The   thematic  inflection in  Proto-Indo-European  conjugation:  a kind of perfect or a kind of aorist?

by Eugen Hill (Universität zu Köln)

In the talk it is argued that in PIE the so-called thematic present stems originally inflected in the same way as the so-called athematic ones. In the singular, the diffe-rence between the ´primary' indicative mood inflectional forms and their ´secondary' injunctive mood counterparts was marked solely by PIE *-i attached to the relevant verb form. This means that this part of the PIE conjugation system was nearly identi-cal with what is attested in ancient Indo-Iranian. The deviating ´primary' inflectional forms in the individual IE languages can be !lausibly explained within the individual history of these languages.

Such explanations become available as soon as one takes into consideration

(a) the laws of final syllables relevant to these !articular languages

(b) the possibility of a recent univerbation of finite verbs with sentence particles and other clitics.

It is essential that in PIE the thematic nouns, adjectives and present tense stems of verbs were inflected - by and large - with the same inflectional endings as their athematic counterparts. In the inflection of nouns and adjectives this simple principle is violated just once by a special form of the ablative singular which is exclusively found in thematic stems. As for  present stems of verbs, the identity of inflectional endings is strongly indicated by the evidence of the Indo-Iranian branch given in (9), where the inflectional paradigm is  particularly well preserved.

1. Introduction: Thematic vs. athematic in Proto-Indo-European

As suggested by the evidence of the Indo-Euro!ean (henceforth IE) languages with most ancient attestations the common parent language of the family Proto-Indo-Eu-ropean (henceforth PIE), must have possessed two major different classes of word-stems both in the domain of nouns, pronouns and adjectives and in the domain of verbs. One of these classes is traditionally called "thematic". Stems belonging to this class ended in the so-called "athematic" vowel which took the form PIE *-e- or *-o- depending on the specific inflectional forms. The second class is called  "athematic". Stems belonging to this class did not contain the thematic vowel. Cs. the table (0) for the contrast thematic vs. athematic in nouns and verbs capable of building a present tense (henceforth present stems), where must a small fraction of possible inflectional forms are given and the inflectional endings are separated from the stems  by -.

Note that in this subset of inflectional forms 1 sg. through 3 sg. the personal endings of the indicative mood (which are traditionally called the "primary" endings) are clear-ly based on their counterparts in the injunctive mood (the "secondary" endings). The difference between these two series is the presence of PIE *-i in the former which might be ultimately related with the ending of the locative case in nouns and adjectives .

2. Stating the problem: inflection of thematic present stems in the individual Indo-European languages

However, this clear and symmetrical picture becomes less clear and symmetrical as soon as the evidence of other IE branches is taken into account more comprehen-sively. It turns out that such IE languages as for instance Greek. ln Church slavonic or Gothic satisfactorily agree with Indo-Iranian only in a part of the paradigms. Cost severe problems have been registered in the thematic inflection, where only the "se-condary" injunctive mood endings (preserved in Greek and Slavonic) clearly match their Indo-Iranian counterparts. By contrast, the "primary' indicative mood endings of thematic stems are rather incompatible with the corres!onding Indo-Iranian endings either in the 1 sg. (in Germanic and Latin), in the 1 sg. and 2sg. (in parts of Slavonic) or even in the whole set (in Greek or Baltic).

The unexpected shape of "primary" thematic inflectional forms like Gk 3sg.φέρει etc. led many scholars to the conclusion that the deviating desinences in such languages as Greek, Latin and Baltic are more original than their more system-conform Indo-Iranian peers. Indeed, the Sanskrit pattern 1sg. inj.  bhára-t  ~ 1sg. ind. bhára-ti ´carry´ etc. can be easily explained by a recent morphological analogy on the model of athematic present stems (cf. 2sg. inj. é-t  ~ 2sg.ind. é-ti etc.).

Aspect and alignment in Indo-European and Proto-Indo-European workshop at Ghent University, 10. - 11. September 2018

By contrast,for Gk 1sg.inj. φέρε ~ 3sg.ind. φέρει etc.an explanation by a recent mor- phological analogy is not available for the lack of a model. If one takes the "primary" thematic inflectional forms like Gk 1sg. φέρω, 3sg. φέρει etc.as more or less directly reflecting the PIE state of affairs, one cannot but accordingly modify the traditional PIE reconstruction in roughly the following way cf. (4). As has been repeatedly poin-ted out, the new PIE "primary" thematic endings such as 1.sg *-H etc. are vaguely similar to the well-known and most uncontroversial endings of a Luite different verbal formation: the PIE perfect.

 

Currently, this new reconstruction of the PIE "primary" thematic inflection is broadly accepted in the field. The differences of opinion are confined to merely two points. The first  point concerns the amount as to which the traditional reconstruction has to be modified (only the 1sg. or the whole set?). The second point is the question of how much to take from the  perfect (a perfect like 2sg. seems unattested in thematic presents of the individual IE languages). Cf.(5) for an overview of the most influential textbooks and papers. For more references Lindquist & Yates (2018: 2148 - 2149) can be consulted.

 

Eugen Hill (Universität zu Köln) The thematic inflection in Proto-Indo-Euro!ean confugation

 

3. Ways to a more principled account: the case of Greek

However, what does it actually mean, a desinence such as that of Gk "primary" 1sg.
φέρει is incompatible with its Indo-Iranian counterpart in Skt bhárati (and in Goth bairiþ, ORu beretĭ, Lat ag-it) The desinences in question are indeed incompatible, if one tries to transform them into each other by merely mechanically applying the rele-vant sound laws (such as Skt t = Goth þ = Gk t or s before ι). We know that this pro-cedure is a necessary first step in any historical analysis. But we also know that me-chanically comparing sounds is often not sufficient all by itself. Especially when we are dealing with inflectional endings, at least two more factors often operative in the individual history of an IE daughter-language should be taken into account.

The first factor is special sound laws which might affect the end of the word resp sound environments frequently found at word-ends. For instance, our knowledge about the secondary loss of stops in word-final position in Greek (cf. Skt tát ´that´ = tό Gk ´that´ etc.) correctly prevents us from considering Skt sg. prs. opt. bháre-t  and its Greek counterpart φέροι-Ø as mutually incompatible.

The second factor is the well-known inclination of finite verb forms towards a secon-dary univerbation with clitics. We know of several instances of such a univerbation with clitics already in PIE but also during the individual history of the IE daughter-languages. Cf. for the latter case, which is more important for the present discussion two inflectional forms of the Gothic optative wood. In the relevant paradigm, the Opt. prs. bairaima, prt. bēreima and the pl.prs. bairaina, prt. bēreina are clear instances of a univerbation with a clitic = a.This univerbation is not shared by the other Germa- nic languages (cf. OHG opt. 1pl. prs. berēm, prt. bārīm, = pl. prs. berēn, prt. bārīn) and must be therefore particularly recent. Note that in many similar cases it is quite difficult to establish both the etymology of the clitic and why the univerbation occurred only in a subset of the relevant paradigmatic forms (and not, for instance, also in the 2pl. of the Gothic optative mood).

It has been successfully demonstrated that taking into account these two factors sometimes leads to the conclusion that two su!erficially incompatible desinences are a perfect match of each other. Perhaps the most instructive case of this kind is the analysis of the Gk "primary" 2sg.φέρεις 3sg.φέρει by Kiparsky (1967).In his seminal paper Kiparsky showed that Gk 2sg.-εις 3sg. -ει,allegedly incombatible with Skt 2sg. -asi, 3sg. -ati, may be explained by a recent metathesis dentals and -i in word-final position (at least after shortt undressed non-high vowels):

Note that, first, metatheses of the assumed kind are typologically not uncommon (cf. Gk bαiνω [baino] < PIE  gʷm-ié-, cf. Lat ueinō etc.), second, the loss of stops at the end of the word is independently established for Greek (cf. Skt tád ´it´ = Gk tό ´it´ etc.).

In the same publication Kiparsky rightfully pointed out that his sound change can be demonstrated also for other words than verbs and is thus established beyond reasonable doubt (cf. Cogill 1985: 99-103, Rix 1992, 151, now Hackstein 2002: 107-109, Willi 2012: 266-269, 2012: 6-7, for more references see 1995, 39).

Aspect and alignment in Indo-European and Proto-Indo-European workshop at Ghent University, 10. - 11. September 2018

 

 It follows that Gk 2.sg.prs.ind. φέρεις, 3sg.prs.ind. φέρει are in fact perfectly compatible with their Indo-Iranian, Italic and Germanic counterparts and may well respectively reflect PIE *bʰéresi and *bʰéreti.
 
4. The "primary" 2sg. in Baltic and Slavonic

The !ur!ose of the !resent talk is to !rovide a similar !honological account for the indicative food or "primary" 2sg. of thematic present steps in Baltic and Slavonic. The Baltic and Slavonic counterparts of the 2sg Skt bhárasi, Goth bairis, Lat agis - all pointing to PIE *bʰéresi - are often viewed as a clear case of deviation from the above stated principle according to which the thematic presents originally used the same inflectional endings as their athematic peers, o% most recently, the 2sg. of thematic present stems has been reconstructed for Proto-Balto-Slavonic as inj. *-e-s ~ ind. *-e-Ɂi and, accordingly, as PIE ind. -e-s ~ ind. -e-h1i by Kortlandt (2015 5-6). Similar views have been expressed in a number of comprehensive treatments such as Stang (1942: 2015, 1966, 407), Watkins (1969: 213-214), Langston (2018: 1552), cf. Hock (2018: 26-27) and Olander (2015: 312-315) for more references. The deviating "primary" 2sg. of Baltic and Slavonic, superficially incompatible with PIE *bʰéresi, is frequently seen, beside the Greek data discussed above, as the second major piece of evidence for a special perfect-style inflection of thematic present stems in PIE. It is more convenient to start the analysis of the Dalto-Slavonic data with Slavonic. In Old Russian - a well-documented and, at the same time, one of the oldest Slavonic languages  -, the verbs reflecting PIE athematic and thematic present stems ewre inflected as given in (8).

It is evident, that Old Russian 3sg. forms of both athematic and thematic verbs perfectly match their Sanskrit (as well as Anatolian, Greek, Latin and Germanic) counterparts and should be, therefore reconstructed as ending respectively in PIE *-ti and *-e-ti.

It is also evident that in the 2sg. the attested Old Russian inflectional forms deviate from the expectation in the following two points.

First, both the athematic 2sg. ending ORu -si and the thematic 2sg. desinence ORu -eši end in an etymologically long vowel which is no match for Skt -i and cannot reflect PIE *-i (which yields ORu -ĭ  as in the athematic 1sg. and 3sg.). At the end of a word, ORu -i can only reflect a long vowel (Proto-Balto-Slav *-i  or *-e or a diphthong (Proto-Dalto-Slav. *-ei; or *-ai, cf. most recently Gorbachov 2015)

Second, the 2sg. desinence of thematic verbs ORu *-eši contains the fricative š instead of s to be expected in the given position as a watch of Skt s and the regular reflex of PIE s. It is known that ORu š can be either a reflex of early-Proto-Sl x secondarily palatalised by a front vowel or reflect the early-Proto-Slav cluster *sj (> late-Proto-Sl *š by the so-called jod-palatilasation').

Eugen Hill (Universität zu Köln) The thematic inflection in Proto-Indo-Euro!ean confugation

The former possibility is, however, difficult to justify because early-Proto-Sl *x - originally an allophone of *s after Proto-Balto-Sl *i, *u, *r or a tectal - would be rather unexpected in the 2sg. ending of thematic verbs, i.e. after PIE *-e- > ORu -e-. 1

In my opinion, these deviations from the theoretically expected reflexes of PIE *-si (Skt *-si) and *-e-si (Skt -a-si) do not make the attested Old Russian desinences *-si and -e-ši incompatible with these reconstructions. On the contrary, ORu 2sg. jě-si and bere-ši would regularly reflect PIE *h₁éd-si and *bʰére-si followed by resp. recently univerbated with a clitic. If this hypothetical clitic is assumed to correspond with the Greek unstressed sentence  particle Ion-Att, Arc εi, Aeol, Dor αi, a series of well-known Slavonic sound changes would lead to the following development. Cf. (9) where, for reasons of simplicity, only the counterpart of Gk αi, i.e. Proto-Balto-Sl *ai, is used.

The different behaviour of the -i in the inherited ending -si before the clitic - vocalic in the athematic stems but consonantal and therefore causing the jod-palatilisation in the thematic ones - conforms to theoretical expectation. All but one Slavonic athematic verbal stems are by one syllable shorter than reflexes of the thematic ones. It is known that in early-Proto-Slavonic the inherited prevocalic *i regularly remained vocalic in word forms with originally no more than three syllables but turned into a palatalising *j in longer word forms. This can be securely inferred from Slavonic reflexes of denominal relational adjectives formed with the suffix PIE *-iHo- > Proto-Dalto-Slav *-io-. Cf. on this derivational pattern the data given in (10).

In Proto-Slavonic, such adjectives display *-io- in trisyllabic words but !alatalising *-o- in words with more than three syllables (cf. Andersen 2017: 87-88 with references). Cf. the data given in (11).

1 It has been suggested (for instance, by Cogill 1985, 107) that Old Russian 2sg. αteεm. si, them. -ši may reflect not the active 2sg. PIE *-si (Skt -si) but its middle voice counterpart PIE *-soi; (cf. Skt -se, Gk Arc -οι). This is improbable for two reasons. First (cf. Olander 2015: 317), all the rest of the paradigm is undoubtedly active, cf. athematic 1sg. ORu -mĭ  (= Skt -mi, Gk -mι), 3sg. ORu ti  (= Skt -ti, Gk -sι). Second, in thematic verbs not ORu š but s should be expected before a reflex of PIE *oi, inde!endently of whether one starts with underlying Proto-Slav *s or, as is usually done, with Proto-Slav *x.