Printer Friendly

On the origin of the Middle Indic future suffix -hi-.


Middle Indic future formations can be classified into the following three types:

1) the future in -(i)ss-: e.g., P[]li karissati 'will do', P[]li dassati 'will give';

2) the future in [degrees] CCh-: e.g., P[]li lacchati 'will obtain';

3) the future in -hi-: e.g., P[]li k[]hiti (and k[]hati) 'will do'.

The first two formations are clearly continuations of the Old Indic future in -sya-l-(i)sya-: e.g., OI d[]syati (d[]-) > MI dassati; OI kartryati (kr-) > MI karissati; OI lapsyati (labh-) > MI lacchati. On the other hand, the origin of the future in -h- is not transparent.

As regards chronology and diffusion, the formation is already documented in the ASokan inscriptions (cf. Bloch 1950: 74f.), where it appears in texts written in the oriental variety (the so-called "chancellery language"): hoharpti (Delhi-Topra Pillar Edict VII) 'they will be'; dahamti (Pillar Edict IV) 'they will give'; ehatha 2p1. from i-'go' (Dhauli Separate Edict II); it also appears in the languages of the major Middle Indic textual traditions, i.e., Pali, Ardham[]gadh[], Jaina-M[]h[]r[]str[], M[]h[]r[]str[] (cf. Pischel 1900 [section] [section] 520-34), Apabhramga (cf. Tagare 1987: 36, 306-11), and Buddhist Sanskrit (cf. Edgerton 1953: 1490f.). (1) Moreover, in several New Indic languages (Bhojpuri, Maithili, Marwari, Awadhi, Bhili, Braj, Bundeli) there are future formations in -h- which are likely to be continuations of the Middle Indic -h- future (cf. Bloch 1965: 247, Masica 1991: 289). (2)

According to common opinion, the element -hi- is a special development of the sequence -sya-. However, adopting this assumption makes it necessary to explain two unexpected formal peculiarities of the morph: i.e., the appearance of a thematic vowel -i- (in competition with the usual -a-), and the presence of the consonant -h- instead of -ss- (from -sy-).


An attempt at a solution that does not involve the problematic development -sya- > -hi- is that offered by Vittore Pisani in his "Noterelle Pali" (1952: 285-87): according to him, the sement -h- of the suffix had been transferred from the 1st person singular of the 01 peri-phrastic future of the type hantaham (< hang-aham) 'I shall kill'. (3)

However, Pisani's argumentation leaves unexplained why the element -h(a)- in h[]ntaham should have been reanalyzed as a tense marker: even if we admit that h[]ntaham was resegmented as hant[]ha-m (or as hant[]h-am), this reanalysis, per se, might simply create a 1st person allomorph -t[]h(a)- of a tense marker -t[]-. Significantly, in Middle Indic the 1st sg. of the indicative can show a special byform in -[]ham (cf. Oberlies 2001: 217f; von Hintiber 2001: 277f.), but such a formation is not associated with the future tense. It should also be noted that Pisani's proposal cannot explain the appearance of the thematic vowel -i-. In conclusion, the hypothesis advanced by Pisani is not satisfactory. but his idea that the -h- of the future must he "di ragione morfologica anziehe thnetican deserves consideration.


The majority opinion (4) holds that the appearance of the vowel -i- has to be explained as the result of the so-called Middle Indic sampras[]rana--a sound change consisting of the transformation of a glide into a syllabic vowel. According to this assumption, the -i- in the suffix would he the continuation of the glide -v- contained in the original OI suffix -sect-. It should be noted, however, that the other future formations do not normally show the development -ya-> -i- even if they are indisputably continuations of the original type in -sya-. Indeed, among the other future formations, a thematic vowel -i- is found only in sporadic cases such as Moka va((I)dhi(s)siti (cf. Bloch 1950: 74), P[] gatichisilgacchisi, dakkhiti, sakkhinti (besides gatichasilgacchasi, dakkhasi, sakkh.ati; cf. Oberlies 2001: 248, 244 n. 1); as for the last three forms, moreover, it should be mentioned that, in competition with those, P[]li also shows double-suffixed or, more precisely. resuffixed forms like gacchiss ati, dakkhissati, sakkhis.vati (cf. Oberlies 2001: 248 n. 2; Bloch 1965: 227). so that the possibility cannot be ruled out that the type dakkhiti owes its 4-to the influence of the type dakkhissati.

As we shall soon see, an attempt at explaining the absence of the samprasarat:ia ya > i in the other future formations was made by resorting to a particular relative chronology of the sound changes affecting y. On the other hand, it should be noted that the very assumption of a samprasurana ya > i is somewhat problematic. Indeed, while such a phenomenon is relatively well documented for the glide w, in the case of y the shift from glide to nucleus seems not to be an expected development. On this issue von Hithiber (2001: 129) writes "Das Samprasarana ya: i ist im Mi. auBer in Futurformen [...] nur selten nachweisbar: Skt. abhvantam > P abbhantara, aber Amg abbhintara; Skt. tirvak: Amg usw. tirikkha, aber: CDIAL 5823 tiras. In alien anderen, in der alteren Literatur hierher gestellten Fallen liegt Sandhi, Analogie, Suffixwechsel and Volksetymologie vor." Worth noticing is the fact that in Pali the existence of this phenomenon seems not to be demonstrable without making reference to the future stems.


In his Zwei Probleme der mittelindischen Lautlehre (1955: 79), Hermann Berger tried to explain the presence of the consonant h by assuming that this sound is the result of a dis-similation process that would have originated in the 2nd person singular: a form like, e.g., *gamissisi would have developed into gamihisi in order to avoid the sequence [degrees] ssVs [degrees]; the stem gamihi-would then have been generalized to the whole paradigm. More recently \Tit Ruben& has also expressed himself in favor of such an explanation (1996: 105).

Nevertheless the supposed dissimilation process has no exact parallels in Middle Indic; moreover, the assumption of an analogical generalization of the dissimilated stem is not so obvious from a psychological point of view: indeed, it may be thought that morphological analogy would have rather tended to restore the undissimilated form in the 2nd singular. To conclude, the dissimilation hypothesis is quite an ad hoc explanation and cannot be considered a satisfactory solution.


Other scholars attempted to explain the -h- as the outcome of a geminated -ss- (which in turn is the expected MI outcome of the OI sequence -sy-) by assuming a series of two sound changes: 1) a degemination -ss- > -s-, and 2) a debuccalization -s- > -h-. But the fact that different scholars invoke different causes for the degemination process reveals the problematic nature of the assumption.

Ralph L. Turner (1931: 531; cf. also 1927: 232ff.), followed by Luise Anna Schwarzschild (1953: 42) and Thomas Oberlies (2001: 245), maintained that "MI. -ss- as a component of a suffix or termination might have, and in most dialects did have, a special development into -s- and -h-"; such a statement, however, disregards the circumstance that, with the partial exception of Magadhi and Apabhramga (cf. Pischel 1900 [section] 366, and below), the geminate -ss- is in fact preserved in Middle Indic even in a terminational element like the genitive ending -assa < -asya. (5)

It should be added here that, as regards Pali, metrical pressure has been held by Oberlies to be an additional factor favoring degemination: thus in the line adisv[] k[]lam | kariss[]mi Ubbarim (Ja III 161,14*) (6) the metrics require the second syllable after the caesura to be light (for other instances, see Oberlies 1996: 1151; 2001: 109, 245). Nevertheless, in our view cases of degemination forced by metrical necessity should not be considered as actual evidence with respect to the discussion of a phonological phenomenon: formal alterations that are due to metrics normally affect the textual tokens and can hardly involve the underlying phonological form of a word in such a way as to result in a true sound change. (7)

Along a different line of reasoning, Manfred Mayrhofer (1951: 42f) hypothesized that the degemination of -ss- took place at first in stems in which the geminated consonant was preceded by a long vowel. Indeed, in this phonological environment, the process could be viewed as a means of complying with the MI two-mora law (on which see von Hiniiber 2001: 117-18). This solution has the advantage of resorting to a true phonological constraint of Middle-Indo-Aryan (for cases of -ss- degemination accompained by compensatory lengthening in Prakrits, see Pischel 1900 [section] 63), and it bears mentioning that it is true that the suffix -hi- appears with relatively high frequency in stems where it is preceded by a long vowel (for a different interpretation, see below).

On the other hand, it should also be noted that forms that could be the intermediate stage of the assumed development -ss-> -s-> h--i.e., future forms with a degeminated s preceded by a long vowel are not as well documented as might be expected. Sporadic occurrences like the k[]sam (P[]li) attested in, Ja Iv 286,21 * . . . 287,15 * and vi 36,20 * (but cf. kassarn Th 381, S I 179,8; kass[]mi Th 1138-39; cf. Geiger 1916: 125f) or d[]s[]mo and esaipti (Ardham[]gadh[]) attested, respectively, in []y[]rangasutta II 5, 11-13 and in S[]yagadangasutta I, 11, 29 (cf. Schwarzschild 1953: 43) are, in fact, a too meager corroboration. Indeed, the normal Middle Indic future stem from a base such as, e.g., d[]-'give' is dassa-(cf. Geiger 1916: 124 [section] 151) with vowel shortening (due to the two-mora law) and preservation of the geminated consonant.


A detailed attempt at explanation concerning the origin of the suffix -hi- has been offered by Oskar von Hinilber (2001: 298f.), developing an idea proposed by Paul Tedesco (1945: 1581; see also Edgerton 1953: 149). He assumes that the consonantal assimilation -sy- > -ss- took place only after the sampras[]rana -ya- > -i-. The advantage of this relative chronology lies in the possibility of dispensing with the assumption of a degemination process: according to this picture, the sequence -sya- with short a undergoes sampras[]rana and becomes -si- so that no assimilation -sy- > -ss- can take place. Therefore the assimilation -sy-> -ss- could only occur in the forms of the paradigm where the thematic vowel was long--i.e., in the first persons--since before a long [] the glide y could not undergo sampras[]rana. As a consequence, a stage is posited in which the first persons had -ss[]- (< -sy[]-), whereas the other persons had -si- (< -sya-). Subsequently the forms with -si- underwent the change -s- > -h-, which, according to von Hinuber, preferably occurred in the position between two identical vowels (2001: 178). Indeed the author assumes that the morph -hi- appeared at first in the future stems of the set type in -isya- through the development -isya- > -isi- (sampras[]rana) > -ihi- (debuccalization of s between identical vowels).

As regards the last point, it should be noted that, if it is true that in a Prakrit language like M[]h[]r[]str[] the -ihi- futures are actually frequent (cf. Weber 1870: 63), the evidence from Pali appears not to confirm the idea of a preferential appearance of the morph -hi- after a vowel -i- (cf. the forms quoted in Oberlies 2001: 246). Moreover, on the basis of the assumption that s between two i's is regularly represented by h, it would be necessary to think, as von Hiniiber suggests, that in the 2nd singular the hypothesized -isisi (< -isyasi) became *-ihihi and that the 2nd person ending -si (cf. the attested forms in -hisi) was restored secondarily in order to avoid allomorphy.

It should be stressed here that, according to the theory we have just described, no -h- is phonologically expected in the first person forms. For this very reason the distribution of -h- and -ss- future stems in classical M[]h[]r[]str[] has been considered a confirmation of the validity of the entire explanation. Indeed, in the Sattasa[] by H[]la the future in -h- is normally used for the 2nd and 3rd persons (ten and six instances, respectively, as against one instance for the 1st person), whereas the -ss- stem is normally used for the 1st person (nine instances, as against zero instances for the 2nd person and one instance for the 3rd person). However, as we shall see below, this situation is to be explained in a different way.


All the attempts at deriving the Middle Indic future suffix -hi- from the Old Indic suffix -sya-I-(i)sya- run into a major difficulty: the problematic nature of the assumed MI sound change s > h. In the Prakrits and Apabhramga this sound change is a sporadic phenomenon. Not only has no sound law capable of predicting the distribution of the forms with s > h been defined, but there is no consensus among scholars as to what phonological enviroment could favor--let alone cause--the occurrence of the sound change. (8) Pischel (1900 [section] 151) and Tedesco (1945: 155-61; cf. also Bhayani 1997: 31) discuss roughly twenty lexical items that show or could show such a development in one or more MI varieties, and it is indeed revealing that Paul Tedesco himself, in an argument aimed at defending the validity of Middle Indic s > h, could not help remarking that "the Middle Indic change of intervocalic s to h is everywhere exceptional" (1945: 157).

A stronger tendency to reduce intervocalic s to a glottal fricative seems to be shown by later Middle Indic (Apabhramga) in grammatical morphemes. However, in this case as well, some of the alleged instances are not unproblematic. So in Apabhramga the -a- stems show a genitive singular ending -aha/-aho/-ahu (cf. Tagare 1987: 129-31; Bubenik 1998: 66), which appears to be a continuation of OI -asya. However, an alternative explanation for this form was proposed by Jules Bloch (1965: 142; cf. also von Hintiber 2001: 178), who assumed a transfer of the -ha contained in the genitive form of the 1st person pronoun maha (from the CH dative mahyam) already attested in the Prakrits (cf. Pischel 1900 [section] 418). (9) Among the Prakrits, M[]gadh[] already showed a similar genitive ending -[]ha in competition with the expected -als'a (< -asya, cf. Pischel 1900 [section] 366); if we adopt Bloch's hypothesis for M[]gadh[] as well, the length of the thematic vowel can be explained not as compensatory lengthening, but as analogical to other forms of the same paradigms such as the ablative -[]do or the dative -[]a (cf. Pischel 1900 [section] 363).

The existence of Apabhramga genitive plural forms in -[]h[] of -[]-substantives (cf. Pischel 1900 [section] 264) has been viewed as another instance of s > h: the ending would have transferred from the pronominal declension, where it would be interpretable as a continuation of 01 -[]s[]m (cf. Tedesco 1945: 156). However it should be noted that in Apabhramga all the oblique plural forms of both the -a- and the -[]- declension show -h- endings (10) (cf. also Bloch 1965: 69), so that the whole inflectional subparadigm of the oblique plural has been restructured in such a way that -h- seems to function as a sort of demarcative segment positioned between stem and ending. (11)

Also etymologically ambiguous is the Apabhramga 2nd sg. indicative ending -hi, which may be considered a reflex of the ending -si showing the change s > h (with Pischel 1900 [section] 264) or as an expanded usage of the imperative ending -hi (with Bloch 1965: 244). (12)

Apart from the assessment of the single grammatical morphemes mentioned so far, the assumption of a change s > h for the etymological explanation of the MI future suffix -hi- shows difficulties not only due to the sporadic nature of this phenomenon but also, and more importantly, due to its chronology: indeed all the alleged evidence is taken from Prakrits and Apabhrarriga, whereas there are, to my knowledge, no instances--leaving the -hi- future out of consideration--belonging to the first stage of Middle Indic, i.e., to the languages of the Agokan inscriptions and P[]li. (13)

Moreover, an observation concerning New Indo-Aryan made by Jules Bloch (1965: 69) should be stressed: "the modern future in -h- [...] is widely distributed, even outside the areas in which sibilants normally open." Indeed, according to the survey by Colin Masica (1991: 206f.), among the New Indo-Aryan languages the change MI s (< OI s, s, s) > h is observed 1.) in Sinhalese-Maldivian; 2) in some languages of the northwestern zone, i.e., in Sindhi, Lahnda dialects, and, among the Dardic languages, sometimes in Kashmiri; 3) in some languages situated in the western part of the central zone, i.e., in North Gujarati, Western Rajasthani (S. Mewari, NW Marwari), the Bhili dialects of the Rajasthan-Gujarat-Madhya Pradesh border areas, and--sometimes--Panjabi; 4) in some languages of the Bengali-Assa-meSe group (eastern zone), i.e., in Assamese, North Bengali, and--sometimes--Bishnupriya.

On the other hand, the languages in which the -17-future is observed cover an area that is predominantly extended over the central zone (cf. Masica 1991: 289); among these are not only Marwari (belonging to the Rajasthani group) and Bhili, but also languages situated in the middle and eastern parts of the central zone, such as Bundeli, Braj, Awadhi, Bhojpuri, and Maithili. Therefore, it may, at least, be said that New Indo-Aryan dialectology provides no evidence supporting the derivation of the -h- future from the -s(s)- future.

We can conclude that the hypothesis that the suffix -hi- is a continuation of -sya- is highly problematic from the point of view of historical phonology and should only be embraced if no other satisfactory explanation can be found. To our view, however, a different account is indeed available.


It can be argued that the Middle Indic future suffix -hi- originated from a reanalysis of the 2nd sg. imperative ending -hi. Reanalysis phenomena concerning 3rd sg. endings are well known and constitute instances of what is often called "Watkins' law" (cf. Collinge 1985: 239f.) on account of the theoretical discussion given by Calvert Watkins (1962; 1969). Thus, in the Old Irish t-preterite the tense marker -t- is the reflex of an original 3rd sg. ending; analogously, the imperfect stem oed of the Middle Welsh verb bod 'be' is the reflex of an original 3rd sg. form (Watkins 1962: 90-96, 171ff.). Other instances of this kind of reanaly-sis have been found in Romance languages: in the dialect of Clermont-Ferrand (Proven al) the past tense stem cantet-(cf. cantete 'I sang') of the verb for 'sing' has been created on the basis of an original third person form cantet in which -t constituted the personal ending; in Swiss Vallader Romantsch an almost identical process of reanalysis created the past definite of the type chan-et-an 'I sang' (see, with references, Koch 1995: 33). Such phenomena result from two cross-linguistic principles: 1) the semantic unmarkedness of the third person, a characteristic that was pointed out already by Roman Jakobson (1932) and Emile Benveniste (1946), and 2) the tendency of languages towards constructional iconism (s. Mayerthaler 1981: 25), according to which an element which is more marked semantically is also likely to be more marked formally. As a consequence, not only do 3rd person forms often exhibit a zero ending, but it is also not rare for a 3rd sg. form that was originally provided with a non-zero personal ending to be reanalyzed as zero-marked for person and number, while the original personal ending is reanalyzed as a tense or mood maker or, more generally, as a part of the stem.

Now, as Harold Koch (1995) has stressed, the distribution of morphological zeroes can be conditioned by so-called local markedness phenomena, i.e., phenomena in which, in certain sectors of a language's grammar, the normal markedness relations can be altered or reversed. In the imperative mood in particular, the unmarked person is the second and not the third (cf. Kurylowicz 1964: 241; Bybee 1985: 77).

Interestingly, an example of reanalysis concerning a 2nd person form is that of Greek [epsilon] [sigma] [theta] [iota] [omega]) 'eat', whose stem [epsilon] [sigma] [theta] [iota] -, according to the explanation provided by Karl Brugmann (1913), may be identified with the athematic 2nd sg. imperative [epsilon] [sigma] [theta] [iota] (Homer). It is note-worthy, that the Greek imperative ending -[theta] [iota], that underwent reanalysis is a continuation of the same PIE ending *-[d.sup.h]i that is also the ancestor of CH -dhil-hi, MI -hi. (14) Another striking parallel is offered by the Slavic verb idQ 'go' (< PIE *[h.sub.i]ei-/[h.sub.i]i-), whose present stem is to be explained (Kortlandt 1979: 52f.) as the outcome of a reanalysis involving the 2nd sg. imperative in *-[d.sup.h]i; cf. OCS. idi (with the secondary addition of a final morph -i probably originating from the PIE optative marker), Gk. i[theta] [iota], OI lhi.

The fact that parallels so exact can be found in the diachronic developments of IE languages is not due to chance. Indeed, in PIE morphology there was a language-specific condition that could favor the occurrence of such phenomena--the fact that the 2nd sg. imperative of the thematic inflectional macroclass already had a zero ending at the proto-stage.

The hypothesis that the future suffix -hi- and the imperative ending -hi can be etymologically identical is supported by the existence of MI homonymous pairs of an imperative form and a future stern: from e-'go' we have Pali chi (Th175) 2nd sg. imp., ehiti (la 153,18*) 3rd sg. fut; from ho-'be' we have Pali hohi (Sn 31, 44 in 134, 14), hohiti (Th 1137, Th[] 465); from ne-'lead' we have P[]li nehi (Ja ir 160,2), Patna Dharmapada (276; s. Cone 1989: 175) nehisi (2nd sg.). (15)

According to our hypothesis, the future stems appearing in these pairs must have constituted most of the original representatives of the -hi- future formation. This assumption is strengthened by the fact that verbs like e-'go' or ho-'be' are among the lexical items characterized by the highest frequency of usage, so that a morphological reanalysis affecting these verbs can easily produce a structural change such as the creation of a new stem class.

Moreover, it hears noting that eha- and hoha- (with thematic vowel -a- instead of -i-) constitute two of the three -h-future formations (the third is d[]ha-from d[]-'give') attested already in the Agoka inscriptions.


It is well known that in later Old Indo-Aryan the imperative mood inherited some of the forms (i.e., the first persons) and the functions of the subjunctive mood, which, at the same time, fell into disuse. (16) With respect to Sanskrit, J. S. Speyer wrote (1886: 273; cf. also 1896: 57): "The imperative is a kind of potential mood, expressive of possibility and doubt" (see also Renou 1961: 412f.; Oberlies 2003: 142 nn. 1-2). This value of the Sanskrit imperative is especially evident in interrogative sentences: compare, e.g., Uttararamacarita 1.190 pratyetu kas tad "who will believe it?" Such usages of the imperative mood are also found in Pali; an instance with a first person imperative (which continues the original subjunctive on the formal level as well) is found, e.g., in fa v 254,32*: atha kena nu vannena ubho loke labhamase "but in which way could we gain both worlds [the human and the celestial]?"

Now the development of a future tense value from a subjunctival usage is not a typo-logically rare phenomenon (cf. also Palmer 1986: 216-18; on related theoretical issues see Jaszcolt 2009: 50ff.); as is well known, there are a number of parallels for it and it will be sufficient here to recall the case of lat. er[] 'will be', which continues the PIE subjunctive. (17)

Yet, one fact still needs explanation: the original usage of the morph -hi as imperative ending was not eliminated by the occurrence of its reanalysis as a mood/tense marker. Indeed, in Middle Indic the imperative ending -hi is still functionally alive. We must conclude that during an earlier period an oscillation between the old and the new interpretation of the morph had been in progress. Subsequently, with the appearance of the forms in -hisi (2nd sg.), -hiti (3rd sg.), a lexical split must have taken place, since at that time the ending -hi and the suffix -hi- had different morphotactic properties.

In order to understand the grounds for the creation of the forms with -hi- plus indicative ending, it is useful to distinguish between illocutionary categories, which concern the sentence type, and true modal categories; indeed, the morphological category of mood may cover both these semantic areas (cf. Hengeveld 2004). Originally the form in -hi was a 2nd singular with an illocutionary imperative function, i.e., basically associated with commands. The addition of the indicative personal ending to the -hi- stem conferred on the form a non-imperatival illocutionary function. This effect can be explained as the result of a proportional analogy: in the 2nd singular of the thematic conjugation, there was a form with illocutionary imperatival function that had a zero personal ending (cf. the OI 2nd sg. imperative bhara 'carry!') and a form with illocutionary non-imperatival function that had a non-zero personal ending (cf. the CH 2nd sg. indicative bharasi 'you carry'); therefore, on the illocutionary level, the analogy bhara: bharasi = ehi: ehisi can be posited. On the other hand, as regards the true modal level, the newly created forms in -hist, -hiti, precisely because they contained the morph -hi- reanalyzed as a modal marker, could preserve the true modal (subjunctival) functions of the late Old Indo-Aryan imperative and accordingly be used for the expression of future time.


On the basis of our etymological proposal, some conclusions may be drawn with respect to the stem-formation rules associated with the future suffix -hi-. First, it should be noted that a stem like P[]li k[]hiti, whose corresponding present and aorist stems are, respectively, karoti and ak[]si, reveals that, as already suggested by Jules Bloch (1965: 228), the sigmatic aorist stem, deprived of its tense markers (i.e., augment and sigmatic suffix), could also serve as a base for the formation of the -hi- future (e.g., a-k[]-si k[]-hi-ti). (18)

The model for such a formation rule is likely to have been provided by verbs whose imperative and, accordingly, -hi- future exhibited a base morph that was identical to that of the corresponding sigmatic aorist. This is, for instance, the case of the verb P[]li ne-'lead' (cf. OI n[]-), which has the aorist form nesi, 3rd sg. (Ja v 281,23, cf. OI anais[]t; see Geiger 1916: 132), and the imperative form P[]li (a) nehi (Ja It 254,19; Geiger 1916: 109). Given such a paradigm, it was possible to interpret the -hi- future stem nehi- (cf. nehisi, Patna Dharmapada 276), originally formed from the imperative, as derived from the base morph contained in the aorist.

Other -hi-stems, nevertheless, seem to contain the same root-allomorph shown by the corresponding -(i)ssa-future, so that it must be assumed that the old -sya-1-(i), syastem type could influence the formal structure of the new -hi-formation: cf., e.g., P[]li karihiti 'he will make' (Th[] 424) compared to karissati.

It is also worth noting that the impression that the suffix -hi- appears preferably after a diphthong or a long vowel (cf., e.g., Woolner 1928 [section] 118; Schwarzschild 1953: 42) can now receive a new and strictly morphological explanation. Indeed, this situation is due, on the one hand, to the actual formal structure of the forms ehi- and kohl-, which must have constituted the starting point for the creation of the new formation, and, on the other hand, to the appearance of a morphological connection between 'Vsi aorists and -hi- future. (19)

As for the thematic vowel, the oscillation between -i- and -a- (e.g., P[]li 3rd sg. 'Gallia and k[]hati, 3rd pl. k[]hinti and k[]hanti: cf. Oberlies 2001: 246) is to be interpreted as due to a tendency toward substituting the usual thematic vowel -a- for the vowel -i- originally contained in the suffix. In this respect, the first persons have a special status: as is well known, in the unmarked inflection class--i.e., the -a- thematic type--these persons are characterized by a long -[]-. This lengthened thematic vowel, precisely because its distribution was restricted to particular forms of the paradigm, could come to be felt as a part of the 1st person ending or, at least, as an element strictly associated with it. This explains the fact that, e.g., in P[]li, while the usual 2nd and 3rd sg. persons of the -hi- formations are, respectively, in -hisi and -hill, the corresponding 1st sg. and 1st pl. persons are usually in -h[]mi (or -ham) and -h[]ma. On the other hand, 1st person forms in -himi are exhibited by Apabhramga (cf. Tagare 1987: 307) and mentioned by the grammarians (cf. Pischel 1900 [section] 520).

Moreover, in the Prakrits the first person in -h[]mi is used in competition with a form in -ham (cf. Pischel 1900 [section] 520) that exhibits a transfer of the secondary 1st sg. ending -am; such a phenomenon occurs in parallel in the paradigm of the future stems in -(i)ssa- and in -CCh-, where it can already be observed in P[]li (cf. Oberlies 2001: 244; also von Hintiber 2001: 297).


It now remains to explain why in classical Maharastri a mixed paradigm seems to be observable, in which the 1st person tends to be expressed by an -issa- form, whereas the other persons normally show -hi- forms. (20) In fact, it can be shown that in Middle Indic the morph -(i)-ssam can be analyzed as a non-segmentable ending morph. Such a final sequence is expected and is indeed found in the 1st person singular of the conditional mood, which is formally equivalent to a future in -(i)-ssa-with augment and secondary endings (cf. Pali abhavissam 'I were' fa 470,15; Obellies 2001: 250). In addition to this, we find the same final sequence in those 1st sg. future forms in -(i)ssa that show the secondary ending -am instead of the expected -a-mi (cf. above).

Even more importantly, the same ending is also observed in the 1st person of the -is- aorist: as the Pail inflectional paradigm of this formation shows (sg: 1 st. aganzisanz, 2nd aganzi 3rd agami; p1.: 1st agamimha, 2nd agamittha, 3d agamisurnlagamirnsu), the 1st person agarnisant is synchronically opposed to forms that do not contain the sequence -is- (with the exception of the 3rd pl. agamisum, which, however, is used in competition with the variant agamlinsu). Therefore, from a synchronic point of view, the element -is- of agamisam, even if it originally constituted an autonomous morph, is easily analyzed as a part of a personal ending -isam. This is confirmed by the fact that, beside the inherited agamisam, a variant agamisam is also found, where the 1st person ending -issam is substituted for the morpho-logically isolated (in that it is restricted to the -is- aorist) -isam (cf. Oberlies 2001: 241, von Hinuber 2001: 298). (21)

These phenomena demonstrate that the ending-issam tended to be extended into different kinds of inflectional paradigms. Therefore, it is by no means surprising that in the Prakrits the same ending could penetrate into the inflectional paradigm of the -hi- future, in this case at the expense of the form in -ham/-hiimi. (22)

The creation of mixed paradigms is not a rare event, in particular in the case of future tenses. Among the typological parallels it is difficult not to mention the well-known case of the Latin future legam, leg[]s, in which the sole 1st sg. form is taken from the -[]-subjuntive (cf. also Wackernagel 1926: 198f.). However, it is even more interesting to note that a typo-logical parallel is offered by the future tense of some New Indo-Aryan Languages of the east central zone (Eastern Hindi and Bihari group; Masica 1991: 289-91), where within the same inflectional paradigm the -h-future forms are in complementary distribution with -b-forms whose stem continues the old gerundive in -tavya-(on the corresponding finitization process, see Bubenik 1998: 193-95). Thus, in Bhojpuri an -h-future form is found with the 3rd person, whereas -b-future forms are used for the 1st and the 2nd person (Shukla 1981: 109).


In this paper we have argued for the derivation of the MI future suffix -hi- from the imperatival ending -hi- (OI -dhi). While the traditional explanation of -hi- as a special outcome of -sya- does not hold from the point of view of historical phonology, the solution advanced here is capable of explaining both the formal and functional properties of the suffix. More-over it allows us to recover a typologically interesting instance of morphological reanalysis that supports the corollary to so-called Watkins' law pointed out by Harold Koch (1995), i.e., the generalization that, in a particular morphological domain such as the imperative mood, a secondary attribution of a zero personal ending to a verbal form--via reanalysis of the original ending--is expected to involve the 2nd person rather than the 3rd.

Author's Note: This paper develops part of the presentation "Zum mittelindischen Futur auf -hi-" read at the 19. GeSuS-Linguistik-Tage 2010 (2nd-4th March 2010, Freiburg i. B.) in the workshop "Verbale Semantik and Syntax aus diachroner Sicht." I wish to thank Martin J. Kiimmel, Reiner Lipp, Luca Lorenzetti, and an anonymous reviewer for useful comments and suggestions. The author is of course responsible for any errors.

(1.) It should be mentioned here that no instance of this stem formation type can be found in the extant dramas of Agvaghoa and Bh'sa (cf. von Hinuber 2001: 298). Moreover, no attestation of the -h- future is present in the post-Agokan Prakrit inscriptions studied by M. A. Mehendale (1948).

(2.) A particularly interesting case also with respect to the argumentation we will propose below is that of Maithili. According to R. Yadav's description (1996: 1641), in that language the suffixal element -ih- is restricted to the imperative mood, where its presence distinguishes the future (e.g., hau sunar 15 ghar dekh-ih-ah "0 Sunar, have an eye on the house") from the present (e.g., (to) cal-ah "(You) go!").

(3.) It should be noted that. while the 01 periphrastic future of the type dat[]sini, formed by the nomen agentis in -tar- plus the verb asti, is actually already univerbited in the Br[]hmana prose (on its functional specification. see Tichy I 992). this is not true for the type formed by nomen agentis plus personal pronoun, which is Ibund especially in the later language of the epic. In the latter type, which is genuinely periphrastic also from a synchronic point of view, the nomen agentis is normally inflected for number: moreover the position of the pronoun is free (cf., e.g., gant[]ro narakam vayam. Mbh 1.88.10: see also Oberlies 2003: 237-39). Such behavior means that in these periphrases the pronouns still function as such and not as morphological markers. On the other hand, as for the 1st sg. in -[]ham. the creation of a corresponding 1st sg. middle form in [degrees] []he (cf. yast[]he 'I shall sacrifice' Titirt[]r 1.11.4) was viewed by Pisani as an indirect argument for the assumption of an early fusion, since the ending [degrees] []he may be explained as constructed on the basis of the active "[degrees] []ham in analogy to abharam: abhare ton this issue, see, with references. Thumb-Hauschild 1959: 328f.).

(4.) This idea was at first proposed by Richard Pischel (1900 [section] 151: cf. Tedesco 1945: 158f). A different view is that of Wilhelm Geiger (1916: 471. [section] 19). who saw the vowel -i- as the result of a stress-conditioned reduction of -a-. This hypothesis is, however, relatively weak, for in Pali a reduction of a to i is not a regular sound change.

(5.) Turner's view (1927: 232) that, in the writing system employed in the Kharosthi inscriptions discovered by Sir Marc Aurel Stein (Gandhari, Niya), the special symbol s used for writing the genitival ending -asa (< -asya) represented a lenited s is untenable (see von Hiniiber 2001: 178).

(6.) Pali texts are cited according to the format used in the Critical Pali Dictionary (Copenhagen: Royal Danish Academy of Science and Letters, 1924-).

(7.) This is confirmed by the fact that, in the metrical passages cited by Oberlies to exemplify his claim that "under strong metrical pressure, geminate -ss- of the termination was simplified" (2001: 109), a non-geminate spelling like dakkhis[]ma (Ja III 99,7*) is the exception rather than the rule. On this issue it is appropriate to quote what Ludwig Alsdorf wrote about the eirya stanzas in the Ther[]-g[]th[] (1966: 235): "Sometimes (and not only at the beginning of a word or second member of compound), a double consonant is metrically treated as single (a practice proved beyond doubt by numerous parallels in Amg poetry)." The genitive ta(s)sa in Th[] 406 (Alsdorf 1966: 240) also belongs to this type of phenomena and no importance can be attached to it with respect to the phonological opposition between simple and geminate s (pace Oberlies 1996: 116).

(8.) In order to limit ourselves to the more recent literature, it will suffice here to recall the idea that the phenomenon occurs preferably after a long vowel (cf. Oberlies 2001: 245) or between identical vowels (cf. von Hinuber 2001: 178).

(9.) The parallel 2nd person form tuha (cf. Pischel 1900-421) is modeled on maha < mahyam or is an outcome of tubhyam with debuccalized labial.

(10.) Cf. the following plural endings: 1) -a- declension: gen. -ah[]; abl. -chit; instr. -ahi and -MI; loc. -ahi (s. Pischel 1900 [section] 363; Bubenfk 1998: 37, 66); 2) -[]- declension: gen./abl. -[]hu; instr./loc. -ahi (s. Pischel 1900 [section] 374; Bubenfk 1998: 37).

(11.) It is difficult to assess whether this seeming morphophonemic function of h can be also invoked for Ap. eho, a demonstrative pronoun considered a continuation of 01 esa via the change s > h (cf. also Bloch 1965 196).

(12.) Another instance of s > h could be represented, according to the testimony of Hemacandra (8.3.162). by the aorist forms th[]h[] (for th[]s[]) and k[]h[] (for k[]s[]). As for Pischel's derivation of the MI -h- locative types from < -asmin (1900 [section] [section] 264, 425, and 366a), see Tedesco's refutation (1945: 156).

(13.) It has been proposed that the Pan verse ak[]mam r[]ja k[]mehi akanto kantam icchasi (Ja v 295,15*) contains an instance of s-debuccalization. Indeed, in Luders-Waldschmidt (1954: 85 n. 4) the form k[]mehi is taken as a variant of k[]mesi (2nd sg. ind. 'you desire') with s > h, and the verse is rendered as follows: "Du liebst, o Konig, die, die keine Liebe hat, ungeliebt begchrst du die. Geliebte." This interpretation is based on a parallel line found in the II/lahapastu 481,12 and m 16,20): akamorp raja kam.esi naitam panaitala4anam "0 king, you love one who does not love you. This is not the mark of a wise man" (tr. J. J. Jones, The Mah[]vastu, vol. 2 [London: Luzac & Co., 19521, 428). In fact, the two passages are similar but not identical: k[]mesi is the only 2nd sg. verbal form occurring in the Sanskrit verse and icchasi may be viewed as its counterpart in the Jataka text; on the other hand, k[]mehi is the regular form of the instr. pl. of the noun k[]ma-'love, desire' and can he interpreted as depending on ak[]mam (cf. k[]mehi analthik[] "for whom desires of love are worthless" Thi 485, ak[]mak[]mi(n)-'passionless' Sn 1096). Therefore. the assumption of an unparalleled 2nd sg. indicative ending -hi -Si is unnecessary.

(14.) Brugmann quotes as a possible parallel Lith. veizdeli 'see, look at', which he connects with the Old Lith. imp. veizd (i) 'see V (di 01 viddhi, Gk. i [sigma] [theta] [iota] with the original zero grade of the root). The Vedic hapax srudh[]yant (RV 6.673), pres. ptc., is based on the imperative srudhi 'listen'. However, as Brugmann notes, the Rigvedic passage in question admits of more than one interpretation and a possible rendering of srudh[]ya- is 'say "frudizi" '; if this is correct, this -ya- formation is to be classified as a delocutive verb and the parallelism with Gk. [epsilon] [sigma] [theta] I [omega] is only illusory.

(15.) For the particular case of the verb e-'go', the plausibility of the assumed reanalysis is supported by the almost perfect Slavic parallel mentioned above. A formal difference is that the MI imp. ehi that underwent reanalysis is not the direct continuation of the PIE imperative, but a secondary formation with full grade of the root.

(16.) The functional proximity between imperative and subjunctive in Old Indo-Aryan is confirmed by the special status of the Vedic "imperatives" in -si, which have been explained as haplologized 2nd sg. aor. subjunctives (Szemeranyi 1966).

(17.) The Vedic (and Avestan) subjunctives also show functional properties which are on the border between mood and tense. For a recent investigation of this subject see Tichy 2006.

(18.) In fact, the same could also be true for the -ss- Pali future kassa- (if we assume a shortening of the -a- due to the two-mora law) and, possibly, for its degeminated variant k[]sa- (cf. k[]sam Ja tv 287,13*), which, however, could also be interpreted as a blend of kassa-and k[]hi-. (On the relationship between aorist and future stems, see also Alsdorf 1936: 324; Thieme 1981: 2991; for a different explanation via analogy, cf. Oberlies 1996: 115 n. 136.) Indeed, in order to explain the form kassati by means of historical phonology scholars have had to assume an unattested OI future variant *karsyati or to imagine an exceptional loss of the vowel -i- of the stem karisya- (cf. Turner 1931: 534). The future stem d[]ha- (with thematic vowel -a-) from MI dad[]ti 'give', already attested in the fourth Pillar Edict of Agoka, also shows the same base allomorph contained in the corresponding aorist: cf. both the P[]li athematic (ad[]) and sigmatic (ad[]si) forms (Oberlies 2001: 231, 236).

(19.) It is also Worth noting that in MI the imperative ending -hi is often extended into the thematic conjugation, where its application is usually accompanied by the lengthening of the thematic vowel; cf., e.g., Pali gacchahi (von HiniTher 2001: 2801). The model of such reshaping is probably to be identified as the imperative of the OI athematic -a- roots (cf. also Oberlies 2001: 220); thus the tendency to be preceded by a long vowel is significantly shared by both the imperative ending -hi and the suffix -hi-.

(20.) According to Schwarzschild (1953: 43), a relatively low frequency of -h- with 1st persons seems to be also shown by the Pali of the Gallas and by the Ardhamagadhi of the Jaina canon. If this is correct--but statistical confirmation would be needed--we must assume that the tendency toward the formation of the mixed paradigm was partially present in these varieties as well.

(21.) Another competing variant is the type againiip 'I went', which is formed after the pattern of the type ak[]sim (Oberlies 2001: 240f.).

(22.) According to Bhamaha's commentary to Vararuci (cf. Schwarzschild 1953: 44; Bubenik 1996: 106f.), -hi- is the usual future suffix in M[]h[]r[]str[] (Bh. VII, 12), but both the I st singular and plural forms can show the suffix -ssa- as well as the suffix -hi- (Bh VII, 13; cf. also Hemacandra III, 167-68). This seems to indicate that in the mixed paradigm in question the I st pl. could be influenced by the 1st sg.


Alsdorf, Ludwig. 1936. The Vasudevahindi, a Specimen of Archaic Jaina-M[]h[]r[]str[]. Bulletin of the School of Oriental and African Studies 8: 319-33.

--. 1966. Appendix II: []ry[] Stanzas in Thera-Ther[]-G[]th[]. In The Thera-and Theri-Gatha, ed. Hermann Oldenberg and Richard Pischel, 2nd ed. with appendices. Pp. 233-50. Oxford: Pali Text Society.

Benveniste, Emile. 1946. Structure des relations de personne dans le verbe. Bulletin de la Socidte de linguistique de Paris 43: 1-12.

Berger, Hermann. 1955. Zwei Probleme der mittelindischen Lautlehre. Munchen: Kitzinger.

Bhayani. Harivallabh C. 1997. Some Topics in the Development of 0IA, MIA, NIA. Ahmedabad: L. D. Institute of Indology.

Bloch, Jules. 1950. Les inscriptions d'Asoka. Paris: Les belles lettres.

--. 1965. Indo-Aryan from the Vedas to Modern Times. (English edition largely revised by the author and translated by Alfred Master.) Paris: Adrien-Maisonneuve.

Brugmann, Karl. 1913. Homerisch [epsilon] [sigma] [theta] [omega] und [epsilon] [sigma] [theta] i [omega]. Indogermanische Forschungen 32: 63-71.

Bubenlk, Vit. 1996. The Structure and Development of Middle Indo-Aryan Dicdacts. Delhi: Motilal B an ars idass.

--. 1998. A Historical Syntax of Late Middle Indo-Aryan (Apabhrainta). Amsterdam: Benjamins.

Bybee, Joan. 1985. Morphology: A Study of the Relation between Meaning and Form. Amsterdam: Benjamins.

Collinge, N. E. 1985. The Laws of Indo-European. Amsterdam: Benjamins.

Cone, Margaret. 1989. Patna Dharmapada I. Journal of the Pali Text Society 13: 101-218.

Edgerton, Franklin. 1953. Buddhist Hybrid Sanskrit: Grammar and Dictionary. vol. 1: Grammar. New Haven: Yale Univ. Press.

Geiger, Wilhelm. 1916. Pali Literatur und Spra he. Strassburg: Trubner.

Hengeveld, Kees. 2004. Illocution, Mood, and Modality. In Morphologic Ein internationales Handbuch zur Flexion und Wortbildung, 2. Halbband (HSK 17.2), ed. Geert Booij et al. Pp. 1190-1201. Berlin: de Gruyter.

Jakobson, Roman. 1932. Zur Struktur des russischen Verbums. In Charisteria Gvilelmo Mathesio. Pp. 74-84. Prague.

Jaszcolt. K. M. 2009. Representing Time: An Essay on Temporality as Modality. Oxford: Oxford Univ.Press.

Koch, Harold. 1995. The Creation of Morphological Zeroes. In The Yearbook of Morphology 1994, ed. Geert Booij and Jaap van Marie. Pp. 31-71. Dordrecht: Kluwer.

Kortlandt, Frederik. 1979. Toward a Reconstruction of the Balto-Slavic Verbal System. Lingua 49: 51-70.

Kurytowicz. Jerzy. 1964. The Inflectional Categories of Indo-European. Heidelberg: Winter.

Luders. Heinrich (and Ernst Waldschmitt). 1954. Beobachtungen fiber die Sprache des buddhistischen Urkanons, aus dem Nachlass herausgegeben von Ernst Waldschmitt. Berlin: Akademie Verlag.

Masica, Colin P. 1991. The Indo-Aryan Languages. Cambridge: Cambridge Univ. Press.

Mayerthaler. Willi. 1981. Morphologische Natiirlichkeit. Wiesbaden: Athenaion.

Mayrhofer, Manfred. 1951. Handbuch des Pali, pt. 1: Gmmtnatik. Heidelberg: Winter.

Mehendale, M. A. 1948. Historical Grammar of Inscriptional Prakras. Poona: Deccan College Postgr. Res. institute.

Oberlies. Thomas. 1996. Stray Remarks on Pali Phonology, Morphology, and Vocabulary (Miscellanea Palica V). Manchener Studien cur Sprachwissenschaft 56: 91-130.

--. 2001. Pali: A Granunar of the Language of the Theravada Tipitaka. Berlin: de Gruyter.

--. 2003. A Grammar of Epic Sanskrit. Berlin: de Gruyter.

Palmer, F R. 1986. Mood and Modality. Cambridge: Cambridge Univ. Press.

Pisani. Vittore. 1952. Noterelle Pali. Rendiconti dell'Istituto Lombardo di Science e Lettere. Classe di lettere 85: 279-88.

Pischel, Richard. 1900. Grammatik der Prakrit-Sprachen. Strassburg: Trilbner.

Renou. Louis. 1961. Grammaire, sanscrite, 2nd ed. Paris: Adrien-Maisonneuve.

Schwarzschild. L. A. 1953. Notes on the Future System in Middle Indo-Aryan. Journal of the Royal Asiatic Society 1: 42-52.

Shukla, Shaligram. 1981. Bhoipuri Grammar Washington: Georgetown Univ. Press.

[Speyer] Speijer, Jacob Samuel. 1886. Sanskrit Syntax. Leiden: Brill.

--. 1896. Vedische and Sanskrit-Syntax. Strassburg: Triibner.

Szemerenyi. Oswald. 1966. The Origin of the Vedic "Imperatives" in -si. Language 42: 1-6.

Tagare, Ganesh V. 1987. Historical Grammar of Apabhramla. Delhi: Motilal Banarsidass.

Tedesco, Paul. 1945. Hindi bhejn[] 'to send'. JAOS 65: 154-63.

Thieme, Paul. 1981. Lexikalische und grammatische Bemerkungen zu den Agoka-Inschriften. In Studien zum Jainismus und Buddhismus: Gedenkschrift far Ludwig Alsdorf, ed. Klaus Bruhn and Albrecht Wezler. Pp. 297-300. Wiesbaden: E Steiner. (Repr. Kleine Schrjften, ed. Renate SOhnen-Thieme. Pp. 907-10, Stuttgart: F. Steiner, 1995.)

Thumb, Albert, and Richard Hauschild. 1959. Handbuch des Sanskrit, pt. 2: Formenlehre. Heidelberg: Winter.

Tichy, Eva. 1992. Wozu braucht das Altindische ein periphrastisches Futur. Zeitschrift der Deutschen Morgenlandischen Gesellschqft 142: 334-42.

--. 2006. Der Konjunktiv und seine Nachbarkategorien. Bremen: Hempen Verlag.

Turner, Ralph L. 1927. The Phonetic Weakness of Terminational Elements in Indo-Aryan. Journal of the Royal Asiatic Society of Great Britain and Ireland 1927: 227-39.

--. 1931. The Future Stem in Agoka. BSOS 6: 529-37. (Repr. Collected Papers 1912-197. Pp. 323-30. London: Oxford Univ. Press, 1975.)

von Hiniiber, Oskar. 2001. Das altere Mittelindisch im Uberblick, 2nd ed. Vienna: VOAW.

Wackernagel, Jacob. 1926. Vorlesungen fiber Syntax, 2nd ed,, vol. I. Basel: Birkhauser.

Watkins, Calvert. 1962. Indo-European Origins of the Celtic Verb. Dublin: Institute for Advanced Studies.

--. 1969. Geschichte der indogermanischen Verhaljlexion. Heidelberg: Winter.

Weber, Albrecht. 1870. Ueber das Saptagatakam des Hdla: Ein Beitrag zur Kenntniss des Prakrit. Leipzig: Brockhaus.

Woolner, Alfred C. 1928. Introduction to Prakrit, 2nd ed. Calcutta: Baptist Mission Press. (Repr. Delhi: Motilal Banarsidass, 1975.)

Yadav, Ramawatar. 1996. A Reference Grammar of Maithili. Berlin: de Gruyter.

COPYRIGHT 2011 American Oriental Society
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2011 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Milizia, Paolo
Publication:The Journal of the American Oriental Society
Article Type:Essay
Geographic Code:4EUIT
Date:Jan 1, 2011
Previous Article:Mir Damad in India: Islamic philosophical traditions and the problem of creation.
Next Article:On the semantic foundation of P[]ninian derivational procedure: the derivation of kumbhak[]ra.

Terms of use | Privacy policy | Copyright © 2020 Farlex, Inc. | Feedback | For webmasters