canevascini as having the ‘old orthography’. It also appears to have older morphophonology, cf. NSm bal[ys]ūñavūysei ‘Bodhisattva’ Sgh[4] 54.3 and āchänei ‘ill’ Sgh[4] 55.2 showing a+ä>ei. The history of this D2 contraction appears to be that it originally was e+ä>ei, that is, mid vowel + mid vowel > low-mid diphthong, phonetically [ǣ+ε>aε̯]. After the merger of ei with ai, the contraction was e+ä>ei > ai. The 3Pp.a spellings deindä ‘they see’ in Z (4.91, 9.11, 21.32) may reflect the suffix -ändä, i.e. de- + -ändä > deindä. This word in Śgs is daindä (3.4r2) which would reflect instead the more common -īndä, i.e. de- + -īndä > daindä.

A point of discussion involves the 2S form dai. It occurs three times in the Saṅghāṭasūtra. There is agreement that two of the instances, Sgh[10] 216.3 and Sgh[22] 256.13 dai thu? ‘Do you see?’, are indicative (SGS 47; Sgh 267a). Emmerick considered optative as a possibility for the third instance, Sgh[4, 10] 53.4 ku buro gyastānu gyasta balysa dai ‘wherever you see Lord Buddhas’. He listed a suffix -ä < OIr *aiš and suggested a derivation < *dayä (SGS 207). The position here is that Emmerick was correct both that the form of the 2Sop.a suffix is -ä, and that the third instance of 2S dai is optative, though his derivation needs to be rewritten synchronically. In the synchronic analysis 2S dai is formally ambiguous deriving either as 2Sp.a dai < de- + -ⁱä, or as 2Sop.a dai < de- + -ä.

Semivocalization

The semivocalization of e+ā>yā /ē+ā>yā/ was the pattern originally used above to separate the D2 from the D1 verbs. These forms may also show glide resolution to iyā, äyā. Here also are included the āmatā-nouns showing -ye- in the GDS and IAS (like ysyemate) since they undergo e+ā>yā before undergoing umlaut of ā>e.

Diphthongization

There are two cases of the diphthongization of e+a>ya /ē+a>i̯a/ as well as one case of e+e> ye /ē+ē>i̯ē/ which is only attested in a diphthong resolved form, i.e., e+e>*ye>iye /ē+ē>*i̯ē>iyē/.