Balto-Slavic accentual mobility: an innovation in contact with Balto-Finnic


Some very specific prosodic innovations affected the Balto-Slavic linguistic community, probably at a time when it already showed internal dialectal differences. Whether those innovations were related to archaic remnants stemming from the parent Proto-Indo-European language, and whether that disintegrating community included different dialects, remains an object of active debate.

“Archaic” Balto-Slavic?

The main question about Balto-Slavic is whether this concept represents a single community, or it was rather a continuum formed by two (Baltic and Slavic) or possibly three (East Baltic, West Baltic, Slavic) neighbouring communities, speaking closely related Northern European dialects, which just happened to evolve very close to each other, i.e. in cultures that were closer to each other than they were to Germanic or Balto-Finnic.

In my opinion, their similarities warrant the reconstruction of a single original central-east European community since the dissolution of Bell Beakers, speaking a North-West Indo-European dialect, and most internal differences between Baltic and Slavic may be explained as innovations. The precise identification of a Proto-Balto-Slavic community remains elusive, although the Unetice-Iwno-Mierzanowice triangle remains the best bet, with Trzciniec showing what seems like an Early Slavic-like population reaching up to the East Baltic.

Bell Beaker expansion in eastern Europe and around the Baltic.

The reconstruction of a common Balto-Slavic proto-language is known to range from difficult to impossible, depending on who you ask, not the least because of the differences that are discussed in this post, and which have been the own battlefield created by Balticists and Slavicists for decades. The old tenet that Balto-Slavic had inherited some traits directly from PIE is – in contrast with e.g. the Italo-Celtic concept – surprisingly vivid still today.

Take, for example, these internal differences and supposedly archaic traits:

  • The ruKi rule, where Baltic shows mostly *is, *us, and Slavic shows *, *; or the different output of Satemization in Baltic compared to Slavic (and both compared to Indo-Iranian). Nevertheless, the Satemization trends in Balto-Slavic and Indo-Iranian are usually explained together and taken as a sign of a traditional three-velar system for PIE.
    • If you consider Satemization as a late trend in Balto-Slavic, affecting each dialect in a different way, and thus Balto-Slavic phonetic evolution clearly distinct from the Indo-Iranian trend, rejecting trictectalism, this problem is solved. This would also solve the impossible Indo-Slavonic problem, and the paradox of Balto-Slavic sharing a genetic phylum with Germanic and Italo-Celtic.
    • If you, however, conflate these differences and North-West Indo-European features with an ad hoc explanation of a hypothetic Centum dialect called Temematic, which intends to solve their (in Holzer’s words) unlösbaren inconsistencies, you essentially add a whole new inconsistency without solving their previous ones. For a full rebuttal of Holzer‘s Temematic etymologies, see Matasović (2014).
  • Kortlandt’s reconstruction of a PIE 3rd singular *-e (Baltic from *-et, Slavic from *-eti) and 3rd plural *-o, which would have been replaced independently in other Indo-European dialects (by *-eti, *-onti), is reminiscent of his own reconstruction of laryngeals almost up to the attestation of all Indo-European dialects, including Baltic. If you consider these traits an innovation, this artificially created problem is immediately solved.
  • Genitive plural Pre-Baltic *-ōm vs. Pre-Slavic *-ŏm is another commonly cited example. However, I would place this difference among other similar differences found within other related IE dialects, hence a common phonetic innovation (see e.g. below for the classicist view of unstable obliques).
  • Kortlandt’s reconstruction of oblique cases in *-m-, shared with Germanic, as stemming from a common Middle PIE *-mus (based essentially on Old Lithuanian *-mus and on a non-existent equivalent Anatolian formation), hence different from those in *-bʰ-. While you can argue for infinite more reasonable alternatives, the most often cited one is the ins.-dat. pl. *-bʰ- as a common NWIE innovation based on ins. sg. *bʰi-, while forms in *-m- (including ins. sg.) as a Northern European phonetic innovation. The simplest, most elegant explanation I’ve read to date (I think by Rémy Viredaz) is the similar bilabial change of Giacobo/Giacomo in Italian…

As you can see, some Balto-Slavicists could have written whole books about how their object of study holds the key to solve problems on common Proto-Indo-European paradigms, some of which wouldn’t need solving if they hadn’t been started by Balto-Slavicists themselves…

While all of these “archaic” traits are easily dismissed without further ado (except for some understandable damaged pride among academics), there is one especially pervasive idea among those willing to find the white whale of laryngeal remnants in Indo-European languages (see here for other examples of dubious laryngeal remains).

The prophecy before the battle, Józef Ryszkiewicz, 1890. Or, how to conjure laryngeal remnants in Balto-Slavic.

Accentual development in contact

Whichever position one prefers, the general argument is that the Balto-Slavic accentual system is non-trivial for the classification of both dialects into a common branch. However, that would only be completely true if it were a common innovation, but not so much if it were a natural laryngeal evolution.

In fact, the broken tone preserving a PIE laryngeal, as proposed by Kortlandt – continuing Meillet’s idea of synchronous PIE-PBS developments – was always very difficult to accept. Even the rising pronunciation is not original, and represents a shift of the accent on the initial syllable in Latvian…

In my opinion, the derivation of a modern phenomenon from a PIE laryngeal must always raise a red flag (see below on archaisms vs. innovations in IE languages). As you can see from my take of the fable in Balto-Slavic, which uses Kortlandt’s reconstruction, I preferred not to take into account the reconstructed accents. The fable remains thus a model of what could have been a common Proto-Balto-Slavic, unlike other reconstructions, which are much less tentative.

NOTE. You could argue that accents may be reconstructed in spite of the wrong theory behind them, but this is not true; at least not of all reconstructed accents, some of which require further assumptions. Think about it this way: I wouldn’t take into account a reconstruction of Germanic accent which used Danish glottalized tone for a hypothetical Proto-Germanic laryngeal, even if most accents seemed correct at first sight. The truth is, I didn’t want to dedicate time to go through each reconstructed word and its explanation, so it was easier to delete them all, even though that’s not an actual solution, either. You will find the same doubts in the description of Balto-Slavic evolution in my old Modern Indo-European grammar. The introduction to IE dialects was partially copied from Wikipedia (which, in the case of Balto-Slavic, essentially summarized data from Kortlandt), but in the grammar I just tried to keep the basics, and not very successfully, because you need a comprehensive and coherent description of a language’s evolution. That’s how messed up the question was, and how it still is, even though 15 years of research have passed…

Despite the idea of an “archaic Balto-Slavic”, especially prevalent among older researchers, the current trend is to consider Balto-Slavic prosodic changes as a natural innovation, even among those who would artificially reconstruct laryngeal remnants up to late Balto-Slavic stages.

NOTE. You can read more about the Proto-Indo-European laryngeal loss and vocalism. While the presence of certain laryngeals up to Late PIE is certain, the loss in many environments is also generally agreed upon. This is especially true of a hypothetical Indo-Slavonic branch, like that supported by Kortlandt: even those supporting multiple laryngeal loss events must admit that Indo-Iranian showed no laryngeals before its disintegration, whether they put this loss as an internal Proto-Indo-Iranian evolution, or they place it earlier. Tocharian attests to an evolution similar to the rest of Late PIE dialects (hence to a quite early laryngeal loss trend), and Balkan dialects (supposedly splitting before Indo-Slavonic) also lost laryngeals in a similar way, except for initial ones, which show vocalic output instead of full loss.

So, where does a laryngeal loss fit in this “Indo-Slavonic” scheme, exactly? Before the Tocharian split? Before the Balkan split? After the Balkan split but before the full loss in Indo-Iranian? And where exactly does this group belong regarding Corded Ware, and where does Germanic? No idea (but you can read Kortlandt try fitting his model with Gimbutas’ “Kurgan peoples”). Because one thing is to reconstruct Proto-Greek, or Proto-Celtic, or Proto-Italic forms without laryngeals and to put them in relation with a purely theoretical three-laryngeal PIE, and a different one is to reconstruct laryngeals (including in environments which were already lost in Tocharian) up to Proto-Baltic and Proto-Slavic, which seems more than just a bit of a stretch…

Indo-European dialectal relationships, from Mallory and Adams (2006).

Thomas Olander offered a summary of the current positions regarding the Balto-Slavic accentual system recently in Indo-European heritage in the Balto-Slavic accentuation system (2013), which also contains a summary of his Mobility Law, to explain this phenomenon as a common Pre-Baltic and Pre-Slavic innovation.

Andersen, an advocate of different Baltic and Slavic dialects developing in contact with Satem dialects, suggested in The Satem Languages of the Indo-European Northwest. First Contacts? (2009), partially based on Olander’s initial proposal, that Baltic and Slavic accentual mobility arose as a result of contact with languages with fixed word-initial ictus: the accent was lost in the word-final mora in pre-Proto-Baltic and, independently, in pre-Proto-Slavic. Hence, the central innovation, the accent loss

technically is not a shared Slavic and Baltic innovation. On the contrary. It shows that the speakers of the Pre-Slavic and Pre-Baltic dialects formed bilingual communities with speakers of contact dialects that were of the same prosodic type, viz. had fixed initial ictus but no free accent.

In the meantime, Olander (2019) has found out about more real-world examples of this same phenomenon:

Prosodic features are known to be susceptible to contact influence (Salmons 1992:1 and passim). While it does not directly influence the evaluation of the Mobility Law as a non-trivial innovation, it is interesting that most of the alleged parallels are indeed considered to be contact-induced changes due to influence from languages with an ictus on the word-initial syllable (Andersen 2009: 11-14; Rinkevičius 2013): Balto-Fennic in the case of the Karelian and (perhaps through Latvian as an intermediary) Žemaitian dialects, and Hungarian in the case of the Slavonian dialects (for Karelian see Jakobson 1938/2002: 239; Veenker 1967: 74; Thomason & Kaufman 1988: 122, 241; Salmons 1992: 41- 42; for Žemaitian see Zinkevičius 1966: 45- 46; for Slavonian see Ivić 1958: 287).

I am not aware of any hypotheses on a contact-induced origin for Greek prosodic innovations, but it is at least worth noting that there is agreement on significant substrate influence on Greek. While we may speculate that these substrate language(s) had word-initial ictus like Balto-Fennic and Hungarian, we do not have any actual information about the prosodic system(s) (thus even Beekes 2014: 9, who in other respects provides a fairly detailed picture of the substrate).

The parallels from other speech varieties show that an accent loss of the type suggested for a pre-stage of Baltic and Slavic is a type of prosodic change that has occurred several times in different various systems. In the context of the present paper this means that the sound law itself cannot be classified as a non-trivial innovation; it may have taken place in already differentiated dialects or languages. Also, the parallels suggest that a loss of the accent may be the result of influence from languages with fixed word-initial ictus.

In this time when even linguists agree that substrate/contact languages have to be related to specific ethnolinguistic groups (see here for Germanic), the fact that Olander stops short of naming this substrate behind Pre-Baltic and Pre-Slavic as being Late Uralic in general, or Balto-Finnic in particular, is surprising.

NOTE. Not the least because Olander is part of the Homeland Timeline map project of the Copenhagen group (their website is not working right now), and they placed Volosovo as Uralians expanding with Netted Ware in contact with the Baltic during the Bronze Age…So what’s to doubt about Balto-Slavic – Balto-Finnic contacts, exactly? Maybe if Balto-Finnic was the substrate language behind Balto-Slavic (as it was in Germanic), it would mean that Uralic languages were previously spoken in territories that became later Germanic- and Balto-Slavic-speaking?

Still image from the Copenhagen Timeline Map (accessed one year ago), showing in green Volosovo hunter-gatherers who, according to the map, later expand to the north-east with Netted Ware…

Archaism vs. Innovation

If we tried to describe these trends of explaining peculiar traits in recent Indo-European dialects as archaism vs. innovation from a purely theoretical point of view, we could roughly distinguish two different positions (with infinite variants, of course) among academics – just like we could find people more inclined to leftist or rightist trends when speaking about economy. When it comes to linguistics, which is the least messed-up field where one can describe Indo-European and Indo-Europeans, I think we can find two alternative basic tenets:

  • One idea would hold that the oldest attested dialects – and those with an older guesstimated proto-language – are the gold standard as to what the original situation may have been, and about what could be described as an archaism. For example, Ancient Greek and Mycenaean or Vedic Sanskrit for old dialects; Tocharian, or Italic dialects for those with quite old guesstimates, each for different reasons; and Anatolian for both, old dialect and attested early.
  • NOTE. Nevertheless, the phonology of Anatolian inscriptions is often difficult to ascertain, and its ancient dialectal nature stemming from a Middle PIE stage may still be disputed by some. The archaic nature of Tocharian seems to be maybe less generally accepted than that of Anatolian, but I would say there is general consensus on the matter today.

  • The other general idea would support that the most isolated dialects are those which may hold the key to the oldest Indo-European traits, somehow hidden from external influences and areal contacts, and thus from generalized innovative trends that have affected the best known ancient dialects. In that sense, languages like Slavic, Baltic, Albanian, or Armenian – as well as some Balkan fragmentary dialects – are quite common aims of study to reveal exceptional PIE traits.

I think the education system in Southern Europe and South Asia is that of formal classicists. In eastern Europe, I’d reckon the education system – especially in regions that were never connected to the Graeco-Roman tradition – favours linguistics as a study of the own and related proto-languages. For northern Europe, I would say it’s 50/50, especially in Scandinavia, depending on whether classicists or linguists dominate over the departments of Indo-European. For example, while Germany or Austria would maybe lean more toward the classics, Copenhagen’s obsession with Germanic as the most archaic IE branch is well known…

A 17th-century birch bark manuscript of Pāṇini’s grammar treatise from Kashmir. Image from Wikipedia.

Both positions, when blindly accepted, are bound to fail at some point or another:

  • If you take Classical Sanskrit, Classical Greek, or Classical Latin as an example of Proto-Indo-European, you are bound to make radical mistakes when reconstructing the parent language, more so if you disregard the oldest attested layers of the languages. An interesting view of the so-called Adradists at the Complutense University of Madrid – apart from their famous 9-laryngeal reconstruction – is that Middle PIE had only 5 cases, with a general (unstable) oblique one in Late PIE that later evolved into the attested 5 to 8 cases in the different dialects. That is, in my opinion, a fairly typical classicist error, which would be easily addressed by taking into account the oldest stages, like those attested in Mycenaean and in Old Latin, instead of focusing on classical grammar. The 8-case system is, in fact, one of the few true Balto-Slavic archaisms, supported by external comparanda.
  • On the other hand, if you take Albanian, Armenian, Baltic or Slavic, or even phonetically dubious data like those from some Anatolian inscriptions, you can eventually argue for anything. And I really mean anything; you are leaving the logic door wide open for any crazy-ass opinion about Proto-Indo-European based on traits found in modern languages: From how many velars evolved (if at all, because you may find all of them in Luwian, or still living in Albanian or in Armenian…) and their nature as ejective consonants in Late PIE (based on Armenian or Germanic); to how many laryngeals and when these laryngeals disappeared (if they actually did disappear, because some may even find them in Modern Lithuanian, in Armenian, or in Danish…); etc. Once you believe your own romantic view of some modern language(s) retaining traits from five thousand years ago, there is no stopping that; not for you, but not for anyone else, either.

NOTE. One of the funniest consequences of this type of ‘worldview’, where one assumes that – the own interpretations of – modern dialects are as reliable (or even more so than) ancient ones, and that Indo-European dialects somehow split at the same time from the parent language (so there was one common “full laryngeal” language, and then all attested dialects evolved from it) are some of the theories that you can easily find posted on Facebook’s group on Proto-Indo-European. Let’s just say, for the sake of simplicity, that you can compare English ‘sunrise’ with Spanish ‘sonrisa’ “smile” all you want, and assert that both reveal a common origin in PIE *sup- hence from the Sun and the smile going “up” or something, but any explanation as to how you reached that conclusion doesn’t make for the why this comparison shouldn’t have even started at all. Now replace English and Spanish with Armenian, Slavic, and/or Albanian, invent some new IE sound law, throw one or two laryngeals in the mix, and somehow this might get a pass among certain linguists…

The Celebration of Svetovid on Rügen, Alphonse Mucha, The Slav Epic. Image from Wikipedia. Were Early Slavs some among a selected few romantic peoples to keep the “true” Indo-European language and traditions? Of course not.

While no one can deny the value of different Indo-European branches for the reconstruction of the parent language, no matter how recently they were attested, the only reasonable solution whenever a difficult case arises is to trust ancient dialects more than recent ones. Using data from fringe theories based on recent dialects to build a Proto-Indo-European paradigm, especially when there is contradictory data from ancient IE dialects, is flawed for two reasons:

  1. Languages attested later – especially after periods of population movements and contacts – would show, in general, a greater degree of change. Preferring Old Slavic or Classical Armenian to reconstruct Indo-European over ancient dialects like Ancient Greek, Vedic Sanskrit, or ancient Italic dialects is, in a way, like taking Byzantine Greek, Pali, or Old French as models, respectively.
  2. Classical languages are indeed modified due to the action of grammarians, but once standardized these “languages behind a state” (or religion) are less prone to change, due to the transmission of oral (and written) literature, education, commerce, etc. Languages left to unorganized tribes are less constrained in their evolution, and their internal (substrate) and external (contact) influences are greater and (what’s worse) unknown.

Baltic and Slavic, like Albanian or Armenian, are dialects attested very recently, which may have undergone complex internal and external influences we may never fully understand. Confronted with controversial or inexplicable traits compared to ancient branches like Greek, Indo-Iranian, or Italo-Celtic (especially if they fit with other Indo-European dialects), the conservative solution that will be right most of the time (and I mean 99.9999% of cases) is to assume they represent an innovation over Late PIE.

The fact that some researchers still use these recent dialects as a blank canvas instead, in order to propose unending new ideas about how to reconstruct IE proto-languages, or even older common PIE stages, is shocking. Not “R1a/Steppe” vs. “N1c/Siberian” haplogroup+ancestry bullshit-level shocking, but still unacceptable in a serious academic environment.

The only reason why Balto-Slavicists have failed so many times in this “unsolvable” question that seems to be Proto-Balto-Slavic reconstruction, apart from the known differences between Baltic and Slavic, is precisely the fixation of many with their object of study as a model for other IE languages (and thus for PIE), instead of taking the rest as a model for the reconstruction of Balto-Slavic (or of Proto-Baltic and Proto-Slavic).

Repeating ad nauseam the popular concept of Balto-Slavic (or Baltic and Slavic) being among the most archaic IE dialects, or the slowest evolving IE dialects, and cheap nationalist slogans of the sort, does not help this aim, and just reading or hearing that should make anyone cringe instantly. Not less than reading or hearing about Sanskrit being essentially equal to PIE, or spoken in the Indus Valley 10,000 years ago. Because we are not living in the 19th century, mind you.


Improving environmental conditions favoured higher local population density, which favoured domestication


New paper (behind paywall) Hindcasting global population densities reveals forces enabling the origin of agriculture, by Kavanagh et al., Nature Human Behaviour (2018)

Abstract (emphasis mine):

The development and spread of agriculture changed fundamental characteristics of human societies1,2,3. However, the degree to which environmental and social conditions enabled the origins of agriculture remains contested4,5,6. We test three hypothesized links between the environment, population density and the origins of plant and animal domestication, a prerequisite for agriculture: (1) domestication arose as environmental conditions improved and population densities increased7 (surplus hypothesis); (2) populations needed domestication to overcome deteriorating environmental conditions (necessity hypothesis)8,9; (3) factors promoting domestication were distinct in each location10 (regional uniqueness hypothesis). We overcome previous data limitations with a statistical model, in which environmental, geographic and cultural variables capture 77% of the variation in population density among 220 foraging societies worldwide. We use this model to hindcast potential population densities across the globe from 21,000 to 4,000 years before present. Despite the timing of domestication varying by thousands of years, we show that improving environmental conditions favoured higher local population densities during periods when domestication arose in every known agricultural origin centre. Our results uncover a common, global factor that facilitated one of humanity’s most significant innovations and demonstrate that modelling ancestral demographic changes can illuminate major events deep in human history.

Path diagram for piecewise-SEM exploring the effects of environmental and cultural variables on population densities of foraging societies. Measured variables are represented by the large boxes and R2 GLMM values (see Methods) are provided for response variables. n = 220. Red arrows depict negative relationships among variables, black arrows positive relationships, and dashed grey arrows depict non-significant paths (P ≥ 0.05). Standardized coefficients are presented for all paths (small boxes) and arrow widths are scaled to reflect the magnitude of path coefficients.

Interesting excerpts:

(…) our results are consistent with the surplus hypothesis, which suggests that improving environmental conditions and the potential for increased population density may have facilitated the domestication of plants and animals in agricultural origin centres4,7 (Fig. 3). Several factors may explain the links between environmental conditions, potential population density and the origin of domestication. For one, rates of innovation may scale positively with the number of potential innovators13,14. In turn, the likelihood of domestication innovations may have increased in environments that could support increasingly higher densities of foraging people.

In addition, foraging societies may have become more sedentary to take advantage of locally abundant resources, some of which were later domesticated35. Our results indicate that residential mobility scales negatively with population density in foraging societies (Fig. 1). Therefore, increasingly sedentary lifestyles may have contributed further to increases in population density and the potential for innovation. Increases in the productivity of wild progenitors of important domesticates may have also facilitated growing population densities and the viability of cultivation for food production15,16.

Predictions of potential population density for foragers. a–c, Predicted population densities at 4,000 (a), 10,000 (b) and 21,000 (c) YBP. Blue hues depict potential population densities below the median population density of observed foraging societies, and red hues depict potential population densities above the median. The second red hue and above are greater than the mean population density of observed foraging societies. Note the increase in area, through time, with potential population densities greater than the mean of observed foraging societies (number of 0.5° × 0.5° cells: 21,000 YBP = 3,027; 4,000 YBP = 4,673). For example, a notable increase in the number of red cells in the Sudanic savannah and Ganges of East India (Northeast India) between panels c and a.

It is also possible that improving environmental conditions may have resulted in a situation where necessity drove the origins of domestication. For example, population densities may have increased in foraging societies that occupied productive, coastal areas, causing an outflow of groups into regions with less ideal conditions where the cultivation of plants and animals was required to secure adequate food resources6,17,18. Our results cannot support, or refute, the possible influence the outflow of people from hospitable locations to less ideal environments may have played. A detailed understanding of the movements of ancient populations is required for more rigorous testing of the role that forced habitation of marginal environments may have played in the origins of domestication at particular sites.

See also:

Enigmatic *-nt-Stems : an investigation of the secondary -t- of the Greek neuter nouns in *-men- and *-r/n-


Interesting Master thesis Enigmatic *-nt-Stems: an investigation of the secondary -t- of the Greek neuter nouns in *-men- and *-r/n-, by Stephanie Stringer, Université de Montréal (2018).


This paper aims to provide an explanation of the secondary -t- found in the oblique stem of ancient Greek neuters such as πρᾶγμα, πράγματος and ἧπαρ, ἥπατος. After a brief overview of the Greek data, and a survey of the relevant nominal classes in Greek and Indo- European, previous hypotheses are evaluated. To this end, several problems of nominal morphology are discussed, including the existence of a PIE suffix *-m(e)ntom, the secondary -t-s of certain animate nouns, the ablatival suffix *-tos, the Hittite ergative; and the ablaut of neuter active participles. Certain phonological issues are also addressed. Since the majority of hypotheses formulated to explain the secondary -nt- inflection of Greek neuters date from the nineteenth century, attempts are made to re-evaluate their conclusions in the light of more recent research, particularly that related to ablaut classes. Also considered are a number of twenty-first century works which purport to explain the Greek data as part of a larger Indo-European phenomenon.

This paper makes no attempt, however, to explain the PIE origins of either the *r/n-, or of the *nt- stems. It concludes that the best explanation of the Greek declensional pattern is to be found in the analogy between stems in -nt- and those in *-mn- or *-r/n-.

Interesting excerpts, from the conclusion (emphasis mine):

In comparison with other proposed solutions, there are relatively few objections to be levelled at Schmidt’s theory, in this slightly modified form. Anghelina (2010) criticises it on the ground that it does not provide an explanation of how the -t- came to be inserted into the r/n-stems, but this criticism has already been addressed. Anghelina also objects to the idea that participles might affect the declension of nouns. Given that participles can function syntactically as nouns, and that their declension is formally identical, except for the distinction of gender, it is difficult to see why the inflection of one might not affect the inflection of the other. Furthermore, it appears to have done so within the history of Greek. In addition to the neuters, a number of masculine n-stems are inflected as nt-stems, although related formations within Greek attest to the secondary nature of the -t-, e.g. δράκων, -οντος, but δράκαινα, λέων, λέοντος, but λέαινα etc. Perhaps Anghelina would prefer to explain these cases also as
developments from the dative plural, before the ablaut was levelled, but in general, the influence of participles is accepted as an explanation. One could also point to the influence of the pronominal declension on the endings of thematic stems in PIE.

Sihler (2008, 297) argued that if one accepted the nt-stems as a model it was “hard to progress beyond a vague likelihood” and “the supposed model paradigm has been everywhere replaced.” The first of these criticisms is valid, in a sense. One cannot conclusively demonstrate that the nt-stems served as the model for the men- and r/n-stems. Only that the model was available, and that the outcome of the change conforms with it. However, it does not seem that Sihler’s caution is more pertinent in the case of this theory than in any other proposed explanation of morphological change.

Silhler’s second criticism, is true as well. The starting point, a NA sg. nt. -n̥, G sg. -n̥t-os is indeed only preserved, at best, in a few relics. All the same, it can be assumed with some confidence to have existed at the right time. Furthermore, the analogy is unobjectionable. The nt. NA sg. ending -n̥ could genuinely belong to an nt-stem as well as to an n-stem. It is no surprise that the neuters were systematically replaced, while the masculines and feminines showed only sporadic transition to the nt-declension. In the neuter both the NA were liable to re-interpretation as a t-stem, while in the animate forms only the nominative was. The m(e)n-stems are a highly uniform group, and it is easy to understand how a change could spread relatively quickly. The connection to the r/n-stems is slightly more tenuous, but they do have more in common with the m(e)n-stems than with any other group. (A m(e)r/m(e)n- suffix does exist, but given that it is quite rare, and given that its only two representatives in early Greek, τέκμαρ and τέκμωρ are attested only in the NA sg., it is hard to see that this subclass can have played a significant role.)

The nt-stem theory provides an adequate explanation of the Greek situation. That was indeed the very limited aim of this paper. In very general terms, it may also provide an explanation for some of the “stray” t’s one finds attached at times to n-stems in PIE or other languages. Given the co-existence, whatever their origin, of both -nt- and -n-, and in fact t-stems, and given that both -n- and -t- were under certain conditions liable to be lost or assimilated to surrounding sounds, one might expect to find a certain degree of erratic fluctuation between the two classes. Such an observation is so vague as to be quite unhelpful, but at least it is not contradicted by known facts.

In opting for a solution that seems to account for the facts in Greek, one is forced to leave many other phenomena unexplained. Although it would be more satisfying if one were able to draw together the -t- of the NA *-r/n- in Sanskrit and the -t- of the nearly synonymous suffixes -man-, -vant-, man, mant, vasanta, gimmant- etc., it seems at present they can only be connected if one ignores many of the details of each specific situation. For the time being, it appears they must be dismissed as similar, but essentially unrelated, or at least only very indirectly related phenomena. It is entirely possible that further research will reverse this conclusion.

See also: