Khotanese orthography and double tt, gg, ss, ss
In the orthographic system of the majority of Old Khotanese texts there are four cases where a double consonant graph is used for a single phoneme, tt=/tl, gg=/gl, ss=/s/, ss-/s/. This doubling avoided ambiguity as it allowed the singly written consonant graphs to have dilferent values, t=/d/, g=/y/, s-lzl, s-lzl. This may be called the classical Old Khotanese orthography. Some documents survive in an older orthographic system first noted by E. Leumann in his 1920 edition of the Eurangamasamadhisutra, where he remarked that, "die Laute g s s nur selten doppelt geschrieben werden" (Neb: 92). This oldest attested spelling system may be called archaic. In it the three graphs mentioned by Leumann are ambiguous, with g=/g, y/, s'=/s, z/, s=/s, z/. But tt for /t/ is already in the oldest Khotanese documents.
The archaic system is amply attested. It is featured in manuscripts 1 to 5 of the Sanghatasutra. In his edition of that text G. Canevascini provides a useful discussion of the relationship between the archaic and classical systems (not his terms). Importantly, he shows that the paleographic age of the script of a particular document parallels the age of its orthographic system. This confirms that the archaic precedes the classical (Sgh: xv-xvii). A unique folio from The Book of Zambasta (Z) found in Sorcuq near Qarasahr, T III S 16, also features old script and archaic orthography. Using L. Sander's paleographic criteria, M. Maggi recognized that the folio is "the oldest witness of Z and altogether one of the earliest Khotanese manuscripts" (Maggi 2004: 184b). (1) No edition of the folio has been published,
Documents in archaic orthography also show the most conservative grammar. Leumann noted that in the Eurangamasamadhisutra the nominative singular masculine (NSm) of a-stems was always -a, and the genitive-dative singular (GDS) was always -i (Neb: 92). In later texts these are confused. Similarly, he noted that the third singular present indicative active ending -ata is kept distinct from middle -ate. R. E. Emmerick later confirmed and expanded these findings in his edition of the text (Sgs: xix-xxi). T III S 16 also shows this conservative morphology, consistently using NSm -a for the a-declension, e.g., vairocana (= vairocani 13.10) and sravaka-yana (= sravaka-yani 13.16), as well as NSm -ei twice for the aa-declension, an archaic feature noted by Emmerick, buddhavalamtsei (= buddhavalamtsai 13.13) and barei (= barai 13.13). (2)
While the oldest attested Khotanese orthography, the archaic, does not use double gg, ss', or ss, it does use double tt. The device was introduced in an earlier, unattested period of orthographic development. The three other double consonant graph spellings entered the system after the archaic period and are distinctive features of the classical system. They were likely designed on the model of tt, which makes determining the origin of tt more important.
Writing Khotanese dental obstruents
Ernst Leumann seems to have been the first to correctly note the functions of single t and double tt in Khotanese orthography. Single t is mostly /d/ except in some clusters like st where it is It/. He thought this was due to sound change within Khotanese wherein intervocalic voiceless consonants became voiced:
so hat t zwischen Vokalen den Wert von d bekommen, ist aber doch, weil bekanntlich die Orthographie leicht hinter der Lautentwicklung zuruckbleibt, weiterhin als t geschrieben worden, worauf man denn, um von dieser Pseudo-Tenuis die wirkliche zu unterscheiden, fur letztere die Verdoppelung des Zeichens einfuhrte. (1912: 38) That is, he proposed that after t had become /d/ intervocalically, tt was introduced to distinguish the voiceless phoneme /t/ from the "pseudo" or orthographic voiceless one.
The use of tt for [t] is easily seen in Khotanese in loan words. The Indian word ttathaggatta-(Z 5.103) from Skt tathagata shows tt for single [f] initially and intervocalically. Leumann also recognized that the signs for Sanskrit voiced stops, b, d, g, could stand for voiced fricatives /[??], [??], [??]/ (his symbols b, d, g; 1912: 42).
These values have essentially been followed by all scholars. In his last statement on the value of the dental graphs, Sten Konow in "A Primer of Khotanese Saka" wrote,
Of the dentals t is a voiceless stop only after s and s; ... Intervocalic t was voiced, i. e. d. Elsewhere tt is written for the voiceless dental stop.... D was the voiced dental stop after consonants and apparently before r.... Elsewhere d was a spirant, English th in "thou." (1949: 15) Konow deviated slightly from the Leumann values with his view that dr was /dr/ rather than /[??]r/. This is not a large difference, as there is no contrast between /d/ and /[??]/ before /r/. It may not be possible to choose between those values for Old Khotanese but it is also not critical to do so.
Gercenberg (1965: 47-51) had tt=[t] and d=[[??]] but t=[[??]], probably because he thought [d] was written dd. For the latter value he gives two examples, both Indian loans but each having a different explanation. The first, ssadda- 'faith', reflects Skt sraddha- and simply had double /dd/ as shown from its metrical use in the The Book of Zambasta (Z) where the first syllable is always heavy. In the examples below the word occurs as H(eavy)L(ight) in cadence 1 HLLHL, first as HLLHL and then as HLLHL:
13.160ab padamjsi I hettu bodhi-citta  u vayslnamjsiya ssadda . (A:5+12+7) (3)
23.370ab ttana ma pratalbimbai vlri ml kye ma udisa I ssadde jsa yande (A:5+7+5+7)
The stem is also spelled ssada- with single -d- six times in Z, a practice I call "hidden double d." The practice is clear from the Khotanese spelling of what is Suddhodana in Sanskrit, which occurs eight times as ssadutana-, where the meter proves that singly written -d- stands for [dd] in every instance (Hitch 2014: 34-35).
The other loan, reflecting Skt mudita- 'joy', appears once in Z with dd as NSf mudditta 10.14 but also once with d as ASf muditto 15.124. In both cases the metrical count is LLL, precluding an internal consonant cluster. (4) Further, the related form pramuddtto occurs twice with single d (10.12, 16.60). Emmerick was certainly correct to observe, "No doubt the doubling of d in mudditta was an attempt to represent [d] unambiguously at a time when d had become  between vowels, but it was a sporadic effort" (1981: 187). Gercenberg's first example has dd=/dd/ while the second is an effort to write [d] between vowels in a loan word where the Skt spelling with d was likely known. Neither example proves Gercenberg's value for Khotanese.
Although Emmerick 1981 (185-88) essentially follows the values of Leumann, he only refers to Konow and Gercenberg. His description for the motivation for the introduction of tt may be seen as a paraphrase of that given above from Leuman: "tt was invented to represent [t] because t had developed in intervocalic position to [d]" (1981: 186). Also like Leumann, but not like Konow, he regarded dr as [Sr] (1981: 188).
Emmerick 1989 (213-14) continues these values. His chart of initial consonants that were retained appears to contain a confusing error. In the line "/t/ /d/ It'/ spelled tt d th" (126.96.36.199.3.3.2) the /d/ should likely be replaced by /8/ as shown by his phonemic transcription "drainu /Srainu/ 'of three'" two paragraphs later (188.8.131.52.184.108.40.206). He does not specifically discuss values in clusters or between vowels but he uses /8/ for intervocalic orthographic d in the transcription "padajs- /pa[??]adz/ 'to burn'" (220.127.116.11.18.104.22.168).
E. Leumann in 1912, Gercenberg, Konow, and Emmerick agree on tt=[t] and, between vowels, t=[d]. All but Gercenberg agree on r/=[S] in initial and intervocalic position (except for Konow's dr-=[dr-]). The consensus view on these segments implies that there was a shift in the values of the Brahmi symbols relative to Sanskrit: Skt i=[t] > Kh t=[d], and Skt d-[d] > Kh d=, This shift made it necessary to adopt or develop a new device for Kh [t].
Earlier ideas on tt
Until now perhaps the most accepted explanation for the introduction of the tt-device has been from Oskar von Hiniiber. In "Die Paisaci und die Entstehung der sakischen Orthographie" (1981) he outlined the evolution through Middle Indie of the intervocalic stops originally written with t, d, and k, g. Although this is not his formulation, the sense is that the segments weaken over time along the hierarchies [t > d > [??] > y > [??]] and [k > g > [??] > y > [??]]. As additional proof he pointed out that g, d, and t in Khotanese can represent lyl, 181, and /d/. He concluded, likely correctly, that this use "kann die Lautwerte des indischen Alphabets in Indien selbst zur Zeit seiner Ubemahme widerspiegeln" (p. 124). About the "-device in Khotanese he suggested the cases of double tt writing in Prakrit could have served as a model (p. 125).
Emmerick 1981 (186-87) proposed a contrasting idea:
tt representing the result of the secondary contact of two OIr. dentals, e.g. butte 'he knows'
The view proposed in the concluding pages of this paper is partly a blend of the ideas of von Hiniiber and Emmerick, which are complementary rather than mutually exclusive. The Middle Indie model provided VttV=fVttV/, which was initially applied by the Khotanese to their /VttV/. From that use it spread to writing single III.
My interest in this problem arose from studying the synchronic morphophonology of Khotanese present stems. In the early 1990s I came to the conclusion that stems traditionally listed as ending in *-d- or *-h- actually ended in a vowel in the synchronic grammar. This development was due to a morphophonological reanalysis or metanalysis that also led to the development of suffixes beginning with /tt/. Briefly, a word like 3Sp.m butte 'he knows', historically *bud- + -te, was reanalyzed to bu- + -tte. It was essential for this theory to find other...