NEW TRENDS IN INDO-EUROPEAN LINGUISTICS Number 3 | August 30, 2021 @FIU 2.0* Transitivity Direction in Proto-Indo-European: Direct, Inverse, and Undirected Verb Forms By Roland A. Pooth Abstract. I present a novel reconstruction of the morphs and categories of Proto-Indo-European verbal inflection. I claim that the common ancestor of the Indo-European languages exhibited hierarchical direct and inverse indexing of transitivity direction (Jacques & Antonov 2014). The novel model deviates from the traditional model in that it is typologically different and based on a combined methodology. I am using general linguistics and the comparative method to complement each other. I evaluate and interpret the results of the comparative method against the background of cross-linguistic findings and a broader typological pattern. This article is part of work in progress towards a revised reconstructed synchronic grammar of Proto-Indo-European. Keywords: Linguistic reconstruction of Proto-Indo-European verbal inflection, transitivity direction, direct-inverse system, hierarchical indexing D R . R O L A N D A . P O O T H was Senior Research Fellow at Ghent University (Belgium). He lives in Cologne, Germany. Born in 1972, he received his Ph.D. from Leiden University several years later [E-mail: roland.pooth@gmx.de]. 1. Introduction 1.1. Aim, Research Questions, and Methodological Dos and Don’ts It may go against a dominant belief among philologists and Indo-European linguists but the aim of this article is to pursue the claim that Proto-Indo-European verbs had hierarchical indexing and were not inflected by attachment of fusional “endings” for person, number, voice, aspect, tense, and mood to verb stems. Even if Indo-European languages like Vedic Sanskrit, Ancient Greek, and Hittite etc. must be analyzed this way, the morphotactic structure of their parent language Proto-Indo-European might have been different. The general research questions pursued in this article thus run as follows. Is it reasonable and possible to apply a proper linguistic morph analysis to the reconstructed Proto-Indo-European verb forms on their own, a bit more independently from Indo-European grammar? And does such a morph analysis yield different results? Are the results typologically sound, even if they are different from the traditional model? And when we finally compare the independent reasoning process to the standard, whose results are linguistically proper and more promising, and which model is to be preferred? For investigating these research questions, I make use of a combined methodology. I believe that methods of general linguistics and comparative philology should always complement each other. The established results of Indo-European comparison are evaluated from the background of a typology-based linguistic morph analysis. There is a long-standing discussion on the interplay of typology and reconstruction in Indo-European Studies. It has yielded intriguing theories, such as the Glottalic Theory. The need of typologically consistent reconstructions has also been widely discussed for phonology (Kümmel 2015). Even if typologically informed reconstructions, ranging from phonology to syntax, have been now around for decades, one can still diagnose that the morphological and syntactic model of the proto-language presented in the recent handbooks deviates from Classical Greek and Vedic Sanskrit grammar only minimally, mostly in phonological features. Here, the Proto-Indo-European word forms are not solely analyzed from the average comparative perspective on Indo-European morph boundaries. My methodological reasons for using combined general linguistic and comparative methods have been outlined in more detail in my paper “Towards a rigorous linguistic reconstruction of morphs and categories” (Pooth 2017). The argumentation given there can be summarized here as follows. I claim that the application of the comparative method to Indo-European morph boundaries and grammatical categories has yielded a highly problematic and anachronistic model of Proto-Indo-European morphology and syntax. When judging from general epistemology and its distinction of reasoning processes, I diagnose that the comparative reasoning is based on a circular deduction and a problematic axiom (see below). The German term “Nullhypothese” (‘zero hypothesis’) has been used to refer to this axiom (Hill p.c.; see Appendix 3): it is the assumption that Proto-Indo-European grammar, morphs, and grammatical categories must or should be identical to the ones occurring in Indo-European languages. It is further based on the principle claim that such an assumption would be “less costly” and “most economical” way of assigning functions to reconstructed forms. However, we can drop the question of economy here because I argue that the Nullhypothese is unlikely. This overrules any argument based on economy. I claim that it is unlikely from the outset assuming that Proto-Indo- *The FIU 2.0 is a homage to the social sculpture established by Joseph Beuys. The FIU logo is quoted from his “Aufruf zur Alternative”, Frankfurter Rundschau, December 23, 1978 © by the author. Published under the Creative Commons Attribution-NonCommercial 4.0 International License CC BYNC 4.0: http://creativecommons.org/licenses/by-nc/4.0 2 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European European grammar had the same categories and features as occurring in Indo-European languages. I further generally argue that a circular back-projection of younger morphs and categories is a corollary of the comparative method, if this is the only method used for the linguistic reconstruction of morphology and syntax. The circular reasoning process can be illustrated as follows. The morphological parsing of the standard model is based on the preconception that PIE morphs and morph boundaries should only be reconstructed by means of a deductive reasoning process, which is based in morph and category equations. The standard reasoning usually implies that if Ancient Greek, Vedic Sanskrit, Hittite, and all the other Indo-European languages agree in exhibiting a particular grammatical category or feature x, we would be allowed to automatically claim on this basis that Proto-Indo-European also had this category or feature x, only in a somewhat different phonological shape. Thus, within an initial inductive step within the standard reasoning, the following “rule” is established on the IndoEuropean correspondences. Within a deductive step of reasoning, the conclusion (C) is drawn. However, I argue here that the whole deduction is based on the highly problematic axiom (B). an unknown language of its own. I think this additional step is really necessary as a linguistic test in order to avoid the mentioned automatism and circularity. Thus, in order to avoid a misleading back-projection of a younger structural analysis to the preceding layer, we need to introduce another step of reasoning: We should now stand back from Indo-European grammar. Analytical inferences concerning the proto-morphology and proto-syntax should be drawn more independently from the younger layer – and this is better done with a general linguistic and typological epistemic background. A third research question that can be pursued here more or less in between the lines thus runs as follows: If an independent morph analysis leads to results that are typologically different from the traditional model, should we give up the assumption that Proto-Indo-European wasn’t too different from Indo-European on the morphosyntactic level? To sum up, from a general epistemological perspective and from a general linguistic background, it is almost self-evident that a proper linguistic update of the traditional comparative analysis is needed. The present article aims at filling this gap. I use the following representation for the steps of reasoning that have been outlined: A. Rule based on facts: All Indo-European languages agree in having a grammatical category or feature x (e.g. person indexing like Vedic 3sg pres. ind. act. -ti). Having this category or feature is thus a general “rule” that is based on a corresponding Indo-European category or feature x. B. Axiom: As an “unwritten law” it is often implied that Proto-Indo-European morphology and syntax was identical to Indo-European grammar and that a proto-language should only exhibit the morphs and categories typically found when comparing its daughter languages. However, this is in fact a speculative assumption that may be probable for some of us, but is not probable for others including me. As outlined, it has been said that this should be our first and only Nullhypothesis (Hill p.c.). In other words, there is even a kind of dogmatic “order” to consider this projection to be reasonable. In my view, however, it is epistemologically problematic and quite probably just a flaw. C. The conclusion is usually drawn as follows: Given that this axiom (B) must be true, we “must deduce” that ProtoIndo-European had the category or feature x. a. TPnégwts. The German term T r a n s p o n a t refers to a word form that is generated by simple back-projection of the respective Indo-European phonological rules, as if the word form and its Indo-European meaning and grammatical function belonged to Proto-Indo-European synchronic grammar without implying that it really did. Transponate are marked by superscript TP (Hittite nēkuz → TPnégwts). If it can be argued on positive internal evidence that such a Transponat should further be interpreted as a potentially realistic and reconstructable Proto-IndoEuropean word form, it is marked by superscript TP before the asterisk (TP*négwts). b. The term E q u a t i o n F o r m u l a is used here for a phonological Transponat that is based on a given Indo-European equation and correspondence. It is important to distinguish two sub-types of equation formulas: ba. “Formally reliable” equation formulas are based on formation patterns that are less frequent, unproductive, not potentially innovative but potential relics in Indo-European languages, e.g. EFgwʰénti ‘he, she is killing him, her, it’, EFḱéioi ‘he, she is lying down, has been laid, put down’ bb. “Formally and functionally less reliable” equation formulas are based on formation patterns that are frequent, productive, and thus potentially innovative in Indo-European languages, e.g. EFbʰéreti (Vedic bhárati etc.). It is not implied that all equation formulas can automatically be reconstructed as a Proto-Indo-European word forms in their very form and function – and without additional re-modelling of the equation formula’s form and function. Equation formulas are marked by superscript EF (Vedic bhárati :: Gothic bairiþ etc. → EFbʰéreti). Especially such equation formulas that are based on frequent, productive, and thus We can immediately see that this reasoning process is problematic because it rather disallows PIE grammar to be different. In order to solve this problem, I suggest that we should test the standard model against a more independent linguistic analysis, which can be applied to reconstructable forms like */gwhént/ ‘he, she, is killed/slew him, her, it’. In my view, this analysis should be applied to the given reconstructed form-function correlates on their own – as if they were potentially real word forms of N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 potentially innovative IE formation patterns are suspicious because they can be Scheingleichungen of parallel but independent innovations. Such equation formulas are marked by superscript LR for “less reliable” here (LRbʰéreti). Such forms are not automatically reconstructed in their corresponding forms and functions for Proto-Indo-European without additional methodological inferences and changes (Pooth 2017). This is a major methodological disagreement between the traditional approach and the one pursued here. Finally, reconstructable PIE word forms are to be generated by means of e x t e r n a l a n d i n t e r n a l reconstruction. E.g., the reconstruction of PIE 3sg *bʰéro(i) is based on a less reliable equation formula LRbʰéreti. The corresponding IE forms are taken as parallel innovations. This EF1 is mapped upon or against the more archaic formation pattern of EFḱéioi. It is claimed that Vedic bhárati, Greek φέρɛι, etc. go back to a form belonging to the formation type of Vedic śáye, etc. This finally leads to the reconstruction of a 3sg detransitive form *bʰéro(i) (Pooth & Orqueda 2021, Pooth 2014, Watkins 1969). For methodological reasons, I thus distinguish among provisional forms (“pro-forms”): Transponate and equation formulas. These “pro-forms” are considered as being artificial products that can only constitute a provisional basis for reconstructing the potentially realistic word forms, morphs, formation patterns, and formal and functional categories of the Proto-Indo-European language. 1.3. Remnants of grammatical distinctions not continued as such in IE languages The axiom outlined above is heavily contradicted by formal and functional discrepancies among Indo-European languages. If it was true that Proto-Indo-European morphology and syntax was identical to Indo-European grammar, many issues would remain inexplicable, e.g. the comparative issue that a form with *-u̯ é- is continued as 1pl form in Anatolian languages but 1dual in Vedic Sanskrit, Greek, etc. This section provides three cases that can serve to illustrate the problems. We all know that there are a few Proto-Indo-European grammatical categories that are maintained in Ancient Greek and Vedic but were abandoned or modified in other Indo-European languages. For instance, the OPTATIVE (-IRREALIS) mood is maintained in Greek and Vedic 1 Abbreviations: EF = equation formula; TP = phonological transponatum. IR = internally reconstructed word form. 2 In general, I follow the Leipzig Glossing Rules and make use of the conventional glosses: ABS = “absolutive” or “anti-agentive” unmarked case, ACC = accusative case, ACT = active voice (agentive with agentive verbs), AGT = agentive case, COL = collective number (group-like sociative-associative plural ‘they are doing something together’), DTR = detransitive or “protomiddle” voice, NDUR = non- 3 but is not continued/not existing in the Anatolian branch. It has been modified to a general subjunctive from PIE to Latin and Proto-Germanic: cf. Vedic Sanskrit 3sg and 3pl 2 imperfective (= “present stem”) optative active syā́ t syúr 3 = Old Latin subjunctive active siēt sient < Proto-IndoEuropean (henceforth PIE) 3sg *h1siéh1t and 3pl *h1sih1ént *h1sih1érs *h1sih1ér → post-PIE *h1sih1r̥ ́ (s) > Vedic syúr, etc. Further below, I return to the question why we can reconstruct three different 3pl forms ending in *-r-, *-rs, *-nt-. On the other hand, a thorough Indo-European comparative investigation also reveals remnants of morphosyntactic distinctions that are NOT continued as such in a single Indo-European language. These grammatical categories are UNKNOWN and ABSENT from old Indo-European languages or from Indo-European languages in general. If these categories existed in the common ancestor grammar, they must have been abandoned or functionally modified by the time the protolanguage developed into a post-PIE variant cluster with posterior dialects in close areal contact. I argue here that, by additional application of the method of internal reconstruction, we are able to detect and identify such parallel independent innovative categories of the post-PIE dialects. These parallel innovations presumably occurred shortly after the common PIE synchronic grammar broke up. 1.3.1. The 1+2|2+x inclusive category A first illustrative example of a category that is non-existing in Indo-European languages but should better be reconstructed for their ancestor, is the 1+2|2+x INCLUSIVE person category ‘we, that is, you and me/us’. It is continued only modified as a first-person DUAL INCLUSIVE category in Early Vedic (RV, AV, e.g. Vedic ganvahi (TPgwm̥ u̯ édhh2) ‘we two, you (sg) and me, go (up) together’, attested in RV 8.69.7b). In later IE languages, it is reflected as a first-person DUAL without clusitivity (that is, it is both inclusive AND EXCLUSIVE) in Old Church Slavonic and Gothic (Pooth 2011). This is a well-known 4 diachronic path. durative (“aorist-like”) aspect, PL = plural number, PROG = progressive aspect, SG = singular number, TOP = topical referent, NTOP = non-topical or “anti-topical” referent. The symbol > generally means ‘… developed into … by sound law’ but here it also symbolizes the transitivity or causation direction (e.g., 2sg>3sg = 2sg acting on/causing an effect in 3sg). 3 Weiss 2009: 416. 4 This development has been described by Dixon 2010 II: 194 for several Australian languages. P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 4 Figure 1. 1dual forms from 1pl inclusive forms 5 PIE Early Vedic OCS, Gothic 1+2inclusive → 1dual inclusive → 1dual (incl|excl) *gʷm-uó(s) → gan-vahi (middle) … The fact that 1dual forms are overwhelmingly INCLUSIVE in the Early Vedic period is a functional trace and positive evidence for this category, and constitutes the decisive argument (Pooth 2011). Another piece of evidence in this puzzle is the comparative fact that the Vedic 1dual nom. pronoun vā́ m (RV 6.55.1) and the 2dual acc. dat. gen. clitic vām are so very similar. The formal identity leads to the following claim of functional identity: It points to a PIE pronoun *(h1)ué(h1), clitic *uo(h1) for both 2dual and 1dual inclusive (Pooth 2016c, section 9). The internal evidence thus supports the hypothesis that PIE had a 1+2 INCLUSIVE that was likewise a 2nd person INCLUSIVE/ASSOCIATIVE (‘you & an underspecified participant’ = ‘you and me’, ‘you and your companion’, ‘you and your (uncountable or countable) company’). This can be glossed as 1+2|2+X. Such a polysemy is not unknown among the languages of the world. The pronominal polysemy fits with the polysemy of the corresponding verbal category reconstructed here. Thus, the PIE suffix -u- can be interpreted as a second person (addressee, oppositional, adversative) suffix underspecified for the other included/inclusive participant. NB. The form *gʷmuó obviously lacked an additional number suffix and was thus less specified, whereas two additional forms 2nd person inclusive sociative-associative *gʷm-uó-h1, plural *gʷm-uó-s had additional number suffixes (they are structured in parallel with the 1pl exclusive forms *gʷm-mé, *gʷm-mé-s). The PIE form *gʷmuó(s) has been interpreted as a DETRANSITIVE (“proto-middle”) form by Pooth (2015) with the detransitive marker being the o-grade here. The active counterpart was *gʷmué(s). 2nd person plus x category to 1st person DUAL by application of the method of internal reconstruction. 1.3.2. The PIE progressive aspect A second illustrative example of a PIE grammatical category whose original function did not survive in any IE language as such is the PROGRESSIVE aspect. Kloekhorst (2017) argues for some functional traces in Old Hittite “present” forms. Its exponent was the suffix *-i. The idea that it encoded present tense cannot explain its overall 6 distribution (Pooth 2009a). This suffix wasn’t generally attachable; only part of the verb forms took it, e.g. 1sg active *gwhén-m-i ‘I am slaying, killing him, her, it’, 1pl active *gwhn-més-i ‘we (excluding you) are ditto’ etc. We can draw the straightforward inference that many PIE progressive forms developed into post-PIE IMPERFEC7, 8 TIVE aspect and PRESENT or IMPERFECT tense forms (Pooth 2009a). It is an implication of this inference that the post-PIE imperfective present category was marked by fusional portmanteau suffixes only at a post-PIE stage, but NOT BEFORE. The emergence of fusional TAM + voice endings with cumulative exponence *-mi, *-si, *-ti, *-mes(i), *-u̯ es(i), *-ten(i), *-énti, etc. (and variants) must thus be a SECONDARY functional and morphotactic innovation. In other words, these fusional suffixes do not go back to PIE in their Greek and Vedic forms and functions. The parallel independent post-PIE development can be illustrated as follows. Figure 2. The origin of IE present forms PIE progressive → *gwhén-t-i To conclude, we are able to identify a post-PIE functional narrowing of the PIE 1st+ 2nd person inclusive and 5 Abbreviations: PIE= Proto-Indo-European, PII = Proto-Indo-Iranian/Iranic, OCS = Old Church Slavonic, etc. 6 This suffix has hitherto been called “particle of the hic et nunc”, but it is neither assured that it was a particle nor is “hic et nunc” an adequate term. If it was a tense suffix, our expectation is that it was attached to all verbs to form a present tense form, but this is not the case (Pooth 2009a). 7 The development of progressive forms to imperfective present forms is common (Bybee, Perkins & Pagliuca 1994). 8 These progressive forms were extended by productive endings yielding new imperfective and present stems (Pooth 2017). 9 post-PIE Hittite imperfective (present) → present wh *g énti kuēn-zi ↓ Vedic Sanskrit imperfective present (later vs. perfective)9 hán-ti (vs. aor. vádhīt)10 The development of the PERFECTIVE (“aorist”) category was an innovation of the so-called “Inner IE” parts of post-PIE dialect continuum, presumably innovated in close areal contact. It is mistaken to reconstruct the traditional array of characterized aorist stems for PIE, except for the athematic reduplicated aorist, but this must have had a more specified (COMPLETIVE) meaning. Even “root aorist” stems cannot be rendered canonical PERFECTIVE stems in PIE because many of them were still compatible with the progressive aspect, e.g. *dhéh1ti (Pooth 2016). 10 García Ramón (1998) noticed the suppletion of Vedic pres. hánti vs. aor. vádhīt. But the lack of corresponding suppletive pairs N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 I return to the emergence of the “primary endings” and “secondary endings” in more detail below. To sum up, the back-projection of the Vedic and Greek so-called “primary endings” onto PIE seems to be an anachronistic mistake (thus already Beekes 1985). 1.3.3. The PIE collective number Another example of a PIE category that was obviously modified in post-PIE times is provided by the PIE COLLECTIVE (UNCOUNTABLE GROUP PLURAL) number category. E.g., PIE *uédor ~ *uédorh2 ‘water(s)’ (>*u̯ édo:(r)) besides PIE *uódr ~ *uóh1r ‘water’ (> Hittite wātar, etc.). It relevant for the methodological purpose pursued here to recognize that this number category is not maintained as a number category in any IE language. Nevertheless, it is one of the rare categories that comparative philologists reconstruct for PIE grammar (see the discussion by Eichner 1985, Litscher 2014), even if this category is no longer existing in this particular function in the old IE languages. The PIE collective number forms are continued as neuter singular or plural, or feminine singular forms in the later period: e.g., PIE *dóru-h2 ‘wood-COL’ → IE ‘woodPL’; Greek has a nom.-acc. SINGULAR n. τό ὕδωρ, gen. ὕδατος ‘water’, whereas Hittite has a PLURAL n. widār ~ witār ~ wedār ‘waters’ (vs. sg. wātar ‘water’). The Greek and Hittite forms seem to go back to PIE *uédor(h2) (*udón, *udnós). The locative and the oblique forms are continued as Vedic udán, udnás, etc. Greek reflects the generalized zero-graded root *ud- but the original accent on the root, whereas Hittite has the accent outside the root. It is thus a well-known comparative fact that the IE outcomes of PIE *uédor(h2) belong to different IE singular vs. plural number categories. Even if it is often assumed that *uédor(h2) was a neuter “singular” form that could be used as a plural number form, it is clear that it must have had a plural-like meaning in PIE. For these reasons, Pooth (2015) has drawn the inference that *uédor- was transnumeral and had both a GENERIC reading ‘water in general, water as type’ and an uncountable COLLECTIVE number reading ‘group of waters’ (but I only mention this here; discussing this hypothesis is not necessary.) The relevant methodological point is that the PIE form *uédor(h2) is continued in different IE languages by two different categories. Therefore, this comparative fact justifies the general assumption that there were major speaks in favor of parallel but independent root suppletion in Ancient Greek vs. Vedic (Kölligan 2007). 5 changes of original PIE derivational and inflectional grammatical categories (e.g. number categories). To sum up, I consider it imperative that within a proper linguistic reconstruction of the common ancestor of all Indo-European languages it must be our task and duty to reveal the semantic nature of all these lost and modified categories by means of internal reconstruction grounded in cross-linguistic patterns and based on general linguistic typology. Also, we must find reasonable motives for why these categories were given up or modified when the protolanguage broke up and developed into a bundle of dialects. To be clear, I think that it is unreasonable, and a kind of “naïve skepticim”, to a priori doubt or neglect the realistic and probable existence of PIE grammatical distinctions that were different from the ones existing in IE languages. All historical linguists are well aware that categories continually change their functional nature and that such changes can happen quickly and radically. Assuming that the opposite was true for the period in between PIE and IE is contradicted by the overall cross-linguistic diachronic evidence. In my view, such a state of mind is not a fruitful skeptical position. 2. Comparative settings Before we can analyze the Proto-Indo-European word forms, we need to make use of the comparative method. It is a comparative fact that the IE correspondences and word or morph equations point to minimally three differently encoded third person finite verb forms, both singular forms (*-o, *-to *-sto), and plural forms (Figure 3). These third person forms are provisionally termed D for the ones with the segment *-t- (the voiceless alveolar stop), S for the ones with the segment *-s- (the sibilant), SD for the ones with a combined *-st-, and Z (zero) for the ones without any of these segments. My further analysis is based on the following endings. Figure 3. Active and “middle” endings active 3SGZ 3SG.D 2PLD 2|3SGS 3SGSD 3PLZ 3PLD 3PLS *-t*-te*-s*-er / -r̥ *-ent- / -n̥ t*-ers / -r̥ s “middle” *-o*-to*-so- (2SG) *-stó*-ro*-ont- / -nto- P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 6 I use the term DETRANSITIVE (DTR) voice instead of “middle” or “mediopassive” for the Proto-Indo-European non-active voice category. I restrict the use of “middle” to the post-PIE stage. The reasons for doing so have been outlined in my dissertation (Pooth 2014). I must disagree with the second anonymous reviewer, who believes that active and detransitive cannot be used as terms for two PIE voice categories. Throughout the paper an opposition between “detransitive” and “active” forms is discussed. These are categories of different type that should not be mixed up. However, this comment is a gross misunderstanding of both, the general linguistic terminology and the semantic nature of crosslinguistic voice categories other than active and passive. The term “middle” is actually the one that is less optimal, and detransitive is to be preferred at minimum in Early Vedic and PIE. I only reconstruct two voice categories for PIE, not more than two. ACTIVE (LESS MARKED) voice vs. DETRANSITIVE (MORE MARKED) voice. The latter is what is sometimes termed “proto-middle” (Jasanoff 2003). It is a superordinate term for the common voice category that preceded the IE middle and perfect active forms, the Anatolian ḫi-conjugation, and many IE thematic active forms. Its original functional scope is outlined in Pooth (2014) and Pooth & Orqueda (2021). The following forms are externally and internally reconstructable for a PIE stage. We can leave away unnecessary non-phonemic diacritics here. *gwhén-s *gwhn-só wh *g én-t *gwhn-tó *gwhn-ó (TPǵénh1sto) wh The analysis is also based on three plural endings, at least in the active voice: *gwhn-ént *gwhn-érs → *gwhn-éːr (via Szemerényi’s law) IR *gwhn-ér *stéu-nt- / *dhughn̥ tó *stéu-rs IR *stéu-r / *dhughró IR 2.1. The alveolar morph *-tThe morph *-t- is usually taken for a regular fusional third person singular portmanteau ending (e.g. *gwhént, *gwhénti > Vedic hán, hánti etc.) However, many more forms were marked by this morph *-t- in PIE; and the alleged status as 3sg ending seem to be doubtful. A segment *-t- is also found in the 3sg detransitive ending *-to- (> Vedic -ta = Greek -το = Hittite –ta etc.). Since we also have an ending *-o- that lacks the *-t-, the segmental string *-to- seems to consist of two separate morphs, *-t- and detransitive *-o- (Pooth 2004a, 2009b, 2015). The third person plural forms displaying *-nt- vs. *-nto- also show such a segment *-t-. Therefore, this segmental string seems to also consist of separable morphs *-n- + *-t- and *-o-, respectively. Figure 4. PIE forms exhibiting *-t- + *g n-ó > Young Avestan °γne (niγre). It can be equated with the first part of Hittite kun-ati. I see no reason to doubt that the latter reflects an old ending (pace the second anonymous reviewer). *gwhn̥ -tó is the expected 3sg with *-t-. It is based on many other forms ending in *-to. This ending is widely attested. *gwhn̥ -só is the expected form with *-s-. For -so cf. Greek ο, -σο (e.g. ἐφέρου ‘you were carried’) = Latin 2sg middle -re (~ ris ~ rus), :: Vedic -se = Gothic -za, among others. TP ǵénh1sto is a Transponat of Vedic (á)jániṣṭa. It points to a pleonastic ending *-sto-, which may be younger. IR *dhughn̥ tói > Vedic duhaté. EF-r̥ s is evidenced by Old Avestan cikōitǝrǝš. The k of Old Avestan cikōitǝrǝš (TPkwikwóitr̥ s) may be analogical or points to variant with o-grade of the root (Jasanoff 2003: 40, fn. 32). Old Avestan -ǝrǝ̄ points to *-r̥ without *-s. TP ǵusró is continued in Vedic ajuṣran. The Vedic 3pl middle endings -ran, -ram go back to Proto-Indo-Iranic *-ra and PIE *-ro. This was extended by *-n or *-m in paradigmatic analogy to the 3pl active ending *-an < PIE *-ont or the 2d and 3du active endings *-tām, *-tam, respectively. IR *stéu-nt- is based on the ablaut pattern of Vedic stáuti and the corresponding Avestan participle stem stauuat- (nom. sg. m. stauuas), which had the same structure. It is also based on the 3pl ending of reduplicated patterns. For all other matters, and for more equation formulas, I generally refer to the currently available handbooks. ACT DTR DTR ACT DTR 3SG *gwhént *gwhntó *stḗut *stéuto 3PL *gwhnént *gwhnntó *h1sónt → MIDDLE *h1sónto and → NEOACTIVE *h1sónt *stéunt *stéuont → *stéu̯ onto ~ *stéu̯ n̥ to The shift of former PIE proto-middle forms to active (neo-active) voice can be well-motivated on syntactic grounds. I refer to my recent publication on this matter (Pooth & Orqueda 2021). The term “neo-active” means “ex-proto-middle form reanalyzed as new active form” (cf. Jasanoff 2003). The PIE 3pl detransitive form *h1sónt (vs. active *h1sént) was reanalyzed as neo-active form from PIE to post-PIE (Pooth 2011; 2014: chapters 7-8; Pooth & Orqueda 2021). N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 7 Thus, it seems quite clear that an innovative pleonastic 3pl middle ending *-onto (> Hittite -anta = Vedic -anta, etc.) developed from *-ont with further attachment of the final *-o in analogy to *-n̥ to. In addition to its occurrence in 3person forms, the segment *-t- is part of the 2pl endings *-te- (e.g. *h1stén(i); *-teni is continued as Hittite -teni). There were a few analogical variants like *-tes. It is further part of the postPIE middle endings 2sg *-th2e- (with various analogical extensions) and 2pl *-th2e(n). It is a synchronic Vedic fact, that Early Vedic sigmatic aorists typically pattern together with optatives that are still asigmatic (Figure 7; Table 1). Early Vedic sigmatic optatives derived from sigmatic aorist stems are extremely infrequent. There are some middle forms, for instance a 1sg aor. opt. mid. masīya from 1man- ‘to think’, etc. However, the other optatives are root optatives and are not derived from the underlying sigmatic stem, but are templatically derived from the Vedic root, which is the basis for verb inflection in Early Vedic. Figure 5. More endings with internal *-t- Figure 8. List of Vedic sigmatic aorist forms 2PL.ACT 2SG.DTR 2PL.DTR 2SG.DTR 2|3COL.DTR *gwhnté*gwhnth2é*gwhnth2é*uóidth2e *gwhntéh2- → 2pl active *-te→ 2sg middle *-thās → 2pl neo-active *-tha → 2sg neo-active *-tha → 2du neo-active *-tā(m) I think that a detransitive ending preceded the Vedic 2pl active primary endung -tha ~-thanā̆. Likewise, the IE 2/3DUAL active endings go back to prior proto-middle 2|3collective (COL) forms (Pooth 2011). 2.2. The sibilant morph *-s- and Vedic evidence for 3sg -s In this section, I argue that we need to reconstruct a third person singular ending *-s in one of the pre-Vedic stages serving as the basis for the emergence of the so-called “precative” optative aorist. This development runs in parallel with the one of the sigmatic aorist stem from 3sg form with an inflectional suffix *-s- (thus already Watkins 1962; Figure 6). Figure 6. The origin of the sigmatic stem (Watkins 1962) 3sg asigmatic 3sg sigmatic *sḗǵh-t *sḗǵh-s → *sḗǵh-st → TPsḗǵh-s-t Thus, it is reasonable to claim that the morph *-s- was not originally a sigmatic stem suffix but an inflectional suffix in PIE. In order to back my claim up with evidence, I provide a relevant example for the morphotactic tendency to generate stem suffixes out of former endings. It should be kept in mind that this is an actually attested case of what has been outlined by Watkins (1962). 3sg apās (root aorist?) aprās ahās hā́ s ájaiṣ yaus akṣār atsār abhār bhā́ r asvār atān ayān áchān ákrān asyān bhāk aprāṭ adhāk dhāk avāṭ ā ́ raik acait aśvait adyaut dyaut asrāk (ānīt akārīt gārīt cārīt árāvīt asāvīt svānīt, Narten 1964: 53f.) 2sg prā́ s (root aorist?) ákrān yāṭ ayās adyaut 1sg ayāsam ajaiṣam (prec. TS jeṣam) stoṣam abhārṣam áspārṣam áhārṣam ayāṃsam 3pl dhāsur ayāsur hāsur ábhaiṣur yauṣur áchāntsur amatsur 2pl naiṣṭa áchānta 1pl ájaiṣma (prec. jéṣma) ábhaiṣma 2du yauṣṭam 3du asvārṣṭām There is only one possible scenario that plausibly helps to understand the Early Vedic situation. It runs as follows: We must conclude that the active forms of the optative were not derived from an underlying stem via linear suffixation in Early Vedic, but were derived templatically from the root as the ultimate inflectable base. Only a few morphotactically younger forms are derived in a de-thematic way from the sigmatic stem, e.g. masīya. This was correctly seen by Jasanoff (2003): The root aorist optative forms ... are simply the forms that Indo-Iranian inherited from the optative of the PIE presigmatic aorist. Since the spread of *-s- across the extended paradigm of the presigmatic aorist had apparently not yet affected the optative in the parent language, Vedic and Avestan continue to associate s-aorist indicatives ... with optatives of ... root aorists. (Jasanoff 2003: 185) The association of Vedic root optatives such as avyā-s, etc. with the sigmatic aorist thus strengthens the hypothesis that the sigmatic aorist itself developed from a sigmatic 3sg belonging to a PIE root formation (Figure 6). Figure 7. Vedic sigmatic stems and root optatives 2.2.1. The Vedic evidence sigmatic stem āvīt, āvīṣ-ur (ví) yauṣ, yauṣ-úr sákṣantmardhīṣ, mardhiṣ-ṭam The scenario of the inflectional prehistory of the sigmatic stem suffix is underpinned by the following evidence in A, B, C, D (see also Kümmel 2015+, 2016, 2018). root optative (see Table 1) av-yā-s (ví) yū-yā-s sah-yā-s, sah-y-ur, sāh-yā-ma mr̥ dh-yā-s P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 8 Table 1. Early Vedic 3sg 2sg root aorist optatives and their 1sg 1pl 3pl forms verb 3SG 2SG avyās avyās r̥ dhyās r̥ dhyās gam gamyā́ s gamyās jñā jñeyā ́ s dah daghyās naś aśyās pā peyās bhū bhūyā́ s mr̥ dh mr̥ dhyās yam yamyās av i r̥ dh 3PL 1SG 1PL r̥ dhyā́ ma kr̥ kriyāma ~ kriyāsma aśyur bhūyā́ s yā aśyā́ m aśyā́ ma bhūyāsam bhūyā́ ma yeṣam yu yūyās vr̥ j vr̥ jyās śru śrūyā́ s sah sahyās vr̥ jyā́ m vr̥ jyā́ ma śrūyāsam (AV) sahyur A. R̥ gvedic active sigmatic aorists frequently occur in the 3sg (Narten 1964, Lubotsky 1997), see the list of forms in Figure 8 (which includes the irregular “precative” jéṣma). The isolated 3sg aor. act. acait ā ́ raik (a)dyaut are attested besides root aorist middles ácidhvam rikthās dyutāná- (Kümmel 2012, 2018), cf. YV (a)mauk besides AV ámok, RV ámugdhvam (Kümmel 2018). Possible relics of a 3sg SE -s are (as per Kümmel 2016, 2018): (1) apās RV 5.29.8, since more forms are root aorists (ápām ápās ápāma) ~ 3sg ápāt (RV 2.37.4, 6.38.1, 8.69.11, 8.92.4). But notice that Hittite pāšš-ḫi and Luwian pašš- ‘swallow’ point to a PIE suffix (enlargement) *-s-. (2) aprās ‘has/have filled’ RV (10x 3sg, 1x 2sg), subj. prā ́ s (= pra.as), imp. prā ́ si, cf. AV aprāt. (3) véṣ RV 1.77.2, 2.5.3, 4.7.7 (vī-) may not be a “Kunstbildung”, but an archaic 3sg pres. inj. (pace Malzahn 2002), cf. 2sg pres. inj. véṣ RV 4.3.13, 4.7.8, 6.15.14; RV 1.77.2 yó adhvaréṣu śáṃtama r̥ tā́ vā hótā tám ū námobhir ā́ kr̥ ṇudhvam | agnír yád vér mártāya devā́ n sá cā bódhāti mánasā yajāti is unlikely to be *vay-ī-ṣ (Kümmel 2018). (4) dhāyīṣ (RV 1.147.5d) utá vā yáḥ sahasiya pravidvā ́ n márto mártaṃ marcáyati dvayéna | átaḥ pāhi stavamāna stuvántam ágne mā ́ kir no duritā ́ ya dhāyīḥ “Ja auch, du Starker, welcher Sterbliche mit Vorsatz einen Sterblichen durch seine zwei (Hände) schädigt, vor dem schütze, du Gepriesener, den sāhyā́ ma Preisenden. Agni, dass nicht irgendwer uns dem Unglück übergebe!” (RP). 3sg is possible, because it usually occurs with mā ́ kiṣ (Kümmel 2018). (5) dhās HirGS 1.13.15 ĀpMB 2.10.17 tan ma ūrjaṃ dhāḥ “[d]as verschaffe mir Kraft!” (Burrow 1957: 64f.) occurs besides a highly archaic 3sg opt. īśīya, cf. 3pl dhāsur RV 7.97.5, subj. dhāsathas RV 1.160.5, dhāsathā RV 1.111.2 (Narten 1964: 151f., Kümmel 2016). Here may belong Messapic hipades ‘ἀνέθηκε’ (lit. ‘κατέθηκε’?) with -des (likely -dēs) < *dhéh1-s – and other such sigmatic forms (D. Kilday p.c.). (6) bhūṣ (optative syās?) in RV 10.11.9 = 12.9cd ā ́ no vaha ródasī deváputre mā ́ kir devā ́ nām ápa bhūr ihá syāḥ = AV 18.1.25 AVP 18.59.5 “fahre herbei zu uns die beiden Welthälften. Sei nicht als einer der Götter weg/Dass keiner der Götter weg sei; mögest du hier sein!” (Kümmel); 3sg for bhūṣ is at least possible, because it mainly occurs with mā ́ kiṣ (Kümmel 2018). (7) abheṣ ‘has feared’ AB 1.20.3 prāṇo vā ayaṃ san nābher iti. tasmān nābhis. tan nābher nābhitvam “The breath being here hath not feared (they say); therefore is it the navel; that is why the navel has its name” (Keith) (Narten 1964: 180, Anm. 515). However, potential grammatical mistakes are (as per Kümmel 2018): TS/TB: TS 2,2,12,6 agnír dā dráviṇaṃ vīrápeśāḥ ~ dād RV 10,80,4 TS 5,6,8,6 pitā ́ mātaríśvā ́ chidrā padā ́ dhāḥ (= AB 2,38,6f.) ~ dhāt KS 40,6 TB 2,5,4,15 ná=atārīr asya sámr̥ tiṃ vadhā ́ nām ~ ná=atārīd RV 1,32,6 KB 27,4 asmāsu nr̥ mṇaṃ dhāḥ ~ ā ́ smā ́ su nr̥ mṇaṃ dhāt MS 1,9,1: 131,9 TA 3,1,1. N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 B. Early Vedic exhibits a 3sg aor. opt. act. secondary ending -s. Table 1 provides a list of the R̥ gvedic 3sg active so-called “precatives” and some corresponding root aor. opt. active forms, cf. also AVŚ 16.2.4 śrūyāsam (for jéṣma see above). R̥ gvedic siṣ-aorist ind. and “precative” forms are attested from gā and 2yā (cf. Narten 1964: 70, Lubotsky 1997: 1149, 1151: 1sg ind. act. ayāsiṣam, 2sg opt. mid. yāsisīṣṭhās (RV 4.1.4b), 3sg ind. act. ayāsīt, opt. mid. yāsīṣṭa, 2pl ind. act. áyāsiṣṭa, 3pl ind. act. agāsiṣur ayāsiṣur, 2du inj. act. yāsiṣṭám (3du ind. act. RVKh 5.7.3c prāyāsiṣṭām); cf. 3sg subj. gāsiṣat yāsiṣat. 3sg opt. forms of thematic stems show -t, e.g. aor. sanet, AV r̥ dhet gamet -yamet vocet, VS vidét, pres. syā ́ t bhávet, likewise 3sg perf. opt. forms, e.g. dadhyā ́ t jagamyāt juguryā ́ t, etc. An irregular “precative” type is RV yeṣam (and jeṣma?), AV+ stheṣam, etc., YV khyeṣam. It exhibits the suffix -īṣ- and should be analyzed as *°a(H)-īṣ- (thus *ja(y)-īṣ-?). A 3sg in *°eṣ is not attested, but this can be accidental (Kümmel 2016, 2018; Jamison 2009; Hoffmann 1967b: 32 = 1976: 472f., 1968 = 1975: 247, fn. 4). The 3sg SE -s was taken as (irregular) innovation by Renou (1952: 291) and Hoffmann (1967b: 28), but such an analysis is implausible and ad hoc: Hier konnte beim Optativ das s-Formans vermißt worden sein [but why? This assumption is not justified, nor plausible]. Es wurde deshalb bei der 2. Sg. die Personalendung s in *yūyās nach der 2. Sg. yaus als Aorist-Formans s interpretiert und demzufolge zur 3. Sg. yaus eine 3. Sg. RV yūyās geschaffen. (Hoffmann 1967: 28; emphasis of “vermißt” mine) Thus Jasanoff (1991: 113f.; 2003: 186ff.) (starting from *-ī-ṣ → -yā-s), Harðarson (1993: 109-112), Gotō (2013: 93f.). However, such an assumption is ad hoc and not plausible because the substitution usually takes the o p p o s i t e d i r e c t i o n , and -s is replaced by -t, e.g. iṣ-aor. *átārHs(t) > *átārīṣ → átārī-t ... ágrabhī-ṣ, ágrabhī-t, likewise ānīt akārīt gārīt cārīt árāvīt asāvīt svānīt (Narten 1964: 53f.). Thus, Hoffmann’s idea does not conform to the attested inner-Vedic tendencies, also because the evident tendencies are the following: 1. RV yāṭ → ind. áyās (RV 3.29.16) confirms a spread of -s as 2sg (not 3sg!) ending; cf. also srās for *srāṭ (AV 11.2.19, 26; Kümmel 2015+, 2018; Narten 1964: 200, 273); abhinas AVP 13.4.2 for RV abhinat < *abʱinat-s (Kümmel 2015+, 2018, Hoffmann 1965: 188f. = 1975: 179f., cf. Pāṇ. 8.2.75). 2. RV bhūyā́ s → AV bhūyā-t confirms a spread of -t as 3sg ending, i.e., -t replaces -s from RV to AV, not vice versa (Hoffmann 1967: 29 = 1976: 469). Likewise, RVKh ájait, AV+ (a)nait aprāt ahāt, etc. for *ájaiṣ 9 *ánaiṣ aprās ahās; cf. also VS asrat for *asras (Kümmel 2015+, 2018). 3. The genesis of the siṣ-aorist (3sg *áyās → áyās-īt) conforms to the general tendency to introduce the endings 2sg -īs and 3sg -īt in the sigmatic aorist (Narten 1964: 71), cf. AV+ avāts-īt, dhākṣ-īt, anaikṣ-īt besides regular 1sg -s-am, etc. For the given reasons, a younger emergence of a 3sg ending -s is highly unlikely. The -s is much more plausibly an (archaic) 3sg ending, and I think Burrow was correct about it from the beginning: Taking the Vedic system where s occurs only in the second and third person singular, it is clear that in the active the suffix is -yā- and that -s is in both cases termination. (Burrow 1954: 40) Recall that “the optative of the sigmatic aorists is still acrostatic, but asigmatic in Gāthic (Narten 1984), whereas the sigmatic stem has been generalized in [the m]iddle (opt. 1sg rāŋ́hē Y 12.3)” (Tremblay 2008: 29). Old Avestan corresponds to Early Vedic in this respect. To sum up, it is evident that the 2sg 3sg -yā́ -s was reanalyzed as *-yā́ s-Ø with 2sg 3sg zero ending -Ø and spread over to a few 1st person forms. The RV provides evidence for a morphological reanalysis of a 3sg aor. opt. -yā-s to a new “precative” stem suffix: RV kri-yās-ma, bhū-yās-am, AV śrū-yās-am (Harðarson 1993: 110, Kümmel 2018). This is an attested case of WATKINS’ TRANSFORMATON, because the 2|3sg form with ending -s is reanalyzed as a stem suffix plus zero ending within Early Vedic. The same is quite likely for *-s → *-s-Ø and other IE stem suffixes in a period before Vedic. It should be acknowledged that the transformation/reanalysis is evident and provable within Early Vedic. This can finally back up the assumption of a similar morphotactic reanalysis in a preceding stage, where pretty much the same thing happened to the ending *-s when it was reanalyzed as sigmatic aorist stem suffix. Figure 9. List of Vedic forms 1sg 2sg 3sg 1pl 3pl 2du diṣīya (3dā ‘distribute’) bhakṣīyá masīya (1man ‘think’) mukṣīya rāsīya maṃsīṣṭhās darṣīṣṭa maṃsīṣṭa mr̥ kṣīṣṭa yāsīṣṭa (2yā), redupl. aor. rīriṣīṣṭa (RV 6.51.7d) ririṣīṣṭa (RV 8.18.13c), root aor. padīṣṭá (5x RV), bhakṣīṣṭa (AB) bhakṣīmáhi maṃsīmáhi vaṃsīmáhi ~ vasīmahi sakṣīmáhi maṃsīrata trā́ sīthām 10 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European C. More sigmatic endings in Early Vedic Early Vedic had corresponding 2sg and 3sg aor. opt. mid. endings -sthās, -sta. The R̥ gveda displays special sigmatic 2sg 3sg aor. opt. endings, cf. also SV bhakṣ-ī-ta ~ AB bhakṣ-ī-ṣṭa (Narten 1964: 43ff.). See the following list of forms (Figure 9). R̥ gvedic iṣ-aorist “precative” middles are 3sg janiṣīṣṭa vaniṣīṣṭa, 1pl tāriṣīmahi, vandiṣīmahi sāhiṣīmahi (cf. Narten 1964: 67), cf. 2sg yāsisīṣṭhās (RV 4.1.4b), AV maṃsīṣṭhā ́ s sāsahīṣṭhā ́ s modiṣīṣṭhās. Thematic aor. videṣṭa (AV 2.36.6), śoceṣṭa (+śuceṣṭa) ĀpMB 1.9.3 (Hoffmann 1967b: 31 = 1976: 472 Fn. 9) show -eṣṭa. The sigmatic endings were interpreted as secondary by Renou 1952: 292 (“donc janiṣīṣṭa de JAN- remplacant *janiṣīta, d’apres l’indicatif ájaniṣṭa”). But -ī-ṣṭa for -ī-ta in parallel with ár-ta : arī-ta = jáni[ṣ]-ṣṭa : X → X = janiṣī-ṣṭa is ad hoc, cf. Kümmel (2015+, 2018: “Sekundäre Entstehung von [...] -ṣṭa nicht leicht verständlich”). The stem of 1sg aor. mid. opt. maṃsī-máhi, etc. is maṃsī- (†maṃsīṣ-). Therefore, the sibilant CANNOT be segmented as part of the opt. suffix, but belongs to the endings (2sg -ṣṭhās, 3sg -ṣṭa). This leads to the reconstruction of Proto-Indo-Iranic 2sg 3sg aor. middle secondary endings *-stHās, *-sta, and 2sg 3sg mid. optatives in *-ī-ʃtHās, *-ī-ʃta. There was also a 2sg in *-īʃa, cf. YAv. °raēxš-īša (raēxš-, raēk- ~ i ric ‘let’, pres. irinaxti). These sigmatic opt. endings further confirm that the sibilant was not due to the alveolar + alveolar rule */TT/ > *-TsT- (pace Jasanoff 2003) because there is no preceding alveolar here: rīriṣ-īṣṭa ririṣīṣṭa (redupl. aor.) and padīṣṭá (root aor.) confirm that -ṣṭa is not restricted to the sigmatic aorist stem and was more independent from it. These forms are relics and clearly contradict the idea of a secondary “precativization”. D. On the origin of RV -jániṣṭa The type RV -jániṣṭa was (unconvincingly) derived from a 1sg (RV ajani →) JB ajaniṣi (ajaniṣṭhās ajaniṣṭa) by Narten (1964: 60). But whereas RV (á)jániṣṭa is attested early, JB ajaniṣi is attested too late to serve as a model (thus Kümmel 2015+, 2018). Insler (1995) suggests that 2sg -ṣṭhās < *-s-th2e-, but as seen by Kümmel (2015+, 2018), there is no Vedic parallel for -s-t° in the 2sg middle. It is more likely that the Vedic iṣ -aorist stem -jániṣajaniṣṭhās ajaniṣṭa and du. act. janiṣṭām janiṣṭam go back to a Proto-Indo-Iranic root aor. mid. 2sg */ʥánHstHās/, 3sg */ʥánH-sta/ < *ǵénh1-sto with 3sg ending *sto, thus likewise PII*/ʥánH-stām/ < *ǵénh1-stah2(m). NB. A proto-middle origin is evident for the 2du 3du forms RV janiṣṭām, AV janiṣṭam, since -jániṣ- is otherwise middle all over (medium tantum). Another otherwise fully middle paradigm (-vr̥ ṇī-) also includes 2du 3du “active” RV 1.180.4b avr̥ ṇītam, AVP 1.92.1 vr̥ ṇītām. Moreover, the Vedic 3rd dual ending -tām < *-tah2(m) is still used as a middle (not active) ending in RV 10.4.6, where 3du adhītām (-tām) even has a passive meaning and belongs to the aorist middle (adhithās, adhita, adhīmahi, etc.), cf. RV 10.4.6 tanūtyájeva táskarā vanargū́ raśanā́ bhir daśábhir abhi àdhītām “So wie zwei den Leib hingebende, im Wald umhergehende Räuber, mit den zehn Zügeln (raśanā́ bhir daśábhir, sc. mit den 10 Fingern) wurden die beiden (sc. Reibhölzer) [...] festgemacht” (Pooth 2011). 2.2.2. Comparative evidence for 3sg *-s There are obvious IE comparanda. These also point to an inflectional 3sg suffix *-s. Hittite shows sigmatic ~ asigmatic 3sg pret. act. endings, e.g. ākiš (a-ak-ki-iš) (OS) → ak-ta (NS, OH/NS NH) (Melchert & Hoffner Jr. 2008: 189, 215; Kloekhorst 2008: 167), cf. also OH a-ša-aš-ta from the ḫi-pres. ašaši (Oettinger 2002: 51, 430, Kloekhorst 2008, s.v.). Hittite further exhibits suffix PLEONASM: pa-iš (OS), pa-iš-ta (OH/NS), pe-e-eš-ta (NH) (Kloekhorst 2008: 614); tar-na-aš (OS), tar-ni-eš-ta (KUB 13.34 iv 14 (NS)), tar-ni-iš-ta (KUB 1.1+ iv 49 (NH)) (Kloekhorst 2008: 846); ši-pa-an-ta-aš (KBo 15.10 iii 59, 66 (OH/MS)), ši-pa-an-da-aš (KBo 15.10 iii 64, 68 (OH/MS)), ši-pa-an-za-aš-ta (KUB 20.59 v 6 (see Groddek, DBH 13, p. 106), KBo 8.68 iv 5), ši-pa-an-da-za (KUB 19.37 ii 24 (NH)) (Kloekhorst 2008: 405, Oettinger 2002: 41, 408); cf. also ḫa-a-az-ta (= /ḫāt-št/) (OH/MH), ḫa-a-az-za-aš-ta (MH/MS), 3pl. ḫa-a-te-e-r (OH/MS) (ḫāt-i, see Oettinger 2002: 408, Kloekhorst 2008: 328). The Proto-Anatolian sigmatic ending *-s(t) p e r f e c t l y e q u a t e s with Proto-Indo-Iranic *-s(t) (and its middle counterpart *-sta). Tocharian AB also show sigmatic 2sg 2pl 3sg act. pret. endings, cf. Figures 10ab (Malzahn 2010: 38ff. and chapters 7-9). Figures 10a (Malzahn 2010: 38ff. and chapters 7-9) 2sg 3sg 2pl ā̆-inflection TB TA -ā̆sta -āṣt -a -Ø/-āṃ -ā̆s(o) -ās non-ā̆-inflection TB TA -asta -äṣt -sa -äs/-sām -as/-so reconstruction PT *-stā *-sā *-sä ~ *-så (NB. Malzahn 2010: 514 reconstructs PT 2pl *-sās) N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 Figures 10b (Malzahn 2010: 38ff. and chapters 7-9) 2sg 3sg 2pl 2pl PT *-stā *-sā *-sä *-så reflecting *-sth2e *-sh2e (?) *-se or *-so *-seh2 internally comparable with *-th2e *-th2e *-te, *-ste, 2sg *-sto *-teh2- (> PII 2/3du *-tām) When taken together, the comparative evidence points to a 2sg proto-middle segmental string *-sth2e(i), cf. Proto-Tocharian 2sg pret. *-stā, Hittite (NH) 2sg pret. ind. -(i)šta (e.g. pa-it-ta (OH/MS), pé(-e)eš-ta (NH); Kloekhorst 2008: 614), Latin 2sg perf. ind. act. -istī < istei, e.g. CIL 10 gesistei, Greek -σθα, e.g. Homeric Greek τίθησθα, etc. It is conclusive that the traditional segmentation into sigmatic stem plus ending †-s-th2e(i) must be an anachronistic mistake. It has been suggested that Hittite -(i)šta is an inner-Anatolian creation (differently Jasanoff 2003: 119, Kloekhorst 2008: 802). But this hypothesis is rather weakened by the IE equations. After subtraction of the “union” vowel -i-, Latin 2pl perf. ind. act. -istis includes a string *-stes corresponding to Hittite 2pl ind. act. pret. -šten :: pres. -šteni (Kloekhorst 2007). This points to PIE sigmatic 2pl active endings *-ste(-), *-ste(n)(i). Therefore, Latin °sti° (-istis) and Hittite -šte° (-šten(i)) make up a comparative equation that is unlikely secondary at least with regard to the sigmatic morph. An interesting and relevant paradigm regarding the sigmatic morph in Early Vedic is the one of 1sg cyávam, 3pl cyavante (:: Old Avestan š́ auuaitē), which has a sigmatic 2sg cyoṣṭhās. These two forms point to a PIE “Narten” (= durative) proto-middle 1sg *kwiéuh2e(i) (TPkwi̯ éu̯ (h2)om), 3sg *kwiéuo(i), 3pl *kwi̯ éu̯ n̥ to(i) ~ *kwi̯ éu̯ onto(i) and a sigmatic 2sg *kwiéusth2e (later + TP-es) (2/3collective *kwi̯ éusteh2). Remarkably, Vedic 2sg cyoṣṭhās has the sibilant exactly in the paradigmatic position where we would expect it from the given comparative perspective – and thus points to an ending *-sth2e(-). 2.2.3. Pleonasm *-s → *-st It is conclusive that PIE had a set of 2sg 3sg (2pl 3pl) sigmatic endings (thus also Kümmel 2016, 2018) with a hitherto UNKNOWN function – one that should have been different from *-t-. Figure 11. Post-PIE pleonasms PIE *-s *-so *-(e)rs post-PIE, dialectal ~ *-st → *-s-t ~ *-sto → *-s-to → *-sn̥ t, *-sn̥ to, *-sonto → *-s-n̥ t, *-s-n̥ to, ... 11 These endings merged with the non-sigmatic ones yielding PLEONASMS, e.g. PIE *klép-t → root aorist (vs. *klḗp-t → “Narten type”), PIE *klép-s → *klép-st → post-PIE sigmatic aor. *klép-s-t, cf. Latin clepsit ~ clēpit (Pike 2009), and likewise *klḗp-s → *klḗp-st → sigmatic aor. *klḗp-s-t. Latin thus shows clear reflexes of both *klḗp-t and *klép-s. The corresponding middle *klépto → post-PIE *klḗptor(i) with analogical lengthened grade > Toch. A klyepträ ‘touches’. Cf. further Vedic (á)stóṣṭa < TPstéusto. The sigmatic segment was transferred to other persons. This assumption is paralleled and thus backed up by the development of sigmatic root aorist optative forms to “precative” stems in Early Vedic. We can conclude that the sigmatic aorist suffix, both its form (as a stem) and its function (as aorist), should NOT be reconstructed for the common protolanguage of Hittite, Tocharian, Vedic, Greek. The reconstruction of a sigmatic aorist stem for the common proto-language of all IE languages seems to be a severe anachronistic mistake. Like the Anatolian branch, PIE obviously lacked a sigmatic aorist stem. Based on the comparative method it seems true that the addition of *-t- to *-s- yielded *-st- Initiually, this was a secondary suffix pleonasm, similar to the one that affected the old so-called “dental-less ending” of the 3sg middle, cf. Hittite -at = Vedic 3sg aor. injunctive middle -at (aduhat) < *-ot with additional morph *-t vs. earlier *-o without *-t. A primary status of *-s- (and *-st-) as ending is evident from Hittite and Vedic. The same status seems to hold for Proto-Tocharian and Toch. A and B (Figures 10ab; Malzahn 2010: 38-44, 192). Cf. also Toch. A -ṣ (< TP-si); Vedic (ŚB) °seci (< *sóikwi) vs. Toch. A 3sg subj. V sekaṣ (TPsóikwHsi). Recall that *-a- < *-H- was generalized as a stem suffix in Proto-Tocharian (Jasanoff 2003, chapter 6-7, §93ff.; Malzahn 2010: 274-316; with further references). The old assumption that the *-s- was lost in the first person and third person plural forms in Proto-Tocharian and Proto-Anatolian is highly unlikely. To conclude, the pre-Vedic “secondary ending” of the 3sg *-s (or *-st) clearly corresponds to Hittite -š, -št(a). This correspondence can be taken for an external comparative proof that *-s (or *-st) occurred in PIE 3sg forms (Figure 12). Figure 12. A Hittite-Vedic equation Hittite -š, -št(a) = Vedic -s PIE *-s P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 12 However, such a comparative equation must still be a mystery and riddle for comparative philology, because “another” third-person singular ending *-s alongside *-t does not make much functional sense when interpreted within the canon of categories expressed by the “secondary” endings. However, the two variants are much more important than hitherto assumed. We can conclude that this suffix was the same sibilant morph *-s- that developed into a sigmatic stem suffix from post-PIE to Proto-Indo-Iranic and Proto-Greek and some other IE languages – either by the time some of them (perhaps) formed a common proto-sub-branch or independently from each other in close areal contact. It is conclusive that several IE languages other than Proto-Anatolian and Proto-Germanic reflect a post-PIE morphological reanalysis of the first part of *-st- as a new sigmatic stem suffix (→ EF-s-t). Based on the given comparative evidence, it is further possible to reconstruct a proto-middle counterpart *-so~ *-sto- via internal reconstruction, e.g. *stéuso ~ *stéusto > Vedic (á)stóṣṭa. It is quite likely that the ending *-so- (> Proto-Indo-Iranic 2sg pres. ind. mid. *-sai, Proto-Greek *-soi, etc.) was not an original PIE second person form, but a PIE second-and-third person form. If true, it was reinterpreted as a post-PIE 2sg middle ending due to the following analogy proportion: 2sg *-si :: 3sg *-ti = X → 2sg *-so- :: 3sg *-to- 2.2.4. Corresponding *-r-s The corresponding sigmatic 3pl forms were obviously marked by a sigmatic 3pl ending *-rs (*-r̥ s), cf. Old Avestan 3pl cikōit-ǝrǝš. The third person plural optative form, e.g. Vedic syúr, seems to go back to *h1s-ih1-r̥ ́ s (or *h1s-ih1-r̥ ́ ). This 3pl form *h1s-ih1-r̥ ́ s seems to be a blend of the endings -r̥ s and -érs, and was thus preceded by PIE *h1sih1érs. This form, in turn, was structured like the one with the alveolar morph *h1sih1ént > Old Latin sient. The same ending *-rs is found with e-grade (*-eːr) reflected as Hittite 3pl preterit active ending -er (~ -īr ~ eir ~ -ēr) < *-eːr. This is a relatively obvious outcome of the PIE wordfinal sequence */érs/ via Szemerényi’s law (Jasanoff 2003). It is continued as the first part of Old Latin -ēre and -ērunt (< *-eːr extended by *-i and *-ont, respectively (Weiss 2009: 393). To conclude, *-rs is separable into two agglutinated morphs: *-r- (3PLURAL) + *-s (category S) 2.3. The zero morph A third 3person finite verb form was the so-called “dental-less stative” form (e.g. Vedic śáye etc.). However, the term “stative” is functionally inappropriate because these do not have an overall stative meaming (Pooth 2000). Such 3sg forms were marked by the absence of the two morphs *-s- or *-t- in the slot to the left of the *-o- in PIE detransitive forms like *stéuo(i). *-Øo- :: *-so- :: *-toCf. Young Avestan +ni-γne ‘is knocked down’ = kuna° of Hittite kunati, Vedic stáve ‘is praised’, etc. The plural forms, e.g. *ǵusró (from *ǵeus- ‘to taste, choose; please someone; like, find pleasure in something’) were marked by *-ro(-) without any sibilant or alveolar morph (cf. RV 1.71.1b ajuṣran). The corresponding active plural forms were marked by the *-r (*-r̥ ) with no further extension, cf. Old Avestan -ǝrǝ̄. The form *-er with e-grade is still found as the first part of Old Latin 3pl perfect ind. -erunt < *-er + (extended 11 by) *-ont/*-n̥ t. The third person singular forms of the *uóide(i) type and a zero-marked 3sg form *ǵónh1(i) of a “pre-passiveaorist inflectional type” also belonged to this category Z, simply because there is no stop *-t- or sibilant *-s- before the vowel in *uóid-e(i) and *ǵónh1(i). Figure 14. PIE forms exhibiting zero 3SG ACT DTR ACT DTR DTR DTR ACT It is not necessary to reconstruct *-is-ont in this particular case (pace Weiss 2009: 393). Jasanoff 2003: 33, fn. 11 claims that -erunt > Vedic duhé, duhré cf. Old Avestan -ǝrǝ̄ > Vedic 3sg śáye > Vedic 3sg véda etc. > Vedic 3sg jani, ájani → Old Latin dederunt ... DTR 11 3PL *gwhnér *dhughó(i) *dhughró(i) *mélh2rA *ḱéio(i) *uóide(i) IRmelh2órB *ǵónh1(i)C *di-dh3ér *dé-dh3oD should go back to *-iront (cf. -imus, etc.), but this is likewise unnecessary: -erunt from *-er is straightforward. N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 Additional notes to this Figure 14. A. This 3pl form belonged to the *stḗuti type. The protoactive forms of the this so-called “Narten type” merged with the PIE *uóide(i) type yielding a post-PIE new imperfective (including habitual) mixed paradigm *u̯ óid-ei, *u̯ óid-e ~ *u̯ ḗids(t) (Jasanoff 2003; Pooth 2016, 2017). B. I think that, e.g., IRmelh2ór ‘they (always) grind’ was the original 3pl form of the *mólh2e(i) type. This form must have been modified to post-PIE *melh2r̥ ́ (s) ~ *molh2r̥ ́ (s) with and without paradigmatically levelled o-grade. Some plural form, e.g. *u̯ id-r̥ ́ (s), had zero-grade in analogy to the other non-reduplicated 3pl forms *gwhnéːr, *gwhnént, etc. The analogical spread of the zero-grade in such 3pl forms is supported, e.g., by Vedic stuvánti ← *stéu̯ n̥ ti. Kloekhorst (2012) thinks that Hittite would confirm that the zero-grade was original in this type, but Jasanoff (2003: §27) is correct by claiming that there is independent evidence for an e-grade in the corresponding reduplicated type, cf. Vedic ádīdhayur, avivyacur, where the guṇa ablaut-grade of the root must be archaic. Therefore, there is Vedic evidence for a 3sg perfect *u̯ eu̯ i̯ ókwe :: 3pl with e-grade *u̯ eu̯ i̯ ekwr̥ ́ (s) besides *u̯ eu̯ idr̥ ́ (s). As just mentioned, I follow the idea that the *u̯ óide(i) type merged with the “Narten type” *stḗus(t) (and *stḗut) in Late PIE (Jasanoff 2003). Within a “mixed paradigm” we can expect the given IE ablaut variants. Recall that Hittite -ar can go back to both *-r̥ and *-or. C. The PIE *ǵónh1(i) type merged with the aorist-like part of the *ǵusó or *dhughó(i) type in post-PIE. This yielded another mixed paradigm which is reflected as Vedic middle root aorist. Vedic abudhran abudhram both go back to *bhudhró. I assume that forms of this shape originally belonged to the *dhughó(i) type and were structured in parallel with *dhughntó(i) (with *-r- instead of *-nt-). I think that the old 3pl form belonging to *ǵónh1(i) was *ǵnh1ór. Its structure matches the one of *h1sónt(i) (again with *-r- instead of *-nt-). The issues with this merger are a bit more intricate; 3pl forms of this type seem to be allomorphs of the type that once corresponded to the amphikinetic active forms, cf. 3pl active *h1sént(i) :: 3pl protomiddle *h1sónt(i). I will have to return to this difficult matter elsewhere. It is not relevant here. D. This zero-marked form was later pleonastically extended by the productive 3sg ending *-to(i) in post-PIE first yielding *dédh3oto(i) → later analogical *dédh3eto(i) > Vedic dáda-te (Pooth 2014). 2.4. Number marking by means of vowel transposition In several published and pre-published papers of mine (Pooth 2004a, 2009b, passim), I have made the analytical claim that the morph *-m- marking the 1sg active (“present” and “aorist”) forms was not specified for number and the very same morph that also marked the 1pl (exclusive) active and middle endings 13 *h1és-m- :: *h1s-m-é(-s) Actually, there is hitherto no alternative explanation for why *-m- occurred in both forms that can explain the concomitant ablaut alternation at the same time. The pattern is paralleled by the relationship between what is usually reconstructed as 3sg and 2pl forms, respectively. *h1és-t- :: *h1s-t-é(-n) On this additional structural basis, I can base my inferences that the morph *-t- was originally the same morph marking both 3sg and 2pl forms with the difference in number being marked by the position of the vowel _é_ within the word form via internal reconstruction. I also reconstruct the following minimal pair now. *dʰéh1-s- :: IR*dʰh1-s-é(-n) We can see that the position of the vowel outside the base (or root) encoded derived forms that were internally derived and internally modified from underlying basic forms. (We all know that similar patterns occurred in the PIE noun inflection and derivation; and I claim that such a system of internal modifications constitutes the origin of IE ablaut.) We have already seen that the same morphs also occur in middle endings. *´-o / *-ó *´-t-o / *-t-ó *´-s-o / *-s-ó (not to forget TPstéusto > Vedic (á)stóṣṭa etc.) We have also seen that the same morphs also occur in 3pl form. It is still a mystery why they are reconstructable as a set of three forms with different shapes. *-ér / *-r̥ *-én-t / *-n̥ -t *-ér-s / *-r̥ -s Given that this is an obvious marking pattern, I must now claim that the morphs *-t- and *-s- cannot be original PIE person-and-number indexing morphs, but must have had a different original function in PIE. It seems obvious to me that they had nothing to do with numbermarking, and I return to this matter below. Number was originally coded either by a different position of the vowel within the word form or by more specific number suffixes (1PL *-s- and 2PL *-n-) or by combining both strategies. Even if I suppose that I am wrong, how can the devil’s advocate explain the pattern differently and in a reasonable way? I cannot see a way that is more reasonable. I am forced to analyze the pattern as I do here. P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 14 3. How to analyze these three morphs? 3.1. Three morphs for two IE functional distinctions? As a given result of the application of the comparative method to the Indo-European material, the word forms were coded by three different morphs *-t-, *-s-, and zero The comparative evidence, when taken together and analyzed this way, rather speaks against the existence of a binary opposition of “primary” vs. “secondary” endings in the common ancestor language. The most straightforward explanation is definitively not to leave the two ending sets the way they are in Greek and Vedic Sanskrit. Not touching any IE functional distinctions would enforce us to functionally reduce the three morphs to just two morphological categories because Greek and Vedic both basically distinguish two opponent sets of “primary” vs. “secondary” suffixes, not three (when we leave away the special proto-middle endings of the perfect stem). Given that the comparison actually leads to one more set, such a reduction would need many troublesome ad hoc explanations. In short, we would have to get rid of one of the three sets. However, the less costly and easiest explanation is to infer that these three morphs once indicated three different verbal categories, not two, simply because their number is three, not two. These verbal categories were presumably given up from PIE to post-PIE. They are provisionally termed categories D, S, and Z here. grammatical categories that have vanished or were modified in between PIE and its daughter languages. Comparative philologists should really recognize that this is an unfortunate blind spot of the comparative method. When it is only this method that is applied to grammatical categories, the results of comparison are unreliable and there is no chance whatsoever to uncover morphosyntactic categories and syntactic constructions that were lost or modified. However, it is undeniable that the application of the comparative method actually results in a formal distinction of three different morphs, not two. Assigning a functional value to these three morphs, therefore, cannot be done by application of the comparative method. It must be done via inductive reasoning against a background of general linguistic diachronic typological knowledge and findings. As illustrated, it is evident that these three morphs functionally merged and were integrated into the bundle of post-PIE so-called “secondary endings” in IE languages. Consequently, I must claim that the development of specific “secondary” endings only followed the emergence of the “primary” imperfective present forms and endings from PIE PROGRESSIVE aspect forms, see Figure 15 (Pooth 2009a). Figure 15. The origin of the IE endings from progressive and non-progressive forms (Pooth 2009a) PROGRESSIVE 3.2. Assigning functions to D, S, Z other than “primary” vs. “secondary” A proper identification of the morphosyntactic functions of these three lost verbal categories of PIE cannot be done on the basis of IE correspondences and the comparative method, because these three categories D, S, Z have vanished without a functional trace, leaving only formal traces. I can admit that a reconstruction of verbal categories that are not in the canon of the IE verbal categories is not provable and cannot be falsified by comparative means. However, I can argue now that the lack of comparative proof is just due to the use of the wrong method for this particular purpose. The lack of comparative falsification is a corollary of the problem that IE category equations (e.g. Greek dual = Vedic dual, etc.) cannot be used for the reconstruction of PIE categories and their exponents because they do not offer reliable results with regard to initial functional shift post-PIE PIE → IMPERFECTIVE PRESENT *-m-i *-s-i *-t-i *-n-t-i *-é-n-t-i etc. *-mi *-si *-ti *-n̥ ti *-énti etc. PIE NON-PROG second shift post-PIE “secondary” → perfective/preterit etc. *-m *-s *-t *-r-s etc. *-m̥ *-s ~ *-st *-t *-r̥ s ~ *-r̥ etc. The IE secondary functions of the “secondary” endings must not be functionally back-projected onto the protolanguage. E.g., the overt coding of perfective (“aorist”) aspect and past tense is clearly secondary. Notice, however, that I exclude the optative, whose function is not secondary in my view. Optative-irrealis forms were already optative (-irrealis) forms in PIE, but they lacked the progressive suffix *-i for obvious reasons. N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 In short, I claim that the distinction of “primary” imperfective present endings from secondary endings is only a post-PIE to post-PIE innovation. This may explain relics like Vedic kr̥ thás (aorist injunctive) with irregular “primary” ending (Pooth 2011: 478, fn. 12). 3.3. We can’t base the assignment on IE categories at this stage As outlined, the comparative reconstruction of morphosyntactic functions and grammatical categories – and their formal exponents – is methodologically unreliable and linguistically insufficient (Pooth 2017). The reasons are quite simple. Conclusions which are solely based on the application of the comparative method to IE morphosyntactic categories and their markers can only be provisional, not only because morphosyntactic functions and grammatical categories that have vanished completely or were modified cannot be compared. The plain fact that the original morphosyntactic categories and their exponents do not exist in their original forms and functions in IE languages does not mean that they did not exist. The issue is a severe pitfall of comparative philology (Pooth 2017). However, even if comparative equations alone and per se cannot prove or disprove a given hypothesis regarding the linguistic reconstruction of morphology, syntax, and lexical categories, this does not mean that a proper linguistic reconstruction of morphosyntax is impossible, nor does it imply that the comparative method cannot be used as a starter. We can use further means of reasoning for our purpose. In my view, it is necessary to first use the comparative method in order to supply the word forms and their potential meanings. At a final stage of our reasoning process, however, it is necessary to think out of the “comparative box”. As for grammatical categories, a plausible and potentially real and realistic descriptive analysis and reconstruction can be established by inductive and deductive reasoning and by combining the results of comparison with diachronic linguistic typology. Such a way of doing linguistic reconstruction implies using ALL the multiple linguistic methods known from general and historical linguistics and comparative philology. The linguistic reconstruction presented in this paper is “polymethodological” in this sense (Pooth 2015, 2016, 2017). 12 I use the terms PROXIMATIVE and OBVIATIVE for the two hierarchically indexed participants or arguments here, although these notions are also often termed proximate and obviate. 15 In short, my further reasoning process is not just based on a single comparative observation – but on two. 1. FIRST OBSERVATION, BASED ON EXTERNAL COMPARATIVE AND INTERNAL PIE EVIDENCE. There is a relevant formal similarity strengthening the conclusion that these three lost categories D, S, Z were verbal DEICTIC-DIRECTIONAL categories. The two morphs *-t- and *-s- are comparable to the deictic morphs *t- vs. *s- which are found in the PIE demonstrative pronoun system, cf. PIE *só > Vedic sá, PIE *tó- > Vedic tá-, etc. 2. SECOND OBSERVATION, BASED ON DIACHRONIC TYPOLOGY There is a relevant grammaticalization or grammaticization path which is of major relevance here. It has become an established insight that “erstwhile deictic-directional markers” constitute a “natural diachronic pathway for the grammaticalization of semantic inverse systems” (Givón 2001 I: 166). A very useful typology of such direct-inverse systems has been provided by Jacques & Antonov (2014). DIRECT marking typically indicates that proxima12 tive/proximate participant or argument higher on the person hierarchy is acting on the lower obviative/obviate participant or argument, while INVERSE marking encodes the opposite direction. 3.4. A possible deduction I am in the position to draw the following conclusion via deductive reasoning: 1. THE GIVEN INTERNAL EVIDENCE (VIZ. THE COMPARATIVE FACTS): It is internally evident that the two morphs *-t- and *-s- look exactly like deictic-directional markers that occur as morphs in the demonstrative stems *só-, *tó-. To be fair, it is not absolutely assured that they were – but at least it is absolutely clear that they exactly look like deictic-directional markers. This formal identity is taken as comparative fact and sufficient internal evidence here. [Reviewer #3 objected against this as follows: The inflectional elements -t- and -s- are said to look “exactly” like the demonstrative stems *só- and *tó-. Actually, there is one phoneme each that matches, but PIE *só … does not match the verbal ending *-s, nor does *tód match the ending *-t. The “formal identity” invoked is rather imprecise and it seems hazardous to invoke grammaticalization as a black box.] 16 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European However, this critique is based on an anachronistic analysis of the given PIE forms. The most basic radical stem (that is, the pronominal root) of the PIE sigmatic demonstrative pronoun, whatever its original function may be, was not *só- but a mono-phonemic *s-. The root and the vowel must be separated; *só- was just a secondary “derived” adnominal stem of that pronoun. There were more stems derived from this root, e.g. *sí-/*sé- and probably an adverb *sú and a clitic *=se. In addition, there was another (I say OBVIATIVE) pronoun *tí-/*té (adnominal *tó-) with an adverb *tú and clitic *=te. The idea that pronominal stems with the structure *Có- were adnominal and corresponding full pronouns (heads of an NP) had the structure *Cí- is not mine, but Warren Cowgill’s (quoted by Ringe 2006 as “in one of his lessons”; but there is no paper on that). Again, what was objected by reviewer#3 is quite problematic. Reviewer#3 may disagree with my functional reconstruction, but what he or she claims to be “actually true” and “imprecisely invoked” on a formal level is rather formally imprecise itself. 2. THE “RULE”/DIACHRONIC PATTERN: A GRAMMATICALIZATION PATH: It is a diachronic “rule” or regularity that “erstwhile deictic-directional markers” constitute a “natural diachronic pathway for the grammaticalization of semantic inverse systems” (Givón 2001 I: 166; as already mentioned). 3. THE CONCLUSION: It is plausible to identify these three lost categories with verbal deictic-directional grammatical categories belonging to the domain of transitivity direction. Typologically speaking, this assumption is plausible and possible, if not even most probable, as outlined below. I must admit that this way of reasoning is only similar to a classical deduction. There is general law, rule, or regularity, i.e. the mentioned grammaticalization (or grammaticization) path; there are given facts, i.e. the given internal evidence, formal identities and functional similarities. Of course, this deduction does not exclude other possibilities because the rule cannot be exceptionless, and the functional identity of formally identical morphs cannot be 100% proven. However, the formal identity of *-s- and *-t- with deictic-directional *s- and *t- and their functional similarity (referent tracking, referent indexing), when taken together with the well-known diachronic path, make other prehistorical possibilities LESS EXPECTED AND A BIT LESS LIKELY in a probabilistic sense. Let us have a look at the possible alternatives. 3.5. Possible alternatives First of all, suggesting that these three lost categories D, S, Z once were aspect categories is actually not very plausible because there is little that speaks in favor of it. *-t- is found in endings of both the IE imperfective and perfective aspect. I admit that *-s- is reflected as marker of perfective aspect in IE languages, although this does not hold true for Anatolian languages and Hittite. As mentioned, it is quite unlikely that an original sigmatic stem suffix was transformed into a second-person and thirdperson past tense suffix in this branch. It is more likely that the opposite happened outside Anatolian. And what about the second person singular *-s-? Should we functionally separate the two morphs, or shouldn’t we? Questions like these are relevant for a proper analysis of the equation formulas and the reconstruction of morphosyntactic categories. The traditional approach (that is, the one operating with fusional person-indexing) cannot answer such questions without making ad hoc assumptions. For instance, many of the suffixes (e.g. *-s- and *-s-) have to be declared to be just homophones, although it is likewise unprovable that they really were. Even worse, such questions are not usually touched at all within the standard theory. As mentioned, it is often taken for a god-given truth that PIE grammar was (almost) identical to Ancient Greek and Vedic Sanskrit grammar, except for its phonology. As argued above, I really find that such a preconception is unlikely from the outset. Thus, I think that it is rather not plausible to suggest that these markers were inflectional portmanteau endings and functionally identical with the IE “primary” vs. “secondary” endings (pace Lundquist & Yates 2018). If we made this assumption, we would not be able to give a diachronic justification for the given variation. We would have to ignore the morphological problems. We would have to claim that there were person-indexing markers that were “just homophones”, although this would be a bit at odds with person-indexing without pronouns otherwise, and we would also not be able to prove it. In that case, we would have to neglect a proper linguistic analysis of PIE word forms, as provided here. We would have to make rather unlikely claims, e.g. claims like a “secondary precativization”, as illustrated above. We would have to ignore the formal and functional identity of reconstructed morphs, as presented above. The internal evidence would remain completely unexplained, and no additional knowledge would be gained. In the end, PIE morphosyntax would remain a decalcomania of Greek and Vedic grammar. We could solve the issue by setting up categories like a “stative voice” to describe forms like *gwhnó(-) ‘someone is, was slain’ and *stéuo(-). But this would be of little plausibility, because the existence of a “stative” is contradicted by the IE comparative evidence. The relevant middle forms do not generally indicate a state, nor are N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 they intransitive, but transitive in several cases (and in Hittite). There is nothing that speaks in favor of the assumption that PIE had a state-indicating “stative” verbal voice category (Pooth 2000). Also, original marking of tense distinctions is unlikely and almost excluded. The Vedic “injunctive” is not a modal category Hoffmann (1967) was wrong. It is still rather an archaic tenseneutral (non-tensed) category that can be termed a polyfunctional OLD INDICATIVE and HORTATIVE-IMPERATIVE category. Its existence strongly points to a preceding nontensed type of language. In other words, we can conclude that PIE lacked marked tense distinctions. Accordingly, a PIE 3sg *gwhént(i) could be used both with present and past tense reference (Pooth 2009a, Kloekhorst 2017). Assuming that the three categories were specific tense categories makes little sense when seen from the perspective of systematic polyfunctionality. Although PIE maybe had a non-obligatory morph, namely the so-called “augment” *(h1)é(-), for particularly specifying something functionally similar to past tense, its potential existence cannot disprove the argument that PIE was non-tensed because its PIE function might have been different from past tense. Moreover, these endings were not just person markers. Such an assumption cannot explain the fact that the very same markers *-s- and *-t- coded both second and third person forms. Finally, these markers were not specified for singular, dual, or plural number. If one suggested this, one would not be capable of explaining the fact that the same segments *-t-, *-s-, and zero occur both in singular and plural forms. See the structural argumentation given in section 2.4. To sum up, our answer of the question regarding a possible original function of these categories D, S, Z must be that original marking of person, number, voice, tense, aspect, and mood distinctions is almost ruled out – or at least rather unlikely. 17 the first to “somehow envision” that the original functional distinction belonged to the domain of transitivity and transitivity direction (‘who is acting on whom’), it is clear that nobody has ever reconstructed a hierarchical direct-inverse system for PIE before. Let me conclude as follows. On the primary basis of the comparative method, secondarily extended by cross-linguistic linguistic diachronic typological patterns, we are able to plausibly reconstruct three different PIE verb forms of the second person and third person singular, plural, and collective number, respectively. The PIE second and third person forms are given in Figure 16. The PIE verb in question is *gwhen- ‘to slay, kill, beat, chase, hunt’. Figure 16. Directed and undirected PIE verb forms D = S = Z = DIR INV UDR *-t*-s*-Ø- D = S = Z = DIR INV *-t*-s*-Ø- UDR ACTIVE voice 3|2SG *gwhént *gwhéns *gwhénA 2PL *gwhnté(n) *gwhnsé(n) *gwhné(n)B 3PL *gwhnént *gwhnérs *gwhnér DETRANSITIVE 3SG voice 2SG *gwhnth2é C *gwhnsh2éC *gwhnh2éA D = S = Z = DIR INV UDR *-t*-s*-Ø- *gwhntó *gwhnsó *gwhnó D = S = Z = DIR INV UDR *-t*-s*-Ø- 2COL *gwhntóm *gwhnsóm *gwhnóm 2|3COL *gwhntéh2(m) *gwhnséh2(m) *gwhnéh2(m) D = S = Z = DIR INV UDR *-t*-s*-Ø- 3PL *gwhnntó 2PL *gwhnth2é(n)C *gwhnsh2é(n)C *gwhnh2é(n)A *gwhnró 3.6. Conclusion It is conclusive that these markers were markers of a different domain, such that is non-existing in IE languages. This functional domain can now be identified as a verbal deictic-directional domain. We can conclude that the domain of TRANSITIVITY DIRECTION is a perfectly suitable candidate and remains the most likely option. NB. I am happy to give some credit to the Leiden School at this point in our discussion. Kortlandt was on the right track when he suggested that *-to- goes back to a “transitive middle” marker (Beekes 1995, 2011). However, even if other scholars may deserve the merit to be Additional notes to Figures 16 A. In my view, the 2pl in *-h2é of the detransitive voice merged with the active in *-é. Both are reflected by the Vedic 2pl perf. ind. act. -á, e.g. Vedic vid-á. B. The zero-marked 2sg form was used in imperative function, cf. Hittite 2sg imp. kuenni < *gwhén-i. C. Parallel to the third-and-second person collective forms in *-téh2(-m) (→ Proto-Indo-Iranic 3-and-2-dual active *-tām), the 2pl detransitive forms were reanalyzed as belonging to the neo-active voice category form PIE to post-PIE. *-sh2e seems to be reflected in Proto-Tocharian (cf. Figures 10a and 10b above), whereas *-th2e is reflected as the Vedic primary 2pl pres. ind. (neo-) active ending -tha (Pooth 2011). P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 18 4. Why is *-s- the inverse marker? [2] I am indebted to Guillaume Jacques for asking me a quite important question that needs special treatment in this section. [3] “Why do you reconstruct *-s- as inverse marker but not *-t-?” (Jacques p.c.) The following internal evidence provides some reasons for doing so. The reconstruction can be based on the following three arguments. Two of them (1. and 3.) are outlined here in more detail. 1. First of all, the suffix *-s- was also used as a SOURCE marker in ablative-genitive case forms (e.g. genitive *dém-s ‘family’s, of the family’, *négwt-s ‘dusk’s, of dusk’, *diéus ‘daylight’s, sky-god’s, of daylight, skygod’, ablative-genitive *diués, etc.). The *-t- had a deictic meaning, e.g. *tód *dóm ‘DEMONSTRATIVE house, to DEMONSTRATIVE house’. I can infer that *-s- coded a direction towards the primary (i.e. topical, proximative/proximate) participant originating from the second participant as the SOURCE of the causation, whereas the *-t- coded causation towards a GOAL, that is, direction from the primary (i.e. topical, proximative/proximate) participant to the second participant. The following examples can serve to illustrate what is implied here: (1) a. *gwhén-t slay:NONDUR|NEUTRAL:ACT:SG-DIRECT\3 literally, ‘agentive slaying of topical participant to second participant’ b. *gwhén-s slay:NONDUR|NEUTRAL:ACT:SG-INVERSE\3 literally, ‘agentive slaying of topical participant (hither) from second participant (as the starting point of the action)’ Inferences like this one must of course be based on possible diachronic sources of direct and inverse markers. It has been proposed that word forms indicating a CISLOCATIVE direction (‘hither’) towards the speaker (or towards the topical participant) are a possible source of inverse markers (Jacques & Antonov 2014: 325ff.). In order to illustrate this diachronic path, let me quote Jacques & Antonov (2014: 325ff.); the numbers of their examples are mine. “The term ‘cislocative’ is used to refer to markers expressing a motion towards the speaker, both directional (‘verb hither’) and associated motion (‘come to verb’) ones. The inverse marker in Nez Percé (Sahaptian) used in local scenarios with second person acting on first person has grammaticalized from an earlier cislocative marker, reconstructed for Proto-Sahaptian as *-im (Rude 1997, 122). héexn-e see-PST I saw you. héxn-ím-e see-CIS-PST You saw me. [PST = PAST, CIS = cislocative] Interestingly, this marker has also grammaticalized into the ergative case suffix which appears on the non-SAP agent in mixed scenarios in cases where one would expect inverse marking on the verb (Rude 1997, 121-2). [4] hi-héxn-e háama-nm 3S/A→SAP-see-PST man-ERG The man saw me/you. (NEZ PERCÉ) [...] In still other languages, we observe a formal similarity between some inverse or direct markers and various types of third person markers including agent, patient, or possessive affixes. The clearest case is the inverse prefix found in Sino-Tibetan (Rgyalrongic and Kiranti) languages. As first noticed by DeLancey (1981b), the Situ Rgyalrong third person possessive prefix wə- is formally identical to the inverse marker. This is also true of other Rgyalrong languages and of some Kiranti languages that have an inverse marker, such as Bantawa [...]. The similarity between the two sets of prefixes is striking and suggests a grammaticalization from a third person marker into an inverse marker. While the exact pathway remains unclear and thus requires further investigation, it is possible that non-finite verb forms carrying a third person possessive prefix were reanalyzed as finite ones. In the case of Sino-Tibetan languages, this scenario probably occurred in the ancestor of both Rgyalrongic and Kiranti languages [...].” (end of quote) Such a bidirectional “poly-grammaticalization” in Nez Percé from CISLOCATIVE both to inverse direction and ergative case marking looks very similar to what is found in PIE morphology. In PIE, the sigmatic agentive (viz. pre-nomimative) case suffix *-s (e.g. PIE *póti-s ‘lord’, *h2nér-s ‘man’ > post-PIE *h2néː(r)) and the ablative-genitive SOURCE suffix *-s (e.g. PIE *négwts) were thus formally identical with the verbal deictic-directional suffix *-s-. This typological parallel strengthens the conclusion that *-s- was a cislocative-like inverse marker and *-tthe corresponding direct marker. 2. A second argument may run as follows. It is perhaps more likely claiming that the PROGRESSIVE DIRECT forms (e.g. *gwhénti), not the inverse forms, were later generalized as a post-PIE third person singular imperfective present forms. However, although this idea may perhaps serve as an argument, it needs an additional investigation on its own. 3. Probably the best indication is the following one. The internal evidence is provided by the neuter (= proto-neuter or “inanimate”) demonstrative form *tód. (The term “inanimate” does not only refer to lifeless things, but is N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 meant as a grammatical cover term, much the same as “masculine”.) It is highly likely that the proximative demonstrative stem *s- (*só-, etc.) could not be used to refer to neuter (= proto-neuter or inanimate) nouns because these referents could only be OBVIATIVE. This conforms to a cross-linguistic typological pattern found in hierarchical marking systems. It speaks in favor of the inference drawn here. Thus, the fact that the demonstrative stem *s- (*só-, etc.) could not be used to refer to proto-neuter or inanimate nouns can strengthen the claim that the stem *tó- was an OBVIATIVE adnominal demonstrative stem, whereas *só- was PROXIMATIVE. I can integrate the obviative function of *-t- into my general hypothesis as follows. The suffixes *-s- and *-tseem to go back to original P-indexing (not A-indexing) suffixes, and might thus ultimately be related to the PIE PT-indexing clitics (*=se, *=te). The abbreviations S, A, P, T(heme), R(ecipient) indicate the respective referents that can be compared with the notions of intransitive subject (S), transitive subject (A = most agent-like argument), and transitive object (P = most patient-like argument; Haspelmath 2011). For my model of the PIE case-marking and alignment system see Pooth & Orqueda (2021). (5) a. *h2nér-s *gwhén-t man-AGT slay:NDUR:ACT:SG-DIR/OBVP\3 ‘man (proximative A) slew obviative P’ 19 clauses, presumably extending its function to the domain of intransitive S-marking. The PIE direct-inverse markers presumably occurred in intransitive clauses as well – either generally or for encoding switched referents, as in (7ab) further below. To round off the typological picture, I can briefly add that it is quite likely that there were two general verb classes: A. agentive or “unergative” verbs, *gwhen- ‘to slay, kill’. With these verb, *-s- encoded the proximative P and the obviative S (= Sa). B. “ambient force verbs” or “unaccusative” verbs, e.g. *bhueh2-/bhuh2- ‘to grow’. With these verbs, *-s- encoded the proximative inagentive S (= Sp). A sub-class, which I call *mérti class, was intransitive and could only be combined with a proto-neuter S. Since these were always obviative, the -t- encoded the obviative S (Spobv) here. *gwhén-t man (P) slay:NDUR:ACT:SG-DIR/OBVP\SG ‘proximative A slew obviative man (P)’ (6) c. *h2nér d. *h2nér *bhuéh2-s ~ *bhúh2-s man (Sp) grow:NDUR:ACT:SG-INV/PROXP\SG ‘proximative P grew to be a man (causally affected by an “obviative” natural force)’ e. *séh2ul *mér-t sun (Spobv) vanish:NDUR:ACT:SG-DIR/OBVP\SG ‘sun (proto-neuter = obviative) vanished’ b. *h2nér *gwhén-s man slay:NDUR:ACT:SG-INV/PROXP\3 ‘obviative A slew proximative man (P)’ The hypothesis that PIE had such verb classes is typologically plausible. However, a further discussion of these issue must be postponed. c. *h2nér-s *gwém-t man-AGT come:NDUR:ACT:SG-DIR/PROXS\3 ‘man (proximative S) came hither/thither’ 5. Implications for PIE syntax w d. *h2nér-s *g ém-s man-AGT come:NDUR:ACT:SG-INV/OBVS\3 ‘man (obviative S) came hither/thither’ (6) a. *h2nér *gwhén-m man slay:NDUR:ACT:SG-1A\SG ‘I (A) slew man (P)’ b. *h2nér *bhuéh2-m ~ *bhúh2-m man grow:NDUR:ACT:SG-1S\SG ‘I (S) grew to be a man, became a man’ As illustrated (6ab), PIE indexing is reconstructed with an active first-person suffix -m- for A|S. This was true 1person indexing, since it indicated 1person A|S only. The idea that the other two suffixes were P-indexing suffixes implies that there was a kind of “split-indexing” system in PIE grammar. The 1person indexing (-m) was a kind of “subject-marking” (with A = S), whereas the 2|3 indexing was hierarchy-triggered P-marking in transitive It is cross-linguistic and typological common knowledge that direct forms are in principle and by definition used to indicate that the participant causing a change of state in another participant is either higher on a grammaticalized animacy or topicality hierarchy – or is in TOPIC function. On the other hand, inverse forms are used to indicate that the causing participant is lower on that hierarchy or non-topical (Givón 2001 I: 166, Jacques 2010, Jacques & Antonov 2014). To be fair, languages with direct vs. inverse systems considerably vary in the use of these forms (Thompson 1989, Zúñiga 2006, Jacques 2010). Nonetheless, it is a typological implication of the hypothesis outlined in this paper that PIE had a person or animacy and topicality hierarchy – and hierarchically triggered direct-inverse indexing. It is further evident that PIE displayed grammatical core case marking. This fact makes it a bit more different from other systems with hierarchical alignment. The system I reconstruct here implies that with third person referents, the direct marking indicated that the P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European 20 causing participant was topical and PROXIMATIVE and that the transitivity direction was towards a second nontopical or OBVIATIVE one. The inverse transitive marking served to code the opposite direction. The following reconstructed sentences can illustrate this difference. (7) a. *h2nér-s *gwén *Kuás-t man-AGENTIVE woman kiss:NDUR:AGT:SG-DIR\3 ‘man (PROX) kissed woman (OBV) (10) *h2nér-s man-AGT (8) a. ʁdɤrʑi kɯ ɬamu Rdorje ERG Lhamo.ABS ‘as for Rdorje, he saw Lhamo’ pɯ-a-mto AOR-3SG>3-see (DIR) (11) *h2nér-s man-AGT *médhu *dáh3-t honey.ABS give/take:NDUR-DIR *páh3-s *=kwe swallow:NDUR-INV =and/with/alike ‘mani gave honey to someonej and that personj swallowed it’ There were perhaps other oppositions encoded by *-svs. *-t-. I can briefly add that third person inverse optative forms might have been used if the speaker was emotionally involved in the event. (12) a. *h2nér *bhuh2-iéh1-s man.ABS grow-OPT-INV\3 ‘I WISH he would grow to be a man, if only he grew to be a man’ b. ɬamu ʁdɤrʑi kɯ pɯ́-wɣ-mto Lhamo.ABS Rdorje ERG AOR-INV-see:3>3SG ‘as for Lhamo, Rdorje saw him’ A similar, but not identical system with direct-inverse distinction is found in Koyukon (Athabascan, Alaska, USA; these are examples (3, 4, 56, 48) of Thompson 1989: 2, 3, 10, 11): *dh3-t-ó give/take:NDUR-DIR-DTR *páh3-t *=kwe swallow:NDUR-DIR =and/with/alike ‘mani took honey and hei swallowed it’ b. *h2nér *gwén-s *Kuás-s man woman-AGENTIVE kiss:NDUR:AGT:SG-INV\3 ‘as for man (PROX), woman (OBV) kissed him’, ‘the man was kissed by the woman’ A typolological comparandum of a language combining ergative vs. absolutive case marking with direct vs. inverse marking is provided by Japhug Rgyalrong (Jacques 2010: 135, his example): *médhu honey b. *h2nér *bhuh2-iéh1-t man.ABS grow-OPT-DIR\3 ‘he would grow to be a man (if this or that happened)’ 6. Answering the Research Questions Cross-linguistically, such direct-inverse systems are often used to encode switch-reference (Thompson 1989: 13). It is thus inferable that PIE also made use of it to code switched referents: At this point in our discussion, I am in the position to give answers to my research questions. 1. We have seen that it is generally possible to apply an independent morph analysis to the reconstructed ProtoIndo-European verb forms on their own. This methodologically step is not intellectually dangerous. 2. The independent analysis has led to results that are typologically different from the traditional model. These results are methodologically proper, typologically plausible, and actually quite promising. In order to decide which model is to be preferred, the differences of the reasoning processes and the differences of the results need to be discussed within our field,14 from the perspective of both, comparison and linguistic typology. 3. I suggest that we should better give up the default hypothesis that Proto-Indo-European grammatical categories were functionally identical to Indo-European ones. This axiom is unjustified. The need of such a “Nullhypothese” is debatable (see Appendix 3). 13 14 (9) a. John yi-nee-ł-’aanh yi-THM-CL(ł)-see13 ‘John is looking at him/her’ b. John bi-nee-ł-’aanh bi-THM-CL(ł)-see ‘as for him/her, John is looking at him/her’ c. ts’i-nee-ł-’aanh INDEF/1PL-CL(ł)-see ‘s.o./we/he/she is looking at him/her/it’ d. nee-l-’aanh THM-CL(li)-see ‘s.o. is watching him/her; she/he is being watched’ Thompson 1989 uses THM for “thematic prefix”, CL for “classifyer” (one of these “classifyers” is identified with middle voice marking). After its first mention in 2002 (Pooth 2004a) within the following 18 years, the novel Templatic Model of PIE morphology has been widely ignored within the field of Indo-European Linguistics. N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 7. Appendix 7.1. A verb paradigm The following Paradigm Grids 1–3 provide an overview of verb forms of the most basic PIE verbal inflectional type or “aspectual binyan” (Pooth 2016, 2017). It’s provisionally termed nondurative-neutral (NDUR) aspect. Grid 1. The PIE nondurative active transitive forms >1 1SG> 1PL.EXCL> 1PL.INCL> 2SG> 2PL> 3SG> 3PL> 3’SG> 3’PL> >2 gwhéns gwhnsé(n) gwhéns gwhnérs gwhéns gwhnérs >3 >3’ gwhénm gwhnmé(s)A gwhnué(s)A gwhént wh g nté(n) gwhént gwhnént 21 Additional notes to Grids 1–3: Again, the PIE verb is *gwhen- ‘slay, kill, beat, chase hunt’ again (> Vedic han-, etc.). I gloss the obviative third-person as 3’. > means ‘acting on/directed to’ here. For the time being, I use the label “nondurative” for this aorist-like aspectual category. One may use the label “aorist” as well, but not in its Greek definition A. This suffix *-s was a partially optiomal, partically obligatory 1pl marker and should not be confused with the inverse suffix here. The partially obligatory 2pl suffix was *-n. The one used in 3pl forms was *-r- before zero and *-s-, and *-n- before *-t-. This 3pl suffix looks like the nominal so-called “heteroclitic” suffix *-r- ~ *-n-. B. Perhaps the direct/inverse forms were used for more consistent subject reference-tracking – as already mentioned above. If true, the unmarked singular forms were mainly used in the imperative mood. C. In the session on my manuscript on academia.edu, I had the following question: Should I use an intransitive verb like *gʷem- here? I think that it is not to be excluded that “intransitive” (unmarked/undirected) forms of transitive verbs had a meaning like ‘subject slew (someone/something)’, whereas the transitive forms had a meaning ‘subject slew referent’. I am grateful to Onno Hovers (p.c.) for pointing me to this matter: “That’s how the subjective conjugation works in the Uralic Grid 2. The corresponding detransitive forms >1 1SG> 1PL.EXCL> 1PL.INCL> 1COL.EXCL> 1COL.INCL> 2SG> 2PL> 2COL> 3SG> 3PL> 3COL> 3’SG> 3’PL> 3’COL> >2 >3 >3’ gwhnh2é gwhnmó(s)A gwhnuó(s)A gwhnmó(h2) gwhnuó(h2) wh g nsh2é gwhnth2é wh wh g nsh2é(n) g nth2é(n) gwhnséh2 gwhntéh2 wh g nsó gwhntó ? gwhnntó wh g nséh2 gwhntéh2 gwhnsó ? gwhnséh2 Grid 3. The PIE nondurative undirected formsC 1SG 1PL.EXCL 1PL.INCL 2SG 2PL 2COL 3/3’SG 3/3’PL 3/3’COL ACTIVE DETRANSITIVE gwhénm gwhnmé(s) gwhnué(s) gwhénB gwhné(n) — gwhénB gwhnér — gwhnh2é gwhnmó(s) gwhnuó(s) gwhnh2é gwhnh2é(n) gwhnéh2(m) gwhnó gwhnró gwhnéh2(m) languages that have this (Ugric, Samoyedic, Mordvinic). But the Uralic subjective can also take an object which is then indefinite. So, subjective ‘I slew’ ~ ‘I slew something’ ~ ‘I slew some reindeer’ versus objective ‘I slew it/him/her’ ~ ‘I slew the reindeer’.” (O. Hovers) 7.2. A Comment I am grateful to Douglas G. Kilday (p.c.) for the following comment (via academia.edu, his original comment is shortened): “I think the view that PIE deictic-demonstrative *-s- was proximative (not specifically 2nd pers. sg.), and *-t- was obviative (not specif. 3sg.), helps explain the evidence for 3sg. verbal *-s in Germanic as well as Messapic, Phrygian, and scattered Vedic injunctives. Old Norse has generally 3sg. pres. -r < *-esi identical to 2sg., with scattered remnants of 3sg. *-ð. The Northumbrian dialect of Old English also has 3sg. pres. -s, against -þ in West Saxon, with 2sg. -st throughout West Gmc. generally explained by epenthesis in collocations with postposed *þū (n.b. however “singes þu” in the Cuckoo Song, ca. 1240). The notion that the simple 2sg. form could have displaced the 3sg. in ON and Nthb. OE isn’t plausible. What is plausible, within the framework of PIE prox. *-s- and obv. *-t-, is that Proto-Germanic inherited both as verbal endings, and used them both as 3sg. pres. suffixes, generally *-s- when the verb preceded the subject and *-t- (later *-þ, *-ð) when it followed.” 7.3. Die “Nullhypothese” I am grateful to Eugen Hill (p.c.) for providing me with the following definition (via academia.edu; with a minor omission (…)). 22 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European “Are you familiar with the concept which in German is called the Nullhypothese (…)? It works roughly the following way: If 3sg. act. opt. of thematic verbs in the IE daughter languages either presupposes *-oid (Skt -ed) or is compatible with *-oid (Gk -oi, Goth -oi, OCS -i) then the easiest assumption = reconstruction is PIE *-oid. This reconstruction can be wrong. But unless this has been shown by reference to facts, PIE *-oid remains the Nullhypothese which everybody prefers to thinkable alternatives (such as PIE *-oi + analogical dental in Indo-Iranian or whatever). I fail to see what is circular [reasoning] or a petitio principii here.” [See https://www.academia.edu/s/d329085630.] I can respond here that such a reconstruction is unlikely and prone to be wrong on the Proto-Indo-European grammatical level, simply because thematic stems are post-PIE innovations and go back to athematic 3sg proto-middle forms. This has been shown by references to comparative and language-internal facts such as voice-marking patterns and is also evident on a comparative morphological and syntactic basis (Pooth & Orqueda 2021, Pooth 2014, Jasanoff 2003, Meillet 1931). Consequently, I find it rather unlikely that PIE had optative forms derived from such thematic stems. (By the way, the post-PIE (or “Proto-Nuclear-Indo-European”) thematic optative morphologically and phonemically had a sequence */-o-ī-t/ with a hiatus between the thematic vowel and the optative suffix.) What I find close to circular here or at minimum highly problematic – as outlined above – is the assumption that PIE “must” have had the same morphs and formal and functional categories as (most of) its daughters “in the default case”. Conversely, I think we need to always base a hypothesis on more positive evidence than just equations. To further illustrate my point, it is okay for me to assume that the athematic optative formation pattern is a PIE modal category because this assumption can be based on what I have termed “reliable equation formulas”. Distinguishing between reliable and unreliable equation formulas is thus based on more positive evidence than just equations (that is, frequency of formation patterns, productivity, etc.). I think that the “Nullhypothese” is unreliable exactly in the moment when it comes to unreliable Scheingleichungen and reliably archaic hapaxes and scattered relics. It is only logical that by the moment we base a hypothesis about the existence of a given PIE category on more evidence than just unreliable equation formulas, the “Nullhypothese” becomes obsolete, simply because we have a more substantial hypothesis now. To my mind, the “Nullhypothese” is more dangerous than helpful because it may very easily lead to more problematic claims that are “nullhypothesis-internal”. E.g., it may lead to the claim that the Anatolian branch must have lost the “thematic optative” or that a specific form (e.g. a hapax) cannot be inherited because there is no perfect equation etc. However, I thank you, Eugen, because you have made a point that is of course relevant for doing linguistic reconstruction. The value of such a “Nullhypothese” needs more methodological discussions in the future. Acknowledgments Acknowledgements go to Christopher Miller for correcting my Denglish. Reviewers’ comments Three reviewers commented on a preceding version of this manuscript, which was rejected. Dear Dr. Pooth, Reviewers’ comments on your work have now been received. Based on these and our own assessment, we regret to have to inform you that your paper cannot be accepted for publication in the Journal of Historical Linguistics. The reviewers’ and editors’ comments are attached below for your information. Thank you for giving us the opportunity to consider your work. Journal of Historical Linguistics It’s only fair and okay for me to provide the reader with all the information the reviewers gave me. Reviewer #1 The paper deals with a potentially interesting topic, that is, the possibility to reconstruct the Proto-Indo-European verbal system as originally reflecting a direct/inverse system. While this hypothesis is in principle intriguing and worth pursuing, the paper is overall poorly written and presents a number of fundamental flaws. The author fails to make a compelling case for the underlying hypothesis and the work is simply not up to the scientific standards required to be considered for publication. All in all, the paper cannot unfortunately be published in this current form and I recommend rejection. [RP: The hypothesis is intriguing and worth pursuing? Thanks for this at least.] General remarks: The paper is overall extremely difficult to read, especially keeping in mind the general readership of the JHL. Most of the notions concerning Proto-Indo-European and Indo-European languages in general introduced in the paper are poorly if not defined. This makes it almost impossible for the reader to follow the author’s main argument. The paper should be rewritten keeping the potential readership in mind (or else it should be submitted to a specialized journal). [RP: I did this. I submitted it to Indogermanische Forschungen (Melanie Malzahn, Vienna), but the paper was rejected without peer review the very next day.] Structure: the paper has an overall reasonable structure, but still presents some flaws. The introduction should be entirely reworked. As of now, it is not clear what the focus of the paper is, nor what the specific research questions to be addressed are. Moreover, it is not clear at all how the three cases presented in 1.1-1.3 are relevant to the rest of the paper. If the goal is to exemplify the methodology, only one such case would be sufficient (see also remark below). Methodology: the methodology of the paper is problematic. The supposed methodological innovation of the paper is first presented in fn. 4 but other methodological remarks are scattered all over the paper, even in pp. 20-21, where a long (and yet somewhat uninformative) methodological discussion is placed. It would have been better to have a single methodological section discussing all these issue in much more detail and N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 in a more accessible way. Concerning the methodology itself, the author makes a number of claims of novelty but all these result uncompelling (see specific notes on the pdf). For example, it is simply not true that reconstructions of PIE do not normally take also into account the findings of linguistic typology, as there is a longstanding discussion in the field on this point (which is entirely omitted by the author in the paper). Data: the data is usually arranged in a particularly confusing way, with little if no explanation of the forms (this is esp. true for Section 2). This concerns both forms in IE languages as well as reconstructions. With respect to the latter, it is also unclear where these come from. Are these the author’s reconstructions? Are they taken from some source? [They are mine.] Argumentation: the argumentation appears to be ill-thought. The paper lacks a clearly defined research question, which makes it difficult to evaluate whether the author has fulfilled their own agenda. In particular, without a clear preliminary presentation of the ‘traditional’ reconstruction of the PIE verbal system (see below), it is difficult to evaluate whether the reconstruction proposed by the author actually constitutes an improvement or what its specific advantages are. [RP: Presenting the ‘traditional’ reconstruction is absolutely unnecessary here because it can easily be found in any of the currently available handbooks.] Use of references: this is probably one of the most problematic aspects of the paper. Overall, the author fails to seriously engage with the existing scholarship. This is especially problematic, because the author repeatedly points out that their paper constitutes a major break from the traditional reconstruction of PIE. However, it is never specified which reconstruction the author has in mind. There are almost no references to the supposedly ‘traditional’ views that the author is criticizing, so that instead of an actual (more or less justified) critique to an existing model, the paper reads as an exposition of the author’s personal beliefs without much support. This also leads to a number of gross misrepresentations of several major points, such as the limits of the comparative methods, the ‘post-mono-comparative’ method and the ‘debunking’ of the Greco-Aryan model (p. 22). The author presents these as novelties, while in fact these are issues that have long been discussed among specialists, but none of the existing scholarship is cited (not even to disprove it). Another puzzling point is that, especially for the non-specialist readership of the journal, it is almost impossible to tell what in the paper represents mainstream IE linguistics and what are instead the author’s own proposals, especially because on many fundamental issues (e.g. middle *-o/*-to p. 7) only work by Pooth is cited. In this connection, it is disturbing that on many fundamental issues the only available reference is to unpublished manuscripts, while there is more than abundant existing scholarship on individual points that is systematically omitted. [RP: Other scholarship is “systematically omitted”, simply because other views are irrelevant for my own analysis that is presented here. I don’t need to give reference to all papers that have treated PIE inflection. Most other scholars agree in reconstructing person indexing instead of hierarchical indexing for PIE. Discussing the whole literature would only blow up the 23 paper to an unreadable size. Everybody engaged in Indo-European Studies should know the comparative facts. I generally refer to all currently available handbooks.] Specific remarks: see the attached pdf. (These were minor points and needn’t be reproduced here.) Reviewer #2 General Comments: The author’s criticism that the historicalcomparative method has traditionally been applied in a manner that tends to project back the synchronic typology of the oldest descendant language to the protolanguage is well-taken. Given the minimal time depth involved in the present case of PIE (by any conception known to this reviewer), one would tend to expect rather that there has been a considerable change in the synchronic typology. One need think only of the history of Greek, or of English for that matter. It is also entirely in order to consider that the synchronic typology of PIE may be of a radically different type from that of Ancient Greek, Sanskrit, or Hittite. [RP: Thanks.] However, one must never lose sight of the goal of language reconstruction: to account for the facts of the attested languages—how have they come to have the grammars that they have? And there are a number of serious problems with the author’s present concrete attempt to elucidate the attested verbal systems from a radically new and different PIE starting point. [RP: No, the goal of language reconstruction is to reconstruct the language as potentially realistic language spoken by native speaks – not just to explain attested facts. I disagree. This may be the ultimate goal of reconstruction for some of us, but it is not my ultimate goal. My goal is to describe Proto-Indo-European synchronic grammar in a proper way.] Reviewer#2: (1) The statement on p. 6 (lines 151-2) that “It is a comparative fact that the IE correspondences and word or morph equations point to minimally three different PIE third person finite verb forms, both for singular and plural” is an egregious overstatement. [No, it is not. The comparative facts do point to 3sg forms in *-o-, -to-, 2sg -so- and -t- and -s-, 3pl form in *-nt- *-rs, *-r. That’s at least three 3pl forms, the rest can be analyzed the way done here.] The author is of course entitled to base further internal reconstruction on whichever interpretation of the comparative results s/he wishes, but the author’s penchant for identifying virtually every *-t- and *-s- component of verbal suffixes and endings far exceeds reasonable bounds. [No, I disagree. It is actually quite reasonable.] Homophony is conceded only in very limited cases (e.g., p. 30, line 922 in the note on the first plural endings). The problem with this unrestrained identification is that such atomism also permits almost unlimited possibilities for functional analysis. To cite merely one example, why should we believe that the *-t- of various second plural endings has anything to do with the *-t- of the third singular? [RP: Because plural or nonsingular number was obviously coded primarily by transposing the vowel to a different vowel slot – and because this makes up a pattern: *gʷʰén-m → NSG *gʷʰn-m-é *gʷʰén-t → NSG *gʷʰn-t-é (see section 2.4.)] 24 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European Predictably, the distribution and very existence of the zero ending is particularly shadowy. As to the *-s- morph (p. 8ff.), as is duly acknowledged, the author is not the first to suppose that the third singular ending *-s is a source of a suffixal *-sappearing in some languages, though this is by no means still a matter of consensus. And this reviewer happens to agree that the *-s- that appears in various endings before *-t- has a far wider distribution than some acknowledge. But there is virtually no agreement regarding the relationship of the suffixal *s’s and the ending *-s. [RP: Now what’s the purpose of this comment? Am I wrong because there is no concensus? Am I wrong because someone must have had the same idea before, so that my idea can be acceptable? But I acknowledge what is marked in bold type.] It is also disingenuous to say (p. 18, 537-40) “Not touching any IE functional distinctions would enforce us to functionally reduce the three morphs to just two morphological categories because Greek and Vedic both basically distinguish two opponent sets of “primary” vs. “secondary” suffixes, not three (when we leave away the special proto-middle endings of the perfect stem).” The “proto-middle/perfect” morphs ARE the third set beside the “primary” vs. “secondary” (NOT the invented zero morph), so one cannot claim that the more standard system tries to operate with just two functional categories (likewise p. 19, 559-61). [Disingenuous? Really? I wrote “when we leave away the special proto-middle endings of the perfect stem”. I did this because the “proto-middle/perfect” obviously could be marked by the morphs -t- (-th2e) and zero before a vowel (-Øe), which is similar to (-Øo). The claim that the standard system tries to operate with just two functional categories “primary” vs. “secondary” is okay because the perfect endings are often said to belong to the “secondary” endings (i.e. the set without *-i). No, the zero morph is not invented. It is evident: *stéuo has it. *gwhnó has it, too. The “third” set of perfect active endings seems to be secondary. I take the stance that the perfect endings are post-PIE innovations (Pooth 2014).] The author may vehemently disagree with all previous attempts to reconstruct a typologically coherent synchronic PIE system (in functional terms) that can account for the attested systems, but cannot pretend that none have been made. [Have I? I am just saying that these are all back-projections of IE morphs and categories, namely, a back-projection of the Greek and Vedic Sanskrit tripartite aspect stems (“present”, “aorist”, “perfect” and their fusional endings.] One must again protest that it is not a FACT “that the very same markers *-s- and *-t- coded both second and third person forms”. This is a claim of the author. [RP: No, this critique is not justified, and my claim is actually descriptively okay, simply because *-s- looks like *-s- and *-tlooks like *-t-. So, I can descriptively claim that these are the “same” markers. I said in the paper that I base the analysis on formal identify of morphs – not only on a back-projection of IE categories. Therefore, it is the personal claim of the reviewer that *-s- in 3sg forms and 2sg forms were “homophonous” and two different morphs in PIE. He or she obviously follows the standard reasoning of back-projecting IE categories onto PIE, thus separating identical morphs for problematic reasons, which I still don’t trust, nor can I consider such a reasoning process as being more reliable than mine. (2) It is of crucial importance to the author’s analysis that *sóbe a “proximative/proximate” demonstrative stem (finally so characterized explicitly p. 26, 802–3). Nowhere does the author give any evidence for this function of *só-, which implicitly seems to be contrasted with that of *tó- (loc. cit.). It certainly has been suggested that *só- may have had second-person deixis (Klein 1986), which would be grist for the author’s mill. However, Klein argues this for the combined *só/tó- and not for a proximal value, but rather that of near the addressee (Latin iste). The author must therefore seriously address the matter of the deictic value of *só-. [RP: Thanks for the reference, but Klein may be wrong. As I see it, PIE had two endophoric demonstrative pronouns *sí(*só- etc.) and *tí- (*tó- etc.). The proximative value can explain its distribution very well. The obviative value of the latter can explain its use as 2sg pronoun (> Proto-Anatolian *ti). The matter is currently being discussed.] (3) However, the principal weakness of the presentation here is that the author utterly fails to spell out remotely adequately how one is to derive from the PIE system sketched and illustrated on pp. 26–29 and in the appendixes to at least the attested systems of Vedic Sanskrit, Ancient Greek, and Hittite (to Latin would also be a desideratum). All of these languages share the features of an active vs. mediopassive, and a new reconstruction of the latter has now been offered by Inglese (2020), although he prudently leaves some points about the PIE function open. [RP: This is done in Pooth & Orqueda 2021. I cannot do everything in one paper. This paper is just an initial outline of the hypothesis. No, it is not “imprudent” to have an idea about the PIE function of the protomidde that is different from Silvia Luraghi’s view that middle inflection was just a lexical feature of a special verb class. I must disagree on that matter because the counterevidence is legion, although we both agree that the IE middle marking does not go back to reflexive marking, see Pooth 2014, Pooth & Orqueda 2021.] One also would like to know something about how anaphora is treated in direct/inverse systems. While the status of subject anaphoric pronouns à la Anatolian is an open question, one can hardly avoid reconstructing anaphoric direct object pronouns for PIE. How did these arise/what was their role in a PIE with a direct/inverse system? [RP: A system with direct/inverse marking doesn’t actually need anaphoric “direct object pronouns” because two participants are typically implied in the verb form, if the verb is a transitive or labile verb. But PIE presumably had a set of patient/theme clitics, e.g. *=me, *=se, *=te etc.] For the author’s scenario to be taken seriously, the path(s) from the reconstruction offered to the facts of the attested languages must be laid out in far more detail. N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 [By the moment the alignment changed to nominative-accusative, the direct/inverse system collapsed, the markers merged, and the fusional morphotactic structure emerged. This is outlined in Pooth & Orqueda 2021.] Minor comments: Page 6, line 260: Hittite kunati ‘was/has been killed’ is a hapax in a New Hittite deposition. While the ending -ati requires that it be older than the date of the composition, its archaicity may be questioned, since in Old Hittite, the passive of ‘to kill, slay’ is supplied by ak(k)- ‘to die’. [RP: I have no doubts about the archaic status of kunati, even if it should be attested in a New Hittite deposition (but see Alwin Kloekhorst’s comment below) – and the reviewer’s argument is weak. It seems to be based on a misunderstanding about verb semantics, because kunati is from ‘to kill, slay’ and ak(k)is a verb meaning ‘to die’, so these are two different verbs in Hittite with a different meaning. As far as I can tell, a typical passive category does not exist in Hittite, so he or she needs to provide at least some examples for what he actually means by “passive” in Hittite. Is it a special construction with neuter ergative? But this is not the same as a passive. In other words, the comment is problematic. When judged from the overall IE correspondences, *gwhéncould form detransitive voice forms. As argued, I reject the idea that voice marking was just a lexical feature of a class of verb in PIE. I am grateful to Alwin Kloekhorst (p.c. on facebook) for the following comment on Hittite kunati. Alwin Kloekhorst: “In this specific case, I don’t see anything wrong with Roland’s argument that kunati is an important form. It is attested in a text that HethKonk dates as “mh.” (so the reviewer is the one that has a non-standard opinion on this matter), and it indeed seems to show the basic shape *CC-ó, which in Hittite is only found in a few middles that form a closed category and that are all of an Indo-European origin. So, if kunati indeed means something like ‘was slain’ (which seems to fit the context) I would agree with Roland that it could well represent a PIE formation *gʷʰn-ó. What further consequences Roland attaches to this idea, I don’t know: I haven’t seen his article.”] p. 22, lines 674–5: while the results are not necessarily incompatible with the author’s larger PIE reconstruction, the author must acknowledge and respond to the serious attack on the very existence of an “injunctive” in the traditional sense for PIE by Hollenbaugh (2018) and (2020). [I think that the Vedic “injunctive” cannot be considered as being a modal category and has been mistreated by Hoffmann (1967), and I have already presented my issues with that 16 years ago and published the argumentation 12 years ago (Pooth 2009a). Due to the Corona-lockdown, I have currently no access to his paper. I wonder if he has cited my work] References: Hollenbaugh, Ian. 2018. Aspects of the Indo-European Aorist and Imperfect: Re-evaluating the evidence of the R̥gveda and Homer and its implications for PIE. Indo-European Linguistics 6.1–68. ——. 2020. A New Approach to Prohibitive Constructions in the R̥gveda and the Atharvaveda. JAOS 140, 777–802. Klein, Jared. 1996. "Sá-figé" and Indo-European Deixis. Historische Sprachforscung 109, 21–39. 25 Reviewer #3 p. 2. “by means of the deductive reasoning of the comparative method” – Shouldn’t this be “inductive”? [RP: The comparative method is inductive in generalizing sound laws from IE form-function correspondences, but sound laws are irrelevant here. Deriving PIE morphs and categories from IE morph and category equations is based on the axiom that PIE and IE morphs and categories must be identical, as outlined.] fn. 1: For readers less acquainted with the Sanskrit data, it would be useful to give a list of the abbreviations of the texts quoted (RV = Rigveda, etc.). p. 3: “the Vedic 1dual nom. pronoun vā́ m ... and the 2dual ... vām are so very similar.” Is similarity a sufficient argument for assuming historical identity? [RP: Of course, it is.] p. 3: Throughout the paper an opposition between “detransitive” and “active” forms is discussed. These are categories of different type that should not be mixed up. RP: No, that’s a misunderstanding of the general linguistic terminology and the PIE voice opposition (Pooth & Orqueda 2021). I only reconstruct two voice categories for PIE: active voice vs. detransitive voice. The latter is what is termed “proto-middle”. p. 3: “The PIE form *gwmmós has been interpreted as a detransitive form by Pooth (2015)” – This is thus taken for granted here, but unavailable for the readers of this paper. At least some discussion of this claim would be useful. [RP: No, this is not necessary here. The paper is available to everyone and downloadable. Or else, send me an e-mail.] p. 4: Why is the development of *-mi etc. as a present tense marker post-PIE? What is the argument for assuming that *-i was an aspectual marker? [RP: See Pooth 2009]. p. 6: Figure 3 presents reconstructions based on one form each. This looks circular, note especially “in Vedic, we can expect *haté”. What is this expectation based on? [RP: Every Indo-European linguist should know that.] p. 7: What warrants the assumption that we are dealing with the same morpheme *-t- in *-to- and *-nto-? Does -n- appear independently elsewhere as a marker for 3pl? [RP: The assumption is warranted by the marking symmetry in the corresponding active forms: *-nt- *-rs- *-r- = *-t- -s- zero. No, -n- doesn’t appear independently elsewhere as a marker for 3pl, but as a plural marker of 2pl, e.g. *gʷʰnté-n.] p. 8 et passim: The Vedic forms should probably be supplied with meanings and references to the respective roots. Not all Sanskrit text examples are translated. It would be recommendable to supply translations. [RP: This is not necessary and would only blow up the paper to twice its size. You need to have some knowledge of Sanskrit to understand a paper in Indo-European Linguistics.] p. 10: It is unclear to me what “potential mistakes” refers to. [RP: Potential grammatical mistakes of the singer or writer. I changed that.] p. 11: “Cf. also Jasanoff (1991: 113f. ...)” This should probably go into a footnote. p. 12 “Regarding these facts, any denial of the importance of Watkins’ Law for IE languages is not reasonable.” What is the background of this remark? Either leave it out or engage in a discussion with the arguments brought forward against it. [RP: The background of this remark is the recent skepticism against Watkins’ law. Eugen Hill has written on this and played 26 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European down its importance, but I can’t find the paper right now (due to Corona).] p. 13: Supply meanings for the Hittite forms. p. 16: note C on the history of -sva: if it is irrelevant for the paper, one should leave it out. [OK, this is not in this version anymore.] p. 18: “D, S and Z have vanished without a functional trace.” But D and S are continued in the daughter languages. What about “source determination” in grammaticalization? Would we not expect to find traces of earlier functions in later uses? [RP: No, that’s a different story. My story is rather about mergers of patterns and markers and a subsequent functional redistribution of *-t- and *-s- in terms of exaptation.] p. 20: “Conclusions which are solely based on the application of the comparative method ...”: How can one compare things that, as the author states, “have vanished completely”. Are we comparing chimaeras? [RP: I have no idea, what this comment is about.] p. 20: “to think out of the comparative box”: hence to use internal reconstruction, this is nothing new as for the method applied in the paper. p. 21: The inflectional elements -t- and -s- are said to look “exactly” like the demonstrative stems *só- and *tó-. Actually, there is one phoneme each that matches, but PIE *só … does not match the verbal ending *-s, nor does *tód match the ending *-t. The “formal identity” invoked is rather imprecise and it seems hazardous to invoke grammaticalization as a black box. [RP: This critique is based on a misleading view on the given forms. I replied to it in the text, as given above.] Do we want to make similar claims regarding the -m- of the 1st person, the oblique stem of the 1st person singular pronoun *me-, the -m used as object marker and e.g. the prefix PIE *me‘toward’? Again, the premise made in the paper that such similarity is meaningful must be accepted without being verifiable. [RP: No, not necessarily. But homophony is not always the preferred option; formal identity of morphs principally points to functional identity – and this must always be taken into account and investigated.] p. 22: “slewn”? “slain” is meant. [RP: Oops, I am not a native speaker of English. I am German. Such things sometimes happen to me. Corrected!] p. 22f.: Is the proposed interpretation of -s- and -t- as direct/inverse markers the only option? It seems a bit like a rabbit pulled out of the hat. [RP: Thanks for the rabbit, but I think it is the most probable option, as argued.] p. 24: The beginning of chapter 4 should probably be rephrased putting “G. Jacques (p.c.)” in a footnote. p. 27: Are there other languages with the kind of split between 1. vs. 2/3. person as proposed here? [RP: Marking splits between 1person and 2|3 are common.] p. 29: What is the hypothesis based on claiming that -s- was used “if the speaker was emotionally involved in the event”. How could this be tested on data of attested IE languages? [RP: It is based on the Vedic “precative”. It sometimes indicates that the speaker is emotionally involved in the event. This is why it has been termed “precative”.] Appendix II: probably reformulate this quotation into a brief discussion with reference as p.c. Kilday in a footnote. References Bauer, B. L. M. 2009: Residues as an aid in Internal Reconstruction, Internal Reconstruction in Indo-European. Methods, Results, and Problems. Ed. by J. E. Rasmussen & T. Olander. Copenhagen, 17-31. Beekes, R. S. P. 1995, second edition 2011: Comparative IndoEuropean Linguistics. Amsterdam & Philadelphia. Second edition. Revised and corrected by Michiel de Vaan. Amsterdam & Philadelphia. Burrow, T. 1954: The Sanskrit precative. In: J. Schubert (ed.), Asiatica. Festschrift Friedrich Weller, zum 65. Geburtstag gewidmet von seinen Freunden, Kollegen und Schülern. Leipzig, 35-42. ― 1957: An archaic verbal termination in early Indo-Aryan. Indo-Iranian Journal 1, 61-76. Bybee, J., R. Perkins & W. Pagliuca 1994: The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago & London. DeLancey, S. 1981: The category of direction in Tibeto-Burman, Linguistics of the Tibeto-Burman Area 6 (1), 83-101. Dixon, R. M. W. 2010–2012: Basic Linguistc Theory. Volume I–II 2010, Volume III 2012. Oxford. Eichner 1985: Das Problem des Ansatzes eines urindogermanischen Numerus ‘Kollektiv’ (‘Komprehensiv’), in: Grammatische Kategorien, Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen Gesellschaft, Berlin 20.25. Februar 1983. Ed. by B. Schlerath & V. Rittner. Wiesbaden, 134-169. García Ramón, J. L. 1998: Indogermanisch *gu̯ hen- ‘(wiederholt) schlagen, töten’, Mír Curad. Studies in Honor of Calvert Watkins. Ed. by Jay H. Jasanoff, H. Craig Melchert and Lisi Oliver. Innsbruck, 139-54. Givón, T. 2001: Syntax. Volume I & II. Amsterdam. Gotō, T. 1987: Die „I. Präsensklasse“ im Vedischen. Wien. ― 2013: Old-Indo-Aryan morphology and its Indo-Iranian background. In co-operation with Jared S. Klein and Velizar Sadivski. Wien. Groddek, DBH 13 = Groddek, D. 2004: Hethitische Texte in Transkription KUB 20. Dresden: Dresdner Beiträge zur Hethitologie 13. Harðarson, J. A. 1993: Studien zum urindogermanischen Wurzelaorist und dessen Vertretung im Indoiranischen und Griechischen. Innsbruck. Haspelmath, M. (2011): On S, A, P, T, and R as comparative concepts for alignment typology. Linguistic Typology 15, 535–567. Hoffmann, K. 1965: Materialien zum altindischen Verbum, 1. añc ‘schöpfen’, 2. chand, 3. du, 4. path, 5. mreḍ, 6. édhate, 7. VS. dīṣva, 8. AB. purāṇi, 9. 2. Sg. abhinas, 10. JB. liliyur. Zeitschrift für Vergleichende Sprachforschung 79, 171-191 [= 1975: 162-182]. ―1967: Der Injunktiv im Veda. Heidelberg. ― 1967b: Der vedische Prekativtyp yeṣam, jeṣma. Münchner Studien zu Sprachwissenschaft 20, 25-37 [= 1976: 465-474]. ― 1968: Zum Optativ des indogermanischen Wurzelaorists. In: J.C. Heesterman et al. (ed.), Pratidānam. Indian, Iranian and Indo-European Studies Presented to Franciscus Bernardus N E W T R E N D S I N I N D O - E U R O P E A N L I N G U I S T I C S Number 3 Jacobus Kuiper on his sixtieth birthday. Den Haag & Paris, 3-8 [= 1975: 245-250]. ― 1970: Das Kategoriensystem des indogermanischen Verbums, Münchener Studien zur Sprachwissenschaft 28, 19-41. ― 1975, 1976: Aufsätze zur Indoiranistik. Hrsg. von J. Narten. Bd. 1, 2. Wiesbaden. Hoffmann, K. & B. Forsman 2004: Avestische Laut- und Flexionslehre. 2., durchgesehene und erweiterte Auflage. Innsbruck. Hoffner, H. A. Jr. & H. C. Melchert 2008: A Grammar of the Hittite Language. Winona Lake. Insler, S. 1995: Vedic ajaniṣṭa and related problems. In: Verba et structurae. Festschrift fur Klaus Strunk zum 65. Geburtstag. Hg. von H. Hettrich et al. Innsbruck, 91-103. Jacques, G. 2010: The inverse in Japhug Rgyalrong, Language and Linguistics 11 (1), 127-57. Jacques, G. & A. Antonov 2014: “Direct/inverse systems”, Language and Linguistics Compass 8/7, 301-318. Jamison, S. W. 2009: Where are all the optatives? In: K. Yoshida, B. Vine (eds.), East and West. Papers in Indo-European Studies. Bremen, 27-45. Jasanoff, J. H. 1991: The ablaut of the root aorist optative in Proto-Indo-European, Münchener Studien zur Sprachwissenschaft 52, 101-122. ― 1997: Gathic Avestan cikōitərəš. In: A. Lubotsky (ed.), Sound Law and Analogy. Papers in Honor of Robert S. P. Beekes on the Occasion of his 60th Birthday. Amsterdam & Atlanta, 119-130. ― 2003: Hittite and the Indo-European verb. Oxford: Oxford University Press. ― 2009: Notes on the internal history of the PIE optative. In: K. Yoshida, B. Vine (eds.), East and West. Papers in IndoEuropean Studies. Bremen, 47-68. Klaiman, M. H. 1992: Inverse Languages, Lingua 88, 227-61. Kloekhorst, A. 2007: The Hittite 2pl.-ending -šten(i). In: A. Archi & R. Francia (eds.), VI Congresso Internazionale di Ittitologia, Roma, 5-9 settembre 2005 (= Studi Micenei ed Egeo-Anatolici 49-50), 493-500. ― 2008: Etymological Dictionary of the Hittite Inherited Lexicon. Leiden & Boston. ― 2017: The Old Hittite and the Proto-Indo-European tenseaspect system. Indogermanische Forschungen 122 (1), 295308, doi: 10.1515/if-2017-0015 Kölligan, D. 2007: Suppletion und Defektivität im griechischen Verbum. Bremen. Kümmel, M. J. 1996: Stativ und Passiaorist im Indoiranischen. Göttingen. ― 1998: Wurzelpräsens neben Wurzelaorist im Indogermanischen, Historische Sprachforschung 111, 191-208. ― 2012: Typology and reconstruction. The consonants and vowels of Proto-Indo-European, The Sound of Indo-European. Phonetics, Phonemics, and Morphophonemics. Ed. by B. Nielsen Whitehead, T. Olander, B. A. Olsen & J. E. Rasmussen. Kopenhagen, 291-329. ― 2012b: Monosyllabic lengthening in Vedic aorists? International Journal of Diachronic Linguistics and Linguistic Reconstruction 9, 51-63. ― 2015+, 2018 (handout, see Kümmel 2018): Anatolisches und indoiranisches Verbum. Erbe und Neuerung: 100 Jahre Ent- 27 zifferung des Hethitischen. Paper given at Morphosyntaktische Kategorien in Sprachgeschichte und Forschung. Arbeitstagung der Indogermanischen Gesellschaft in Marburg 21.-23. September 2015, downloadable from https://unijena.academia.edu/MartinKümmel ― 2015: The Role of Typology in Historical Phonology. In: The Oxford Handbook of Historical Phonology. Ed. by Patrick Honeybone and Joseph Salmons. Oxford: OUP. ― 2018: Anatolisches und indoiranisches Verbum. Erbe und Neuerung. In: 100 Jahre Entzifferung des Hethitischen. Morphosyntaktische Kategorien in Sprachgeschichte und Forschung. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 21. bis 23 September 2015 in Marburg. Herausgegeben von Elisabeth Rieken. Unter Mitwirkung von Ulrich Geupel und Theresa Maria Roth. Wiesbaden, 239ff. ― 2016: Is Ancient Old and Modern New? Fallacies of Attestation and Reconstruction (with Special Focus on Indo-Iranian). In: D.M. Goldstein, S.W. Jamison, & B Vine (eds.). 2016. Proceedings of the 27th Annual UCLA Indo-European Conference. Bremen, 79-96, downloadable from https://unijena.academia.edu/MartinKümmel Litscher, R. 2014: Voraussetzungen für ein feminines Genus und Implikationen für das Kategoriesystem des frühindogermanischen Nomens, in Neri & Schuhmann 2014: 137-165. LIV = Rix et al. 2001: Lexikon der indogermanischen Verben. 2. erw. und verbesserte Auflage. Wiesbaden. Lubotsky, A.M. 1997: A R̥ gvedic Word Concordance. 2 Volumes. New Haven, Connecticut. ― 2013: The Vedic paradigm for ‘water’, in: Multi Nominis Grammaticus Studies in Classical and Indo-European linguistics in honor of Alan J. Nussbaum on the occasion of his sixty-fifth birthday, ed. by A.I. Cooper, J. Rau & M. Weiss. Ann Arbor: Beech Stave Press, 159-164. Lundquist, J. & A. D. Yates 2018: The Morphology of ProtoIndo-European. In: Comparative Indo-European Linguistics: An International Handbook. Eds. M. Fritz, B. Joseph, J. Klein, see http://www.pies.ucla.edu/resources/ady/papers/IEmorph-F.pdf Malzahn, M. 2002: Auf der Spur von ai. veḥ. Indo-Iranian Journal 45/3, 197-204. ― 2004: 3:3, 5:1, or 4:2? On the Ablaut of the Root Aorist in Greek and Indo-European, Historische Sprachforschung 117/1, 50-75 ― 2010: The Tocharian Verbal System. Leiden & Boston. Meillet, A. (1931): Caractère secondaire du type thématique indo-européen. Bulletin de la Société de linguistique de Paris 32, 194–203. Melchert, H. C. 1998: Traces of PIE Aspectual Contrast in Anatolian, Incontri Linguistici 20, 83-92. ― 2012 (ed.): The Indo-European Verb. Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles 13-16 September 2010. Wiesbaden. ― 2015: The Tocharian s-preterite. In: Tocharian Texts in Context. Ed. by M. Malzahn, M. Peyrot, H. Fellner, T.-S. Illés. Bremen, 127-136. Narten, J. 1964: Die sigmatischen Aoriste im Veda. Wiesbaden. ― 1968: Zum „proterodynamischen“ Wurzelpräsens. In: J.C. Heesterman et al. (ed.), Pratidānam. Indian, Iranian and Indo-European Studies Presented to Franciscus Bernardus 28 P O O T H , R O L A N D A . 2 0 2 1 . Transitivity Direction in Proto-Indo-European Jacobus Kuiper on his sixtieth birthday. Den Haag & Paris, 9-19 (= 1995: 97-107) ― 1984: Optativ und Tempusstamm im Altavestischen. Die Sprache 30, 96-108. (= 1995: 270-282). ― 1995: Kleine Schriften. Band I. Hrsg. von Marcos Albino und Matthias Fritz. Wiesbaden. Neri, S. & R. Schuhmann 2014 (eds.): Studies on the Collective and Feminine in Indo-European from a Diachronic and Typological Perspective. Leiden & Boston. Oettinger, N. 2002 (1979): Die Stammbildung des hethitischen Verbums. Nachdruck mit einer kurzen Revision der hethitischen Verbalklassen. Dresden. Pike, M. 2009: The Indo-European long vowel preterites: new Latin evidence, Internal Reconstruction in Indo-European. Ed. by J. E. Rasmussen & T. Olander. Copenhagen, 205-212. **Pooth, R. A. 2000: Stativ vs. Medium im Vedischen und Avestischen, Historische Sprachforschung 113, 88-116. — 2004a: Ablaut und autosegmentale Morphologie: Theorie der urindogermanischen Wurzelflexion, Indogermanistik – Germanistik – Linguistik. Ed by. M. Kozianka, R. Lühr & S. Zeilfelder. Hamburg, 401-471. — 2004b: Zur Genese der späturidg. thematischen Konjugation aus frühuridg. Medialformen, Indogermanische Forschungen 109, 31-60. — 2009a: Der urindogermanische Progressiv, Protolanguage and Prehistory. Akten der XII. Fachtagung der Indogermanischen Gesellschaft ... in Krakau. Ed. by R. Lühr & S. Ziegler. Wiesbaden, 381-406. — 2009b: Proto-Indo-European Ablaut and Root Inflection, Internal Reconstruction in Indo-European. Ed. by J. E. Rasmussen & T. Olander. Copenhagen, 229-254. — 2011: Die 2. und 3. Person Dual und das Medium, Indogermanistik und Linguistik im Dialog. Akten der XIII. Fachtagung der Indogermanischen Gesellschaft ... in Salzburg. Hg. von T. Krisch & T. Lindner. Wiesbaden, 473-83. — 2012: Zum Aufkommen transitiver Verben in frühen Vedischen am Beispiel 1r̥ , The Indo-European Verb. Proceedings of the Conference of the Society for Ind-European Studies, Los Angeles 13-16 September 2010. Ed. H. Craig Melchert. Wiesbaden, 267-84. — 2014: Die Diathesen Aktiv vs. Medium und die Verbsemantik im Vedischen der R̥ gveda-Saṃhitā. Proefschrift, PhD dissertation. Defended 2014-10-23. Universiteit Leiden, Leiden University Repository, https://openaccess.leidenuniv.nl — 2014b (ms.): More evidence for Proto-Indo-European transfixes: Two types of “lengthened grades”, ms. version 201412-05 — 2015: Proto-Indo-European Nominal Morphology. Part 1. The Noun, Language Arts 1, issue version 2015-12-23, ms. version 2015-12-23. — 2016: Proto-Indo-European Verb Morphology. Part 1. Inflection, Language Arts 2, issue version 2016-03-11, ms. version 2016-03-11 — 2016b (talk, ms.): “Is the “tēzzi principle” a plausible inference”, The precursors of Proto-Indo-European: the IndoHittite and Indo-Uralic hypotheses. Workshop at the Leiden University Centre for Linguistics, 9-11 July 2015 — 2017a: Proto-Indo-European Nominal Morphology. Part 2. Adjectives, Language Arts 5, issue version 2017 Apr 22, author manuscript version 2017 Apr 22 — 2017: Towards a Rigorous Linguistic Reconstruction of Morphs and Categories, Language Arts 6, issue version 2018 May 22, author manuscript version 2018 May 22 — 2018. Linguistic Reconstruction of Inflectional Categories: The Three Original Functions of Indo-European -i-Presents. FIU Indo-European Lingustics Report 1 (ms. version 28 February 2018) — 2018b (ms.): Linguistic Analysis and Comparative Philology: The Case-Marking System of Proto-Indo-European. Ms. version 17 October 2018 — 2018c. A gender issue at ‘hand’: Hittite ki-eš-šar and its prehistory. Research Comment, ms. 27 March 2018 — 2019 (ms.). On the origin of the Vedic subjunctive, pre-published manuscript Pooth, R., P.A. Kerkhof, J. Barddal, L. Kulikov The Origin of Non-Canonical Case Marking of Subjects in Proto-Indo-European: Accusative, Ergative, or Semantic Alignment. Indogermanische Forschungen 124, 245–263. Pooth, R., and V. Orqueda (2021): Alignment change and the emergence of the thematic conjugation from Proto-Indo-European to Indo-European: A wedding of hypotheses. In: Transactions of the Philological Society. In press. Renou, L. 1952: Grammaire de la langue védique. Lyon/Paris. Rude, N. 1997: On the history of nominal case in Sahaptian, International Journal of American Linguistics 63, 113-43. Thompson, C. L. 1987: An Introduction to Athabascan languages. Yukon-Koyukuk School District, https://uafanlc.alaska.edu/Online/CA975T1987b/CA975T19 87b.pdf — 1989: Pronouns and Voice in Koyukon Athapascan: A textbased study, International Journal of American Linguistics 55 (1), 1-24. — 1996: The Na-Dene middle voice: An impersonal source of the D-Element, International Journal of American Linguistics 62 (4), 351-378. Tremblay, X. 2008: Iranian Historical Linguistics in the Twentieth Century – Part Two. Indo-European S Bulletin 13/1, 1– 51. Watkins, C. 1962: Indo-European Origins of the Celtic Verb I. The Sigmatic Aorist. Dublin. — 1969: Indogermanische Grammatik III/1. Geschichte der indogermanischen Verbalflexion. Heidelberg. Weiss, M. 2009: Outline of the Historical and Comparative Grammar of Latin. Ann Arbor & New York. [I haven’t read the second edition yet: http://www.beechstave.com/weiss.html] Willi, A. 2018: Origins of the Greek Verb. Cambridge U.P. Zúñiga, F. 2006: Deixis and Alignment. Inverse Systems in Indigenous Languages of the Americas. Amsterdam & Philadelphia. [**You can find most of my work on https://leidenuniv.academia.edu/RolandPooth] Köln-Nippes, zuletzt in Sülz am Montag, den 30. August 2021 

1 Transitivity Direction in Proto-Indo-European: 2 Direct, Inverse, and Undirected Verb Forms 3 Abstract: I present a novel reconstruction of the system of the Proto-Indo-European verbal inflection by arguing 4 that the common ancestor of all Indo-European languages exhibited direct vs. inverse vs. undirected verb forms 5 and thus hierarchical direct vs. inverse indexing of transitivity direction (viz. ‘who is acting on/directed to whom’; 6 see Jacques & Antonov 2014). The presented model deviates from the traditional comparative model in that it is 7 typologically different and based on a combined methodology. I am using general linguistics and the comparative 8 method to complement each other. I evaluate and interpret the results of the comparative method against the back- 9 ground of cross-linguistic findings and a broader typological pattern. This article is part of work in progress to- 10 wards a novel synchronic grammar of Proto-Indo-European. 11 Keywords: Linguistic reconstruction of Proto-Indo-European verbal inflection, transitivity direction, direct-in- 12 verse system, hierarchical indexing 13 1. 14 It may go against a dominant belief among philologists and Indo-European linguists, but I argue here 15 that Proto-Indo-European verbs were NOT inflected by attachment of fusional portmanteau “endings” 16 for person, number, voice, aspect, tense, and mood to verb stems that exhibited or lacked ablaut distinc- 17 tions. Even if Indo-European languages like Vedic Sanskrit, Ancient Greek, and Hittite must be ana- 18 lyzed this way, it is claimed here that the morphotactic structure of their parent language Proto-Indo- 19 European was different. In the present article, I use a combined methodology; methods of general lin- 20 guistics and comparative philology are complementing each other. The established results of Indo-Eu- 21 ropean comparison are thus evaluated from the background of a typology-based linguistic analysis. The 22 Proto-Indo-European word forms are NOT solely analyzed from an Indo-European comparative perspec- 23 tive on Indo-European morph boundaries. The reasons for using combined general linguistic and com- 24 parative methods follow the ones recently laid out in a paper on the methodology of a more rigorous 25 linguistic reconstruction of morphs and categories (Pooth 2017). In that paper, it is argued that anach- 26 ronistic morphosyntactic segmentations and categorizations are a cogent corollary of the application of 27 the comparative method to Indo-European morph boundaries and categories, if this is the only method 28 used for morphosyntactic linguistic reconstruction. It is argued that a more independent morphosyntactic 29 analysis needs to be applied to the proto-forms and their functions: one that is grounded on the given 30 reconstructed form-function correlates ON THEIR OWN – as if they were potentially real word forms of 31 an unknown language of its own – in order to avoid a misleading linguistic analysis of the parent lan- 32 guage. Analytical inferences concerning the proto-morphology and proto-syntax should thus be drawn 33 more independently from the younger Indo-European layer. This needs to be done with a general lin- 34 guistic and typological epistemic background. That paper finally diagnoses that this internal analysis 35 has not been done adequately yet because the traditional morphosyntactic model of the protolanguage 36 is still predominantly based on classical Greek grammar and thus on a typological knowledge of the Introduction 2 37 nineteenth century. It is further based on the preconception that PIE morphs and morph boundaries 38 should only be reconstructed by means of the deductive reasoning of the comparative method (viz. by 39 morph and category equations). From a general linguistic background, it is almost self-evident that a 40 linguistic update of the traditional morphosyntactic comparative analysis is urgently needed. The present 41 article aims at filling this gap. 42 There are a few Proto-Indo-European grammatical categories that are maintained in Ancient 43 Greek and Vedic but were abandoned or modified in other Indo-European languages. E.g., the OPTA- 44 TIVE-IRREALIS 45 branch and has been modified to a general subjunctive from PIE to Latin and Proto-Germanic: cf. Vedic 46 Sanskrit 3sg and 3pl imperfective (= “present stem”) optative active1 syā́ t syúr = Old Latin subjunctive 47 active siēt sient 2 < Proto-Indo-European (henceforth PIE) 3sg *h1siéh1t and 3pl *h1sih1ént *h1sih1érs 48 *h1sih1ér → post-PIE *h1sih1r̥ ́ (s) > Vedic syúr, etc. Further below, I return to the question why we can 49 reconstruct three different 3pl forms ending in *r *rs *nt. modal category is maintained in Greek and Vedic but is not continued in the Anatolian 50 On the other hand, a thorough Indo-European comparative investigation also reveals remnants 51 of morphosyntactic distinctions that are NOT continued as such in a single Indo-European language. 52 These grammatical categories are UNKNOWN and ABSENT from old Indo-European languages or from 53 Indo-European languages in general. If these categories existed in the common ancestor grammar, they 54 must have been abandoned or functionally modified by the time the protolanguage developed into a 55 post-PIE variant cluster with posterior dialects in close areal contact.3 I argue here that, by additional 56 application of the method of internal reconstruction, we are able to detect and identify such parallel 57 independent innovative categories of the post-PIE dialects. These parallel innovations presumably oc- 58 curred shortly after the common PIE synchronic grammar broke up. 59 1.1. 60 An illustrative example of a category that is non-existing in Indo-European languages but should better 61 be reconstructed for their ancestor, is the 1+2|2+x INCLUSIVE person category ‘we, that is, you and 62 me/us’. It is continued only modified as a first-person DUAL INCLUSIVE category in Early Vedic (RV, 1 The Proto-Indo-European 1+2|2+x inclusive category In general, I follow the Leipzig Glossing Rules and make use of the conventional glossing: ABS = “absolutive” or “anti- agentive” unmarked case, ACC = accusative case, ACT = active-agentive voice, AGT = agentive case, COL = collective number (group-like sociative-associative plural ‘they are doing something together’), DTR = detransitive voice, NDUR = nondurative (“aorist-like”) aspect, PL = plural number, PROG = progressive aspect, SG = singular number, TOP = topical referent, NTOP = nontopical or “anti-topical” referent. The symbol > generally means ‘… developed into … by sound law’ but here it also symbolizes the transitivity or causation direction (e.g., 2sg>3sg = 2sg acting on/causing an effect in 3sg). 2 Weiss 2009: 416. 3 For the notion of a “Late” or “Vulgar” Proto-Indo-European dialect continuum and variant cluster see Pooth 2015. 3 63 AV, e.g. Vedic ganvahi (TPgwm̥ u̯ édhh2)4 ‘we two, you (sg) and me, go (up) together’, attested in RV 64 8.69.7b). In later IE languages, it is reflected as a first-person DUAL without clusitivity (that is, it is both 65 inclusive AND EXCLUSIVE) in Old Church Slavonic and Gothic (Pooth 2011). This well-known5 dia- 66 chronic path can be illustrated as follows. 67 Figure 1. The development of 1dual forms from 1pl inclusive forms 68 PIE6 Early Vedic 69 1+2inclusive → 1dual inclusive 70 *gʷm-uó(s) → OCS, Gothic → 1dual inclusive|exclusive gan-vahi (middle) 71 The fact that 1dual forms are overwhelmingly INCLUSIVE in the Early Vedic period is a func- 72 tional trace and positive evidence for this category, and constitutes the decisive argument (Pooth 2011). 73 Another piece of evidence in this puzzle is the comparative fact that the Vedic 1dual nom. pronoun vā́ m 74 (RV 6.55.1) and the 2dual acc. dat. gen. clitic vām are so very similar. This points to a PIE pronoun 75 *(h1)ué(h1), clitic *uo(h1) for both 2dual and 1dual inclusive (Pooth 2016c, section 9). The internal 76 evidence thus supports the hypothesis that PIE had a 1+2 INCLUSIVE that was likewise a 2nd person 77 INCLUSIVE/ASSOCIATIVE 78 panion’, ‘you and your (uncountable or countable) company’). This can be glossed as 1+2|2+X. Such a 79 polysemy is not unknown among the languages of the world. The pronominal polysemy fits with the 80 polysemy of the corresponding verbal category reconstructed here. Therefore, the PIE suffix -u- can be 81 interpreted as a second person (addressee, oppositional, adversative) suffix underspecified for the other 82 included/inclusive participant. The form *gʷmuó obviously lacked an additional number suffix and was 83 thus less specified, whereas two additional forms 2nd person inclusive sociative-associative *gʷm-uó-h1, 84 plural *gʷm-uó-s had additional number suffixes (they are structured in parallel with the 1pl exclusive 85 forms *gʷm-mé *gʷm-mé-s). The PIE form *gʷmuó(s) has been interpreted as a DETRANSITIVE (“proto- 86 middle”) form by Pooth (2015) with the detransitive marker being */o/ here. The active counterpart was 4 (‘you & an underspecified participant’ = ‘you and me’, ‘you and your com- Abbreviations: EF = equation formula; TP = phonological transponatum. IR = internally reconstructed word form. The term “transponatum” is pseudo-Latinized German Transponat and refers to a form that is generated by simple back-projection of the respective IE sound laws, as if the form and its function belonged to PIE synchronic grammar without implying that it really did. Equations formulas, especially such based on productive (viz. potentially innovative) IE formation patterns (e.g. 3sg EF bʰéreti), can principally be secondary Scheingleichungen due to independent parallel innovations (e.g., an independent par- allel innovation of a simple thematic verb stem EFbʰére/o- in areal contact). Therefore, equation formulas should not automatically be reconstructed in their corresponding forms and functions for PIE without additional methodological inferences (Pooth 2017). This is a major methodological disagreement between the traditional comparative approach and the one pursued here. 5 This development has been described by Dixon 2010 II: 194 for several Australian languages. 6 Abbreviations: PIE= Proto-Indo-European, PII = Proto-Indo-Iranian/Iranic, OCS = Old Church Slavonic, etc. 4 87 *gʷmué(s). To conclude, we are able to identify a post-PIE functional narrowing of the PIE 1st+ 2nd 88 person inclusive and 2nd person plus x category to 1st person DUAL by application of the method of 89 internal reconstruction. 90 1.2. 91 Another illustrative example of a PIE grammatical category whose original function did not survive in 92 any IE language and was functionally modified (most probably immediately after PIE broke up) is the 93 PROGRESSIVE 94 marker of this PIE aspect category was the suffix *-i. This suffix was attachable to many non-progres- 95 sive verb forms in word final position, e.g. *gwhénm-i ‘I am slaying, killing someone, him, her, it’, 96 *gwhnmés-i ‘we (exclusive) are ditto’, etc. 7 We can draw the inference that most PIE progressive forms 97 developed into post-PIE IMPERFECTIVE aspect and PRESENT tense forms8, 9 (Pooth 2009a, Kloekhorst 98 2017). It is an important implication of this inference that the post-PIE imperfective present category 99 was then, but NOT BEFORE (!), marked by fusional portmanteau suffixes. The emergence of fusional 100 TAM + voice portmanteau endings *-mi, *-si, *-ti, *-mes(i), *-u̯ es(i), *-ten(i), *-énti, etc. must thus be 101 a SECONDARY functional morphotactic innovation. In other words, these new fusional suffixes do not 102 necessarily go back to the PIE proper parent language in their Greek or Vedic forms and functions. The 103 parallel independent post-PIE development can be illustrated as follows. The Proto-Indo-European progressive aspect 104 aspect (but see Kloekhorst 2017 for its functional traces in Old Hittite; Bauer 2009). The Figure 2. The development of IE imperfective present or general present forms 105 PIE 106 PROGRESSIVE wh 107 *g én-t-i 108 post-PIE Hittite → IMPERFECTIVE (PRESENT) → PRESENT *gwhén-ti kuēn-zi → Vedic Sanskrit 109 IMPERFECTIVE PRESENT (later vs. PERFECTIVE)10 110 hán-ti (vs. aor. vádhīt)11 7 This suffix has hitherto been called “particle of the hic et nunc”, but it is neither assured that it was a particle nor is “hic et nunc” an adequate functional term. It is unlikely that it was a tense suffix (Pooth 2009a). 8 The development of progressive forms to imperfective present forms is common (Bybee, Perkins & Pagliuca 1994). 9 These progressive forms were extended by productive endings yielding new imperfective and present stems (Pooth 2017). 10 The development of the PERFECTIVE (“aorist”) category was an innovation of the so-called “Inner IE” parts of post-PIE dialect continuum, presumably innovated in close areal contact. It is mistaken to reconstruct the traditional array of characterized aorist stems for PIE, except for the athematic reduplicated aorist, but this must have had a more specified (COMPLETIVE) meaning. Even “root aorist” stems cannot be rendered canonical PERFECTIVE stems in PIE because many of them were still compatible with the progressive aspect, e.g. *dhéh1ti (Pooth 2016). 11 García Ramón (1998) noticed the suppletion of Vedic pres. hánti vs. aor. vádhīt. But the lack of corresponding suppletive pairs speaks in favor of parallel but independent root suppletion in Ancient Greek vs. Vedic (Kölligan 2007). 5 111 I return to the emergence of the “primary endings” and “secondary endings” in more detail below. To 112 sum up, the back-projection of the Vedic and Greek so-called “primary endings” onto PIE seems to be 113 an anachronistic mistake.12 114 1.3. 115 Another example of a PIE category that was obviously modified in post-PIE times is provided by the 116 PIE COLLECTIVE (UNCOUNTABLE GROUP PLURAL) number category. E.g., PIE *uédor ~ *uédorh2 ‘wa- 117 ter(s)’ (>*u̯ édo:(r)) besides PIE *uódr ~ *uóh1r ‘water’ (> Hittite wātar, etc.). It relevant for the meth- 118 odological purpose pursued here to recognize that this number category is not maintained as a number 119 category in any IE language. Nevertheless, it is one of the rare categories that comparative philologists 120 reconstruct for PIE grammar (see the discussion by Eichner 1985, Litscher 2014), even if this category 121 is no longer existing in this particular function in the old IE languages. The PIE collective number 122 The PIE collective number forms are continued as neuter singular or plural, or feminine singular 123 forms in the later period: e.g., PIE *dóru-h2 ‘wood-COL’ → IE ‘wood-PL’; Greek has a nom.-acc. SIN- 124 GULAR n. τό ὕδωρ, gen. ὕδατος ‘water’, whereas Hittite has a PLURAL n. widār ~ witār ~ wedār ‘waters’ 125 (vs. sg. wātar ‘water’). The Greek and Hittite forms seem to go back to PIE *uédor(h2) (*udón, *udnós). 126 The locative and the oblique forms are continued as Vedic udán, udnás, etc. Greek reflects the general- 127 ized zero-graded root *ud- but the original accent on the root, whereas Hittite has the accent outside the 128 root. It is thus a well-known comparative fact that the IE outcomes of PIE *uédor(h2) belong to different 129 IE singular vs. plural number categories. Even if it is often assumed that *uédor(h2) was a neuter “sin- 130 gular” form that could be used as a plural number form, it is clear that it must have had a plural-like 131 meaning in PIE. Therefore, Pooth (2015) has drawn the inference that *uédor- was transnumeral and 132 had both a GENERIC reading ‘water in general, water as type’ and an uncountable COLLECTIVE number 133 reading ‘group of waters’, although discussing this hypothesis is not necessary here. The relevant meth- 134 odological point of our concern is that the PIE form *uédor(h2) is continued in different IE languages 135 by two different categories and that this comparative fact can be read as comparative evidence justify- 136 ing the general assumption of some major changes in PIE derivational and inflectional number catego- 137 ries – and major changes in morphosyntactic (e.g. number) categories more generally. 138 To sum up, I consider it imperative that within a proper linguistic reconstruction of the common 139 ancestor of all IE languages it must be our task and duty to reveal the semantic nature of all these lost 140 and modified categories by means of internal reconstruction grounded in cross-linguistic patterns and 141 based on general linguistic typology. Also, we must find reasonable motives for why these categories 142 were given up or modified when the protolanguage broke up and developed into a bundle of dialects. I 12 Thus Beekes 1995. 6 143 would like to make this methodological point even stronger here. To my mind, it is unreasonable, or 144 naïve, to a priori doubt or neglect the existence of PIE grammatical distinctions that were different from 145 the ones reflected by the IE languages. All historical linguists are well aware that categories normally 146 continually change their functional nature. Assuming that the opposite was true for the period in between 147 PIE and IE is contradicted by cross-linguistic diachronic evidence. In my view, such a state of mind is 148 not a reasonable skeptical position. 149 2. 150 Before we can analyze the PIE word forms more properly, we need to first make use of the comparative 151 method. It is a comparative fact that the IE correspondences and word or morph equations point to 152 minimally three different PIE third person finite verb forms, both for singular and plural. These third 153 person forms are provisionally termed D for the ones with the segment -t- (the voiceless alveolar stop), 154 S for the ones with the segment -s- (the sibilant), SD for the ones with a combined -st-, and Z (zero) for 155 the ones without any of these segments. I operate with the notions of “equation formulas” and “transpo- 156 nata” here to make sure that this is a preliminary methodological step but not yet the ultimate recon- 157 struction (see footnote 4). Comparative settings Figure 3. IE equation formulas (EF) and transponata (TP) 158 EF 159 3SG 160 Z EF -Ø- (2SG) 161 D EF 162 163 EF MIDDLE example IE forms EF -o- TP wh Young Avestan +°γne :: Hittite kun-ati -t- EF -to- EF wh in Vedic, we can expect *ha-té S EF -s- EF -so- (2SG) EF wh g n̥ só(i) (2SG)A SD EF -st- EF -sto- TP ǵénh1sto Vedic (á)jániṣṭa ǵusró Vedic ajuṣranB ACTIVE g nó g n̥ tó 164 3PL 165 Z EF -ér/-r̥ EF -ro- TP 166 D EF -ént-/-n̥ t- EF -(o)nto- TP h S EF 167 -érs/-r̥ s d ughn̥ tó TP w w k ik óitr̥ s Vedic duhaté Old Avestan cikōitǝrǝšC 168 169 Additional notes to Figure 3: 170 A Cf. Greek -(σ)ο (e.g. ἐφέρου ‘you were carried’< *ephere-so) = Latin 2sg middle -re (~ ris ~ rus), etc. :: 171 Vedic -se = Gothic -za, etc. 172 B 173 in paradigmatic analogy to the 3pl active ending *-an < PIE *-ont or the 2du and 3du active endings *-tām, Cf. Vedic -ran, -ram go back to Proto-Indo-Iranian/Iranic *-ra (< *-ro) extended by *-n and *-m (or both) 174 *-tam, respectively. 175 C 176 to variant with o-grade of the root. Cf. Jasanoff (2003: 40, fn. 32). The k of Old Avestan cikōitǝrǝš (TPkwikwóitr̥ s) may be analogical or points 7 177 2.1. 178 The morph *-t- is usually taken for a regular fusional PIE third person singular portmanteau ending, e.g. 179 *gwhént, *gwhénti > Vedic hán, hánti, etc. However, many more forms were marked by this alveolar 180 suffix *-t- in PIE. A segment *-t- is also found in the 3sg middle (= detransitive) ending *-to- (> Vedic 181 -ta = Greek -το = Hittite -ta, etc.). This string *-to- can be split up into two separate morphs, *-t- and 182 detransitive *-o-, respectively (Pooth 2004a, 2009b, 2015). Furthermore, the third person plural forms 183 displaying *-nt- vs. *-nto- also show such a segment *-t-. Thus, these segmental strings are separable 184 (sic!) into *-n- + *-t- and *-o-, respectively. 185 The alveolar morph *-t- Figure 4. PIE forms exhibiting *-t- 186 3SG 3PL 187 ACT *gwhént *gwhnént 188 DTR *gwhntó *gwhnntó, *h1sónt → post-PIE *h1sónt ~*h1sóntoA 189 ACT *stḗut *stéunt 190 DTR *stéuto *stéuont → post-PIE *stéu̯ onto ~ *stéu̯ n̥ toA 191 Note to Figure 4: AThe PIE 3pl detransitive form *h1sónt (vs. active *h1sént) was reanalyzed as neo- 192 active13 form from PIE to post-PIE (Pooth 2011; 2014: chapters 7-8). An innovative pleonastic 3pl middle 193 ending *-onto (> Hittite -anta = Vedic -anta, etc.) developed from *-ont + *-o in analogy to *-n̥ to. 194 In addition to its occurrence in third-person forms, the segment *-t- is part of the 2pl endings *-te(n) ~ 195 *-tes (vel sim.), e.g. *h1stén(i), etc. It is further part of the post-PIE second person middle endings sg. 196 *-th2e (vel sim.), pl. *-th2e(n).14 197 Figure 5. More PIE forms exhibiting the segment *-t- 198 2SG 2/3COL (→2/3DUAL15) *gwhnté(n) 199 ACT 200 DTR *gwhnth2é 201 DTR *uóidth2e 202 DTR 13 2PL *gwhnth2é(n) (→ Vedic 2pl neo-active -tha ~-thanā̆) *gwhntéh2 → neo-active *-tā(m) Cf. Jasanoff 2003. The term “neo-active” means “ex-proto-middle form reanalyzed as new active form”. A syntactic moti- vation for the proto-middle to neo-active voice shift is provided in Pooth (2004b); see also Pooth & Orqueda forthc. 14 Cf. Pooth 2011, 2015. 15 Cf. Pooth 2011. 8 203 2.2. 204 Other third person finite verb forms were marked by a sibilant suffix *-s-. In this section,16 I argue that 205 we need to reconstruct a third person singular ending *-s in one of the pre-Vedic stages serving as the 206 basis for the emergence of the so-called “precative” optative aorist. I thus conclude that the sibilant 207 morph *-s- was an inflectional suffix, not originally a sigmatic stem suffix. The sibilant morph *-s- 208 I can provide a relevant example for the morphotactic tendency to generate stem suffixes out of 209 former endings. Recall that Early Vedic sigmatic optatives derived from sigmatic aorists are extremely 210 infrequent. There are some middle forms, for instance a 1sg aor. opt. mid. masīya from 1man- ‘to think’, 211 etc. However, the other optatives are root optatives which are not derived from the underlying sigmatic 212 stem, but are templatically derived from the Vedic “root” (= the ultimate inflectable base). 213 Figure 6. Vedic sigmatic stems and root optatives 214 sigmatic stem root optative (see Table 1) 215 āvīt, āvīṣ-ur av-yā-s 216 (ví) yauṣ, yauṣ-úr (ví) yū-yā-s 217 sákṣant- sah-yā-s, sah-y-ur, sāh-yā-ma 218 mardhīṣ, mardhiṣ-ṭam mr̥ dh-yā-s 219 Figure 7. The origin of the sigmatic stem (Watkins 1962) 220 3sg asigmatic *sḗǵh-t 221 3sg sigmatic *sḗǵh-s → *sḗǵh-st → TPsḗǵh-s-t 222 Figure 8. List of Vedic sigmatic aorist forms 223 3sg apās (root aorist?) aprās ahās hā́ s ájaiṣ yaus akṣār atsār abhār bhā́ r asvār atān ayān áchān ákrān 224 asyān bhāk aprāṭ adhāk dhāk avāṭ ā ́ raik acait aśvait adyaut dyaut asrāk (ānīt akārīt gārīt cārīt 225 árāvīt asāvīt svānīt, Narten 1964: 53f.) 226 2sg prā́ s (root aorist?) ákrān yāṭ ayās adyaut 227 1sg ayāsam ajaiṣam (prec. TS jeṣam) stoṣam abhārṣam áspārṣam áhārṣam ayāṃsam 228 3pl dhāsur ayāsur hāsur ábhaiṣur yauṣur áchāntsur amatsur 229 2pl naiṣṭa áchānta 230 1pl ájaiṣma (prec. jéṣma) ábhaiṣma 231 2du yauṣṭam 232 3du asvārṣṭām 16 The argumentation is partially based on Pooth (2017). 9 233 There is only one possible explanation that plausibly helps to understand the Early Vedic situation. It 234 runs as follows. The active forms of the optative were not derived from an underlying stem via linear 235 suffixation in Early Vedic, but were derived templatically from the root as the ultimate inflectable base. 236 Only a few morphotactically younger forms are derived the de-thematic way from the sigmatic stem, 237 e.g. masīya. This was correctly seen by Jasanoff: 238 The root aorist optative forms ... are simply the forms that Indo-Iranian inherited from the optative of the 239 PIE presigmatic aorist. Since the spread of *-s- across the extended paradigm of the presigmatic aorist had 240 apparently not yet affected the optative in the parent language, Vedic and Avestan continue to associate s- 241 aorist indicatives ... with optatives of ... root aorists. (Jasanoff 2003: 185) 242 The association of root optatives such as avyā-s, etc. with the sigmatic aorist strengthens the hypothesis 243 that the sigmatic aorist itself developed from a sigmatic 3sg form of the PIE root formations, as illus- 244 trated in Figure 7. This is underpinned by the following evidence (cf. Kümmel 2015+, 2016, 2018): 245 246 a. R̥ gvedic active sigmatic aorists frequently occur in the 3sg (cf. Narten 1964, Lubotsky 1997), see the list of forms in Figure 8 (which includes the irregular “precative” jéṣma). 247 The isolated 3sg aor. act. acait ā ́ raik (a)dyaut are attested besides root aor. mid. ácidhvam rikthās 248 dyutāná- (cf. Kümmel 2012, 2015+, 2018), cf. YV (a)mauk besides AV ámok, RV ámugdhvam (cf. 249 Kümmel 2015+, 2018). Possible relics of a 3sg SE -s are (as per Kümmel 2015+, 2016, 2018): 250 (1) apās RV 5.29.8, since more forms are root aorists (ápām ápās ápāma) ~ 3sg ápāt (RV 2.37.4, 251 6.38.1, 8.69.11, 8.92.4). But notice that Hittite pāšš-ḫi and Luwian pašš- ‘swallow’ point to a PIE suffix 252 (enlargement) *-s-. 253 (2) aprās ‘has/have filled’ RV (10x 3sg, 1x 2sg), subj. prā ́ s (= pra.as), imp. prā ́ si, cf. AV aprāt. 254 (3) véṣ RV 1.77.2, 2.5.3, 4.7.7 (vī-) may not be a “Kunstbildung”, but an archaic 3sg pres. inj. (pace 255 Malzahn 2002), cf. 2sg pres. inj. véṣ RV 4.3.13, 4.7.8, 6.15.14; RV 1.77.2 yó adhvaréṣu śáṃtama r̥ tā́ vā 256 hótā tám ū námobhir ā́ kr̥ ṇudhvam | agnír yád vér mártāya devā́ n sá cā bódhāti mánasā yajāti is unlikely 257 to be *vay-ī-ṣ (cf. Kümmel 2015+, 2018). 258 (4) dhāyīṣ (RV 1.147.5d) utá vā yáḥ sahasiya pravidvā ́ n márto mártaṃ marcáyati dvayéna | átaḥ 259 pāhi stavamāna stuvántam ágne mā ́ kir no duritā ́ ya dhāyīḥ “Ja auch, du Starker, welcher Sterbliche mit 260 Vorsatz einen Sterblichen durch seine zwei (Hände) schädigt, vor dem schütze, du Gepriesener, den 261 Preisenden. Agni, dass nicht irgendwer uns dem Unglück übergebe!” (RP). 3sg is possible, because it 262 usually occurs with mā ́ kiṣ (cf. Kümmel 2015+, 2018). 263 (5) dhās HirGS 1.13.15 ĀpMB 2.10.17 tan ma ūrjaṃ dhāḥ “[d]as verschaffe mir Kraft!” (cf. Burrow 264 1957: 64f.) occurs besides a highly archaic 3sg opt. īśīya, cf. 3pl dhāsur RV 7.97.5, subj. dhāsathas RV 265 1.160.5, dhāsathā RV 1.111.2 (cf. Narten 1964: 151f., Kümmel 2016). Here may belong Messapic 266 hipades ‘ἀνέθηκε’ (lit. ‘κατέθηκε’?) with -des (likely -dēs) < *dhéh1-s – and other such sigmatic forms. 10 267 (6) bhūṣ (opt. syās?) in RV 10.11.9 = 12.9cd ā ́ no vaha ródasī deváputre mā ́ kir devā ́ nām ápa bhūr 268 ihá syāḥ = AV 18.1.25 AVP 18.59.5 “fahre herbei zu uns die beiden Welthälften. Sei nicht als einer der 269 Götter weg/Dass keiner der Götter weg sei; mögest du hier sein!” (Kümmel); 3sg for bhūṣ is at least 270 possible, because it mainly occurs with mā ́ kiṣ (cf. Kümmel 2015+, 2018). 271 (7) abheṣ ‘has feared’ AB 1.20.3 prāṇo vā ayaṃ san nābher iti. tasmān nābhis. tan nābher nābhitvam 272 “The breath being here hath not feared (they say); therefore is it the navel; that is why the navel has its 273 name” (Keith) (cf. Narten 1964: 180, Anm. 515). 274 However, potential mistakes are (as per Kümmel 2015+, 2018): TS/TB: TS 2,2,12,6 agnír dā 275 dráviṇaṃ vīrápeśāḥ ~ dād RV 10,80,4 TS 5,6,8,6 pitā ́ mātaríśvā ́ chidrā padā ́ dhāḥ (= AB 2,38,6f.) ~ 276 dhāt KS 40,6 TB 2,5,4,15 ná=atārīr asya sámr̥ tiṃ vadhā ́ nām ~ ná=atārīd RV 1,32,6 KB 27,4 asmāsu 277 nr̥ mṇaṃ dhāḥ ~ ā ́ smā ́ su nr̥ mṇaṃ dhāt MS 1,9,1: 131,9 TA 3,1,1. 278 b. Early Vedic exhibits a 3sg aor. opt. act. secondary ending -s. Table 1 provides a list of the R̥ gvedic 279 3sg active so-called “precatives” and some corresponding root aor. opt. active forms, cf. also AVŚ 16.2.4 280 śrūyāsam (for jéṣma see above). 281 Table 1. Early Vedic 3sg 2sg root aorist optatives and corresponding 1sg 1pl 3pl forms verb av i r̥ dh 3SG 2SG avyās avyās r̥ dhyās r̥ dhyās 3PL 1SG 1PL r̥ dhyā́ ma kr̥ kriyāma ~ kriyāsma gam gamyā́ s jñā jñeyā ́ s dah daghyās naś aśyās pā peyās bhū bhūyā́ s mr̥ dh mr̥ dhyās yam yamyās gamyās aśyur bhūyā́ s yā aśyā́ m aśyā́ ma bhūyāsam bhūyā́ ma yeṣam yu yūyās vr̥ j vr̥ jyās śru śrūyā́ s sah sahyās vr̥ jyā́ m vr̥ jyā́ ma śrūyāsam (AV) sahyur sāhyā́ ma 282 R̥ gvedic siṣ-aorist ind. and “precative” forms are attested from gā and 2yā (cf. Narten 1964: 70, Lubotsky 283 1997: 1149, 1151: 1sg ind. act. ayāsiṣam, 2sg opt. mid. yāsisīṣṭhās (RV 4.1.4b), 3sg ind. act. ayāsīt, opt. 284 mid. yāsīṣṭa, 2pl ind. act. áyāsiṣṭa, 3pl ind. act. agāsiṣur ayāsiṣur, 2du inj. act. yāsiṣṭám (3du ind. act. 285 RVKh 5.7.3c prāyāsiṣṭām); cf. 3sg subj. gāsiṣat yāsiṣat. 286 287 3sg opt. forms of thematic stems show -t, e.g. aor. sanet, AV r̥ dhet gamet -yamet vocet, VS vidét, pres. syā ́ t bhávet, likewise 3sg perf. opt. forms, e.g. dadhyā ́ t jagamyāt juguryā ́ t, etc. 11 288 An irregular “precative” type is RV yeṣam (and jeṣma?), AV+ stheṣam, etc., YV khyeṣam. It 289 exhibits the suffix -īṣ- and should be analyzed as *°a(H)-īṣ- (*ja(y)-īṣ-?). A 3sg in *°eṣ is not attested, 290 but this can be accidental (cf. Kümmel 2015+, 2016, 2018, Jamison 2009, Hoffmann 1967b: 32 = 1976: 291 472f., 1968 = 1975: 247, fn. 4). 292 293 The 3sg SE -s was taken as (irregular) innovation by Renou (1952: 291) and Hoffmann (1967b: 28) – but such an analysis is implausible and ad hoc: 294 Hier konnte beim Optativ das s-Formans vermißt worden sein [but why? This assumption is not justi- 295 fied, nor plausible]. Es wurde deshalb bei der 2. Sg. die Personalendung s in *yūyās nach der 2. Sg. yaus 296 als Aorist-Formans s interpretiert und demzufolge zur 3. Sg. yaus eine 3. Sg. RV yūyās geschaffen. (Hoff- 297 mann 1967: 28; emphasis of “vermißt” mine) 298 Cf. also Jasanoff (1991: 113f.; 2003: 186ff.) (starting from *-ī-ṣ → -yā-s), Harðarson (1993: 109-112), 299 Gotō (2013: 93f.). This assumption is ad hoc and not plausible because the substitution usually takes 300 the opposite direction, and -s is replaced by -t, e.g. iṣ-aor. *átārHs(t) > *átārīṣ → átārī-t ... ágrabhī-ṣ, 301 ágrabhī-t, likewise ānīt akārīt gārīt cārīt árāvīt asāvīt svānīt (cf. Narten 1964: 53f.). All in all, Hoff- 302 mann’s idea does not conform to inner-Vedic tendencies: 303 (1) RV yāṭ → ind. áyās (RV 3.29.16) confirms a spread of -s as 2sg (not 3sg!) ending; cf. also srās 304 for *srāṭ (AV 11.2.19, 26; cf. Kümmel 2015+, 2018, Narten 1964: 200, 273); abhinas AVP 13.4.2 for 305 RV abhinat < *abʱinat-s (cf. Kümmel 2015+, 2018, Hoffmann 1965: 188f. = 1975: 179f., cf. Pāṇ. 306 8.2.75). 307 (2) RV bhūyā́ s → AV bhūyā-t confirms a spread of -t as 3sg ending, i.e., -t replaces -s from RV to 308 AV, not vice versa (cf. Hoffmann 1967: 29 = 1976: 469). Likewise, RVKh ájait, AV+ (a)nait aprāt 309 ahāt, etc. for *ájaiṣ *ánaiṣ aprās ahās; cf. also VS asrat for *asras (Kümmel 2015+, 2018). 310 (3) The genesis of the siṣ-aorist (3sg *áyās → áyās-īt) conforms to the general tendency to introduce 311 the endings 2sg -īs and 3sg -īt in the sigmatic aor. (cf. Narten 1964: 71), cf. AV+ avāts-īt, dhākṣ-īt, 312 anaikṣ-īt besides regular 1sg -s-am, etc. 313 314 For the given reasons, a younger emergence of a 3sg ending -s is unlikely. It is more plausibly an (archaic) ending, and Burrow was correct: 315 Taking the Vedic system where s occurs only in the second and third person singular, it is clear that in the 316 active the suffix is -yā- and that -s is in both cases termination. (Burrow 1954: 40) 317 Recall that “the optative of the sigmatic aorists is still acrostatic, but asigmatic in Gāthic (Narten 1984), 318 whereas the sigmatic stem has been generalized in [the m]iddle (opt. 1sg rāŋ́hē Y 12.3)” (Tremblay 319 2008: 29). Old Avestan corresponds to Early Vedic in this respect. 320 It is evident that the 2sg 3sg -yā́ -s was reanalyzed as *-yā́ s-Ø with 2sg 3sg zero ending -Ø and 321 spread over to a few 1st person forms. Thus, the RV provides evidence for a morphological reanalysis 12 322 of a 3sg aor. opt. -yā-s to a new “precative” stem suffix: RV kri-yās-ma, bhū-yās-am, AV śrū-yās-am 323 (cf. Harðarson 1993: 110, Kümmel 2015+, 2018). 324 This is an attested case of WATKINS’ LAW, because the 3sg form with ending -s is reanalyzed 325 as a stem suffix plus zero ending within Early Vedic. The same is quite likely for *-s and other IE stem 326 suffixes in a period before Vedic. It should be acknowledged that this law is provable within Early 327 Vedic, as it can be based on an evident inner-Vedic morphotactic reanalysis. Regarding these facts, any 328 denial of the importance of WATKINS’ LAW for IE languages is not reasonable. 329 c. Early Vedic had corresponding 2sg and 3sg aor. opt. mid. endings -sthās, -sta. The R̥ gveda displays 330 special sigmatic 2sg 3sg aor. opt. endings, cf. also SV bhakṣ-ī-ta ~ AB bhakṣ-ī-ṣṭa (cf. Narten 1964: 331 43ff.). See the following list of forms (Figure 9): 332 Figure 9. List of Vedic forms 333 1sg diṣīya (3dā ‘distribute’) bhakṣīyá masīya (1man ‘think’) mukṣīya rāsīya 334 2sg maṃsīṣṭhās 335 3sg darṣīṣṭa maṃsīṣṭa mr̥ kṣīṣṭa yāsīṣṭa (2yā), redupl. aor. rīriṣīṣṭa (RV 6.51.7d) ririṣīṣṭa (RV 8.18.13c), 336 root aor. padīṣṭá (5x RV), 337 bhakṣīṣṭa (AB) 338 1pl bhakṣīmáhi maṃsīmáhi vaṃsīmáhi ~ vasīmahi sakṣīmáhi 339 3pl maṃsīrata 340 2du trā́ sīthām 341 R̥ gvedic iṣ-aorist “precative” middles are 3sg janiṣīṣṭa vaniṣīṣṭa, 1pl tāriṣīmahi, vandiṣīmahi sāhiṣīmahi 342 (cf. Narten 1964: 67), cf. 2sg yāsisīṣṭhās (RV 4.1.4b), AV maṃsīṣṭhā ́ s sāsahīṣṭhā ́ s modiṣīṣṭhās. The- 343 matic aor. videṣṭa (AV 2.36.6), śoceṣṭa (+śuceṣṭa) ĀpMB 1.9.3 (cf. Hoffmann 1967b: 31 = 1976: 472 344 Fn. 9) show -eṣṭa. The sigmatic endings were interpreted as secondary by Renou 1952: 292 (“donc 345 janiṣīṣṭa de JAN- remplacant *janiṣīta, d’apres l’indicatif ájaniṣṭa”). But -ī-ṣṭa for -ī-ta in parallel with 346 ár-ta : arī-ta = jáni[ṣ]-ṣṭa : X → X = janiṣī-ṣṭa is ad hoc, cf. Kümmel (2015+, 2018: “Sekundäre Ent- 347 stehung von [...] -ṣṭa nicht leicht verständlich”). The stem of 1sg aor. mid. opt. maṃsī-máhi, etc. is 348 maṃsī- (†maṃsīṣ-). Therefore, the sibilant CANNOT be segmented as part of the opt. suffix, but belongs 349 to the endings (2sg -ṣṭhās, 3sg -ṣṭa). This leads to the reconstruction of Proto-Indo-Iranian 2sg 3sg aor. 350 middle secondary endings *-stHās, *-sta, and 2sg 3sg mid. optatives in *-ī-ʃtHās, *-ī-ʃta. There was also 351 a 2sg in *-īʃa, cf. YAv. °raēxš-īša (raēxš-, raēk- ~ iric ‘let’, pres. irinaxti). These sigmatic opt. endings 352 further confirm that the sibilant was not due to the alveolar + alveolar rule */TT/ > *-TsT- (pace Jasanoff 353 2003) because there is no preceding alveolar here: rīriṣ-īṣṭa ririṣ-īṣṭa (redupl. aor.) and padīṣṭá (root 354 aor.) confirm that -ṣṭa is not restricted to the sigmatic aorist stem and was more independent from it. 355 These forms are relics and clearly contradict the idea of a secondary “precativization”. 13 356 d. The type RV -jániṣṭa was (unconvincingly) derived from a 1sg (RV ajani →) JB ajaniṣi 357 (ajaniṣṭhās ajaniṣṭa) by Narten (1964: 60). But whereas RV (á)jániṣṭa is attested early, JB ajaniṣi is 358 attested too late to serve as a model (thus Kümmel 2015+, 2018). Insler (1995) suggests that 2sg -ṣṭhās 359 < *-s-th2a, but as seen by Kümmel (2015+, 2018), there is no Vedic parallel for -s-t° in the 2sg middle. 360 It is more likely that the Vedic iṣ -aorist stem -jániṣ- ajaniṣṭhās ajaniṣṭa and du. act. janiṣṭām janiṣṭam 361 go back to a Proto-Indo-Iranian root aor. mid. 2sg */ʥánH-stHās/, 3sg */ʥánH-sta/ < *ǵénh1-sto with 362 3sg ending *-sto, thus likewise PII*/ʥánH-stām/ < *ǵénh1-stah2(m). 363 A proto-middle origin is evident for the 2du 3du forms RV janiṣṭām, AV janiṣṭam, since -jániṣ- 364 is otherwise middle all over (medium tantum). Another otherwise fully middle paradigm (-vr̥ ṇī-) also 365 includes 2du 3du “active” RV 1.180.4b avr̥ ṇītam, AVP 1.92.1 vr̥ ṇītām. Moreover, it is evident that the 366 Vedic 3rd dual ending -tām < *-tah2(m) is still used as a middle (not active) ending in RV 10.4.6, where 367 3du adhītām (-tām) even has a passive meaning and belongs to the aorist middle (adhithās, adhita, 368 adhīmahi, etc.), cf. RV 10.4.6 tanūtyájeva táskarā vanargū́ raśanā́ bhir daśábhir abhi àdhītām “So wie 369 zwei den Leib hingebende, im Wald umhergehende Räuber, mit den zehn Zügeln (raśanā́ bhir daśábhir, 370 sc. mit den 10 Fingern) wurden die beiden (sc. Reibhölzer) [...] festgemacht” (Pooth 2011). 371 e. There are the following IE comparanda. Hittite shows sigmatic ~ asigmatic 3sg pret. act. endings, 372 e.g. ākiš (a-ak-ki-iš) (OS) → ak-ta (NS, OH/NS NH) (cf. Melchert & Hoffner Jr. 2008: 189, 215; 373 Kloekhorst 2008: 167), cf. also OH a-ša-aš-ta from the ḫi-pres. ašaši (Oettinger 2002: 51, 430, 374 Kloekhorst 2008, s.v.). 375 Hittite further exhibits suffix PLEONASM: pa-iš (OS), pa-iš-ta (OH/NS), pe-e-eš-ta (NH) 376 (Kloekhorst 2008: 614); tar-na-aš (OS), tar-ni-eš-ta (KUB 13.34 iv 14 (NS)), tar-ni-iš-ta (KUB 1.1+ 377 iv 49 (NH)) (Kloekhorst 2008: 846); ši-pa-an-ta-aš (KBo 15.10 iii 59, 66 (OH/MS)), ši-pa-an-da-aš 378 (KBo 15.10 iii 64, 68 (OH/MS)), ši-pa-an-za-aš-ta (KUB 20.59 v 6 (see Groddek, DBH 13, p. 106), 379 KBo 8.68 iv 5), ši-pa-an-da-za (KUB 19.37 ii 24 (NH)) (Kloekhorst 2008: 405, Oettinger 2002: 41, 380 408); cf. also ḫa-a-az-ta (= /ḫāt-št/) (OH/MH), ḫa-a-az-za-aš-ta (MH/MS), 3pl. ḫa-a-te-e-r (OH/MS) 381 (ḫāt-i, see Oettinger 2002: 408, Kloekhorst 2008: 328). 382 The Proto-Anatolian sigmatic ending *-s(t) perfectly equates with Proto-Indo-Iranian *-s(t) (and 383 its middle counterpart *-sta). Tocharian AB show sigmatic 2sg 2pl 3sg act. pret. endings, cf. Figures 384 10ab (Malzahn 2010: 38ff. and chapters 7-9). 14 385 386 387 388 389 390 Figures 10a (Malzahn 2010: 38ff. and chapters 7-9) ā̆-inflection non-ā̆-inflection reconstruction TB TA TB TA PT 2sg -ā̆sta -āṣt -asta -äṣt *-stā 3sg -a -Ø/-āṃ -sa -äs/-sām *-sā 2pl -ā̆s(o) -ās -as/-so *-sä ~ *-så (NB. Malzahn 2010: 514 reconstructs PT 2pl *-sās) 391 Figures 10b (Malzahn 2010: 38ff. and chapters 7-9) 392 393 394 395 396 PT reflecting internally comparable with 2sg *-stā *-sth2e *-th2e 3sg *-sā *-sh2e (?) *-th2e 2pl *-sä *-se or *-so *-te, *-ste, 2sg *-sto 2pl *-så *-seh2 *-teh2- (> PII 2/3du *-tām) 397 When taken together, the comparative evidence points to a 2sg proto-middle segmental string *-sth2e(i), 398 cf. Proto-Tocharian 2sg pret. *-stā, Hittite (NH) 2sg pret. ind. -(i)šta (e.g. pa-it-ta (OH/MS), pé(-e)eš- 399 ta (NH), cf. Kloekhorst 2008: 614), Lat. 2sg perf. ind. act. -istī < -istei, e.g. CIL 10 gesistei, Greek -σθα, 400 e.g. Homeric Greek τίθησθα, etc. It is conclusive that the traditional segmentation into sigmatic stem 401 plus ending †-s-th2e(i) must be a mistake. 402 It has been suggested that Hittite -(i)šta is an inner-Anatolian creation (cf. differently Jasanoff 403 2003: 119, Kloekhorst 2008: 802). But this hypothesis is rather weakened by the IE equations. After 404 subtraction of the “union” vowel -i-, Latin 2pl perf. ind. act. -istis includes a string *-stes corresponding 405 to Hittite 2pl ind. act. pret. -šten :: pres. -šteni (Kloekhorst 2007). This clearly points to PIE sigmatic 406 2pl act. endings *-ste(-), *-ste(n)(i). Therefore, Latin °sti° (-istis) = Hittite -šte° (-šten(i)) is a compara- 407 tive equation that is unlikely secondary at least with regard to the sigmatic morph. An interesting and 408 relevant paradigm regarding the sigmatic morph in Early Vedic is the one of 1sg cyávam, 3pl cyavante 409 (:: Old Avestan š́ auuaitē) alongside a sigmatic 2sg cyoṣṭhās. These two forms point to a PIE “Narten” 410 (= durative) proto-middle 1sg *kwiéuh2e(i) (TPkwi̯ éu̯ (h2)om), 3sg *kwiéuo(i), 3pl *kwi̯ éu̯ n̥ to(i) ~ 411 *kwi̯ éu̯ onto(i) and a sigmatic 2sg *kwiéusth2e (later + 412 Vedic 2sg cyoṣṭhās has the sibilant exactly in the paradigmatic position where we would expect it from 413 the given comparative perspective – and thus points to an ending *-sth2e(-). 414 415 TP -es) (2/3collective *kwi̯ éusteh2). Remarkably, It is conclusive that PIE had a set of 2sg 3sg (2pl 3pl) sigmatic endings (Kümmel 2015+, 2016, 2018) with a hitherto UNKNOWN different function. 416 Figure 11. Post-PIE pleonasms 417 PIE post-PIE, dialectal/not in all dialects (partially) 418 *-s ~ *-st → *-s-t 419 *-so ~ *-sto → *-s-to 420 *-(e)rs → *-sn̥ t, *-sn̥ to, *-sonto → *-s-n̥ t, *-s-n̥ to, ... 15 421 These endings merged with the non-sigmatic ones yielding PLEONASMS, e.g. PIE *klép-t → root aorist 422 (vs. *klḗp-t → “Narten type”), PIE *klép-s → *klép-st → post-PIE sigmatic aor. *klép-s-t, cf. Latin 423 clepsit ~ clēpit (Pike 2009), and likewise *klḗp-s → *klḗp-st → sigmatic aor. *klḗp-s-t. The sigmatic 424 segment was transferred to other persons – as typologically confirmed by the development of sigmatic 425 root aorist optative forms to “precative” stems in Early Vedic. We can conclude that the sigmatic aorist 426 suffix, both its form (as a stem) and its function (as aorist), should NOT be reconstructed for the common 427 protolanguage of Hittite, Tocharian, Vedic, Greek. To be clear, the reconstruction of a sigmatic aorist 428 stem for the common proto-language of these languages must be a severe anachronistic mistake. 429 Based on the comparative method it seems true that the addition of *-t- to *-s- yielding *-st- 430 was a secondary suffix pleonasm, similar to the one that affected the old so-called “dental-less ending” 431 of the 3sg middle, cf. Hittite -at = Vedic 3sg aor. injunctive middle -at (aduhat) < *-ot with additional 432 morph *-t vs. earlier *-o without *-t. A primary status of *-s- (and *-st-) as ending is thus evident from 433 Hittite and Vedic. The same status seems to hold for Proto-Tocharian and Toch. A and B, see Figures 434 10ab (Malzahn 2010: 38-44, 192). The assumption that the *-s- was lost in the first person and third 435 person plural forms in Proto-Tocharian and Proto-Anatolian is highly unlikely. 436 437 To conclude, the pre-Vedic “secondary ending” of the 3sg *-s or *-st clearly corresponds to, and equates with, Hittite -š, -št(a).17 438 Figure 12. A Hittite-Vedic equation 439 Hittite 440 -š, -št(a) = Vedic PIE -s < *-s 441 This comparative equation is a riddle for comparative philology, presumably because “another” third- 442 person singular ending *-s alongside *-t does not make much functional sense when interpreted within 443 the canon of categories expressed by the “secondary” endings. Clearly, however, the two variants are 444 much more important than hitherto assumed. As initially outlined by Watkins (1962), this suffix was 445 the same sibilant morph *-s- that developed into a sigmatic stem suffix from post-PIE to Proto-Indo- 446 Iranian/Iranic and Proto-Greek and some other IE languages – either when some of them perhaps formed 447 a common proto-sub-branch or independently from each other in close areal contact. It is conclusive 448 that several IE languages other than Proto-Anatolian and Proto-Germanic reflect a post-PIE morpholog- 449 ical reanalysis of the first part of *-st- as a new sigmatic stem suffix (→ EF-s-t). 450 Based on the given comparative evidence, it is further possible to reconstruct a proto-middle 451 counterpart *-so- ~ *-sto- via internal reconstruction, e.g. *stéuso ~ *stéusto > Vedic (á)stóṣṭa, etc. 17 This correspondence can be taken for an external comparative proof that *-s (or *-st) occurred in PIE 3sg forms. 16 452 Thus, it is likely that the ending *-so- (> Proto-Indo-Iranian/Iranic 2sg pres. ind. mid. *-sai, Proto-Greek 453 *-soi, etc.) was not a PIE second person form, but a PIE third person form. If true, it was reinterpreted 454 as a post-PIE 2sg middle ending due to the following analogy proportion: 455 2sg *-si :: 3sg *-ti = X (→ *-so(i)) :: 3sg *-to(i) 456 The corresponding sigmatic 3pl forms were marked by *-rs (*-r̥ s), cf. Old Avestan 3pl cikōit-ǝrǝš, etc. 457 The third person plural optative form, e.g. Vedic syúr, goes back to *h1s-ih1-r̥ ́ s (or *h1s-ih1-r̥ ́ ). This form 458 *h1s-ih1-r̥ ́ s reflects a blend of the endings -r̥ s and -érs, and was preceded by PIE *h1sih1érs. This form, 459 in turn, was structured like the one with the alveolar morph *h1sih1ént > Old Latin sient. The same 460 ending *-rs is found with e-grade (*-eːr) reflected as Hittite 3pl pret. active ending -er (~ -īr ~ -eir ~ -ēr) 461 < *-eːr. This is an outcome of PIE word-final */érs/ via Szemerényi’s law (Jasanoff 2003). It is still 462 continued as the first part of Old Latin -ēre and -ērunt (< *-eːr extended by *-i and *-ont, respectively 463 (Weiss 2009: 393). To conclude, *-rs is finally separable into two agglutinated morphs *-r- (PLURAL) 464 + *-s (category S). 465 Figure 13. Some PIE forms and a transponatum exhibiting *-s- 466 467 468 ACT DTR 3SG(/2SG) 3PL *kléps *klpérs h *b ud só cf. Vedic 1sg ábhutsi 469 ACT *klḗps 470 DTR *stéusoBC 471 DTR 472 Additional notes to Figure 13: 473 A 474 B 475 476 cf. Latin 3sg clepsit h TP *kléprs B cf. Latin 3sg clēpitA sóikws(i)D Cf. Pike 2009: Latin shows reflexes of both *klḗp-t and *klép-s. Cf. Vedic (á)stóṣṭa – and also *klépto → post-PIE *klḗptor(i) with analogical lengthened grade > Toch. A klyepträ ‘touches’. C In my view, the Vedic 2sg imperative middle ending -sva goes back to a corresponding sigmatic 2sg 477 form with a suffix *-u (e.g. *séikwsh2u) that was extended by *-a → *-su̯ a before Proto-Indo-Iranian/Iranic. 478 But the question of the prehistory of -sva is irrelevant here. 479 D Cf. Toch. A -ṣ (< TP -si) and Vedic (ŚB) °seci (< *sóikwi) and Toch. A 3sg subj. V sekaṣ (TPsóikwHsi). 480 Recall that *-a- < *-H- was generalized as a stem suffix in Proto-Tocharian (Jasanoff 2003, chapter 6-7, 481 §93ff.; Malzahn 2010: 274-316, both of them with further references). 482 2.3. 483 A third 3rd person finite verb form was the so-called “dental-less stative” form. However, the term “sta- 484 tive” is semantically inappropriate and should not be used anymore (Pooth 2000). It was marked by a 485 zero morph to left to the *o in the middle voice: *-Øo :: *-so :: *-to, cf. Young Avestan +ni-γne ‘is The zero morph 17 486 knocked down’ = kuna° of Hittite kunati, Vedic stáve ‘is praised’, etc. The plural forms, e.g. *ǵusró 487 (from *ǵeus- ‘to taste, choose; please someone; like, find pleasure in sth.’) were marked by *-ro(-) 488 without any sibilant or alveolar morph (cf. RV 1.71.1b ajuṣran). The corresponding active plural forms 489 were marked by the *-r (= *-r̥ ) with no further extension, cf. Old Avestan -ǝrǝ̄. The form *-er with e- 490 grade is still found as the first part of Old Latin 3pl perfect ind. -erunt < *-er extended by *-ont/*-n̥ t.18 491 The third person singular forms of the *uóide(i) type and a zero-marked 3rd person singular form 492 *ǵónh1(i) of a “pre-passive-aorist inflectional type” also belonged to this category Z because there is no 493 stop -t- or sibilant -s- before the vowel in *uóid-e(i). 494 Figure 14. PIE forms exhibiting zero 495 3SG 3PL *gwhnér 496 ACT 497 DTR 498 ACT 499 DTR *ḱéio(i) 500 DTR *uóide(i) *dhughó(i) *dhughró(i) cf. Vedic 3sg duhé, pl. duhré *mélh2rA cf. Old Avestan -ǝrǝ̄ cf. Vedic 3sg śáye IR melh2órB C 501 DTR 502 ACT 503 ... 504 DTR 505 Additional notes to this figure: 506 A *ǵónh1(i) cf. Vedic 3sg véda etc. cf. Vedic 3sg jani, ájani *di-dh3ér cf. Old Latin dederunt *dé-dh3oD This 3pl form belonged to the *stḗuti type. The proto-active forms of the this so-called “Narten type” 507 merged with the PIE *uóide(i) type yielding a post-PIE new imperfective (including habitual) mixed para- 508 digm *u̯ óid-ei, *u̯ óid-e ~ *u̯ ḗid-s(t) (Jasanoff 2003; Pooth 2016, 2017). B 509 I think that, e.g., IRmelh2ór ‘they (always) grind’ was the original 3pl form of the *mólh2e(i) type. This 510 form must have been modified to post-PIE *melh2r̥ ́ (s) ~ *molh2r̥ ́ (s) with and without paradigmatically lev- 511 elled o-grade. Some plural form, e.g. *u̯ id-r̥ ́ (s), had zero-grade in analogy to the other non-reduplicated 3pl 512 forms *gwhnéːr, *gwhnént, etc. The analogical spread of the zero-grade in such 3pl forms is supported, e.g., 513 by Vedic stuvánti ← *stéu̯ n̥ ti. Kloekhorst (2012) thinks that Hittite would confirm that the zero-grade was 514 original in this type, but Jasanoff (2003: §27) is correct by claiming that there is independent evidence for an 515 e-grade in the corresponding reduplicated type, cf. Vedic ádīdhayur, avivyacur, where the guṇa ablaut-grade 516 of the root must be archaic. Therefore, there is Vedic evidence for a 3sg perfect *u̯ eu̯ i̯ ókwe :: 3pl with e-grade 18 It is not necessary to reconstruct *-is-ont in this particular case (pace Weiss 2009: 393). Jasanoff 2003: 33, fn. 11 claims that -erunt should go back to *-iront (cf. -imus, etc.), but this is likewise unnecessary: -erunt from *-er is straightforward. 18 517 *u̯ eu̯ i̯ ekwr̥ ́ (s) besides *u̯ eu̯ idr̥ ́ (s). As just mentioned, I follow the idea that the *u̯ óide(i) type merged with the 518 “Narten type” *stḗus(t) (and *stḗut) in Late PIE (Jasanoff 2003). Within a “mixed paradigm” we can expect 519 the given IE ablaut variants. Recall that Hittite -ar can go back to both *-r̥ and *-or. C 520 The PIE *ǵónh1(i) type merged with the aorist-like part of the *ǵusó or *dhughó(i) type in post-PIE. This 521 yielded another mixed paradigm which is reflected as Vedic middle root aorist. Vedic abudhran abudhram 522 both go back to *bhudhró. I assume that forms of this shape originally belonged to the *dhughó(i) type and 523 were structured in parallel with *dhughntó(i) (with *-r- instead of *-nt-). I think that the old 3pl form belong- 524 ing to *ǵónh1(i) was *ǵnh1ór. Its structure matches the one of *h1sónt(i) (again with *-r- instead of *-nt-). 525 The issues with this merger are a bit more intricate; 3pl forms of this type seem to be allomorphs of the type 526 that once corresponded to the amphikinetic active forms, cf. 3pl active *h1sént(i) :: 3pl proto-middle 527 *h1sónt(i). I will have to return to this difficult matter elsewhere. It is not relevant here. D 528 This zero-marked form was later pleonastically extended by the productive 3sg ending *-to(i) in post- 529 PIE first yielding *dédh3oto(i) → later analogical *dédh3eto(i) > Vedic dáda-te (Pooth 2014). 530 3. How to analyze these three morphs? 531 3.1. Three morphs for just two IE functional distinctions? 532 As a given result of the application of the comparative method to the IE material, the IE 533 formulas and TPtransponates very clearly continue three different morphs *-t-, *-s-, and zero. There- 534 fore, the comparative evidence, taken together, rather speaks against the existence of a binary opposition 535 of “primary” vs. “secondary” endings in the common ancestor language. The most straightforward ex- 536 planation is definitively not to leave the two ending sets the way they are in Greek and Vedic Sanskrit. 537 Not touching any IE functional distinctions would enforce us to functionally reduce the three morphs to 538 just two morphological categories because Greek and Vedic both basically distinguish two opponent 539 sets of “primary” vs. “secondary” suffixes, not three (when we leave away the special proto-middle 540 endings of the perfect stem). Given that the comparison actually leads to one more set, such a reduction 541 would need many troublesome ad hoc explanations. In short, we would have to get rid of one of the 542 three sets. However, the less costly and easiest explanation is to infer that these three morphs once 543 indicated three different verbal categories, not two, simply because their number is three, not two. These 544 verbal categories were presumably given up from PIE to post-PIE. They are provisionally termed cate- 545 gories D, S, and Z here. 546 3.2. 547 A proper identification of the morphosyntactic functions of these three lost verbal categories of PIE 548 cannot be done on the basis of IE correspondences and the comparative method, because these three 549 categories D, S, and Z have vanished without a functional trace. I admit that a reconstruction of verbal 550 categories that are not in the canon of the IE verbal categories is not provable and cannot be falsified by 551 comparative means. However, the lack of comparative proof is due to the use of the wrong method for EF equation Assigning three functions to D, S, and Z 19 552 this particular purpose. It should be recognized that the lack of comparative falsification is a corollary 553 of the problem that category equations (e.g. Greek dual = Vedic dual, etc.) cannot be used for the recon- 554 struction of categories and their exponents because they do not offer reliable results with regard to gram- 555 matical categories that have vanished or were modified in between PIE and its daughter languages. 556 Comparative philologists should recognize that this is an unfortunate blind spot of the comparative 557 method. When it is only this method that is applied to grammatical categories, the results of comparison 558 are unreliable and there is no chance whatsoever to uncover morphosyntactic categories and syntactic 559 constructions that were lost or modified from PIE to a post-PIE dialect continuum. However, it is unde- 560 niable that the application of the comparative method actually results in a formal distinction of three 561 different morphs, not two. Assigning a functional value to these three morphs, therefore, cannot be done 562 by application of the comparative method to morphosyntactic categories any longer. It must be done via 563 inductive reasoning against a background of general linguistic diachronic typological knowledge and 564 findings. 565 Figure 15. The origin of the IE endings from progressive and non-progressive forms 566 initial shift 567 PIE 568 PROGRESSIVE 569 *-m-i *-mi 570 *-s-i *-si 571 *-t-i *-ti 572 *-n-t-i *-n̥ ti 573 *-é-n-t-i *-énti 574 etc. post-PIE → IMPERFECTIVE PRESENT 575 second shift 576 PIE post-PIE 577 NON-PROG “secondary” 578 *-m *-m̥ that is, secondary (viz. innovative) functions 579 *-s *-s ~ *-st (but not including the optative19) 580 *-t *-t 581 *-r-s *-r̥ s ~ *-r̥ 582 etc. 19 post-PIE → perfective/preterit, etc., The optative-irrealis mood is not secondary. Optative-irrealis forms were already optative-irrealis forms in PIE. But they lacked the progressive suffix *-i. 20 583 As illustrated, it is evident that these three morphs functionally merged and were integrated into the 584 bundle of post-PIE so-called “secondary endings” in IE languages. Consequently, I must claim that the 585 development of specific “secondary” endings only followed the emergence of the “primary” imperfec- 586 tive present forms and endings from PIE PROGRESSIVE aspect forms, see Figure 15 (Pooth 2009a). The 587 IE secondary functions of the “secondary” endings (e.g., the coding of perfective (“aorist”) aspect and 588 past tense) must not be functionally back-projected onto the protolanguage. To conclude, I claim that 589 the distinction of “primary” imperfective present endings from secondary endings is only a post-PIE to 590 post-PIE innovation. This may explain relics like Vedic kr̥ thás (aorist injunctive) with irregular “pri- 591 mary” ending (Pooth 2011: 478, fn. 12). 592 As mentioned above, the traditional comparative reconstruction of morphosyntactic functions 593 and grammatical categories – and their formal exponents – is methodologically unreliable and linguis- 594 tically insufficient (Pooth 2017). Conclusions which are solely based on the application of the compar- 595 ative method to morphosyntactic categories and their markers can only be very tentative and provisional, 596 not only because morphosyntactic functions and grammatical categories that have vanished completely 597 or were modified in parallel within a post-PIE variant cluster cannot be compared. The plain fact that 598 the original morphosyntactic categories and their exponents do not exist in their original forms and 599 functions in IE languages does not mean that they did not exist. The issue is a severe pitfall of compar- 600 ative philology (Pooth 2017). However, even if comparative equations alone and per se cannot 100% 601 prove or disprove a given hypothesis regarding the linguistic reconstruction of morphotactic structures 602 and morphological, syntactic, and lexical categories, this does not entail that a proper linguistic recon- 603 struction of morphosyntax is impossible – nor does it imply that the comparative method cannot be used 604 as a start. We can use a less monolithic methodology and further means of reasoning for this purpose. 605 In my view, it is necessary to first use comparative means, but finally think out of the “comparative 606 box”. As for morphosyntactic functions and grammatical categories, a plausible and potentially real and 607 realistic descriptive analysis and reconstruction can be established by inductive and deductive reasoning 608 and by combining the results of comparison with morphosyntactic diachronic linguistic typology. Such 609 a way of doing linguistic reconstruction implies using ALL the multiple linguistic methods which have 610 been supplied so far by general and historical linguistics and comparative philology. The linguistic re- 611 construction presented in this paper is “post-mono-comparative” and “poly-methodological” in this 612 sense (see Pooth 2015, 2016, 2017). Its reasoning is based on more than one kind of observations: 613 1. FIRST OBSERVATION, BASED ON EXTERNAL COMPARATIVE AND INTERNAL PIE EVIDENCE: There 614 is a relevant formal similarity strengthening the conclusion that these three lost categories D, S, Z were 615 verbal DEICTIC-DIRECTIONAL categories. The two morphs *-t- and *-s- are comparable to the deictic 616 morphs *t- vs. *s- which are found in the PIE demonstrative pronoun system, cf. PIE *só > Vedic sá, 617 PIE *tó- > Vedic tá-, etc. 618 2. SECOND OBSERVATION, BASED ON DIACHRONIC TYPOLOGY: There is a relevant grammaticaliza- 619 tion/grammaticization path which is of major relevance here. It has become an established insight that 21 620 “erstwhile deictic-directional markers” constitute a “natural diachronic pathway for the grammaticali- 621 zation of semantic inverse systems” (Givón 2001 I: 166). A useful typology of such direct-inverse sys- 622 tems has been provided by Jacques & Antonov (2014). DIRECT marking typically indicates that proxi- 623 mative/proximate20 participant/argument higher on the person hierarchy is acting on the lower obvia- 624 tive/obviate participant/argument, while INVERSE marking encodes the opposite direction. 625 3.3. 626 I am in the position to draw the following conclusion via deductive reasoning: A possible deduction 627 1. THE GIVEN INTERNAL EVIDENCE (VIZ. THE COMPARATIVE FACTS): It is internally evident that the 628 two morphs *-t- and *-s- look exactly like deictic-directional markers that occur as morphs in the 629 demonstrative stems *só-, *tó-. To be fair, it is not absolutely assured that they were – but at least it is 630 absolutely clear that they exactly look like deictic-directional markers. This formal identity is taken as 631 comparative fact and sufficient internal evidence here. 632 2. THE “RULE”/DIACHRONIC PATTERN: A GRAMMATICALIZATION PATH: It is a diachronic “rule” or 633 regularity that “erstwhile deictic-directional markers” constitute a “natural diachronic pathway for the 634 grammaticalization of semantic inverse systems” (Givón 2001 I: 166; as already mentioned). 635 3. THE CONCLUSION: It is plausible to identify these three lost categories with verbal deictic-direc- 636 tional grammatical categories belonging to the domain of transitivity direction. Typologically speaking, 637 this assumption is plausible and possible, if not even most probable, as outlined below. 638 I must admit that this way of reasoning is only similar to a classical deduction. There is general 639 law, rule, or regularity, i.e. the mentioned grammaticalization/grammaticization path; there are given 640 facts, i.e. the given internal evidence, formal identities and functional similarities. Of course, this de- 641 duction does not exclude other possibilities because the rule cannot be exceptionless, and the functional 642 identity of formally identical morphs cannot be 100% proven. However, the formal identity of *-s- and 643 *-t- with deictic-directional *s- and *t- and their functional similarity (referent tracking, referent index- 644 ing), when taken together with the well-known diachronic path, make other prehistorical possibilities 645 LESS EXPECTED AND A BIT LESS LIKELY in a probabilistic sense. 646 Thus, let us have a look at the possible alternatives. First of all, suggesting that these three lost 647 categories D, S, Z once were aspect categories is actually not very plausible because there is little that 648 speaks in favor of it. *-t- is found in endings of both the IE imperfective and perfective aspect. I admit 649 that *-s- is reflected as marker of perfective aspect in IE languages, although this does not hold true for 650 Anatolian languages and Hittite. It is unlikely that an original sigmatic stem suffix was transformed into 20 I use the terms PROXIMATIVE and OBVIATIVE for the two hierarchically indexed participants or arguments here, although these notions are also often termed proximate and obviate. 22 651 a second-person and third-person past tense suffix in this branch. It is more likely that the opposite 652 happened outside Anatolian. And what about the second person singular *-s-? Should we functionally 653 separate the two morphs, or shouldn’t we? Questions like these are relevant for a proper analysis of the 654 equation formulas and a subsequent reconstruction of morphosyntactic categories. The traditional ap- 655 proach cannot answer such questions without making ad hoc assumptions. Even worse, such questions 656 are not usually touched at all because it is often taken for granted that PIE grammar was (almost) iden- 657 tical to Ancient Greek and Vedic Sanskrit grammar, except for its phonology. At least in my view, this 658 preconception is unlikely from the outset. It is rather not plausible to suggest that these markers were 659 inflectional portmanteau endings and functionally identical with the IE “primary” vs. “secondary” end- 660 ings (pace Lundquist & Yates 2018). If we made this assumption, we would not be able to give a dia- 661 chronic justification for the given variation. We would have to ignore the morphological problems. We 662 would have to neglect a proper linguistic analysis of PIE word forms, as provided here. We would have 663 to make rather unlikely claims. We would have to ignore the formal and functional identity of recon- 664 structed morphs, as presented above. The internal evidence would remain completely unexplained, and 665 no additional knowledge would be gained. In the end, PIE morphosyntax would remain a decalcomania 666 of Greek and Vedic grammar. 667 We could set up categories like a “stative” to describe forms like *gwhnó(-) ‘someone is/was 668 slewn’ and *stéuo(-). But this is problematic, if not just wrong, because this functional assignment is 669 descriptively wrong. The existence of a “stative” is contradicted by the IE comparative evidence: the 670 relevant middle forms do not indicate a state, nor are they intransitive (but are even transitive in several 671 cases). There is nothing that speaks in favor of the assumption that PIE had a state-indicating “stative” 672 verbal voice category (Pooth 2000). 673 Also, any assumption of original marking of tense distinctions is unlikely and almost excluded. 674 The Vedic “injunctive” is an archaic tense-neutral/nontensed category. Its existence strongly points to a 675 preceding tense-neutral/nontensed type of language. Accordingly, a PIE 3rd person singular form 676 *gwhént(i) could be used both with present and past tense reference (Pooth 2009a, Kloekhorst 2017). 677 Assuming that the three categories were specific tense categories makes no sense when seen from this 678 systematic perspective. Although PIE maybe had a non-obligatory morph, namely the so-called “aug- 679 ment” *(h1)é(-), for particularly specifying something functionally similar to past tense, its potential 680 existence cannot disprove the argument that PIE was nontensed because its PIE function might have 681 been different from past tense. 682 Moreover, these endings were definitively not just person markers. Such an assumption cannot 683 explain the fact that the very same markers *-s- and *-t- coded both second and third person forms. 684 Finally, these markers were definitively not specified for singular or plural or whatever number. If one 685 suggested this, one would not be capable of explaining the fact that the same segments *-t-, *-s-, and 686 zero occur both in singular and plural forms. 23 687 To sum up, our answer of the question regarding a possible original function of these categories 688 D, S, Z must be that original marking of person, number, voice, tense, aspect, and mood distinctions is 689 almost ruled out – or at least rather unlikely. 690 It is conclusive that these markers were markers of a different domain, such that is non-existing 691 in IE languages. This functional domain can now be identified as a verbal deictic-directional domain. 692 We can conclude that the domain of TRANSITIVITY DIRECTION is a perfectly suitable candidate and 693 remains the most likely option. I am happy to give some credit to the Leiden School at this point in our 694 discussion because Kortlandt was on the right track when he suggested that *-to- goes back to a “tran- 695 sitive middle” marker (Beekes 1995, 2011). However, even if other scholars may deserve the merit to 696 be the first to “somehow envision” that the original functional distinction belonged to the domain of 697 transitivity and transitivity direction (‘who is acting on whom’), it is clear that nobody has ever recon- 698 structed a hierarchical direct-inverse system for PIE before. 699 Let me conclude as follows. On the primary basis of the comparative method, secondarily ex- 700 tended by cross-linguistic linguistic diachronic typological patterns, we are able to plausibly reconstruct 701 three different PIE verb forms of the second person and third person singular, plural, and collective 702 number, respectively. The PIE second and third person forms are given in Figure 16. The PIE verb is 703 *gwhen- ‘to slay, kill, beat, chase, hunt’. 704 Figure 16a. Directed/transitive and undirected PIE verb forms 705 suffix 706 DETRANSITIVE (PROTO-MIDDLE) 3SG 2SG 707 D = DIRECT (DIR) *-t- *g ntó *gwhnth2é A 708 S = INVERSE (INV) *-s- *gwhnsó *gwhnsh2éA 709 Z = UNDIRECTED (UDR) *-Ø- *gwhnó *gwhnh2éB 2COL 2/3COL 710 wh wh 711 D = DIRECT (DIR) *-t- *g ntóm *gwhntéh2(m) 712 S = INVERSE (INV) *-s- *gwhnsóm *gwhnséh2(m) 713 Z = UNDIRECTED (UDR) *-Ø- *gwhnóm *gwhnéh2(m) 714 ACTIVE-AGENTIVE 715 3SG/2SG wh voice 2PL 716 D = DIRECT (DIR) *-t- *g ént *gwhnté(n) 717 S = INVERSE (INV) *-s- *gwhéns *gwhnsé(n) 718 Z = UNDIRECTED (UDR) *-Ø- *gwhénC *gwhné(n)B voice 24 719 Figure 16b. Transitive and intransitive PIE verb paradigm 720 suffix 721 722 D = DIRECT (DIR) *-t- 723 S = INVERSE (INV) *-s- 724 Z = UNDIRECTED (UDR) *-Ø- DETRANSITIVE voice 3PL 2PL *gwhnntó *gwhnth2é(n)A *gwhnsh2é(n)A *gwhnró *gwhnh2é(n)B 725 ACTIVE-AGENTIVE 726 3PL 727 D = DIRECT (DIR) *-t- *gwhnént 728 S = INVERSE (INV) *-s- *gwhnérs 729 Z = UNDIRECTED (UDR) *-Ø- *gwhnér 730 Additional notes to Figures 16a and b. 731 A voice Parallel to the third-and-second person collective forms in *-téh2(-m) (→ Proto-Indo-Iranic 3-and-2-dual 732 active *-tām), the 2pl detransitive forms were reanalyzed as belonging to the neo-active voice category form 733 PIE to post-PIE. *-sh2e seems to be reflected in Proto-Tocharian (cf. Figures 10a and 10b above), whereas 734 *-th2e is reflected as the Vedic primary 2pl pres. ind. (neo-) active ending -tha (Pooth 2011). B 735 736 In my view, the 2pl in *-h2é of the detransitive voice merged with the active-agentive in *-é. It is reflected by the Vedic 2pl perf. ind. act. -á, e.g. Vedic vid-á. C 737 The zero-marked 2sg form was used in imperative function. 738 4. 739 I am indebted to Guillaume Jacques (personal communication via internet) for asking me the following 740 important question: 741 Why do I reconstruct *-s- as inverse marker? “Why do you reconstruct *-s- as inverse marker but not *-t-?” 742 The following internal evidence provides some reasons for doing so. The reconstruction can be based 743 on the following three arguments. Two of them are outlined here in more detail. 744 1. First of all, the suffix *-s- was also used as a SOURCE marker in ablative-genitive case forms (e.g. 745 genitive *dém-s ‘family’s, of the family’, *négwt-s ‘dusk’s, of dusk’, *diéus ‘daylight’s, sky-god’s, of 746 daylight, sky-god’, ablative-genitive *diués, etc.). The *-t- had a deictic meaning, e.g. *tód *dóm 747 ‘DEMONSTRATIVE house, to DEMONSTRATIVE house’. I can draw the inference that *-s- coded a direc- 748 tion towards the primary (i.e. topical, proximative/proximate) participant originating from the second 749 participant as the SOURCE of the causation, whereas the *-t- coded causation towards a GOAL, that is, 750 direction from the primary (i.e. topical, proximative/proximate) participant to the second participant. 751 The following examples can serve to illustrate what is implied here: 25 752 (1) a. *gwhén-t 753 slay:NONDURATIVE/NEUTRAL.ASPECT:ACT(AGENTIVE):SG-DIRECT\3 754 literally, ‘agentive slaying of topical participant to second participant’ 755 b. *gwhén-s 756 slay:NONDURATIVE/NEUTRAL.ASPECT:ACT(AGENTIVE):SG-INVERSE\3 757 literally, ‘agentive slaying of topical participant (hither) from second participant (as the 758 starting point of the action)’ 759 Inferences like this one must of course be based on possible diachronic sources of direct and inverse 760 markers. It has been proposed that word forms indicating a CISLOCATIVE direction (‘hither’) towards 761 the speaker (or towards the topical participant) are a possible source of inverse markers (see Jacques & 762 Antonov 2014: 325ff.). To illustrate this diachronic path, let me quote Jacques & Antonov (2014: 325ff.) 763 (the numbers of their examples are mine): 764 “The term ‘cislocative’ is used to refer to markers expressing a motion towards the speaker, both 765 directional (‘verb hither’) and associated motion (‘come to verb’) ones. The inverse marker in Nez 766 Percé (Sahaptian) used in local scenarios with second person acting on first person has grammatical- 767 ized from an earlier cislocative marker, reconstructed for Proto-Sahaptian as *-im (Rude 1997, 122). 768 [2] héexn-e 769 see-PST 770 I saw you. 771 [3] héxn-ím-e 772 see-CIS-PST 773 You saw me. [PST = PAST, CIS = cislocative] 774 Interestingly, this marker has also grammaticalized into the ergative case suffix which appears on the 775 non-SAP agent in mixed scenarios in cases where one would expect inverse marking on the verb (Rude 776 1997, 121-2). 777 [4] hi-héxn-e háama-nm 778 3S/A→SAP-see-PST man-ERG 779 The man saw me/you. (NEZ PERCÉ) 780 [...] In still other languages, we observe a formal similarity between some inverse or direct markers 781 and various types of third person markers including agent, patient, or possessive affixes. The clearest 782 case is the inverse prefix found in Sino-Tibetan (Rgyalrongic and Kiranti) languages. As first noticed 783 by DeLancey (1981b), the Situ Rgyalrong third person possessive prefix wə- is formally identical to 784 the inverse marker. This is also true of other Rgyalrong languages and of some Kiranti languages that 26 785 have an inverse marker, such as Bantawa [...]. The similarity between the two sets of prefixes is strik- 786 ing and suggests a grammaticalization from a third person marker into an inverse marker. While the 787 exact pathway remains unclear and thus requires further investigation, it is possible that non-finite 788 verb forms carrying a third person possessive prefix were reanalyzed as finite ones. In the case of 789 Sino-Tibetan languages, this scenario probably occurred in the ancestor of both Rgyalrongic and 790 Kiranti languages [...].” (End of quote.) 791 This bidirectional “poly-grammaticalization” in Nez Percé from CISLOCATIVE both to inverse direction 792 and ergative case marking looks very similar to what is found in PIE morphology. In PIE, the sigmatic 793 agentive (viz. pre-nomimative) case suffix *-s (e.g. PIE *póti-s ‘lord’, *h2nér-s ‘man’ > post-PIE 794 *h2néː(r)) and the ablative-genitive SOURCE suffix *-s (e.g. PIE *négwts, see above) are formally identical 795 with the verbal deictic-directional suffix *-s-. This typological parallel strengthens the conclusion that 796 *-s- was a cislocative-like inverse marker and *-t- the corresponding direct marker. 797 2. A second argument may run as follows. It is perhaps more likely claiming that the PROGRESSIVE forms (e.g. *gwhénti), not the inverse forms, were later generalized as a post-PIE third person 798 DIRECT 799 singular imperfective present forms. However, although this idea may perhaps serve as an argument, it 800 needs an additional investigation on its own. 801 3. Probably the best indication is the following one. The internal evidence is provided by the neuter 802 (= proto-neuter or inanimate) demonstrative form *tód. It is highly likely that the proximative/proximate 803 demonstrative stem *s- (*só-, etc.) could not be used to refer to neuter (= proto-neuter or inanimate) 804 nouns because these referents could only be obviative/obviate. This conforms to a cross-linguistic typo- 805 logical pattern that is found in hierarchical marking systems. It speaks in favor of the inference drawn 806 here. Thus, the fact that the demonstrative stem *s- (*só-, etc.) could not be used to refer to proto- 807 neuter/inanimate nouns can very much strengthen that the stem *tó- was an obviative/obviate demon- 808 strative stem. I can integrate the obviative/obviate function of *-t- into my general hypothesis as follows. 809 The suffixes *-s- and *-t- seem to go back to original P-indexing (not A-indexing) suffixes, and might 810 thus ultimately go back to pre-PIE P-indexing (en)clitics. The abbreviations S, A, P indicate the respec- 811 tive referents that can be compared with the notions of intransitive subject (~ S), transitive subject (~ 812 most agent-like argument A), and transitive object (~ most patient-like argument P). For my model of 813 the PIE case-marking and alignment system see Pooth & Orqueda forthc. 814 (5) *h2nér-s *gwhén-t 815 man-AGT slay:NDUR:ACT:SG-OBVIATIVEP\3 816 ‘man (proximative A) slew obviative P’ a. *h2nér *gwhén-s 818 man.ABS slay:NDUR:ACT:SG-PROXIMATIVEP\3 819 ‘obviative A slew proximative man (P)’ 817 b. 27 *h2nér *gwhén-m 821 man.ABS slay:NDUR:ACT:SG-1A\SG 822 ‘I (A) slew man (P)’ 820 (6) a. *h2nér *bhuéh2-m ~ *bhúh2-m 824 man.ABS grow:NDUR:ACT:SG-1S\SG 825 ‘I (S) grew to be a man, became a man’ b. 823 *h2nér *gwhén-t 827 man.ABS slay:NDUR:ACT:SG-DIRECT\SG 828 ‘proximative A slew man’ 826 c. *h2nér *bhuéh2-s ~ *bhúh2-s 830 man.ABS grow:NDUR:ACT:SG-INVERSE\SG 831 ‘proximative S grew to be a man (causally affected by an “obviative” natural force)’ 829 d. 832 As illustrated by the reconstructed sentences (6ab), PIE person-indexing is usually reconstructed with 833 an active subject-indexing first-person suffix -m- for A = S. The idea that the other two suffixes were 834 object-indexing suffixes implies that there was a kind of “split-indexing” system in PIE grammar. The 835 first-person indexing (-m) was subject-marking (with A = S), as in (6ab), whereas the second and third 836 person indexing was object-marking, as in (5ab), presumably extending its function to the domain of 837 intransitive subject-marking. This would imply that the PIE direct-inverse system was a mixed subject- 838 marking and object-marking hierarchical system. The PIE direct-inverse markers presumably occurred 839 in intransitive clauses as well – either generally or for encoding switched referents, as in (7ab) further 840 below. If true, there might have been a “nominative/accusative” first-person A = S indexing vs. “erga- 841 tive/absolutive” O = S indexing split-system (in a more abstract sense of these typological notions). 842 Since the evidence points to unmarked plural forms, such a split was restricted to singular forms. To 843 round off the typological picture, I can briefly add that it is quite likely that there were two general verb 844 classes: A. agentive including so-called “unergative” verbs, *gwhen- ‘to slay, kill’ vs. B. “ambient force 845 verbs” or so-called “unaccusative” verbs, e.g. *bhueh2-/bhuh2- ‘to grow’. The hypothesis that PIE had 846 these two verb classes is typologically plausible. However, its discussion must be postponed. 847 5. 848 It is cross-linguistic and typological common knowledge that direct forms are in principle and by defi- 849 nition used to indicate that the participant causing a change of state in another participant is either higher 850 on a grammaticalized animacy or topicality hierarchy – or is in TOPIC function. On the other hand, 851 inverse forms are used to indicate that the causing participant is lower on that hierarchy or non-topical Typological implications for the reconstruction of PIE syntax 28 852 (Givón 2001 I: 166, Jacques 2010, Jacques & Antonov 2014). To be fair, languages with direct vs. 853 inverse systems considerably vary in the use of these forms (Thompson 1989, Zúñiga 2006, Jacques 854 2010). Nonetheless, it is a typological implication of the hypothesis outlined in this paper that PIE had 855 a person or animacy and topicality hierarchy – and hierarchically triggered direct-inverse indexing. It is 856 further evident that PIE displayed grammatical core case marking. This fact makes it a bit more different 857 from other systems with hierarchical alignment, since additional case markers were used to distinguish 858 the core syntactic functions and/or semantic roles (S, A, P; agent, patient, etc.). This implies that the PIE 859 topicality hierarchy and direct-inverse marking was not necessarily solely used to distinguish semantic 860 participant roles of third person referents. With third person referents, the direct marking indicated that 861 the causing participant was topical and PROXIMATIVE and that the transitivity direction was from this 862 first referent to a second non-topical or OBVIATIVE one. The inverse transitive marking served to code 863 the opposite direction. The following two reconstructed PIE sentences can illustrate this difference. 864 (7) *gwén a. *h2nér-s 865 man-AGENTIVE woman.ZERO 866 ‘man (PROX) kissed woman (OBV)’ 867 b. *h2nér *Kuás-t kiss:NDUR:AGT:SG-DIR\3 *gwén-s *Kuás-s woman-AGENTIVE kiss:NDUR:AGT:SG-INV\3 868 man.ZERO 869 ‘as for man (PROX), woman (OBV) kissed him’, 870 ‘the man was kissed by the woman’ 871 A typolological comparandum of a language combining ergative vs. absolutive case marking with direct 872 vs. inverse marking is provided by Japhug Rgyalrong (Jacques 2010: 135, his example): 873 (8) a. ʁdɤrʑi kɯ ɬamu pɯ-a-mto ERG Lhamo.ABS AOR-3SG>3-see (DIR) 874 Rdorje 875 ‘as for Rdorje, he saw Lhamo’ b. ɬamu 876 ʁdɤrʑi kɯ 877 Lhamo.ABS 878 ‘as for Lhamo, Rdorje saw him’ pɯ́-wɣ-mto Rdorje ERG AOR-INV-see:3>3SG 879 A similar, but not identical system with direct-inverse distinction is found in Koyukon (Athabascan, 880 Alaska, USA; these are examples (3, 4, 56, 48) of Thompson 1989: 2, 3, 10, 11): 881 (9) a. John yi-nee-ł-’aanh yi-THM-CL(ł)-see21 882 21 Thompson 1989 uses THM for “thematic prefix”, CL for “classifyer” (one of these “classifyers” is identified with middle voice marking). 29 883 884 ‘John is looking at him/her’ b. John bi-nee-ł-’aanh 885 bi-THM-CL(ł)-see 886 887 ‘as for him/her, John is looking at him/her’ c. ts’i-nee-ł-’aanh 888 INDEF/1PL-CL(ł)-see 889 ‘s.o./we/he/she is looking at him/her/it’ 890 d. nee-l-’aanh 891 THM-CL(li)-see 892 ‘s.o. is watching him/her; she/he is being watched’ 893 Cross-linguistically, such direct-inverse systems are often used to encode switch-reference (Thompson 894 1989: 13). It is thus inferable that PIE also made use of it to code switched referents: 895 (10) *h2nér-s 896 *médhu man-AGENTIVE honey.ABS *dh3-t-ó give/take:NDUR-DIR-NDUR:DTR 897 *páh3-t *=kwe 898 swallow:NDUR:AGT:SG-DIR\SG and/with/alike 899 900 901 ‘mani took honey and hei swallowed it’ *médhu (11) *h2nér-s man-AGENTIVE honey.ABS *dáh3-t give/take:NDUR:ACT-DIR 902 *páh3-s *=kwe 903 swallow:NDUR:AGT:SG-INV\SG and/with/alike 904 ‘mani gave honey to someonej and that personj swallowed it’ 905 There were probably other oppositions encoded by -s- vs. -t-. I can briefly add that third person inverse 906 optative-irrealis forms might have been used if the speaker was emotionally involved in the event. 907 (12) a. *h2nér *bhuh2-iéh1-s 908 man.ABS grow-OPT:SG-INV\3 909 ‘I WISH he would grow to be a man, if only he grew to be a man’ *h2nér *bhuh2-iéh1(-t) 911 man.ABS grow-OPT:SG(-DIR/ITR.SUBJ)\3 912 ‘he would grow to be a man (if this or that happened)’ 910 b. 30 913 6. Appendix: A paradigm 914 The following Paradigm Grids 1–3 provide an overview of verb forms of the most basic PIE verbal 915 inflectional type (or “aspectual binyan”, see Pooth 2016, 2017). The PIE verb in question is *gwhen- ‘to 916 slay, kill, beat, chase hunt’ again (> Vedic han-, etc.). I gloss the obviative/obviate third-person as 3’. > 917 means ‘acting on/directed to’ here. For the time being, I use the label “nondurative” for this aorist-like 918 aspectual category (one may use the label “aorist” as well, but not in its Greek definition). 919 Grid 1. The PIE nondurative active-agentive transitive forms >1 >2 >3 >3’ gwhénm 1SG> 1PL.EXCL> gwhnmé(s)A 1PL.INCL> gwhnué(s)A 2SG> gwhéns gwhént 2PL> gwhnsé(n) gwhnté(n) 3SG> gwhéns gwhént 3PL> gwhnérs gwhnént 3’SG> gwhéns 3’PL> gwhnérs Grid 2. The PIE nondurative undirected formsC 920 ACTIVE-AGENTIVE DETRANSITIVE 1SG gwhénm gwhnh2é 1PL.EXCL gwhnmé(s) gwhnmó(s) 1PL.INCL gwhnué(s) gwhnuó(s) 2SG gwhénB gwhnh2é 2PL gwhné(n) gwhnh2é(n) 2COL — gwhnéh2(m) 3/3’SG gwhénB gwhnó 3/3’PL gwhnér gwhnró 3/3’COL — gwhnéh2(m) 921 Additional notes to these Grids: 922 A This suffix *-s was a partially obligatory first-person plural marker and should not be confused with the 923 inverse suffix here. The partially obligatory second-person plural suffix was *-n. The one used in third-person 924 forms was *-r- before zero and *-s-, and *-n- before *-t-. This 3pl suffix looks like the nominal so-called 925 “heteroclitic” suffix *-r- ~ *-n-. 31 926 927 928 B Perhaps the direct/inverse forms were used for more consistent subject reference-tracking – as already mentioned above. If true, the unmarked singular forms were mainly used in the imperative mood. C Question: Should I use an intransitive verb like *gʷem- here? I think that it is not to be excluded that 929 “intransitive” (unmarked/undirected) forms of transitive verbs had a meaning like ‘subject slew (some- 930 one/something)’, whereas the transitive forms had a meaning ‘subject slew referent’. I am grateful to Onno 931 Hovers for pointing me to this matter (personal communication via academia.edu): 932 “That’s how the subjective conjugation works in the Uralic languages that have this (Ugric, Samoyedic, 933 Mordvinic). But the Uralic subjective can also take an object which is then indefinite. So, subjective ‘I 934 slew’ ~ ‘I slew something’ ~ ‘I slew some reindeer’ versus objective ‘I slew it/him/her’ ~ ‘I slew the rein- 935 deer’.” (O. Hovers) 936 Grid 3. The corrssponding PIE nondurative detransitive forms >1 >2 >3 >3’ wh 1SG> g nh2é wh 1PL.EXCL> g nmó(s)A 1PL.INCL> gwhnuó(s)A 1COL.EXCL> gwhnmó(h2) 1COL.INCL> gwhnuó(h2) 2SG> gwhnsh2é gwhnth2é 2PL> gwhnsh2é(n) 2COL> gwhnséh2 gwhnth2é(n) gwhntéh2 3SG> gwhnsó gwhntó 3PL> ? gwhnntó gwhnséh2 gwhntéh2 3COL> 3’SG> gwhnsó 3’PL> ? 3’COL> gwhnséh2 937 Appendix II: I am grateful to Douglas G. Kilday for the following comment (via academia.edu, shortened): 938 “I think the view that PIE deictic-demonstrative *-s- was proximative (not specifically 2nd pers. sg.), and *-t- was 939 obviative (not specif. 3sg.), helps explain the evidence for 3sg. verbal *-s in Germanic as well as Messapic, Phryg- 940 ian, and scattered Vedic injunctives. Old Norse has generally 3sg. pres. -r < *-esi identical to 2sg., with scattered 941 remnants of 3sg. *-ð. The Northumbrian dialect of Old English also has 3sg. pres. -s, against -þ in West Saxon, 942 with 2sg. -st throughout West Gmc. generally explained by epenthesis in collocations with postposed *þū (n.b. 943 however “singes þu” in the Cuckoo Song, ca. 1240). The notion that the simple 2sg. form could have displaced 32 944 the 3sg. in ON and Nthb. OE isn’t plausible. What is plausible, within the framework of PIE prox. *-s- and obv. 945 *-t-, is that Proto-Germanic inherited both as verbal endings, and used them both as 3sg. pres. suffixes, generally 946 *-s- when the verb preceded the subject and *-t- (later *-þ, *-ð) when it followed.” (D. G. Kilday) 947 Acknowledgements go to Christopher Miller for correcting my Genglish/Denglish. 948 References 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 Bauer, B. L. M. 2009: Residues as an aid in Internal Reconstruction, Internal Reconstruction in Indo-European. Methods, Results, and Problems. Ed. by J. E. Rasmussen & T. Olander. Copenhagen, 17-31. Beekes, R. S. P. 1995, second edition 2011: Comparative Indo-European Linguistics. Amsterdam & Philadelphia. Second edition. Revised and corrected by Michiel de Vaan. Amsterdam & Philadelphia. Burrow, T. 1954: The Sanskrit precative. In: J. Schubert (ed.), Asiatica. Festschrift Friedrich Weller, zum 65. Geburtstag gewidmet von seinen Freunden, Kollegen und Schülern. Leipzig, 35-42. ― 1957: An archaic verbal termination in early Indo-Aryan. Indo-Iranian Journal 1, 61-76. Bybee, J., R. Perkins & W. Pagliuca 1994: The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago & London. DeLancey, S. 1981: The category of direction in Tibeto-Burman, Linguistics of the Tibeto-Burman Area 6 (1), 83-101. Dixon, R. M. W. 2010–2012: Basic Linguistc Theory. Volume I–II 2010, Volume III 2012. Oxford. Eichner 1985: Das Problem des Ansatzes eines urindogermanischen Numerus ‘Kollektiv’ (‘Komprehensiv’), in: Grammatische Kategorien, Funktion und Geschichte. Akten der VII. Fachtagung der Indogermanischen Gesellschaft, Berlin 20.-25. Februar 1983. Ed. by B. Schlerath & V. Rittner. Wiesbaden, 134-169. García Ramón, J. L. 1998: Indogermanisch *gu̯ hen- ‘(wiederholt) schlagen, töten’, Mír Curad. Studies in Honor of Calvert Watkins. Ed. by Jay H. Jasanoff, H. Craig Melchert and Lisi Oliver. Innsbruck, 139-54. Givón, T. 2001: Syntax. Volume I & II. Amsterdam. Gotō, T. 1987: Die „I. Präsensklasse“ im Vedischen. Wien. ― 2013: Old-Indo-Aryan morphology and its Indo-Iranian background. In co-operation with Jared S. Klein and Velizar Sadivski. Wien. Groddek, DBH 13 = Groddek, D. 2004: Hethitische Texte in Transkription KUB 20. Dresden: Dresdner Beiträge zur Hethitologie 13. Harðarson, J. A. 1993: Studien zum urindogermanischen Wurzelaorist und dessen Vertretung im Indoiranischen und Griechischen. Innsbruck. Hoffmann, K. 1965: Materialien zum altindischen Verbum, 1. añc ‘schöpfen’, 2. chand, 3. du, 4. path, 5. mreḍ, 6. édhate, 7. VS. dīṣva, 8. AB. purāṇi, 9. 2. Sg. abhinas, 10. JB. liliyur. Zeitschrift für Vergleichende Sprachforschung 79, 171-191 [= 1975: 162-182]. ―1967: Der Injunktiv im Veda. Heidelberg. ― 1967b: Der vedische Prekativtyp yeṣam, jeṣma. Münchner Studien zu Sprachwissenschaft 20, 25-37 [= 1976: 465-474]. ― 1968: Zum Optativ des indogermanischen Wurzelaorists. In: J.C. Heesterman et al. (ed.), Pratidānam. Indian, Iranian and Indo-European Studies Presented to Franciscus Bernardus Jacobus Kuiper on his sixtieth birthday. Den Haag & Paris, 3-8 [= 1975: 245-250]. ― 1970: Das Kategoriensystem des indogermanischen Verbums, Münchener Studien zur Sprachwissenschaft 28, 19-41. ― 1975, 1976: Aufsätze zur Indoiranistik. Hrsg. von J. Narten. Bd. 1, 2. Wiesbaden. Hoffmann, K. & B. Forsman 2004: Avestische Laut- und Flexionslehre. 2., durchgesehene und erweiterte Auflage. Innsbruck. Hoffner, H. A. Jr. & H. C. Melchert 2008: A Grammar of the Hittite Language. Winona Lake. 33 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 Insler, S. 1995: Vedic ajaniṣṭa and related problems. In: Verba et structurae. Festschrift fur Klaus Strunk zum 65. Geburtstag. Hg. von H. Hettrich et al. Innsbruck, 91-103. Jacques, G. 2010: The inverse in Japhug Rgyalrong, Language and Linguistics 11 (1), 127-57. Jacques, G. & A. Antonov 2014: “Direct/inverse systems”, Language and Linguistics Compass 8/7, 301-318. Jamison, S. W. 2009: Where are all the optatives? In: K. Yoshida, B. Vine (eds.), East and West. Papers in Indo-European Studies. Bremen, 27-45. Jasanoff, J. H. 1991: The ablaut of the root aorist optative in Proto-Indo-European, Münchener Studien zur Sprachwissenschaft 52, 101-122. ― 1997: Gathic Avestan cikōitərəš. In: A. Lubotsky (ed.), Sound Law and Analogy. Papers in Honor of Robert S. P. Beekes on the Occasion of his 60th Birthday. Amsterdam & Atlanta, 119-130. ― 2003: Hittite and the Indo-European verb. Oxford: Oxford University Press. ― 2009: Notes on the internal history of the PIE optative. In: K. Yoshida, B. Vine (eds.), East and West. Papers in IndoEuropean Studies. Bremen, 47-68. Klaiman, M. H. 1992: Inverse Languages, Lingua 88, 227-61. Kloekhorst, A. 2007: The Hittite 2pl.-ending -šten(i). In: A. Archi & R. Francia (eds.), VI Congresso Internazionale di Ittitologia, Roma, 5-9 settembre 2005 (= Studi Micenei ed Egeo-Anatolici 49-50), 493-500. ― 2008: Etymological Dictionary of the Hittite Inherited Lexicon. Leiden & Boston. ― 2017: The Old Hittite and the Proto-Indo-European tense-aspect system. Indogermanische Forschungen 122 (1), 295-308, doi: 10.1515/if-2017-0015 Kölligan, D. 2007: Suppletion und Defektivität im griechischen Verbum. Bremen. Kümmel, M. J. 1996: Stativ und Passiaorist im Indoiranischen. Göttingen. ― 1998: Wurzelpräsens neben Wurzelaorist im Indogermanischen, Historische Sprachforschung 111, 191-208. ― 2012: Typology and reconstruction. The consonants and vowels of Proto-Indo-European, The Sound of Indo-European. Phonetics, Phonemics, and Morphophonemics. Ed. by B. Nielsen Whitehead, T. Olander, B. A. Olsen & J. E. Rasmussen. Kopenhagen, 291-329. ― 2012b: Monosyllabic lengthening in Vedic aorists? International Journal of Diachronic Linguistics and Linguistic Reconstruction 9, 51-63. ― 2015+, 2018 (handout, see Kümmel 2018): Anatolisches und indoiranisches Verbum. Erbe und Neuerung: 100 Jahre Entzifferung des Hethitischen. Paper given at Morphosyntaktische Kategorien in Sprachgeschichte und Forschung. Arbeitstagung der Indogermanischen Gesellschaft in Marburg 21.-23. September 2015, downloadable from https://uni-jena.academia.edu/MartinKümmel ― 2018: Anatolisches und indoiranisches Verbum. Erbe und Neuerung. In: 100 Jahre Entzifferung des Hethitischen. Morphosyntaktische Kategorien in Sprachgeschichte und Forschung. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 21. bis 23 September 2015 in Marburg. Herausgegeben von Elisabeth Rieken. Unter Mitwirkung von Ulrich Geupel und Theresa Maria Roth. Wiesbaden, 239ff. ― 2016: Is Ancient Old and Modern New? Fallacies of Attestation and Reconstruction (with Special Focus on Indo-Iranian). In: D.M. Goldstein, S.W. Jamison, & B Vine (eds.). 2016. Proceedings of the 27th Annual UCLA Indo-European Conference. Bremen, 79-96, downloadable from https://uni-jena.academia.edu/MartinKümmel Litscher, R. 2014: Voraussetzungen für ein feminines Genus und Implikationen für das Kategoriesystem des frühindogermanischen Nomens, in Neri & Schuhmann 2014: 137-165. LIV = Rix et al. 2001: Lexikon der indogermanischen Verben. 2. erw. und verbesserte Auflage. Wiesbaden. Lubotsky, A.M. 1997: A R̥ gvedic Word Concordance. 2 Volumes. New Haven, Connecticut. 34 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 ― 2013: The Vedic paradigm for ‘water’, in: Multi Nominis Grammaticus Studies in Classical and Indo-European linguistics in honor of Alan J. Nussbaum on the occasion of his sixty-fifth birthday, ed. by A.I. Cooper, J. Rau & M. Weiss. Ann Arbor: Beech Stave Press, 159-164. Lundquist, J. & A. D. Yates 2018: The Morphology of Proto-Indo-European. In: Comparative Indo-European Linguistics: An International Handbook. Eds. M. Fritz, B. Joseph, J. Klein, see http://www.pies.ucla.edu/resources/ady/papers/IEmorphF.pdf Malzahn, M. 2002: Auf der Spur von ai. veḥ. Indo-Iranian Journal 45/3, 197-204. ― 2004: 3:3, 5:1, or 4:2? On the Ablaut of the Root Aorist in Greek and Indo-European, Historische Sprachforschung 117/1, 50-75 ― 2010: The Tocharian Verbal System. Leiden & Boston. Melchert, H. C. 1998: Traces of PIE Aspectual Contrast in Anatolian, Incontri Linguistici 20, 83-92. ― 2012 (ed.): The Indo-European Verb. Proceedings of the Conference of the Society for Indo-European Studies, Los Angeles 13-16 September 2010. Wiesbaden. ― 2015: The Tocharian s-preterite. In: Tocharian Texts in Context. Ed. by M. Malzahn, M. Peyrot, H. Fellner, T.-S. Illés. Bremen, 127-136. Narten, J. 1964: Die sigmatischen Aoriste im Veda. Wiesbaden. ― 1968: Zum „proterodynamischen“ Wurzelpräsens. In: J.C. Heesterman et al. (ed.), Pratidānam. Indian, Iranian and IndoEuropean Studies Presented to Franciscus Bernardus Jacobus Kuiper on his sixtieth birthday. Den Haag & Paris, 9-19 (= 1995: 97-107) ― 1984: Optativ und Tempusstamm im Altavestischen. Die Sprache 30, 96-108. (= 1995: 270-282). ― 1995: Kleine Schriften. Band I. Hrsg. von Marcos Albino und Matthias Fritz. Wiesbaden. Neri, S. & R. Schuhmann 2014 (eds.): Studies on the Collective and Feminine in Indo-European from a Diachronic and Typological Perspective. Leiden & Boston. Oettinger, N. 2002 (1979): Die Stammbildung des hethitischen Verbums. Nachdruck mit einer kurzen Revision der hethitischen Verbalklassen. Dresden. Pike, M. 2009: The Indo-European long vowel preterites: new Latin evidence, Internal Reconstruction in Indo-European. Ed. by J. E. Rasmussen & T. Olander. Copenhagen, 205-212. Pooth, R. A. 2000: Stativ vs. Medium im Vedischen und Avestischen, Historische Sprachforschung 113, 88-116. — 2004a: Ablaut und autosegmentale Morphologie: Theorie der urindogermanischen Wurzelflexion, Indogermanistik – Germanistik – Linguistik. Ed by. M. Kozianka, R. Lühr & S. Zeilfelder. Hamburg, 401-471. — 2004b: Zur Genese der späturidg. thematischen Konjugation aus frühuridg. Medialformen, Indogermanische Forschungen 109, 31-60. — 2009a: Der urindogermanische Progressiv, Protolanguage and Prehistory. Akten der XII. Fachtagung der Indogermanischen Gesellschaft ... in Krakau. Ed. by R. Lühr & S. Ziegler. Wiesbaden, 381-406. — 2009b: Proto-Indo-European Ablaut and Root Inflection, Internal Reconstruction in Indo-European. Ed. by J. E. Rasmussen & T. Olander. Copenhagen, 229-254. — 2011: Die 2. und 3. Person Dual und das Medium, Indogermanistik und Linguistik im Dialog. Akten der XIII. Fachtagung der Indogermanischen Gesellschaft ... in Salzburg. Hg. von T. Krisch & T. Lindner. Wiesbaden, 473-83. — 2012: Zum Aufkommen transitiver Verben in frühen Vedischen am Beispiel 1r̥ , The Indo-European Verb. Proceedings of the Conference of the Society for Ind-European Studies, Los Angeles 13-16 September 2010. Ed. H. Craig Melchert. Wiesbaden, 267-84. — 2014: Die Diathesen Aktiv vs. Medium und die Verbsemantik im Vedischen der R̥ gveda-Saṃhitā. Proefschrift, PhD dissertation. Defended 2014-10-23. Universiteit Leiden, Leiden University Repository, https://openaccess.leidenuniv.nl — 2014b (ms.): More evidence for Proto-Indo-European transfixes: Two types of “lengthened grades”, ms. version 2014-1205, https://leidenuniv.academia.edu/RolandPooth 35 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 — 2015: Proto-Indo-European Nominal Morphology. Part 1. The Noun, Language Arts 1, issue version 2015-12-23, ms. version 2015-12-23, https://leidenuniv.academia.edu/RolandPooth — 2016: Proto-Indo-European Verb Morphology. Part 1. Inflection, Language Arts 2, issue version 2016-03-11, ms. version 2016-03-11, https://leidenuniv.academia.edu/RolandPooth — 2016b (talk, ms.): “Is the “tēzzi principle” a plausible inference”, The precursors of Proto-Indo-European: the Indo-Hittite and Indo-Uralic hypotheses. Workshop at the Leiden University Centre for Linguistics, 9-11 July 2015 — 2017a: Proto-Indo-European Nominal Morphology. Part 2. Adjectives, Language Arts 5, issue version 2017 Apr 22, author manuscript version 2017 Apr 22, https://leidenuniv.academia.edu/RolandPooth — 2017: Towards a Rigorous Linguistic Reconstruction of Morphs and Categories, Language Arts 6, issue version 2018 May 22, author manuscript version 2018 May 22, https://leidenuniv.academia.edu/RolandPooth — 2018. Linguistic Reconstruction of Inflectional Categories: The Three Original Functions of Indo-European -i-Presents. FIU Indo-European Lingustics Report 1 (ms. version 28 February 2018) — 2018b (ms.): Linguistic Analysis and Comparative Philology: The Case-Marking System of Proto-Indo-European. Ms. version 17 October 2018 — 2018c. A gender issue at ‘hand’: Hittite ki-eš-šar and its prehistory. Research Comment, ms. 27 March 2018 — 2019 (ms.). On the origin of the Vedic subjunctive, pre-published manuscript Pooth, R., P.A. Kerkhof, J. Barddal, L. Kulikov The Origin of Non-Canonical Case Marking of Subjects in Proto-Indo-European: Accusative, Ergative, or Semantic Alignment. Indogermanische Forschungen 124, 245–263. Pooth, R., and V. Orqueda (forthc./ms.): Alignment change and the emergence of the thematic conjugation from Proto-IndoEuropean to Indo-European: A wedding of hypotheses. Submitted to ThPS; ms. available via request. Renou, L. 1952: Grammaire de la langue védique. Lyon/Paris. Rude, N. 1997: On the history of nominal case in Sahaptian, International Journal of American Linguistics 63, 113-43. Thompson, C. L. 1987: An Introduction to Athabascan languages. Yukon-Koyukuk School District, https://uafanlc.alaska.edu/Online/CA975T1987b/CA975T1987b.pdf — 1989: Pronouns and Voice in Koyukon Athapascan: A text-based study, International Journal of American Linguistics 55 (1), 1-24. — 1996: The Na-Dene middle voice: An impersonal source of the D-Element, International Journal of American Linguistics 62 (4), 351-378. Tremblay, X. 2008: Iranian Historical Linguistics in the Twentieth Century – Part Two. Indo-European Studies Bulletin 13/1, 1-51. Watkins, C. 1962: Indo-European Origins of the Celtic Verb I. The Sigmatic Aorist. Dublin. — 1969: Indogermanische Grammatik III/1. Geschichte der indogermanischen Verbalflexion. Heidelberg. Weiss, M. 2009: Outline of the Historical and Comparative Grammar of Latin. Ann Arbor & New York. [I have not read the second edition, see http://www.beechstave.com/weiss.html Willi, A. 2018: Origins of the Greek Verb. Cambridge U.P. Zúñiga, F. 2006: Deixis and Alignment. Inverse Systems in Indigenous Languages of the Americas. Amsterdam & Philadelphia. Thursday, 18 February 2021