sign in sign up
<<
Home , Finnish language, Mass comparison, Mordvinic languages

Peter S. Piispanen Stockholm University

Statistical Dating of Finno- through and Sound Laws

Peter Sauli Piispanen

ABSTRACT multi-language actually being genetically affiliated cognates and not only Through comparison of Swadesh-200 word list separate inventions or borrowings. 1 cognates and the employment of Traditional lexical comparisons of this kind are , accompanied by detailed done only through classification, i.e. sound changes, the branching of multilateral lexical comparison. On a deeper some have been statistically second level, only some sound laws have been determined. Assuming linear branching from a proposed for some cognates. Clearly, the line originating in Proto-Uralic and leading to addition of sound laws to the studies would modern Finnish, Moksha (Mordvinic) deepen them and connect them to the originated from 3423 BP, Northern Saami linguistic mainstream. Thus, in this report (Finno-Saamic) from 3038 BP and Estonian great care has been taken to find acceptable (Balto-Finnic) from 1058 BP. The resulting sound laws for cognates. cognacy rates with Finnish (35.6 %, 40.0 % and 72.7 % respectively) and acquired dates are 1.2. No one has ever directly attempted the well in accordance with previous estimates as dating of the Uralic languages through acquired by other methods. lexicostatistics while also taking into account sound laws to the best of my knowledge. This Keywords: Dating, Estonian, Moksha, study therefore aims to fill in some dating Northern Saami, , Swadesh, gaps by proposing statistical results from Comparative Linguistics comparative linguistics and also from a mainstream linguistic standpoint by employing Running short title: Dating Finno-Mordvinic by sound laws. This study also serves as a test of statistics & sound laws the lexicostatistical method per se as employed on the Uralic languages. 2 1. Introduction Interestingly, Sammallahti studied the 1.1. The linguistic analysis of lexical items is relationships of the Saamic languages to each often employed to find and trace the other by employing the Swadesh word list development of genetically related languages. (Sammallahti, P. 1998), while Janhunen That two languages actually have cognates, addressed the problems of analyzing i.e. words originating from the same etymon languages and time depth, including for the in their common proto-language, can be 1 statistically proven by finding the same For example, one particularly important argument is that, statistically speaking, basic words, of similar phonologic form cognates in multiple languages. While a and often identical semantics, can so often be identified with such a high number of common lexical roots between some of majority of scholars have criticized lexical the prospective genetically related languages that they simply methods (cf. Campbell, L. cannot be mere look-alikes or accidents; instead they must represent either a valid distant genetic language relationship or, 1986:488, 2001:45 & 2004:348), others, like at the very least, have extensively borrowed lexicon from the for example Bengtson and Ruhlen (1994) have same sources. 2 Merlijn De Smit, Jenny Larsson and two anonymous reviewers presented some very interesting and thought- are gratefully acknowledged for their valuable input on the provoking statistical arguments in favor of manuscript during preparation. This paper is a much updated and detailed version of old work performed during the author’s Master’s thesis in Finnish.

Fenno-Ugrica Suecana Nova Series • 15 (2016) • 1-58 • © Piispanen, P.S., 2016

Peter S. Piispanen

Uralic languages, elsewhere (Janhunen, . methods is rooted in its usage as a 2008:223-239). replacement for normal comparative work, and, in particular, its application to possible 1.3. The lexicostatistical method was first language families where no comparative work described by Morris Swadesh (cf. Swadesh, M. has yet been done (for example Pama- 1950, 1952 & 1955), and later significantly Nyungan, as studied by Dixon). Deployment of improved, among others, by Starostin, S. the methods used herein into Uralic and Indo- (2000). The lexicostatistical method remains European studies can be argued to be controversial and discredited in some circles “control” experiments. The question should (see for example: Dixon, . 1997 & Renfrew, C. be: do we get sensible results and what does et al. 2000). However, improvements (cf. this imply for the validity of lexicostatistical Starostin, S. 2000) and quite promising results results in general? (for example: Indo-European: Gray, R.D. & Atkinson, Q.D. 2003, Hamito-Semitic: Fleming, 1.4. More recently, different statistical tools .C. 1973, Semitic & Afro-Asiatic: Militarev, A., and methods have been employed and Austronesian: Sirk, . and also within Chinese presented for dating the separation of languages and Native American languages) – languages. A very recent example pertaining which are well in accordance with the results to the Uralic languages includes Syrjänen, K. et from both archeology and genetics – may al . 2013 & Honkola, T., et al . 2013 (using have increased the viability and accuracy of Bayesian phylogenetic analysis). An example the method to a level where it is now ready as from the dating of Indo-European languages is a proper tool for dating. 3 This should be found in Bouckaert, et al . 2012, while the particularly true if the method were also Uralic and Indo-European languages were accompanied by the study of sound laws of statistically compared to each other in terms the compared languages. 4 As a matter of fact, of the Swadesh word list in Kassian, A., et al . a lot of the criticism towards lexicostatistical 2015 (also using Bayesian phylogenetic analysis). In particular, borrowings – and how

3 In all fairness it has to be mentioned that while the method they are handled in the analysis – often cause appears to work with “average” languages, certain languages huge discrepancies in the results. Such results, are uniquely conservative and cannot at this point be dated using traditional lexicostatistics. There are known issues when while often lauded by news media, often have comparing the time-depth of Indo-European languages (IE) problems of their own, one of which is the acquired through lexico-statistical methods to the results from mainstream comparative IE studies; some accept the Kurgan arrival at very much older datings for various hypothesis while others instead accept the Anatolian hypothesis, and the debate is still ongoing. A noteworthy – but proto-language stages than what is usually far from uncontroversial – example of ultra-conservatism is accepted by mainstream linguistics. In the Kusunda, a language in Nepal, which appears to have a rather clear genetic language relationship, to Indo-Pacific languages dating of Indo-European languages, for (cf. Whitehouse, P. , Usher, T., Ruhlen, M. and Wang, .S-Y. example, different methods and results are 2004) that would go back perhaps 40-50 000 years (!)(cf. Bowler, J.M., Johnston, H., Olley, J.M., Prescott, J.R., , often presented, agreeing either with the R.., Shawcross, W. & Spooner, N.A. 2003) – according to Anatolian homeland hypothesis or the Kurgan lexicostatistical methods all lexical traces should have completely vanished at this point, and yet the relationship hypothesis, while for example the work of seems to remain strangely and unexpectedly clear. I postulate that other exceptions, of “too rapid” change, may be expected Syrjänen’s team arrived at higher time depths from languages that have experienced extremely frequent for various proto-languages than what is language contacts, social upheaval, (artificial) scholastic reforms and spread to groups of speakers of completely different currently accepted by most Uralists. With such original languages, and which may include English, Bulgarian, techniques, systematic sound laws are rarely Albanian, and Norwegian Bokmål. 4 This second stage, that is the changes by sound laws, consists employed for the determination of true of the traditional comparative historical method (and could here also be called Modified Etymostatistics or, perhaps, cognates, thus differentiating them from Phono-). 2

Dating Finno-Mordvinic by statistics & sound laws

borrowings, the difference obviously being 1.6. In this study, Finnish was chosen as the very important for the outcome of the results. token language to which to compare the To this author, the determination of actual others. The compared languages were chosen cognates seemingly often boils down to too specifically to have separated in a tree-like much guesswork in the absence of clearly model at different time depths – Estonian, defined sound correspondences, with the Northern Saami and Moksha – and these were resulting lack of proof leading to unacceptably thus limited to before the Finno-Volgaic high margins of error in many studies. While period.6 Sound law analyses complemented sound laws are indeed an after-construction the lexical comparison of the Swadesh-200 list derived from lexical comparisons 5 such laws for these languages in order to find the exact should regularly be employed in finding and semantically unchanged cognates between comparing additional cognates. I note that a them. This places the results within a more specific problem with this kind of analysis is broadly accepted linguistic methodology (as the presence of synonyms, one of which may has often been suggested for lexicostatistics be vernacular; e.g. Fin. missä and kussa (see but has actually not been carried out in detail Appendix). In order to solve this and eliminate as far as this author can ascertain). Naturally, internal diversity as much as possible, I have the sound changes are more numerous the explicitly based the comparison on literary, higher the time depth. The obtained root standard Finnish. glottochronologic results were then, if possible, compared to the dating results 1.5. Hence, the study performed in this paper obtained through paleontology, archaeology takes great care, as far as is reasonable for this and genetics. large a data set, in formulating all cognates in terms only of what is provable both by regular 1.7 It must be noted that with this and irregular, but often observable, sound methodology and these results I do not claim changes. The modern starting point in the that all languages change at a constant rate; dating is set to the year 2010 when most of rather, logically, there are periods of faster the lexical data was collected. Any lexical, change and of slower change, which orthographical or semantic errors are mine eventually do result in an average speed of since I am unfortunately only a fluent speaker change, and which is what the proposed of Finnish and not of Estonian, North Saami or retention factor of lexical change should Moksha Mordvin. Indeed, some lexical items represent. I imagine that the socio-linguistic are very difficult to analyze regarding cognacy conditions for language change are very according to the given lexicostatistical rules different for hunter-gatherers and and principles. Some new possible cognate farmers/city dwellers. R.M.W. Dixon’s suggestions between the Uralic languages will punctuated equilibrium hypothesis – referring also be given in the appendix and some to his large body of work, particularly 1997, possibly hitherto non-discussed sound correspondences are given in the tabulated 6 I generally regard Proto-Finno-Volgaic merely as a lexical layer. This limited time depth was chosen to avoid any potential sound changes of Moksha. problems with the proposed alternative bush (cf. Häkkinen, K. 1984), linear comb and rake models (cf. Salminen, T. 2002) of the Uralic group. It also eliminates any possible 5 That is to say they are in the heuristic phase of the work: we problems that might arise from a case where there never has find “similar” words and construct sound changes, but these been a Proto-Uralic language, but merely a large Lingua Franca sound changes are then used “predictably” to find cognates not (cf. Wiik, K. 2002). On the other hand, others have instead at all similar, e.g. Finnish hiiri ‘mouse’ and Hungarian egér divided Proto-Uralic into Western and Eastern Uralic (cf. ‘mouse’. Itkonen, E. 1966; Häkkinen, J. 2009), an agreeable take. 3

Peter S. Piispanen

2002 – may support this hypothesis as it 1953 BCE BCE BCE BCE BCE Décsy 1965 4000 2500 1500 400 1 CE 1000 suggests that stationary hunter-gatherer BCE BCE BCE BCE CE groups will experience very slow grammatical Hajdú 4000 2000 1500 500 1 CE 1975 BCE BCE BCE BCE convergence often with lexical retention (e.g. Korhonen 4000 3000 2000 1500 1000 1 CE Australia, New Guinea) whereas expansive 1981 BCE BCE BCE BCE BCE Anttila 7000 5000 3000 1500 1250 farmer/pastoralist groups will show a typical 1989 BCE BCE BCE BCE BCE “language tree” with substratum effects from Taageperä 4000 2100 1500 1400 1000 1994 BCE BCE BCE BCE BCE 7 displaced languages. The patterns and speed Kallio 3650 1000 1 CE 8 of language may no doubt change depending 2006 BCE BCE Janhunen 3000 2500 1500 1000 500 1 CE on social structure as well as geographic 2009 9 BCE BCE BCE BCE BCE Honkola et 5300 3900 3700 3200 1200 800 surroundings (for example, compare the al. 2013 10 BCE BCE BCE? BCE BCE? BCE? Eurasian plainlands vs. the New Guinean Abbreviations: PU (Proto -Uralic), PFU (Proto -Finno -Ugric), PFP (Proto-Finno-Permic), PFV (Proto-Finno-Volgaic), EPF (Earlier jungle and mountains; these factors has been Proto-Finnic or Proto-Finno-Saamic), LPF (Later Proto-Finnic), discussed by Nichols, J. 1992, passim ). These BCE (Before Common Era), CE (Common Era) factors probably mean different rates of language change for the language groups The table well shows the contrasts between studied here with the exception of the various estimates and methods. 11 Clearly, the nomadic Saami population; the languages dating of the various proto-languages is studied in this work should mostly relate to difficult for several reasons. The perhaps settled populations. As such, the majority of currently accepted dating of the Proto-Uralic results by the methodology used here may language places it at around 3000 - 4000 BCE. represent an approximate statistical dating result of when the studied proto-language was 2.2. A modified version of Janhunen’s (2009) spoken somewhere. previously published Uralic time depth table is presented below. It gives the representative 2. Traditional dating of the Uralic languages with the number of speakers language tree 8 The datings of Kallio are mainly based on archeological/paleolinguistic findings. 2.1. Throughout the years a number of datings 9 Janhunen notes that the south-to-north dimension of the and places of origin of the Uralic languages Uralic language belt has a chronological depth of less than two millennia, but the geographic length of the east-to-west chain employing several different methods have and the Mesolithic cultural level reflected by Proto-Uralic been presented. A select few of these are suggest a very early dating for the language family as a whole. The first split must have taken place before any contact with the presented in the table below (summarized Indo-Europeans due to the lack of loanwords in the earliest branches. Proto-Uralic itself is likely at a level earlier than the from: Anttila, R. 1989:301, Kallio, P. 2006, earliest stages of Proto-Indo-European (namely Indo-Hittite). It Janhunen, J. 2009 & Honkola, T. et al . is noted that Janhunen’s dating estimates appear to be based under the assumption that a fully formed protolanguage forms 2013:1248 – here adding further materials to every five hundred years. 10 the table presented by Kallio). The datings of Honkola et al . are based on the principles of Bayesian phylogenetics . It also very tentatively estimates the breakup of Hungarian and the Ob- from Proto- Ref. PU PFU PFP PFV EPF LPF Uralic to 3300 BCE, while the Ob-Ugric languages separated in Kettunen 4000 2500 1000 1000 1900 BCE. The study notes that Mari separated from Proto- & Vaula BCE BCE BCE BCE Finno-Volgaic in 3200 BCE, while Erzya broke up in 2900 BCE. 1938 The great difference between these dating estimates and those Toivonen 3500 2500 1500 1000 500 of mainstream linguists is noteworthy and seemingly suggest a scenario something akin to the suggested breakup of Proto- Uralic first into West and East Uralic as has been suggested in 7 However, it must be noted that this hypothesis has been recent years (Häkkinen, J. 2009). 11 heavily criticized from various angles since its conception in Furthermore, one could divide up the estimates into 1997, see, for example: Crowley, T. 1999:109-115; Watkins, C. different schools of thoughts, for example: M. Korhonen 2001:44-63; Janda, R. & , . 2003: 3-180; Koch, H. 2004: (traditional), K. Wiik (long time-depth), J. Janhunen (somewhat 17-60; Campbell, L. & Poser, W.J. 2008:318-329. long time-depth), J. Häkkinen (shallow time-depth), and so on. 4

Dating Finno-Mordvinic by statistics & sound laws

(mostly as given at the World 7. Uralic PU – 7000 BCE – 3500 BCE website) and previously estimated time of 8. Pre-Uralic Pre-PU separation from the Proto-language. The Tundra Yukaghir (150 spk.) – Para-Uralic languages compared in this study have been Kolyma Yukaghir (50 spk.) – Para-Uralic underlined in bold for clarity, and represent Additionally, prospective Para-Uralic entities the time depths in the rightmost columns of have been found in the Yukaghiric languages the table above. (cf. Collinder, B. 1940 & 1957, Fortescue, M.

0. Finnish + Para-Finnish 1998 (also additionally arguing for Eskaleut Modern time – 2010 CE and Chukchi-Kamchatkan as Para-Uralic Finnish (~5.7 millions) (+ Meänkieli and Kven) 1. Finnic entities), Rédei, K. 1999, Wurm. S.A. 2001:27, (LPF) – 1 CE –1000 CE Estonian (1.05 million) Piispanen, P.S. 2013, 2015 & 2016) and Võro (70 000 spk.) perhaps also in the Chukchi-Kamchatkan Ingrian (360 spk.) Karelian Proper (45 000 spk.) languages (cf. Blažek, V. 2006). Ludic (3 000 spk.) Olonetsian (19 270 spk.) Livonian (2 spk.) 3. The lexicostatistical method Veps (5 750 spk.) Votic (15 spk.) 2. Finno-Saamic 3.1. The basis of the method used in this (EPF) –1250 BCE – 1 CE paper is according to the postulates of Western Saamic: Southern Saami (600 spk.) Starostin’s etymological statistics (Starostin, S. Ume Saami (20 spk.) Pite Saami (20 spk.) 2000) given below verbatim: Lule Saami (2 000 spk.) Northern Saami (20 700 spk.) 1. In every language there are some roots that are Eastern Saamic: Kainu Saami (extinct) original, i.e. not borrowed during the period of Saami (extinct in the 19th century) separate existence of this language. According to Inari Saami (300 spk.) preliminary estimates, there are not much more Akkala Saami (extinct in the 21st century) than two or three hundred roots of this type in any Kildin Saami (500 spk.) 12 Skolt Saami (400 spk.) modern language. Ter Saami (2-10 spk.) 2. These roots have different frequencies of 3. Finno-Mordvinic occurrence, in other words they have different Erzya (696 630 spk.) probabilities of being found in a chosen text. 13 Moksha (614 000 spk.) 3. The frequency of occurrence (as just defined) of a 4. Finno-Volgaic PFV – 1500 BCE – 400 BCE given root in some language at a fixed period of Hill Mari (30 000 spk.) time, t, is stable, and does not depend (or hardly Meadow Mari (460 090 spk.) depends) on the type of text. Merja (extinct in the 17th century) 4. All roots can ‘age’ – their frequency of occurrence Muroma (extinct in the 16th century) gradually approaches zero, after which the root is Meshcherian (extinct in the 16th century) 5. Finno-Permic considered to have disappeared from the language. PFP – 3000 BCE – 1000 BCE At the same time, however, the rate of loss of Komi-Zyrian (217 000 spk.) different roots is not identical: roots, like words, may Komi-Permyak (94 300 spk.) 14 be divided into stable and less stable. Udmurt (479 800 spk.)

6. Finno-Ugric 5. The loss of roots from a language proceeds at a PFUg – 5000 BCE – 2100 BCE steady rate – that is, from some set of roots, Hungarian (12.5 millions) characterized by a fixed frequency, over a given 6b. Ob-Ugric period, Δt , a fixed number of roots will be lost. Khanty (13 600 spk.) Mansi (2 750 spk.) 6c. Northern Samoyed Forest Enets (20 spk.) 12 Tundra Enets (10 spk.) Here it must be assumed that Starostin referred to very old Nenets (31 300 spk.) protolanguages such as Proto-Uralic and Proto-Indo-European Nganasan (500 spk.) or perhaps languages from even higher time depths. Yurats (extinct) 13 6d. Southern Samoyed i.e. there are differences between, for example, the rate of Kamassian (extinct in the 20th century) borrowing for cultural, technical, social and basic lexical items. 14 Mator (extinct in the 19th century) The Swadesh-200 word list is considered to be a list of more Selkup (1 640 spk.) stable word items. 5

Peter S. Piispanen

Also, according to Starostin, the shorter semantically different from each other, and Swadesh-100 word list will have a cognacy hence considered different words in the study. rate of 90 % or more for , 70-80 % for closely related languages (such as Slavic, 3.3 Further clarifications of the method of Romance, Germanic and Turkic), 25-30 % for cognate selection are required. Acceptable cognates in the study are the words that Indo-European, 10-20 % for Uralic and 5-9 % clearly share a common origin in an earlier for macro-families such as Nostratic. The rate of word loss from the has been proto-language, are traceable through sound calculated to 14 of 100 items per 1000 years changes and retain the same semantic meaning still today in the two languages, i.e. (giving a retention constant of 0.86). The formula to calculate the point of divergence semantically unchanged cognates as reasonably proven by sound laws. As such, a for two genetically related languages is thus: word can have been subject even to Δt = log c / 2 log r unexplainable or non-categorized phonemic changes, since not all sound changes in the ; where Δt = point of language divergence studied languages are yet known (categories before present (in years), c =current rate of 9, 34 and 36 in the respective tables for cognacy between two languages (0-1.00) and Estonian, North Saami and Moksha sound r is the so-called glottochronological constant changes), but still be considered a cognate. In of 0.86. In this study the larger Swadesh-200 contrast, words that are clearly borrowed only word list has been employed for increased into that particular and narrow branch of 15 granularity and accuracy. languages, invented into that language or a 3.2 It must be noted that while the Swadesh- recent proto-language or semantically 200 list is perhaps not completely culturally or changed words from what appears to have faunistically neutral it does seem to work well been the original meaning in the older proto- to describe Uralic cultural vocabulary. 16 As a language (categories 10, 35 and 38 in the non-speaker of Estonian, North Saami and respective tables for Estonian, North Saami Moksha it has been difficult to determine the and Moksha sound changes) are disqualified most neutral lexical item appropriate for the as cognates in the study since they have been comparisons; however, in order to not miss subject to the mechanisms of change. Finally, cognates between any of the languages, lexical items that are cognates with the synonymic items, if they are deemed common Finnish form are marked in blue semantically identical, have been included in color. Assuming uniformitarianism, items with the analysis. Likewise, dialectal items are a group of words in the entry of both Finnish sometimes presented. In other cases and the compared language are particularly dictionaries have been able to provide original revealing in their comparison; if at least half of cognates, but which are now clearly such word groups are judged cognates that word entry gets checked as a cognate set, indicating close genetic relationship. Relevant 15 Clearly the word choices consist of excellent ’eternal’ words proto-items for the comparisons are given that well fulfill the postulates of lexicostatistics. Should the list (usually as UEW refs.). Relevant, proposed ever be extended, good candidates to add could be ‘to go’ (Fin. mennä ), ‘home dwelling/nest/hole’ (Fin. pesä ), ‘a lot’ (Fin. sound changes for each language pertain only paljon ), ‘to swallow’ (Fin. niellä ) and ‘ear’ (Fin. korva ). 16 to the upmost lexical item if several are For example, since there are no snakes in New Zealand, Iceland, Greenland, Antarctica and Ireland a word for snake presented. (found in the Swadesh-200) would not be a good basic word to include in the studies of languages in those areas, in contrast to all “Uralic” lands where snakes are plentiful. 6

Dating Finno-Mordvinic by statistics & sound laws

3.4. In addition to the references given in the Finnish likely separated sometime after Later appendix of each respective language Proto-Finnic, whose homeland may have comparison, the following basic etymological included southern , and dictionaries have also been employed for 19 (cf. Kallio, P. 2009), and which further tracing the : Collinder, occurred, according to the various estimates B. 1955, Häkkinen, K 2011, SSA 1992-2000, in the previous table, around 1 A.D.-1000 A.D. SKES 1955-1981 and UEW 1988-1991. 4.2. While the languages may be similar from 4. Estonian – Finnish a syntactic standpoint, they differ somewhat from a phonologic standpoint. The known 4.1. Estonian is known to be one of the closest sound changes from Later Proto-Finnic into relatives of Finnish. Finnish has at least modern Estonian 20 are summarized below: 205 000 words, and if all adverbials and archaic words are included the total is 300 000 Sound Change 1 Exchanging second or last voiceless into words (cf. Turunen, A. in Sinor, D. 1988:79). voiced plosives The majority of the modern language (80 %) Example: Fi. jalka – Est. jalg ‘foot’ 2 The loss of harmony was derived from the Earlier Proto-Finnic Example: Fi. suolet – Est. söakus ‘guts’ 3 Assimilations (usually progressive) of the types: language with the remaining being borrowings Example: -tk- -> -kk- ; -lt- -> -ll- ; -ns- -> -s(s)- and inventions. 17 Estonian too has the same 4 Only roots are compared, completely ignoring the infinitive endings amount of original words, borrowings and Example: Fi. -tV(k) – Est. -mV inventions. Most borrowings in Finnish are 5 Other sound changes, such as elimination through gradation from Swedish, while in Estonian they are Example: Fi. sidon – Est. seon ’I bind’ mostly from High and Low German (between 6 The loss of certain 21 th Example: Fi. punainen – Est. punane ‘red’ the 13 century to the 1940s), but also from 7 Elimination of end consonant Finnish (since the 1870s), Old and Modern 8 Elimination of end vowel 18 9 Unexplainable sound changes and non -categorized Russian and Latvian. Still, the majority of the sound changes. Example: iśä -> isa ‘father’ languages’ words originate from Proto-Uralic, 10 Borrowed, invented or semantically changed word Notes: The sound laws are summarized from Turunen, A. in and the number of similar lexical items and Sinor, D. 1988:65-69. the cognacy rate can be expected to be 4.3. By comparing Finnish and Estonian relatively high. The distribution and diachronic cognates on the Swadesh-200 list (Appendix 22 properties of the word borrowings suggest a A) while taking into account the sound laws, geographic movement of the ancestral forms the cognacy rate has been determined. of Finnic (and Estonian and Saami) across the Proposed sound changes are given in the last forest belt between the Urals and the Baltic column for each word. Sea (cf. Janhunen, J. 2009). Estonian and

17 Although there are only about 5 500-6 000 undivided stems in the Balto- (cf. Turunen, A. in Sinor. D. 1988:81) 19 As is rather strongly suggested by the various strata of the rest being changes acquired through local changing loanwords in the different languages: Finnish, Estonian and circumstances of life. Of course Finnish is not a discrete Proto- Northern Saami. A list of loanwords from various eras appearing Finnic branch, but rather a merger of at least three or so in Finnish (and closely related languages) can be found in branches, with SW Finnish dialects closer to Estonian/Livonian Itkonen, E. (1966). and East Finnish dialects closer to Karelian (Sammallahti, P. 20 i.e. after the took place, ex: *tulka > sulka 1977). (feather), *käti > käsi (hand). 18 Lexical borrowings: Low German 770-850 words, High 21 It should be noted that while Estonian has 25 diphthongs and German 490-540 words, Balto-German 60 words, Swedish 100- Finnish only 16, these are much more commonly used in 150 words, Russian 300-350 words, Latvian 30-45 words and Finnish. Finnish 100 words (cf. Sinor, D. 1988). All in all Finnic has about 22 Lexicon in Estonian-Finnish cognate appendix is collected 200 loanwords of apparently Baltic origin (cf. Kallio, P. 2008). from Greenberg, J.H. (2002) common Estonian dictionaries. 7

Peter S. Piispanen

The cognacy rate, i.e. words derived from the BCE – 700 CE (see Table above). More same original word, is 149 words out of 205, precisely, Proto-Saamic is believed to have i.e. 0.7268. The so-called glottochronologic disintegrated into a very diverse by the formula gives: middle of the first millennium A.D. while less than perhaps a millennium earlier (Pre-)Proto- Log 0.7268 / 2 log 0.86 = 1.058 = 1058 BP Saamic had been a dialect of Proto-Finno- This places the split between Estonian and Saamic (cf. Kallio, P. 2009:38). Finnish at around 952 A.D. As such, the result Interestingly, Proto-Finno-Saamic seems to is reasonable and fits rather well with previous have taken Pre- and Paleo-Germanic 23 dating results (i.e. LPF: 1 AD – 1000 AD). The loanwords that spread from about 1700 BCE cognacy rate between Estonian and Finnish is and onwards (cf. Kallio, P. 2009). Importantly, higher than expected perhaps suggesting a Aikio’s research includes a language exchange few hard-to-determine borrowings already hypothesis showing substratum vocabulary th before the 18 century. apparently, based on , as being contemporaneous with Scandinavian 5. Northern Saami – Finnish loanwords, which must have consequences for 5.1. The Saami languages are spoken in a very the dating of the break-up of (Pre-)Proto- large area ranging down from south of Idre in Saamic (Aikio, A. 2004, 2006, 2007). Dalarna (Dalecarlia in English) in central Further, some Proto-Indo-European, 27 Proto- to the tip of the Kola Peninsula in Indo-Iranian or Proto-Balto-Slavic loanwords 28 along the coast in an area 150-300 km are found in the Saamic languages (cf. Kallio, wide. The representative among the Saamic P. 2009), but which are missing from Finnish, languages was chosen to be Northern Saami and thus suggest old origins, branching and which has the most speakers. 24 While three northwards coastal language contacts for Pre- different hypotheses have been presented on Proto-Saamic speakers, 29 likely in southern the origin of the Saamic languages 25 the Finland during the Iron Age. While researchers currently believed theory is that Proto-Saamic such as Aikio and Junttila clearly contend that developed from Earlier Proto-Finnic (cf. all Baltic items in the Saamic stem from Korhonen, M. 1981:23). 26 As such, Saamic indirect borrowings through Finnic, the apparently branched off much earlier than viewpoint is probably correctly countered, for Estonian (as Pre-Proto-Saamic), which example, by the research of L-G. Larsson probably happened during the period of 1000 (2001:237-254).

23 Common Proto-Balto-Slavic loanwords that It is perhaps noteworthy that this places the split at perhaps approximately the same time, BP, as the tentative split between exist in Finnish (cf. Sammallahti, P. 1990), but Catalonian and Spanish ( strict cognacy rate 0.72; Harris, M. 1997). only partly in Saamic are dated to ca. 1000 24 More specifically, Northern Saami belongs to the Western BCE (cf. Kallio, P. 2008), suggesting Finno- Saamic languages and is one of ten current Saami languages. 25 Namely these: (a) Proto-Saamic developed when Samoyed people exchanged their original language with Proto-Finnic at 27 Example: Proto-Indo-European *ḱṷōn > Proto-Balto-Slavic some stage ( language exchange hypothesis ), (b) The speakers of *ś(u/v)ōn ’dog’ -> Early Proto-Saamic *śa/ōvonji > Proto-Saamic Proto-Saami had originally spoken an unknown language, but *śuovunjë > Northern Saami šūvon ’well-trained dog’. started to heavily borrow from Finnish lexicon and morphology 28 (contact borrowing hypothesis ) and (c) Proto-Saamic developed Of the about 40 Baltic loanwords from this era, 30 also have from Earlier Proto-Finnic, which is the main, current theory for cognates in Finnic. Finnic, however, has at least five times more which evidence can be readily found. Baltic borrowings than Saamic. 29 26 Namely: Earlier Proto-Finnic -> Middle Proto-Finnic + Proto- Which is also implied by recent genetic studies of Saami Saamic. Proto-Saamic then produced Western Saamic and populations in comparison to other European and Asian subsets Eastern Saamic. (for example in: Tambets, K. et al. (2004)). 8

Dating Finno-Mordvinic by statistics & sound laws

Saamic linguistic uniformity, but areal Proto - Saamic Proto - Saami Finnic Finnic divergence at that time. Other interesting 1 *-š(t,k,n) - -*jh(t,k,n) - 16 *ä á, i(e), ea borrowing and language contact phenomena 17 *a -,* -a- á, uo Earlier Proto - 18 *o uo, o, oa in Northern Saami are also indicated Proto- Saamic (Piispanen, P.S. 2012). Finnic 2 *-k(l,ń,j) - *-v(l,ń,j) - 19 *ō uo 3 *-kj - *-kš' - 20 *e a, ie, ea While the modern words originate from the 4 *-mp - *-mb - 21 *ē ie same roots, numerous language innovations 5 *-nt - *-nd - 22 *u o 6 *-ŋk- *-ŋg- 23 *ū u and sound changes render cognate 7 *ś - / -ś- *ć - / -č- 24 *ü a recognition and understanding difficult. Still, 8 *-š- *-s- 25 *i a 9 *-- *-k- 26 *ī i the languages no doubt have a relatively close 10 *-č- *-c- 27 *-VKV - -VKKV - genetic relationship. Additionally, semantic 11 *-t(n,v) - *-r(n,v) 28 *p -,t -,k - *b -,d -,g - 12 *-tj - *-rš - 29 *u(o) -a o-i changes sometimes make understanding more 13 *-p(δ,l,r) - *-b(δ,l,r) - difficult. 30 The Saamic languages have around 14 *-kη *-gη - 15 *-č'm - *-зˇ'm - 550 words completely lack (cf. 30 Assimilations (usually regressive) of the type: Sammallahti, P., 1998:125) 31 – as such, these Example: -nt - -> -nd - -> -dd - 31 Only verb roots are compared, completely ignoring the words have no cognates in the other Uralic endings languages or in any other, known language. 32 Example: Fi. -tV(k) – Saam. -Vt 32 Other sound changes, such as insertions It can be assumed that the ancestral speakers Example: -au - -> -av - & - - -> -ohk - of Saami have been into language contact with 33 Elimination of end consonant 34 Unexplainable, non -categorized sound changes some other, now likely extinct language (cf. 35 Borrowed, invented or semantically changed word Notes: V = vowel. The Western Saamic languages are Southern, Aikio, A. 2006). At this time depth intelligibility Ume, Pite, Lule and Northern Saami. The sound laws are between Finnish and Saamic can no longer be summarized from Sammallahti, P. 1998 & Korhonen, M. 1981. expected, and the cognacy rate will be relatively low. 5.3. Comparing Finnish and Northern Saami 5.2. The known sound changes from Earlier cognates on the Swadesh-200 list (Appendix 33 Proto-Finnic into various stages of Saamic are B), while taking into account both known summarized below: word etymology (mainly Sammallahti, P. 1998) and, importantly, the Àlgu tietokanta and the Earlier Western Earlier No rthern sound laws (as outlined above and presented in Appendix B), yields a cognacy rate of 82 30 An example would be Fin. pakkanen ‘frost, cold’ <> N. Saami báhkas ‘warm’ (< Proto-Saamic *pakka- ‘hot, cold’). words from 205 available words, i.e. 0.40 or 31 Examples: atnit (to use), bivvat (keep warm), coagis (low, 40.0 %. 34 The proposed sound changes are shallow), čáhppat (black), čiekčat (to kick), heavdni (spider), jalŋŋis (stump), jorrát (to purr), láhppit (to loose, to spend), nagir (to sleep), njivli (mucus), ohca (lap), oakti (downpour), 33 ravgat (to fall), sarrit (blackberry), šiehttat (to agree) and uhcci Lexicon in Northern Saami-Finnish cognate appendix is (small). A few of these words are present in the Swadesh-200 collected from: Sammallahti, P. (1990), Sammallahti, P. (1998), word list. Korhonen, M. (1981), Greenberg, J.H. (2002), and various 32 Proto-Saamic speakers may have assimilated several layers of dictionaries and from the Álgu tietokanta: earlier languages in from the early hunter-gatherers http://kaino.kotus.fi/algu/index.php?t=etusivu . Further, they encountered (cf. Aikio, A. 2006). However, it has also been Professors Erling Wande and Mikael Svonni are gratefully theorized that when speakers of Uralic languages arrived in the acknowledged for help with Northern Saami glossary. eastern region ( Textile Ware , ca. 1900 BCE) they may have met people already there and in Finland who spoke 34 versions of the Pre-Uralic languages they had brought with It has been suggested that there are a number of Finnic them during the Stone-Age waves from the -Oka region loanwords in Northern Saami, i.e. not inherited words, some of (Sperring Ware , ca. 5100 BCE and/or Combed Ware Style 2 , ca. which are found on the Swadesh-200 list (cf. Aikio, A. 2007 and 3900 BCE). These factors, of course, would make the Sammallahti, P. 1990). The proposed loans are items 74, 146, identification of non-Uralic substrate items in Finnic more 161, 186 & 204. This presents a possible conundrum where a difficult to identify (cf. Kallio, P. 2009). language may have borrowed words from another branch of a 9

Peter S. Piispanen

given in the last column for each word. The so- have separated around 1000 CE and are no called glottochronologic formula gives: longer mutually intelligible.

Log 0.400 / 2 log 0.86 = 3.038 = 3038 BP Mordvinic vocabulary consists of around 30 % each of inherited, invented and loanwords, This places the split between Northern Saami while 10 % has no known etymology. Inherited and Finnish at around 1028 BCE The result fits vocabulary, of course, is inherited from Proto- rather well with earlier estimates (i.e. EPF: Finno-Ugric. Among the invented vocabulary 35 1250 BC – 1 AD). This split concretely onomatopoetic words can naturally be found; represents Pre-Proto-Saamic, which much all languages have them, but sometimes, such later became Proto-Saamic, the ancestor of all as in Estonian, onomatopoetic words are Saamic languages. created consciously, like was done by Aavik who removed awkward compounds in favor of 6. Moksha – Finnish new Finnish borrowings and artificial new 6.1. Moksha (and Erzya) are Mordvinic sounds to revitalize standard Estonian (Aavik, languages that originated from Proto-Finno- J. 1919). Loanwords in the language 37 can be Mordvinic, perhaps as early as 1500 BCE. A dated to Finno-Volgaic, 38 Finno-Permic, 39 third of the current Mordvinic speakers live in Finno-Ugric and even Uralic times. 40 In the Republic of , a federal subject of general, most loanwords are derived from Russia. The languages have fairly certainly Germanic, Baltic, 41 Indo-Iranian, 42 Turkic, 43 been spoken in the area already since at least Tatar 44 and Russian words. Further, Mordvinic 1 CE, and the ancestors of the Gorodets has been heavily influenced by Russian and it culture of 500 CE are believed to have been can be concluded that earlier speakers have speakers of the Mordvinic languages. There been in intense language contact with are currently 614 000 Mokshan and 696 630 speakers of Uralic and other languages (cf. Erzyan speakers, most living in the old RSFSR area. 36 The two languages are assumed to and Temnikov. Erzyan speakers live in Northern and Eastern Mordva, and its , written with the Cyrillic genetically related language, or perhaps even adjusted its own , is based on the dialect of Kozlovska. Previously there form of the word with the form of the related language. were also speakers of the extinct Muroman and Meshcherian. 37 Borrowing itself is usually one of the mechanisms of language Examples: meš (bee) & mije - (to sell), both possibly Indo- change that would make its inclusion as a cognate in European items. lexicostatistics non-valid. However, there may be no way of 38 Example: ĺišmä (horse, from *lešmä, cf. Finnish lehmä (cow)). knowing what the original Saamic word would have been 39 before borrowing, or if Saami-speakers have merely corrected Example: kunda (cover, lid). 40 their speech to more closely resemble that of with whom For exhaustive lists of loanwords arrived in various eras and they were in close contact. If the proposed items are indeed protolanguages, see Itkonen, E. (1966) and Häkkinen, J. (2009), true loanwords into North Saami at a stage later than Proto- and for the layers and sources of loanwords in Uralic, see Saamic and their inclusion as cognates were to be non-allowed, Comrie, B. (2008:482) and Dahl, . et al (2001). 41 the cognacy rate would instead change to 77 words out of 205 Example: karda , kardo (sheep; Lithuanian gardas ). (37.56 %), yielding a dating of 3246 BP. 42 There are 18 such, known words. Mokshan examples: pavas 35 It is interesting to note that the Saamic languages between (God, luck), vaŕgas (wolf), śada (a hundred) and śura (horn), as them have cognacy rates of 80-90 % on the Swadesh-100 word given in Bartens, R. (1999). A curiosity is that there are only 6 list (cf. Sammallahti, P. 1988:37). For this reason the Finnish- such items in Erzya, but 30 in Hungarian. Additionally, there is Northern Saami cognate comparison is quite representative as a also at least one apparently pre-Ossetic loanword, namely whole for the Saamic languages. The loanwords refer to various lomań (person, Modern Ossetian lymæn /limæn )(found in the proto-linguistic stages, eg. Indo-Iranian loanwords may go back dialects, i.e. Mordvinisches Wörterbuch by H. Paasonen, 1990- to PFU times, etc. Further, curiously, the UEW reports several 1999, presented in Lexica Societatis Fenno-Ugricae XXIII). A North Saami words that seem to have cognates only in the useful source for Moksha Mordvin has been Herrala, E. & Samoyed languages, but nowhere else, for example North Feoktistov, A. 1998. Saami čävddë (skin, ) & Nenets śāpt (bark), North Saami 43 čâllât (rub the antlers against something to get the skin off) & There are perhaps around 190 such loanwords in Mordvinic Nenets śelā - (id.) and North Saami čoaw'je (stomach, belly) & from either Bolgar-Turkic or Tatar, as given in Paasonen, H. Kamassian šшj ǝ (id.). (1897). 44 36 The speakers of Moksha live in Western Mordva, and its Example: baška (except, Tatar baška , id.). Many more literary language is based on the dialects of Krasnoslobodskin examples can be found in Hasselblatt, C. et al . (2011). 10

Dating Finno-Mordvinic by statistics & sound laws

Abondolo, D. 1998:211-217). For these laws. 45 In addition to the sound changes reasons, and due to the high time depth presented in the table, several other not involved, a low cognacy rate between Finnish completely regular sound and Moksha can be expected. correspondences between Finnish and Moksha cognates could be observed 46 (such 6.2. The known sound changes from Proto- cognates are marked with 37 among the Uralic and Proto-Finno-Ugric to the Moksha suggested sound changes). language are summarized below: 6.3. Comparing the Swadesh-200 word list for PU or Moksha PU or PFU Moksha 47 PFU the two languages, while taking into account 1 *-i- -e- 16 *-kk - -k- the outlined sound laws, gives 73 cognates out 2 *-ü- -e- 17 *t - d- 3 *-ä- -e- 18 *-t- -d- of 205 words, i.e. a cognacy rate of 0.356 or 4 *-u- -o- 19 *-tt - -t- 35.6 %. The so-called glottochronologic 5 *( -)e -ä (-)i -e 20 *-p- -v- 6 *-ī- -i- 21 *-kp - -kb - formula gives: 7 *-ē- -e- 22 *-pp - -p- 8 *-ō- -a- 23 * ń- n- Log 0.356 / 2 log 0.86 = 3.423 = 3423 BP 9 *-ū- -u- 24 *-ŋ- -v- or –ø- 10 *-a ä 25 *č - š- - This would place the Finno-Mordvinic split at a 11 *-ä -ə 26 *sä -, se -, śV -, śV -, si- śe- rather high time depth of around 1413 BCE 12 *-e -ä or -ə 27 *-es - -iz - 13 *ü - v- 28 *-ś- sometimes and chronologically shortly after the Finno- -ź- Volgaic split (a lexical layer estimated by most 14 *-VkV - sometimes 29 *w -,-w, sometimes -VvV- -w- v-, v-, -v- at around 1500 BCE). It is an interesting note 15 *-eke - -ij ä 30 *- δ- -d-, -t- that the cognates in Moksha Mordvin are 31 Elimination of the first or last syllable Examples: *šukšna –> šna & *śüdämi –> śeďi more often found among the than 32 Synkope, i.e. elimination of end vowel among the . Indeed, the sound changes 33 Only verb root s are compared, completely ignoring the endings of verbs appear to have been complicated. Example: Fi. -tV(k) – Moks. -Vms Moksha Mordvin (MM) words are often very 34 Other changes, such as the voicing of consonant clusters Example: *tulka –> tolga long, which may perhaps be due to more 35 Palatalizations, such as t - into t’ -. recent grammaticalizations (keeping the Examples: *tä- –> t’ä & *näke- –> ńäj ǝms 36 Un categorized sound changes such as insertion, cognates; example: PFU *jäŋe ‘ice’ (UEW 93) > suffixation or irregular vowel changes. Words for which Fin. jäätyä ‘to freeze’, MM äj ǝnda(kšń ǝ)ms ‘to most - but not all - sound changes are known are also placed into this category. freeze’), 48 radical semantic changes or 37 Further sound correspondences between Finnish and Moksha (see the text and footnote 41 for details) 38 Borrowed, invented or semantically changed word 45 39 Observed common change counter to the established For example: *o- ->u- (noted as rule 39) is observed sound laws (i.e rule 4): *-o- –> -u- numerous times. Likewise, the sound k behaves oddly and Not es: V = v owel. The sound laws are summarized from the occasionally changes into unexpected sounds (i.e. other than j, v works of Gábor Bereczki in Sinor, D., 1988:316–331. For or g) or completely disappears. Also, the change *ś- -> s-, and example, *t- means that the sound change pertains to the first other palatalizations, seem to be rather common (noted as rule syllable of the word, and *-ä pertains to the last syllable. 35), as is *-e(-) -> -a(-) and perhaps not fully unexpected. While uncategorized sound changes (marked as rule 36) cannot be fully explained, such words can still be cognates. 46 Such sound correspondences would tentatively include It is further noted that, in certain cases, words (Finnish <-> Moksha) and word item number given in brackets: (-)t- <-> (-)t’- (9, 149), s- <-> ś/sť- (68, 86, 125), (-)l(C)- <-> (- - particularly often verbs - seem to have )lj(C)- (74, 96, 160, 180, 181), k <-> kj (78, 83), -si <-> -dj (83, changed contrary to the established sound 150, 183) and (-)h- <-> (-)sh- (35, 157). Further, I note that Moskhan -ə- appears in prosodically predictable positions. 47 Lexicon in Moksha-Finnish cognate appendix is collected from Abondolo, D. (1998), Greenberg, J.H. (2002) and Moksha and Erzya glossaries found on the Internet. 48 As can be inferred from Bartens, R. (1999:122-161). 11

Peter S. Piispanen

borrowings from both Uralic and non-Uralic and POUg, respectively, would need to be sources (exchanging the cognates; example: verified by word comparisons on the larger PU *ićä ’father, big’ (UEW 78) > Fin. isä Swadesh-200 word list while employing sound ‘father’, but MM oćä ‘brother of father’ – law changes, as well as complemented by a instead MM al’ä ‘father’ is used). Clearly, the new dating of Proto-Finno-Ugric itself by the time depth here is great enough that the methods outlined in this paper. similarities between the languages are quickly disappearing, rendering the recognition of 8. Results and Conclusions true cognates a daunting task. All of these Compared Cognacy rate Point of Proto - factors lead to a higher margin of error in the Languages divergence language analysis and dating of Moksha in this study. Finnish - 72. 7 % 1058 BP LPF Estonian ~ 952 CE 51 Likewise, the known, proposed sound changes Finnish - 40 .0 % 3038 BP EPF Northern ~ 1028 are not complete and in some cases not even 52 Saami BCE exhaustive. Finnish - 35.6 % 3423 BP PFM~ Moksha PFV ~ 1413 7. Khanty – Mansi - Hungarian BCE Khanty -Mansi 45 % 2647 BP POUg 7.1. Here it is also of interest to discuss the ~ 637 BCE Hungarian - 28 % 4220 BP PUg question of Khanty (Ostyak in older literature), Khanty ~ 2210 Mansi (Vogul in older literature) and BCE Hungarian in greater detail. These Ugric 8.1. The here obtained results are in rather languages originate from the Proto-Ugric good accordance with previously established (PUg) branch of Proto-Finno-Ugric, and later datings done by linguists, suggesting that from Proto-Ob-Ugric (POUg). The cognacy earlier linguists were on the right track from rates on the Swadesh-100 word list between the start. All obtained points of divergence these languages have been reported as a mere are, of course, only approximate years, leaving 53 45 % for Khanty and Mansi, 34 % for Mansi room for a certain margin of error. and Hungarian and only 28 % for Hungarian More noteworthy, the very brief time lapse and Khanty (cf. Sammallahti, P. in Sinor, D. between the apparent break-up of Finnic- 1988:499). This suggests that Hungarian split Saamic-Moksha unity and Finnic-Saamic unity 49 off from Proto-Ugric (and what would may actually cast doubt about the existence of become Proto-Ob-Ugric) 4220 BP, i.e. in 2210 a discrete Proto-Finno-Saamic language. Such

BCE – a figure that actually fits quite well with a scenario would instead support the idea later dating estimates of the somewhat earlier presented by Itkonen (Itkonen, T. 1997) that Proto-Finno-Ugric of 2000 BCE-3000 BCE there was no such proto-language. At the Much later the Proto-Ob-Ugric entity broke up into Khantic and Mansic 2647 BP, i.e. in 637 Sammallahti, P. in Sinor. D. 1988:502). The result above does 50 BCE. However, these tentative dates for PUg agree with this hypothesis. 51 Alternatively, due to a possible loanword conundrum (see notes in the Northern Saami section), the cognacy rate may be 49 only 37.6 %, which would yield a somewhat earlier dating of And later turning into Proto-Finno-Permic, although 3246 BP, and EPF ~1236 BCE. Hungarian may have prior to this developed from a branch 52 tentatively called Proto-Finno-Hungarian (cf. Sammallahti, P. in This would surprisingly suggest the existence of a Finnic Sinor, D. 1988:491) - as suggested by etymological difficulties starting with Early Proto-Finnic to change when comparing Hungarian to Proto-Ob-Ugric and Proto-Finno- over two thousand years through a tentative Middle Proto- Permic - which somewhat complicates the picture. Finnic into Late Proto-Finnic, from which sprung a multitude of 50 It has been noted by Sammallahti that, while some have languages closely related to modern Finnish. 53 dated the disintegration of the Ob-Ugric unity to about the year Since most of the used lexical data was collected in 2010, this 500 A.D., the relatively small number of common is the year used with BP to calculate the year of respective do suggest a considerably earlier disintegration (cf. proto-language breakup. 12

Dating Finno-Mordvinic by statistics & sound laws

same time, however, the relatively late break- References up of Finnish and Estonian is in accordance with recent estimates (eg. Kallio) of a Aavik, Johannes 1919: Uute sõnade sõnastik. relatively “late” Proto-Finnic language. Sisaldab üle 2000 uue ja haruldasema sõna ühes tuletuslõppude tabeliga [Glossary of new The results suggest that the lexicostatistical words. Contains over 2000 new and lesser method may work with the Uralic languages used words with tables of derivational back to at least 4000 BP. (North) Saami and ]. : Istandik. Moksha Mordvin appear to have separated from the proto-language that led to Finnish at Abondolo, Daniel 1998: The Uralic Languages, only a few centuries in between. London & New York: Routledge.

8.2 The accuracy of some lexicostatistical Aikio, Ante 2004: An essay on substrate methods has been estimated to have an error studies and the origin of the Saami, margin of up to 10 percent (Gell-Mann, M., Etymologie, Entlehnungen und Entwicklungen: Peiros, I. and Starostin, G. 2009:16). I suggest Festschrift für Jorma Koivulehto zum 70. that the proper inclusion of sound laws for Geburtstag, Hyvärinen, I., Kallio, P. & cognate recognition reduces this margin of Korhonen, J. (eds.), Mémoires de la Société error – as it eliminates human guesswork of Néophilologique de , 63, p. 5-34. look-alikes – possibly at best cutting it down Helsinki: Société Néophilologique. by half or even two thirds. Thus, the error Aikio, Ante 2006: On Germanic-Saami contacts margin in this work is estimated to be and Saami prehistory, Journal de la société approximately 3-5 %, which can be set as the Finno-Ougrienne , 91 , p. 9-55. +/- value to any of the above acquired BP- values. In conclusion, basing comparison on a Aikio, Ante 2007: Etymological nativization of steadfast phonetic ground probably instead loanwords: a case study of Saami and Finnish, makes the semantic classification the largest in: Saami Linguistics , Toivonen, I. & Nelson, D. problem and source of error of the method. (eds.), John Benjamins, pp. 17–52.

8.3 Lexicostatistics is often used for languages Anttila, Raimo 1989: Historical and on which virtually no historical-comparative Comparative Linguistics , 2 nd edition, work has been done. The herein used Amsterdam: Philadelphia: John Benjamins combination of traditional lexicostatistic Publishing Company. methods with detailed traceable, sound change correspondences of cognates should Bartens, Raija 1999: Mordvalaiskielten be considered a new and quite innovative rakenne ja kehitys , Helsinki: Finno-Ugrian methodology, which seemingly allows for Society. more sensible dating result to be obtained. Bengtson, John D. & Ruhlen, Merritt 1994: 14. This implicitly also partially confirms the Global Etymologies,in: On the Origin of reliability of lexicostatistical methods. Language Studies in Linguistic Taxonomy , 8.4 These results do encourage a future Stanford: Stanford University Press, p.277- direction of research: the dating of higher 336. and/or lower time depths by select, Bereczki, Gábor 1988: Geschichte der representative Uralic languages. wolgafinnischen Sprachen, In: The Uralic

13

Peter S. Piispanen

Languages. Description, history and foreign Comrie, Bernard 2009: The World’s major influences , 1988, Sinor, D. (ed), Leiden, p. 315- languages , 2 nd Ed, London and New York: 350. Routledge.

Blažek, Václav 2006: Chukcho-Kamchatkan Crowley, Terry 1999: Book reviews – The Rise and Uralic: lexical evidence of their genetic and Fall of Languages , Australian Journal of relationship, Orientalia et Classica XI. Aspects Linguistics, 19 , p. 109-115 of Comparativistics, 2 , Moscow, p.197-212. Dahl, Östen & Koptjevskaja-Tamm, Maria Bouckaert, Remco & Lemey, Philippe & Dunn (eds)(2001) The Circum-, Michael & Greenhill, Simon J., Alekseyenko, volume 1: Past and Present , Amsterdam & V. & Drummond, Alexei J. & Gray, Philadelphia: John Benjamins. Russell D. & Suchard, Marc A and, Atkinson, Décsy, Gyula 1965: Einführung in die finnisch- Quentin D. 2012: Mapping the Origins and Expansion of the Indo-European Language ugrische Sprachwissenschaft . Wiesbaden: Family, Science, 337 , p. 957–960. Harrassowitz.

Bowler, James M. & Johnston, Harvey & Olley, Dixon, Malcolm Ward 1997: The Rise and Fall of Language , Cambridge: Cambridge Jon M. & Prescott, John R. & Roberts, G. & Shawcross, Wilfred & Spooner, Nigel A. University Press, UK. 2003: New ages for human occupation and Dixon, Robert Malcolm Ward 2002: Australian climatic change at Lake Mungo, Australia, Languages: their nature and development Nature , 421 , p.837-840. (Cambridge Language Surveys), Cambridge: Campbell, Lyle 1986: Comment on Greenberg, Cambridge University Press, UK, xlii. Turner, and Zegura, Current Anthropology , 27 , Fleming, H.C. 1973: Sub-classification in p. 488. Note: This follows directly after the Hamito-Semitic, In: Lexicostatistics in genetic reviewed by Greenberg, J. et al . linguistics: proceedings of the Yale conference , Campbell, Lyle 2001: Beyond the Comparative Yale University, April 3-4, 1971, Dyen, I. (ed), Method, In: 2001 15th The Hague & Paris: Mouton & Co., p. 85-88. International Conference on Historical Fortescue, Michael D. 1998: Language Linguistics , Blake, B.J., Burridge, K. & Taylor, J. Relations Across Bering Strait , London: Cassell (eds), Melbourne, 13–17 August 2001 & Co.

Campbell, Lyle 2004: Historical Linguistics: An Gell-Mann, M., Peiros, I. and Starostin, G. nd Introduction , 2 Edition, Cambridge, (2009) Distant Language Relationship: The Massachusetts: MIT Press. Current Perspective, Journal of Language Campbell, Lyle & Poser, J. 2008: Relationship, 1, p. 13-30. Language Classification , Cambridge, p. 318- Gray, Russell D. & Atkinson, Quentin D. 2003: 329. Language-tree divergence times support the Collinder, Björn 1940: Jukagirisch und Anatolian theory of Indo-European origin, Nature , 426 , p.435-439. Uralisch , Uppsala: Almqvist & Wiksell.

Collinder, Björn 1957: Survey of the Uralic Languages . Stockholm.

14

Dating Finno-Mordvinic by statistics & sound laws

Greenberg, Joseph H. 2002: Indo-European Itkonen, Terho 1997: Reflections on Pre-Uralic and Its Closest Relatives, volume 2. Lexicon. and the “Sami-Finnic proto-language”, Stanford: Stanford University Press, California. Finnisch-Ugrische Forschungen , 54 , p. 229- 266. Greenhill, S.J., Atkinson, Q.D., Meade, A. and Gray, R.D. 2010: The shape and tempo of Janda, Richard & Joseph, Brian 2003: language evolution, Proc. Biol. Sci., Aug. 22, Handbook of Historical Linguistics, Oxford, p. 277(1693), p. 2443-2450. 3-180.

Hajdú, Peter 1975: Sukulaisuuden kielellistä Koch, Harold, 2004: Australian languages: taustaa, In: Suomalaisugrilaiset , Hajdú, P. (ed), Classification and the , Helsinki: Suomalaisen Kirjallisuuden Seura, p. Bowern, Claire & Koch, Harold (eds.), 11-51. Amsterdam, p. 17-60.

Harris, M. & Vincent N. 1997: The Romance Janhunen, Juha 2009: Proto-Uralic – what, Languages . Oxford University Press. where, and when? Suomalais-Ugrilaisen Seuran Toimituksia = Mémoires de la Société Hasselblatt, Cornelius & Houtzagers, Peter & Finno-Ougrienne , 258 , p. 57-78. Pareren, Remco van (eds) 2011: Language contact in times of globalization . Amsterdam Janhunen, Juha 2008: Some Old World & New York: Studies in Slavic and General experience in linguistic dating, in: In Hot Linguistics. Pursuit of Language in Prehistory: Essays in the four fields of anthropology , Bengtson, John Herrala, Eeva & Feoktistov, A. 1998: D. (ed.), Amsterdam/Philadelphia: John Mokšalais-Suomalainen Sanakirja, . Benjamins Publishing Company, p. 223-239.

Honkola, T., Vesakoski, O., Korhonen, K., Kallio, Petri 2006: Suomen Kantakielen Lehtinen, J. Syrjänen, K. & Wahlberg, N. 2013: Absoluuttista Kronologiaa, Virittäjä, 1, p.2-25. Cultural and climatic changes shape the evolutionary history of the Uralic languages, J. Kallio, Petri 2008: On the ”Early Baltic” of Evol. Biol., 26, p. 1244-1253. Loanwords in Common Finnic, In: Evidence and Counter-Evidence: essays in honour of Häkkinen, Kaisa 1984: Wäre es schon an der Frederik Kortlandt , v1, SSGL 32, Amsterdam, p. Zeit, den Stammbaum zu fallen? Theorien 265-277. über die gegenseitigen Verwandtschaftsbeziehungen der finnish- Kallio, Petri 2009: Stratigraphy of Indo- ugrischen Sprachen, Ural-Altaische European Loanwords in Saami . In: Tiina Äikäs Jahrbücher, neue folge 4 , 1-4, Wiesbaden. (ed.), Máttut - máddagat: The Roots of Saami Ethnicities, Societies and Spaces / Places. Häkkinen, Jaakko 2009: Kantauralin ajoitus ja : Publications of the Giellagas Institute paikannus: perustelut puntarissa, Journal de la 12, p.30-45. société Finno-Ougrienne, 92, p. 9-56. Kassian, A., Zhivlov, M. & Starostin, G. 2015: Itkonen, Erkki 1966: Suomen suvun esihistoria, Proto-Indo-European-Uralic Comparison from Tietolipas , 20 , p.5-47. the Probabilistic Point of View, The Journal of Indo-European Studies , 43 :3-4, p. 301-392.

15

Peter S. Piispanen

Kettunen, Lauri & Vaula, Martti 1938: Suomen Piispanen, Peter 2016: A semivowel sound kielioppi sekä tyyli- ja runo-opin alkeet change rule in Yukaghir, Journal of Historical oppikouluille ja seminaareille . Helsinki: WSOY. Linguistics, Accepted.

Korhonen, Mikko 1981: Johdatus lapin kielen Rédei, Károly 1999: Zu den uralisch- historiaan . Helsinki. jukagirischen Sprachkontakten, Finnisch- Ugrische Forschungen , 55 , p.1-58. Larsson, Lars-Gunnar 2001: Baltic influence in the Finnic languages , in: Circum-Baltic Renfrew, C. & McMahon, A & Trask, L. 2000: Languages, Volume I: Past and Present, Time-Depth in Historical Linguistics, Studies in Language Companion Series, Dahl, Cambridge: The McDonald Institute for Ö. & Koptjevskaja-Tamm, M. (eds.), John Archaeological Research. Benjamins Publishing Company, Stockholm Salminen, Tapani 2002: Problems in the University. taxonomy of the Uralic languages in the light Nichols, Johanna 1992: Linguistic Diversity in of modern comparative studies. In: Space and Time . Chicago: University of Лингвистический беспредел: сборник Chicago Press. статей к 70-летию А. И. Кузнецовой. Москва: Издательство Московского Paasonen, Heikki 1990-1999: Mordwinisches университета, 2002. 44–55. Wörterbuch. Mordovskij slovar' , Lexica Societatis Fenno-Ugricae XXIII:1-6, Heikkilä, Sammallahti, Pekka 1977: Suomalaisten Kaino (ed.), Suomalais-Ugrilainen Seura. Also esihistorian kysymyksiä, Virittäjä , 81 , p. 119- found online at: 136. http://www.sgr.fi/lexica/lexicaxxiii.html Sammallahti, Pekka 1988: Historical Paasonen, Heikki 1897: Die türkische Phonology of the Uralic Languages, In: The lehnwörter im mordwinischen, Journal de la Uralic Languages. Description, history and société Finno-Ougrienne , 15 , 2. foreign influences , 1988, Sinor, D. (ed), Leiden, p. 478-554. Piispanen, P.S. 2012: Statistical Dating of Uralic Proto-Languages through Comparative Sammallahti, Pekka 1990: The Sami Language: Linguistics with added sound change law Past and Present, In: Arctic Languages , Collis, analyses, Fenno-Ugrica Suecana Nova Series , D. R. F. (ed), Paris: Unesco, p. 437-458. 14 , p. 61 – 74. Sammallahti, Pekka 1998: The Saami Piispanen, Peter 2013: The Uralic-Yukaghiric languages – An Introduction . connection revisited: Sound correspondences Kárášjohka/. of Geminate clusters, SUSA~JSFOu, 94 , p. 165- Sinor, Denis 1988: The Uralic Languages. 197. Description, history and foreign influences , Piispanen, Peter 2015: Evaluating the Uralic - Leiden. Yukaghiric word-initial, proto-sibilant correspondence rules, SUSA~JSFOu, 95, p. Starostin, Sergei 2000: Comparative-Historical Linguistics and Lexicostatistics, Time Depth in 237-273. Historical Linguistics, v1 , Cambridge: The McDonald Institute for Archaeological Research, p.223-265.

16

Dating Finno-Mordvinic by statistics & sound laws

Swadesh, Morris 1950: Salish internal Mitochondrial DNA and Y Chromosomes, relationships, International Journal of Am.J.Hum.Genet. , 74 , p. 661-682. American Linguistics , 16 , p.157-167. Toivonen, Yrjö Henrik 1953: Suomalis- Swadesh, Morris 1952: Lexicostatistic dating ugrilaisesta alkukodista, Virittäjä, p. 5-35. of prehistoric ethnic contacts, Proceedings Turunen, Aimo 1988: The Balto-Finnic American Philosophical Society , 96 , p.452-463. languages, In: The Uralic Languages. Swadesh, Morris 1955: Towards greater Description, history and foreign influences , accuracy in lexicostatistic dating, International 1988, Sinor, D. (ed), Leiden, p. 58-83. Journal of American Linguistics , 21 , p.121-137. Watkins, Calvert 2001:Areal Diffusion and Syrjänen, Kaj, Honkola, Terhi, Korhonen, Kalle, Genetic Inheritance, Aikhenvald & Dixon Lehtinen, Jyri, Vesakoski, Outi & Wahlberg, (eds.), Oxford, p. 44-63. Niklas 2013: Shedding more light on language Whitehouse, Paul, Usher, Timothy, Ruhlen, classification using basic vocabularies and phylogenetic methods, Diachronica , 30:3 , p. Merritt and William S-.Y. Wang 2004: Kusunda: An Indo-Pacific language in Nepal, 323-352. PNAS, vol. 111, no.15, April 13. Taageperä, Rein 1994: The linguistic distances Wiik, Kalevi 2002: Suomalaisten juuret . Atena between Uralic languages, Linguistica Uralica , Kustannus Oy. 30 , p. 161-167. World Ethymology website: Tambets, Kristiina & Rootsi, Siiri & Kivisild, Toomas & Help, Hela & Serk, Piia & Loogväli, www.ethnologue.com Eva-Liis & Tolk, Helle-Viivi & Reidla, Maera & Wurm, Adolphe 2001: Atlas of the Metspalu, Ene & Pliss, Liana & Balanovsky, World’s Languages in Danger of Disappearing , Oleg & Pshenichnov, Andrey & Balanovska, Unesco Publishing. Elena & Gubina, Marina & Zhadanov, Sergey & Osipova, Ludmila & Damda, Larisa & Voevoda, (Finnish and Saamic) etymological dictionaries: Mikhail & Kutuev, Ildus & Bermisheva, Marina Àlgu tietokanta: & Khusnutdinova, Elza & Gusar, Vladislava & http://kaino.kotus.fi/algu/index.php?t=etusiv Grechanina, Elena & Parik, Jüri & Pennarun, u&kkieli=fi Erwan & Richard, Christelle & Chaventre, Andre & Moisan, Jean-Paul & Barać, Lovorka & Collinder, Björn 1955: Fenno-Ugric Peričić, Marijana & Rudan, Pavao & Terzić, Vocabulary. An of the Rifat & Mikerezi, Ilia & Krumina, Astrida & Uralic Languages . Uppsala. Baumanis, Viesturs & Koziel, Slawomir & Richards, Olga & De Stefano, Gian Franco & Häkkinen, Kaisa 2011: Nykysuomen Anagnou, & Pappa, Kalliopi I. & etymologinen sanakirja, Sanoma Pro. Michalodimitrakis, Emmanuel & Ferák, SSA: Suomen sanojen alkuperä. Etymologinen Vladimir & Füredi, Sandor & Komel, Radovan sanakirja . Helsinki: Kotimaisten kielten & Beckman, Lars & Villems, Richard 2004: The tutkimuskeskus / Suomalaisen kirjallisuuden Western and Eastern Roots of the Saami – the seura. 1992-2000. story of Genetic “Outliers” Told by

17

Peter S. Piispanen

SKES: Suomen kielen etymologinen sanakirja . Lexica Societatis Fenno-Ugricae XII. Helsinki: Suomalais-ugrilainen seura. 1955-1981.

UEW: Rédei, Károly 1988-1991: Uralisches Etymologisches Wörterbuch . Budapest: Akadémiai Kiadó.

Abbreviations

PU=Proto-Uralic, PS=Proto-Samoyed, PFU=Proto-Finno-Ugric, PUg = Proto-Ugric, POUg = Proto-Ob-Ugric, PFP=Proto-Finno- Permic, PFV=Proto-Finno-Volgaic, PFM = Proto-Finno-Mordvinic, EPF = Earlier Proto- Finnic (i.e. Proto-Finno-Saamic), MPF = Middle Proto-Finnic and LPF = Later Proto-Finnic, Fin. = Finnish, Kar. = Karelian, Est. = Estonian, N. Saami = North Saami, K. Saami = Kildin Saami, EM = Erzya Mordvin, MM = Moksha Mordvin, KZ = Komi-Zyrian, Udm. = Udmurt, Hung. = Hungarian, Rus. = Russian, Eng. = English, Latv. = Latvian, Lith. = Lithuanian, Swe. = Swedish, PIE = Proto-Indo-European.

18