oe is expressed by the word 7 na: aaeft navadati he does not say, he is not saying Sandhi a ee a word that we spell only i in one way may be pro- jounced differently according to its position in a sentence. Thus the definite article ‘the’ is pronounced with a neutral vowel (30) before consonants, ‘the man, the hill’, and with a short i vowel (81) before vowels, ‘the owl, the end’. Moreover, in separating this word out in order to talk about it, we may use another, a lengthened vowel, and say ‘the definite article 51. We have examples like the English non-standard ‘doam be stupid’. Here the word ‘don’t’, having lost its final t, changes its 2 to m, which is more like the following b (put more technically, the alveolar nasal # is replaced by the bilabial nasal m before the bilabial stop b). A differe nt sort of example involves not a separate word but a ‘morpheme’, for example ‘plural s’: so, the plural of ‘cat’ is ‘cats’, ue the plural of ‘dog’ is ‘dogz’, although in standard spelling of the latter the same letter, s, is used—the general rule is that voiceless s is added to voiceless consonants (‘cats’, ‘pups’) and voiced z to voiced consonants and vowels (‘dogs’, ‘toes’), except that if the word itself ends in a sibilant the suffix takes the form ‘-iz’ (‘bases’, ‘phrases’). Similar sound changes occurred in Latin words, as the spelling of their English derivatives will indicate—compare for instance ‘induce’ and ‘conduce’ with ‘im- press’ and ‘compress’. The reason underlying such variations is one of euphony or ease of utterance, the fact that what is a convenient sound in one en- vironment may not be at all convenient in another. The phe- nomenon is referred to (very often even in talking about languages other than Sanskrit) by the term sandhi, a Sanskrit word meaning ‘juncture’. What we are concerned with at pres- ent, as in the first two examples above, is external sandhi, i.e. the changes in the appearance of complete words when they come together, or make a juncture, in a sentence. In Sanskrit these changes were particularly widespread and striking, which is why the word sandhi has become generally current among phoneti- cians. They were fully analysed by the ancient grammarians and are extensively reflected in the orth thography. This is not neces- sarily a good thing. Writing ‘tho’ and ‘t or ‘cats’ and ‘dogz’ in English would obscure the fact that a single word or mor- pheme is in question—though it would be marginally helpful to eo nN.zserdeyo 1B 20 foreigners in learning to pronounce the language. Beginners in Sanskrit, being more concerned with reading and writing than with pronunciation, will find the operation of the rules of san- dhi a considerable obstacle in the earliest stages of learning the language, but one that is fairly quickly surmounted because met with at every turn. Use of the sandhi grids The approach to the problem adopted in this book is primarily a practical one. Instead of attempting to master in one go all the phonetic principles involved in euphonic combination, you are encouraged to make use of Table 2.1, where all the relevant combinations of final and initial sounds are set out in tabular form. In addition, certain preliminary remarks to aid you in using the tables are here offered. When sandhi is made between two words, the first may end in a vowel or a consonant and the second may begin with a vowel or a consonant. Four main classes of sandhi are thus distinguished. 1 Vowel + vowel ‘When two vowels come together they coalesce, often into a single vowel. The body of the vowel grid represents the combination of the two vowels. Thus @ tatra followed by ¥@ iva is written as ‘We tatreva ‘as if there’. Most of the features of vowel sandhi will make sense if you remember that historically the Sanskrit diphthongs e, o represent ay, av (or ai, 4u) and the diphthongs ai, au represent ay, av (or ai, au). Hiatus is not permitted, in the sense that when the original vowels come together the appropri- ate rule of sandhi must be applied; but secondary hiatus is per- mitted, in that the resultant sandhi may contain two distinct vowels, So @% vane + ¥ iva results in @4 ¥@ vana iva ‘as if in the forest’ (by way of vanayiva, with elision of the y), and this remains and does not further combine into *vaneva, 2 Vowel + consonant The simplest of all possibilities. The words remain unchanged, with one very minor exception: if the vowel is short and the following consonant is ch, this ch changes to ®&.cch: so 7 feaaftt na cchinatti ‘he does not cut’. If the vowel is long, the change is optional (except after the words 81 4 and @T ma, when it is again obligatory): Wt fat s4 chinatti or Ht fEsate s4 cchinatti ‘she cuts’.“$4.0 yo 28x anp ur adaoxe ‘pou udeq seq suo ang ‘ojqisstuLIad ose sIypuEs aAREUTATE sNOLIEA :370N “seme + fe sem adooxg , “wOYs st famoa Suypasord ays wayan ssm200 Zuyjqnop ILL ¢ “waseqqe] [e) manfun = easeyge] + UF “39 , ‘Wye 03 suatpSuaj sp ‘sapscasd “nyye jt pue ‘sreaddestp 10 @ , © | & q a a a 2 g 3 os muy WY, P ° fe 1 a a g PP ae = 4 a a ¥ a i 4 ae 4 a (Pre 8 q i 4 ° g 1 & a g q - 8 ° g 049% a w a q Pe P 8 ° g x wo u e q P e 3 ° ® 2 a a a a o g ° g 2 a u a q Pp e 8 re 7 4 ao u ¥ d 2 i 4 ° ' 1 i a ¥ q Pp ? a se = s a six v 4 2 a 1 ° g a o a e q & é 3 ee ; ar ia v a i i 4 ° g a a u g q { e 3 e ® 5 a sis v d 2 i x o e a oo w v q P é 8 ie | iw i a 8 ¥ 4 1 i 4 ssuona pump ete afi (fete adzorq) a v a 3 i 1 alana spoil] poggnuiag -— — ~~~ ~~~ == =~ an (eran Bxmogo) » ha uoryrn wasol og axrzipin sinizy porzyooug) swseuosueD spu6 jypues 42 eiqeLmea mek ne 0 ok ae mw wh we a 2k r i jk 3B u wa B 1 a we q| a wh z (-aduoe znpu age us 11200 pynom sovds » qons a1oqum seuouiape oma u2amag ya} 5112048 Y) SMOA spuB jypues 12 eel3 Consonant + consonant Here, the body of the consonant grid represents the form that the last letter of the first word assumes before the following con- sonant. In Sanskrit a word may end only in a vowel or in k, t, t, p, i, n, m, r or h (on these last two, see below), which is why the grid is not even more complicated than it is. In the body of the grid a bracketed sound indicates a change in the form of the fol- lowing initial: thus 4@,tat + Yt sariram becomes Test tac 4 Consonant + vowel The possibilities of this are represented by the penultimate hor- izontal column in the consonant grid. §& Zero There is one further possibility. One word instead of being fol- lowed by another may occur at the end of a phrase or sentence, i.e, ‘before zero’. In this position the basic form of the word re- mains without change. To put the matter the other way round, the form that a word assumes by itself or at the end of a sentence has been selected as the basic form: so ramaniyam vanam ‘the forest is pleasant’. There is, unfortu- nately, one exception to this rule: Sandhi of final r or h (If the following account seems discouragingly complicated, re- member that it is only provided as background explanation; what is important at this stage is simply knowing how to use the sandhi grid.) Visarga (bh) is the last letter of many Sanskrit words as they appear at the end of a sentence. It may represent one of two original (Indo-European) sounds, s and r. Thus, from s, 3%. eee cf equ juus) ‘horse’; HM: asthah (cf. éstés) ‘you stood’; aft: is) ‘going’. But, from original r, At: matah ees ih ‘O mother’; gf: dvah (cf. the English cognate) ‘door’; ‘Bq: catuh (cf. quattuor) ‘four’. The sandhi of these words is complicated by two factors: first, by whether the visarga originates from s or from 7; secondly, by the vowel that precedes the visarga. We can eliminate the first factor, and so reduce the confusion, by taking r as the basic let- ter in the comparatively few cases where h derives from 7 and re- serving h for the cases where it represents original s (so aévah, asthah, gatih; but m&tar, dvar, catur). (In practice, however, final r may be reserved for instances of ar or ar alone, since its sandhi i20 1 when preceded by any other vowel is identical with the sandhi of final h, and therefore a distinction in these cases could be a only after an etymological inquiry, and not always even nt hered basset 0 esaneti fnthnc sae (see grid). ly, F appears. re a wore with a voiced sound, s or sore other unvoiced sound before a word beginning with an unvoiced sound. Furthermore, this is the sandhi of final r even after a and 4 (matar, dvar). But after a, final h is lost before voiced sounds, and words ending in ah change ah to 0 before voiced conso- nants. Before all vowels except short a, ah becomes a: thus 3: + ¥@ aévah + iva becomes 8 ¥4 aéva iva ‘like a horse’. In com- bination with an initial a, ak becomes 0: thus 3t@: + fet aévah + asti becomes Sift asvosti ‘there is a horse’. In modern print- ing this last sandhi is generally represented as ara sf& avo Ssti, faith die aeacraba (5) representing the disappearance of an ini- tial short a. Lr are some further examples of the operation of sandhi les: aft api+ asraresfa avagacchasi = stereaTesfa apy avagac- chasi do you understand? WY onanu + sufeena: upavisamah = TAMA: nandpavisa- mah well, we are sitting down ‘svit_ubhan + arrest: agacchatah = SwraTresa: ubhavagac- chatah both are coming katham + BRA smarati =e orf katham smarati t, he remembers? WL tat + Walt jayati = aft taj jayati he is winning that Re dvi + rae hasati = fgyaft dvid dhasati the enemy Us AFL tan+ J tu= Ay tamstu them however We: narah + Taft raksati = 72 Taft naro raksati_ the man protects ‘FR punar + Taft raksati = FT Tat pun’ raksati again he protects +ammesft agacchati = wearTestt Eee! Sagas bal Goes LagiIt will be observed in the above examples that frequently the nagari script cannot show where the first word ends and the sec- ond begins. Spellings such as 80] arerreste apy avagacchasi are never found, except in one or two texts intended for beginners. In transcription, on the other hand, the words can usually be sepa- rated out. But they still cannot be so where two vowels coalesce into a single vowel, and in such a case, furthermore, there may be considerable ambiguity as to the original vowels: 4, for instance, might represent a+ a, a+4, 4+ or 4+4. In the system of tran- scription used in this book, these difficulties are overcome by the use of the signs » and », which for convenience may be thought of as marks of elision. The former stands in the place of an original short vowel and the latter of an original long vowel. They always stand in place of the first of the two original vowels, except that > is used like the avagraha in the nagari script after ¢ and o and also after . A circumflex over the sandhi vowel indicates that it is not the same as the original second vowel (see Table 2.2). Table 2.2 and vowel aot ft wu @ e wf 0 aw a rae 0 € 6 00 ai ai adam a eA »& 98 »6 95 vat wai vai » au i i ot oT T >to» Bu oa 8 a ou ood Instead of a circumflex, a macron is used over e and 0 to dis- tinguish instances where the second original vowel was long. So Tnat icchati = Teaft w écchati ‘he does not want’, but ‘Ana + FE iksate = TW mw ekgate ‘he does not see’. Note that » always represents a and » @ except in the union of two like simple vowels (namely, f+ fi+t,f+#, the last being very rare). In the early lessons, where Sanskrit is given both in nagari and in transliteration, the sandhis of the nagari text are usually resolved completely in the transliterated version. When this is done, the transliteration is put within brackets to show that it is an analysis and not an equivalent: e.g. @#eSt tat na icchati for tan m écchati ‘he doesn’t want that’. No account is taken, how- ever, of a mere change of final m to anusvara.