Kommentoidaan myöhemmin


Rasmus Hedegaard Bjørn
University of Copenhagen
June 2015

Aspects of Proto-Indo-European Indo-European Morphology

Introduction .............................................................................................................. 2
Methodology.............................................................................................................. 3
Linguistic material relevant for the discussion of aspect in PIE .............. .............. ....3
Terminology................................................................................................................. 4
Material ...................................................................................................................... 5
The Graeco-Aryan system .......................................................................................... 5
Stem formation (intra-radical morphology) ................................................................ 5
Endings and other extra-radical morphology ............................................................ 6
Systemic inconsistencies ........................................................................................... 7
The Hittite system ...................................................................................................... 9
PIE .............................................................................................................................. 9
Typology .....................................................................................................................10
Excourse #1: The Uralic system ...................................................................................10
Discussion ...................................................................................................................12
How PIE eventive became core-IE imperfect and aorist ............................................14
The systemic phases .................................................................................................16
Excourse #2: Adding causative-iterative and molō verbsto  the  perfect  category 16
Excourse #3: Contemporary interpretations of the PIE system .................................17
Conclusion ..................................................................................................................19
Bibliography ................................................................................................................20


Rasmus Hedegaard Bjørn
University of Copenhagen
June 2015
Aspects of Proto-Indo-European
Indo-European Morphology


Being presented over and over again with the three-fold tense-aspect system of im-perfect, aorist, and perfect (e.g. Fortson 2010:88), sense begged to be drawn from the shared traits of the former two in opposition to the latter. The primary concern to the present paper is thus to discern whether PIE had an aboriginal grammatical distinc-tion between imperfect and aorist. Inevitably, the nature of the perfect will be touched upon and to a large extent discussed, but I make no intentions of reaching any con-clusions on this particular topic. There is a scientific imperative of approaching the earliest stages of the dialect continuum possibly attainable through internal recon-struction, and there are only the comparative linguists to humbly approach this pre-historical frontier. The present paper is merely an opportunity to explore some of the glaring cracks in the verbal system traditionally reconstructed for Proto-Indo-European (PIE).


Analyzing the earliest stages of PIE invariably stresses the importance of the compa-rative method of historical linguistics, leaving the researcher to reconstruct a system based solely on the attestations of much younger stages of the language. Despite allegations of purely theoretical and highly speculative approaches, the method has several times over proven its merits, e.g. by predicting elements otherwise eradicated from the attested languages that appeared in the Anatolian material. Confidence must then necessarily be placed in this very founding method of the field, and it is, of course,only with the secure moorings of centuries of meticulous philological research that enquiries into the nature of the proto-language can be made. In focusing this paper, a number of caveats must be made: Linguistic material relevant for the discussion of aspect in PIE

The classical 19th century description of the PIE verbal system is almost exclusively based on Vedic and Greek material. The systems are strikingly similar and can to a large extent explain the further developments in the remaining Indo-European daugh-ter languages (Hoffmann 1970:540). The unearthing of the Hittite tablets demanded, however, that this earliest attestation of the family be taken into consideration, and the impasses tied to uniting the systems are usually mended by assuming Anatolian to bear the brunt of systemic change (cf. e.g. Beekes 1995: 223), which of course is plausible, yet internal irregularities of the Graeco-Aryan (or core-PIE) system point to a developing, or rather, recently developed, system for their part. It is thus inevitable to discuss the ramifications of the Anatolian split from the family tree, and the present paper will consequently recognize three distinct systemic phases of which the latter two are subsequent and are treated on par genealogically

1. Proto-Indo-European (PIE), a purely theoretical stage deduced from the comparative treatment of its primary constituents treated below.

2. Anatolian, primarily Hittite.

3. Core-Indo-European (core-IE),partly theoretical and primarily based on the Graeco-Aryan system. Many discussions subsumed for this paper deserve much deeper discussion, but the scope and aim simply does not allow for in-depth analysis of each given step,just like many languages only will be briefly touched upon, if at all. An inherent danger to the mode of discussion employed here is that of unwarranted extreme reductionism. Hopefully, the perspective this method provides will provide some new insights that can be back-traced to some of the discussions and further questions will arise. Importantly, we need not assume to know every semantic intri-cacy of a given PIE verb, simply qualifying the typological processes and distinctions necessary for a functioning verbal system will suffice.



The terminology of aspect is notoriously imbued with overlapping, confusing, and even contradictory terms. Below, I have given the definitions that best suit the present discussion (although it by no means sets out to define the phenomena on a broader scale!), and I must in advance apologize for any inconsistencies on my own behalf and hope that the context provide the necessary information to discern the proper meaning.

Aktionsart: Lexical, or intrinsic, aspect of a given verb (cf. Weiss 2009:377-378 and
Hoffmann 1970:531).

Aspect: Grammatical, i.e. overtly morphematically marked, - or syntactic - aspect.

Eventive: Sihler’s very appropriate grouping of the Graeco-Aryan imperfective and aorist, as opposed to the perfect (stative) (1995:442ff.).

Imperfect: The opposite of perfect (see below).

Imperfective: An action viewed from within, ‘open’ (Comrie 1974:24), as opposed to

Noem: A linguistic quality expressed by various markers, e.g. the concept of plurality
expressed both with -(e)s and stem gradation (men) in English, represent a single
noem (Hoffmann 1970:524ff.).

Perfect: Stative or resultative. Although inherently prior to reference time, the perfect
aspect is not a tense;rather,it describes the inherent (whether ‘perfected’,i.e. comple- ted, or not) nature of an action in relation to the speech act’s deictic present, whether present,preterite or future;in the terminology used by Kiparsky 1 :E- R 2 (P irrelevant).

Perfective: An action viewed as an indivisible whole, ‘closed’ (Comrie 1974: 16ff.), as opposed to imperfective.

Telicity: A verb with both an intrinsic process and end can be defined as telic; compare

John is singing (atelic) with John makes a chair (telic) (Comrie 1976:44ff.).

Stative: Sihler’s  perfect (1995:564ff.) and one of the two semantic constituents of the
core-IE perfect verb stock, the other being resultative.

Tense: A grammatical function that places a given verb in relation to the speech act,
i.e. E in relation to P (R = P) (Kiparsky 1998).

1  E = event time, i.e. of the verb; R = reference time, whether past or present; and P = perspective time, i.e. that of the speech act (Kiparsky 1998), e.g. when (P) I say: Tomorrow at 5 PM (R) I will have eaten (E).'

2  I.e. E occurs before R.




Greek and Indo-Iranian present the most copious and archaic material within core-IE, and, historiographically, the key evidence for the traditional reconstruction of the PIE verbal system. With the unearthing of the genuinely archaic Hittite material, however, the entire system demanded revision.Below the relevant parts of the attested material is presented in the least polemic way possible for a discussion that invariably must ignore some contented central topics.

The Graeco-Aryan system

The Graeco-Aryan verbal system is comprised of its two constituents, Greek, aboun-dingly attested already from Mycenaean and, more copiously, the epic Homeric tales, on one hand, and Indo-Iranian, most notably represented by the Old Indic Vedas and the subsequent extant Sanskrit material, but flanked by the informative Old Iranian language of the Avesta,on the other. Lexical as well as grammatical similarities justly identified them as early heirs of the proto-Indo-European language, and the first cen-tury of scientific enquiry into the language family was consequently spent looking into their relationship to each other and to reconcile their later attested relatives on the fa-mily tree with the obvious Graeco-Aryan unity. The development of the core-IE verbal systems is almost unanimously away from the root system, most significantly with the implement of the thematic vowel (LIV 2001: 12), and undoubtedly the root formations represent an older stratum of verbal morphology. The employment of the ablaut alter-nation in the temporal markup of various languages, e.g. the Germanic strong verbs, is, of course, due to the fact that it is a salient feature (Comrie 1998:85).

Stem formation (intra-radical morphology)

To start with the heart of any given verb, the root itself presents the observer with the first oddity, viz. that there is no radical morphology to distinguish imperfect from aorist stems: both categories unanimously brand the e/Ø alternation in the root - their forma-tions are identical (cf. Sihler 1995:447).This lack of differentiation is especially salient when compared to the perfect,whose prominent features include the o-graded root. Only secondary derivation thus qualifies the distinction between imperfect and aorist on the stem-level, viz. reduplication (relevant for both classes 3 ), the sole exclusive aorist marker *-s-, and a host of imperfective markers 4, e.g. *-sḱ -e/o-,*-i-e/o-,*-n-, etc. Whatever their exact semantic value, they are (literally) peripheral to the radical dis-tinction, and will appropriately play a secondary role in the present paper. The stem formation thus warrants the historical grouping of the aorist and imperfect into an e- graded eventive category in opposition to the o-graded perfect. Hence, the morpholo- gy that ultimately sets imperfect apart from the aorist is limited to secondary formations and can thus not be considered archaic.

LIV (2001:14ff.) identifies 20 different imperfect (“Präsens-”) stem types, three aorist,   and just one perfect 5. Of primary interest for the present paper is the root forms, of which there has been

Or rather, all three, as one of the hallmark identifiers of the perfect in core-IE is the reduplicated onset of the stem.

There can be no doubt that the different affixes carried distinct meanings, but gram-matically they evidently only qualified to be treated on par to form the allophonic expressions of the imperfect. On the derivative affixes (as they include prefixes, e.g. reduplications, infixes, notably the n-infix, and suffixes en masse), see Sihler (1995: 494ff).
The obviously secondary constructions with suffixes (causative-iterative, desidera-tive, intensive, fientive, and essive) are omitted here for the reasons stated above.


Indo-European Morphology tallied 149(+72) 6 imperfect 7 and 265(+144) aorist, i.e. there is no overt morphological markup that identifies whether one of more than 300 - and possibly 200 more - verbs in the parent language is imperfect or aorist. In addition, 144 (+137) perfect stems are identified, but they do not, as we have seen, overlap morphematically with the imperfect and aorist.

Endings and other extra-radical morphology

Three sets of endings are of relevance to this section: the primary, secondary, and perfect endings.Primary and secondary endings exist among the aorist and imperfect to the emphatic exclusion of the perfect 8, that contrarily only takes the perfect endings (Fortson 2010:91).

Having isolated the perfect system, it is appropriate to term the primary-secondary complex the eventive endings 9. The primary endings are synchronically analyzable as the secondary with the suffixed hic et nunc-marker *-i, cf. Greek νυν-í ‘now’ (Fort-son 2010:93), and can thus be said to enter into a grouping distinct from the perfect endings, which thus leaves the observer with the extrication of the relationship between the primary and secondary endings on one hand and the aorist-imperfect distinction on the other.

Reaching these outer layers of the verb, however, the first apparent morphological
distinction between imperfect and aorist presents itself in the primary endings which only incur with the imperfect in the present tense, i.e. the imperfect carries a set of endings not found in the aorist.These primary endings, and consequently the hic et nunc-marker, only occurs with the imperfect in the present tense. Two possible dicho-tomies are thus possible: either imperfect versus aorist, substantiating the consti-tuents’ disparate provenance, or present versus non-present, stressing the marked-ness of tempus, which would have been a innovation secondary to an original even-tive unity. The problem with the former hypothesis, however, is that the relationship is not mutually exclusive; although the aorist only takes secondary endings, the imper-fect takes both, implying that a line has to drawn within imperfect (indeed, with the exclusion of the aorist), and not between imperfect and aorist. The mere existence of the “unmarked” (Ringe 2006:27) (i.e. of  time 10 ) injunctive sheds further light on the eventive endings, viz. that the temporal markings are peripheral and consequently that the primary endings only represent this later distinction.

The use of primary endings is the one most significant feature setting the imperfect apart from the aorist, but with the use of the secondary endings in all eventive verbs, the original function of the primary endings cannot have been to set the imperfect apart from the aorist, but rather to identify imperfect forms emphatically set in the deictic present of the speech act 11, which apparently is a characteristic unattainable for the aorist verb stock. This is further substantiated by the presence of the injunctive forms that lack any of these secondary temporal markers, so, as Hoffmann observes, the secondary (i.e. unmarked) endings cannot carry temporal meaning (1970: 533).

Secure(+insecure), within the inherent spectrum of insecurity of reconstructional linguistics, cf. LIV (2001:14).

Variant accent paradigms combined; amphikinetic 106(+46), acrodynamic 32(+20), schwundstufig 2(+3), and vollstufig 9(+3), respectively.

The perfect notably lacked the primary-secondary distinction of the imperfect and aorist, and similarly, the perfect was indifferent to diathesis, i.e. there is just one perfect paradigm (Ringe 2006:33).


Vel sim., cf. Kortlandt (2009:164)


The injunctive significantly distinguishes person, number, and aspect.

On the topic of temporal markers, the augment - notably only employed by a handful of IE languages, but importantly by both Greek and Vedic - also deserves mention. Meier-Brügger traces its origins to the temporal particle *h 1 é ‘at  that time’  that  was  prefixed  a  clear marker  of  past  tense  (2003:167).


The prohibitive *meh 2 sentences are instructive with regards to the inherited dis-tinction between the aorist and the imperfect. Imperatives,including their negated va-riants, are inherently tenseless, and when in Vedic the constructions are employ-ed, there is a clear aspectual difference between the use of the imperfect vis-a-vis the aorist. The injunctive of the latter has preventive meaning, i.e. ‘do not begin to   ...’, as opposed to the imperfect injunction that is inhibitive in the sense  ‘stop  doing ...’  (Hoffmann 12 1970:534).

Systemic inconsistencies

Although the Graeco-Aryan verbal system is described as having three tenses / as-pects, some paradigms are not fully inflected. Not all Vedic verbs are inflected in the aorist, and aktionsart is key to understanding why only certain verbs qualify for this category; it is not in the nature of an atelic verb (e.g. be, dwell, lie, shine, weep, cf. above under Terminology) to reach completeness in the same sense that a telic verb (e.g. become, take, leave) does, wherefore the semantics of the aorist only applies to the latter (Kiparsky 1998:45). Similar restrictions apply to Ancient Greek (Hirt 1912: 475), and a number of IE verbs did not form perfects (Fortson 2010:88ff.).

Suppletion is significant in both Greek and Vedic, yet there appear to be none or few lexical overlaps between the two languages. The phenomenon is described for Vedic by MacDonnel (1916:339):

 . . . . . . . . . . .‘be’ . .   ‘run’ . . . . . .       ‘see’ . . . . . .  ‘speak’ . . . . .   ‘slay’
Imperfect .... as- . . . .sísarti . . . . . . . . paś- . . . . . . . brū- . . . . . .. . . han- (+ perf.)
Aorist . bhū- (+ fut.) perf. ádadhāvat  dr̨ś- (+ perf.) vac- (+ perf.) . vadh-

Similarly, Hirt identifies six such suppletive paradigms in Greek (1912:475-76). The distinction may separate either aorist or imperfect from the perfective, but it never isolates both aorist and imperfect on the same side of the distinction. Osthoff already noted the discrepancy most prominently found between imperfect (‘präsens’) and   aorist stems (1899:44), and provides a collection of the amassed Indo-European material (1899:7ff):


(a) ‘essen,  verzehren’ Greek . . . Vedic . . . Armenian
Imperfect   . . . . . . . . . . ἐδω . . . .   ádmi . . . . utem
Aorist  . . . . . . . . . . . . . .ἐ-φαγον . .á-ghas . . keri, e-ker

(b) ‘gehen,  kommen’ Greek . . . Vedic  Old Irish . . . . .  . Gothic
Imperfect . . . . . . . . . . ἐρχομαι . . éti . . . tiagim . . . . . . . . . gagga
Aorist . . . . . . . . . . . . . ἠλ(υ)θον . ágāt . pret. do-chuaid . . pret. iddja



Noted also by Meier-Brügger (2003:255ff.) and Kiparsky (1998:46).


(c) ‘laufen,  rennen’ Vedic Greek
Aorist dhāvati
á-sarat τρέχω

(d) ‘geben, darreichen’ Vedic Old Irish
Aorist pra-yam
fut. prá-dā- do-biur
pret. doratus

(e)  ‘nehmen,  tragen,  bringen,  führen’
Aspects of Proto-Indo-European
Indo-European Morphology
ἠνεγκον vī- 13

(f) ‘sagen,  sprechen’ Greek Avestan
Aorist λεγω
εῖπον mraoiti

(g) ‘sehen, schauen’ Greek Old Irish
Aorist ὁραω
εῖδον ad-ciu
ad-con-dairc 14
Latin Greek 2
pret.  tulī αἱρέω
pret. εἱλον

In Proto-Indo-Iranian, *derḱ - is attested abundantly, but never in the imperfect.

(h) ‘sein,  werden’ core-IE
Imperfect *h1es-
Aorist (*bhuh2-) 16

Suppletive paradigms exist in all branches for the imperfect *h1es- , most often with the perfect *buh2-.

Where the forms on a broad scale fail to correspond formally (with notable excep-tions,treated right below),the sheer number of suppletive stems should alert the spec- tator of a deficient system. When adding the evidence of *derḱ -, a verb form demon-strably confined outside the imperfect, and the ubiquitous opposition of the existential verb *h1es- in the imperfect and the perfect *bhuh2 - elsewhere (five separate bran-ches,according to Frantíková (2014:60)),deep-rooted suppletive paradigms emerges.'

Cf. the Greek future οἰσω.

Osthoff treats con-dercar as an innovation (1899:14).


This may be extended to include the Greek material, where Beekes considers δερκομαι an innovation (2010:318),
and Celtic (see fn. below).
Note that the publication (1899) predates the discovery and acceptance of the Anato-lian languages as Indo-European (1902 for the first suggestions; 1915 before Hrozny settled it). It is now clear that the only branch where the root does not figure in a suppletive paradigm is Anatolian (Frantíková 2014:60).


Indo-European Morphology

The Hittite system

Having thus assembled the Graeco-Aryan evidence, it is now time to look at the Ana-tolian branch, which will be done by examining its most prominent member, Hittite. It is subsumed for this paper that Anatolian was first to leave the PIE dialect continuum, and that the material encountered here should be weighed against that of core-IE to approach the structure of the proto-language. Hittite does not have the same distinc-tion of aorist, imperfect, and perfect as core-IE does, but brands a simple tense sys-tem, present and past, with an enigmatic division of the present into two classes: The mi- and hi-conjugations. The former is easily superimposable with the eventive core-IE verbs, viz. with similar form and function, but the latter has been the center of much heated controversy over the years 17 . What is beyond contention, however, is that the hi-conjugation shares formal correspondences with the perfect in terms of both intra- and extra-radical morphology, viz. that the stem formation is unanimously trace-able to the o-root vocalism, with the ablauting zero-grade in all but a handful of verbs that brand an inherent o-e alternation,and the endings that are similar, if not identical, to those of the core-IE perfect. The main points of criticism for the perfect provenance of the hi-conjugation is in part formal (the lack of ubiquitous reduplication (although core-IE indubitably had one such, *uoid-/uid- ‘know’))  and in part functional (the Hit-tite hi-verbs show no considerable difference in function to the mi-verbs, and certain-ly not the stative/resultative function commonly ascribed to the Graeco-Aryan perfect).

Anatolian is notably devoid of verbal suppletion as seen above in the core-IE stock, and is instructively the only branch of IE that does not have suppletion in the para-digm for the copula *h 1 es- (Frantíková 2014:60), demanding an explanation to why this characteristic feature is absent. Two possible scenarios present themselves:

1. Anatolian eliminated verb suppletion completely. This is epitomizes the schwund
hypothesis by assuming that the very widespread suppletive character of all other IE
languages were completely eradicated here. Typologically, this is a viable explana-tion, albeit radical, as analogy would be an obvious impetus to align the system.

2. Core-IE went through a reformation of the verbal system that created a defect paradigm in which Anatolian did not partake. This hypothesis demands further qualification through direct and/or circumstantial evidence.


The main pieces that need to be fitted together to attempt a reconstruction of the inherited aspect system thus amount to the following:

For Graeco-Aryan:
eventive and perfect

For Hittite: mi- and hi-conjugation

17 For a thorough discussion and research history, see Søborg (2015:88-145).




Having thus observed what can be readily inferred from Anatolian and core-IE, res-pectively, on their verbal systems with regards to aspect, it may be instructive to look at aspectual systems as they exist in a broad range of languages across the globe, including two reconstructed geographical neighbors of PIE.

Among the native American languages, the Northern Iroquoian language of Seneca is described (Mithun 1999:165-66) as a having a tripartite aspectual system akin to that reconstructed for core-IE,viz. perfective (single events, cf. the aorist), imperfective (ongoing or habitual events), and perfect (resultant state, ie. stative). Some verbs only have stative forms as they never denote events; contrarily,other verbs have no perfect as they inherently express events without meaningful consequences, e.g. to sing and talk.

A more elaborate system is found in Koyukon (Mithun 1999:166-68) of the Athabas-kan-Eyak-Tlingit language family where change in aspect is indicated both through ablaut and suffixation (along with phonological processes in connection herewith), and it is possible to express a host of different aspects, e.g. perambulative, semelfac-tive, durative, transitional, etc. In modern standard Chinese, an aorist (‘action’) verb may similarly transfer to imperfect (here: ‘stative’) by the addition of the suffix -zhe, e.g. chuān yīfu ‘puts on clothing’, with the original telicity of the lexeme, as opposed to the marked chuān-zhe yīfu ‘is wearing clothes’ (from Norman 1988:164).

The verbal system of Wolof of the Niger-Congo language family is inherently aspec-tual(where the aspect marker actually connects to the pronoun rather than the verb), and temporal information outside the intrinsic aspectual references must be overtly marked on the verb, e.g. with the preterite suffix -(w)oon (Stewart & Gage 1970:371). In the reconstruction of the Semitic languages, Moscati describes the verbal system as one of two aspects, imperfect and perfect (1964:131),  where  “semantics connections may be somewhat fluctuating” (1964:122).  

Excourse #1: The Uralic system

For proponents of the Indo-Uralic hypothesis, describing the earliest stages of Indo-European, otherwise only attainable through internal reconstruction (as practiced in the present paper), entails comparison with the Uralic language family and thus allows for additional external evidence. On a general note, the Uralic system comes across as extremely simple in relation to the multifaceted PIE system traditionally reconstructed, cf. above. But like, and in a seemingly diametrically opposite manner from, the development of the nominal systems (cf. Bjørn 2014), the verbal paradigms appear to coalesce in the prehistoric perspective. The descriptions by Collinder and Abondolo of the proto-Uralic verbal system comes eerily close to that proposed for PIE, and, if willing to concede Indo-Uralic genealogical affinities,supporting evidence for a very simple PIE (and further, Proto-Indo-Uralic) system thus amasses. Even for non-adherents to the genetical relationship, the typological parallels are informative:


"In Samoyed, the present tense - called aorist by some grammarians - refers to the past time (preferably the immediate past) when the verb has a non-durative mode of action or is used in a terminative aspect[. ...] In Selkup, the present tense can refer to the past time or the future as well as the actual time." (Collinder 1960:245)

"There is no clear past-tense marker reconstructable for [proto-Uralic],and the original tense system may have been one similar to that found in Samoyedic,where the lexi-cal, i.e.intrinsic, aspect [=Aktionsart,RHB] of verb roots determines the semantic force of their finite form. Thus an inherently non-prefective, stative verb such as 'lives' normally had present or non-past meanings, while a verb such as 'dies', with inherent punctual, perfective aspect, normally had past-tense meaning." (Abondolo 1998: 27)



Formally, the Graeco-Aryan system is easily reduced to the eventive-stative oppo-sition. What remains to discuss is the development of the aspectual split within the eventive.But before this is embarked upon with internal evidence alone,a comparison with the Hittite material deserves scrutiny, and it is, in fact, difficult to find arguments heavy enough to dissuade the simple addition of two and two (the mi-conjugation with eventive and the hi-conjugation with stative).

The weightiest candidates would be:

1. The hi-conjugation is an innovation.

2. The hi-conjugation is old, but does not correspond with the core-IE perfect.
a. lack of reduplication
b. semantics

(1) The obliteration of the Hittite material as being wholly innovative would be devas- tating to the system proposed here, and it is evidently a minority view that the forma- tion should be secondary within the Anatolian sub-branch,as Szemerényi does when he passes off the Hittite hi-conjugation based on Luwian material (1995: 245 - 46). Again,reference must be made to Søborg’s recent and thorough thesis (2015:71- 75)  in  defence  of  Hittite’s  antiquity.

(2) This question has captivated many a Indo-European scholar for the past century,
and the two key points of criticism are treated here, albeit briefly:


For the perfect hypothesis to work out, the lack of reduplication in Hittite must be ac- counted for, and two solutions present themselves: Either, Hittite has removed the prefix, as is similarly observable within the Germanic branch 18, or core-IE hasgene- ralized it. The fact that reduplication does exist in a few hi-verbs (cf. Søborg 2015: 115) betray that the formant has not vanished completely in Anatolian and shifts the gravity of the argument slightly towards generalization in the Graeco-Aryan branch. Reduplication is also known in the imperfect and aorist, and either the semantics of reduplication (whatever that may be) have made it popular in the perfect word class or the formant itself was enough as a sign of markedness to spur a generalization to aid the ablaut and special endings 19 . At any rate, the mere fact that core-IE perfect is triple marked with reduplication, ablaut (the two former could be argued to be mor-hophonemically linked), and distinct endings from the imperfect complex, should hint to superfluous material.

Cf. LIV (2001:10ff.)


Cf. Hoffmann who considers the phenomenon analogical (170:540).




The functional gap from core-IE resultative/stative to the neutral Hittite hi- conjugation is similarly addressed by Søborg:

“The perfect is traditionally regarded as the third aspect of the Indo-European verbal system, opposing the imperfective and the perfective aorist. But actually, a resultative stative is semantically imperfective; the perfect is thus probably better described as an imperfective aktionsart with resultative-stative   function.”  (2015: 118)

“The resultative nuance can be seen as a developing force within the perfect func-tion, starting as a denotion of the result in the current state and gradually focusing more and more on the implied completed action leading up to the state,virtually deve- loping into a pure past tense. This transition is easy to understand from verbs like *h1 ge-h1 gór-e ‘is  awake’ (Ved. jāgara, Gk. ἐγρήγορε) > ‘is awake (from  being awoken)’ > ‘has awoken’ > ‘awoke’. [...] Thus, the perfect can be reconstructed as a fairly pure stative, the resultative function having probably arisen secondarily in Core Indo-European and continuing to do so in the individual branches, often leading to a functional  shift  into  pure  preterite.”    (2015:119)

and further:

“Thus, the meaning of some imperfective and perfect verbs were very similar, if not
identical. Especially meanings such as ‘say’, ‘yell’, ‘shine’, etc., do not have a very  
strong stative nuance, and the verba dicendi hardly any at all. I believe that this
could have been a factor in the neutralization of function in the development from
the Proto-Indo-European perfect to the Anatolian *Hai̯ -verbs.”  (2015:120).

Cyclical diachronic shifts from stative to general active past tense and vice versa is
actually a common phenomenon, e.g. observable from Latin to Italian (Comrie 1998: 87). The weight of the semantic argument is severely lessened, and although the above treatments by no means prove that core-IE perfect and the Hittite hi-conju- gation are related,it is sufficient to denounce the negative claim once propounded.

The obvious formal correspondences in endings and ablaut speak for perfect as the purveyor of the hi-conjugation, cf. Kloekhorst who also considers the hi-conjugation parallel with the PIE perfect (2008:167ff.). The two systems can thus be schematized as parallel in form and function,

Graeco-Aryan Eventive Stative

Hittite   mi-conjugation hi-conjugation

and are thus superimposable as to create

PIE Eventive mi-conjugation (e-stems) Stative hi-conjugation (o-stems).



How PIE eventive became core-IE imperfect and aorist

Having established the PIE eventive-stative opposition, all that is left is to qualify the core-IE split of the eventive into two distinct paradigmatic aspects and to discuss the material presented above in this new light. As there is no overt morphological dis-tinction between the two, the solution rests with the lexical meaning of the individual verb (aktionsart) in combination with the extra-radical aspectual markers.

Hoffmann (1970:531-32) makes a basic four-way aktionsart distinction in the durative (e.g. search), punctual (find), closely related hereto the momentative (get up), and lastly the terminative (bring,as opposed to durative carry). While the imperfect is dura- tive, the root aorist is by nature punctual, and definitely not durative,and it can thus be deduced that root aorists aboriginally had a punctual meaning   unlike the root imper-fects (“präsens”),regardless of the more or less problematic semantics in the synchro- nic readings of the attested daughter languages, i.e. the system is neat, but individual words may present themselves as problematic. As treated above, the imperfect had a nature that made it available for the deictic present, and Hoffmann’s  identification  of  an  inherent lexical opposition suffices to find the seed of the split in the eventive category.

Another impetus is found in the temporal hic et nunc-marker which introduced a new dimension to the verbal system 20 . This marker is definitely old, it must be noted, as both Anatolian and core-IE employ it, but the paradigmatic tense is an innovation, which is resonated by more than a few prominent Indo-Europeanists, e.g. Weiss (2009: 379), and the specific development from an aspect language to a tense language is still visible in Vedic (cf. Kiparsky 1998:29) and Greek as we have seen above 21 . The use of the injunctive in prohibitive mā  sentences in Vedic all but proves the aspectual nature of the aorist-imperfect distinction. Comrie consequently identifies the two innovations that spurred on the development of a temporal distinc-tion in PIE, viz.on the one hand the primary endings (chronologically secondary) with the deictic hic et nunc marker *-i as opposed to the original secondary endings un-marked for time (as expected in an aspectual system); and on the other the augment, serving to place a verb firmly in the past (1998:85).


Comrie recognizes tense as a plausible areal feature of the European languages, i.e. including non-Indo-European (1998:87). It is probably too far fetched to search for a reason for the temporal innovations in the rise of complex societies connected with the agricultural expansion from the Middle East, but the conjecture is nonetheless tantalizing.


Hittite is, of course, innovative in this regard having neutralized the stative aspect of their hi-conjugation to enter into an equal relationship with the mi-conjugation (cf. Søborg 118ff.) to form a tense language.


“The injunctive testifies to the relatively recent grammaticalization of mood and tense as inflectional categories in Indo-European, and how its loss goes hand in hand with the enrichment of the tense system through the conversion of the original  aspects  into  tenses.”  (Kiparsky 2005:15)

Vedic Sanskrit definitely represents an aspect language, despite Szemerenyi’s attempt to muddle the evidence (1990:307). Characteristic for grammars of Sanskrit is the synchronic description of the verbal system as past-present oriented (e.g. Macdo-nell 1916:339ff.), partly due to the fact that the augment evidently introduces the con-cept of the preterite in a purely temporal sense, yet it does by no means imply that the provenance of the system is temporal.

The augment is notably only relevant for the aspectual Graeco-Aryan languages. If tempus was the inherent meaning of the distinctions, there would be no need to introduce the morpheme. It may  thus  be  a  valuable  insight  to  the  inner  workings  of  a  language’s  shift from aspect to tense.

Whether the augment existed outside of the languages in which it is attested is evolu-tionarily unclear; it depends on whether the creation of the primary endings was enough to split the inherent aspects into a temporal system.

Meier-Brugger posits verbal suppletion as possible evidence of an older system (2003: 253), and already Osthoff recognized that different aktionsarten must be the provenance of the phenomenon (1899:44ff). Frantíková asserts that the most likely interpretation of the voluminous core-PIE suppletive complex compared to the non-existing situation in Hittite is that the former have acted on the paradigmatic vacuum that incurred with the formalization of three distinct aspects (2014:59), an innovation absent from Anatolian. Opposing this view is Szemerenyi who denies the relevance of suppletion and refers to its commonality in modern languages, cf. modern English go versus went (1990: 301ff), but suppletivism necessarily has an impetus, and it is obviously encountered en route to core-IE from PIE. Suppletive paradigms are thus lexical items that were not easily reconciled with some or other of the aspect invento-ry in the new system, and they retained their aktionsart as primary means of aspect rather than the newly formalized grammatical markup.

The aorist-imperfect distinction is thus nothing more than an inherited telic lexical dif-ferentiation that was redefined in a variety of ways in the daughter languages.The im- perfect-perfect split hence represent the original aspect,visible in archaic ablaut, later supplied with the characteristic reduplication of the perfect, whereas the imperfect-aorist complex is a vestige of a central telic differentiation, overlapping in morphema-tic markup that was aspectually neutral (root forms,reduplication,and thematic vowel), yet differing emphatically in the secondary derivational markers that irreversibly allot-ted either imperfect or perfect (aorist) aktionsart that was mutually incompatible, e.g. the exclusively imperfect *-sḱ -e/o- and *-n- as opposed to the equally exclusive aorist *-s-, whatever their provenance. The subsequent use of the imperfect and aorist within a single paradigm may be hard to fathom, but the defect paradigms, re-stored with suppletion, actually epitomizes this state. The Hittite system, on the other hand, is thus the original system, in terms of grammatical aspect, without the splitting of the eventive class as in core-IE.


The systemic phases

Anatolian and Graeco-Aryan thus co-operate to illuminate the earliest stages of PIE attainable through internal reconstruction. A sketch chronology of the developments from the aboriginalsystem is given below with typological examples to testify to the very plausible nature of the different stages.

*PIE stage 1

- The primary grammatical verbal opposition is marked by ablaut and distinguishes eventive from stative.
- Tense is not marked morphologically.
Compare the reconstructions of the Semitic and Uralic language families.

*PIE stage 2

- The temporal hic et nunc-marker may be suffixed to the personal endings of certain verbs
- Semantically detailed aspectual affixes appear

Compare how Wolof shows how an aspectual system may introduce concrete tense through the use of overt morphology distinct from that of the aspect, and the host of different formations that may alter the aspectual character of a given lexeme more or less emphatically in Koyukon. This is the last step reconstructable for both Anatolian and Graeco-Aryan. From here on they are treated separately.

*Anatolian step
*Graeco-Aryan step 1
- hic et nunc is spread by analogy to all
- aspectual affixes are grammaticalized
non-past verb forms
- the augment is introduced
- semantic depletion of the stative
Compare Mandarin Chinese where a suffix
may change the aspect of a given verb.
*Graeco-Aryan step 2
- aspectual affixes qualify a split of the
eventive verb class
- stative reduplication is generalized
Greek and Vedic
- hi- and mi-conjugations exist on par
- tripartite aspectual distinction between
with no marked difference
imperfect, aorist, and perfect
- unmarked personal endings denote
- dawning temporal system
past tense

Compare the identical system in Seneca.



Excourse #2: Adding causative-iterative and molō verbs to the perfect category If the only truly archaic verbal distinction in PIE is the quality of the root vowel, extra-radical elements cannot qualify the pertinence of to one group over another. Therefore, the fact that o-stems also occur in causative-iteratives that belong to the core-IE imperfect class must originally originate in the stative class;The semantics are self-explanatory, viz. ‘make [state]’, derived  from ‘[state]’ with the suffixation of *-éie-, yet the endings differ. Something is deficient about the perfect endings so as to preclude their use in the causative that then applies the eventive endings. The answer, again, seems straight forward if the medial connection with the perfect endings is resuscitated, an idea that Sihler rejects (1995:567). The opposition is clear when the stative verbs use one set of endings and the active the same very ones known from the eventive class. This ultimately removes the superfluous markings of the perfect completely: The re-duplication is analogous from some aspect marker also found in the eventive verbs, and the endings are simply the medial endings, leaving the true stative marker as the root vowel o. The molō verb class may then represent an older less (or un-) marked causative in the sense *ghrobh - ‘dug’  > Goth. graban ‘make dug = dig’;;  *h2woks- > Goth.wahsan ‘make grown = grow’,instructively in opposition to the e-grade (imper- fect stem) Greek (w)έξομαι (examples from Fortson 2010:181);and *mólh 1 - > Goth. malan and Lat. molō (Kroonen 2013:351).The later *-éie- formant of the causa- tive is thus in the same stratum as the imperfect and aorist markers (and likely the re-duplication syllable in the perfect). This interpretation opposes that of Søborg, who infers that this class was incorporated into the inflection at a later stage (2015:120).

The iterative semantics of Greek and some Slavic *-éie- formations demand that this phenomenon also be traced back to at least core-IE (Fortson 2010:99), but the se-mantic shifts, as we have seen, does not carry sufficient weight to damage the theory. Instructively,the reverse semantic shift of causative to iterative happens in the reverse in Tocharian with the iterative suffix *-sḱ - (Sihler 1995: 503 - 05). Whether formal af-finities exist between perfect and middle must remain unsettled. But if molo are inflec-ted with imperfect endings, then the transition from medial stative,  “ghrobh-e”,  to  active  “ghrobh-(éie-)ti”  (vel sim.) is qualified.

Excourse #3: Contemporary interpretations of the PIE system

Sihler refrains from adducing anything about the PIE system as whole, but prudently lingers with the Graeco-Aryan system itself. Here, he famously makes the dichotomy of eventive (imperfect and aorist) versus stative (perfective) verbs (1995:442ff.), an inference possible already from evidence available in the 19th century! It is thus a bit anachronistic when most handbooks will echo the Graeco-Aryan system as that of PIE, e.g. Fortson who identifies the system as tense-based, recognizing three diffe-rent stems: present, aorist, and perfect (2010:88ff.). More typologically correct is Com-rie when he posits a three-way aspect distinction devoid of tense: imperfective (pre-sent), perfective (aorist), and perfect (1998:84ff.), but the temptation of projecting the defect Graeco-Aryan system back to PIE is evidently too great to resist. The interpre-tation by Kloekhorst, who is an asset in the field Anatolian and consequently for the “final  frontier”  of  IE  linguistics,  exemplifies  the  classical  analysis:

“On the basis of the fact that the formation of the PIE imperfect (ablaut *e/Ø‘, seconda-ry endings) was identical to the formation of the root-aorist  (also *e/Ø‘-ablaut and secondary endings), the two categories easily fell together in Hittite.


On the basis of the root-aorist a new inflection with primary endings (= addition of -i) was created which had presentic meaning and was formally identical to the root-present.”  (2008:148, my underlining)

LIV similarly suggests that the perceivable aspect opposition is an innovation that occurred when aktionsarten in the perfective aorist collapsed with the imperfective presence (2001:10ff.).

It is striking that the reverse conclusion to the material, i.e. that the identity of the im-perfect vis-a-vis the aorist is aboriginal, is so hard to come by when even the Graeco-Aryan evidence, albeit perhaps circumstantially, point in the same direction. Prudent scholarship is seemingly steeped in 19th century tradition. Not so prudent scholar-ship, on the other hand, can be extremely adventurous in positing new systems, e.g. Jasanoff (2003) who produces an elaborate and intricate system of several verbal classes for PIE to account for the Anatolian hi-conjugation.



The original research question for the present paper was evidently the least problematic part:

There is no indication pointing to an aboriginal tripartite aspect system in PIE. Aorist and imperfective must so obviously have come from a single verb class that the ma-jority of the present paper has been spent around the topic, especially the aboriginal distinction between the eventive and stative systems, rather than directly on it.

Synchronic descriptions are all proper and fitting, and the inferences made by some the field’s most esteemed scholars are by and large correct 22 ; only the courage to accept that the Graeco-Aryan system is not truly original and further the audacity to formulate a common verbal system for Indo-Hittite has been generally wanting. When embarked upon, hugely complex systems (cf. Jasanoff 2003) seem to be favored to simpler core constituents with morphematic markup available for further detailed dis-tinctions. Occam’s razor would cut itself blunt if applied to Jasanoff’s hypothesis.  The  system presented above is neat and typologically sound, yet there is evidently much more work to be done.

Although the system may seem overly reduced,it only appears so in light of the highly complex inflectional system known from the oldest attestations of IE languages. Much of the inflectional morphology has been successfully reconstructed back to an analyzable stage where the language rather must gain the epithet agglutinative - or even isolative -, and there is thus no reason to assume a high number of stem for-mations for the proto-language. That is not to say that  the  language  was  simple;;  the  constituents  simply  lived  “freer”  lives  as  particles (vel sim).

Sihler’s  prudent  reconstruction  of  the  eventive  vs.  stative  deserves  honorable  mention  (1995:442ff.).