Comparisons with previous divergence time estimates
Our phylogenetic hypothesis follows the traditional view of the Uralic language divergences in that the first split occurs between Samoyed and Finno-Ugric languages and is followed by further divergences of the Finno-Ugric group leading eventually to smaller groups such as Finnic, Saami, Permian and Ob-Ugric (e.g. Korhonen, 1981). Thus, our results do not support the highly polytomous views of the Uralic language tree (Häkkinen, 1983; Salminen, 1999). However, our phylogenetic hypothesis does not specify a completely clear branching pattern; that is, the ambiguity in the order of the intermediate divergences seen in some previous linguistic studies (e.g. Kulonen, 2002; Michalove, 2002) can also be seen in our results.
Our results suggest that the first divergence of the Uralic language occurred ca. 5300 YBP, which is close to the average calculated from previous linguistic studies (5600 YBP) (Fig. 3; Kettunen & Vaula, 1943; Toivonen, 1953; Décsy, 1965; Korhonen, 1981, 1991; Sammallahti, 1988; Janhunen, 2000, 2009; Kallio, 2006; Häkkinen, 2009). The intermediate level divergences of Finno-Ugric, Finno-Permian and Finno-Volgaic occurred during ca. 3900–2900 YBP, which is largely what previous linguistic studies also suggest (from Finno-Ugric to Finno-Volgaic ca. 4750–3000 YBP) (Fig. 3). Within this period of multiple divergences, our results propose a slightly more recent divergence of Finno-Ugric languages than what is suggested in the earlier studies. Instead, the divergence time estimates proposed in this study and in earlier studies are rather similar for Finno-Permian and Finno-Volgaic protolanguages (Fig. 3). It is noteworthy that our results do not support the entity of Finno-Permian as the posterior probability value for this group is low (0.29; Fig. 2). Similarly, the results do not support an Ugric branch containing Hungarian and Ob-Ugric exclusive of other languages (posterior probability 0.47) (Fig. 3). Due to these low posterior probability values for the Ugric and Finno-Permian protolanguages, a polytomous branching of the Finno-Ugric clade to Hungarian, Ob-Ugric, Permian and Finno-Volgaic can be posited. Similar polytomous branching is also suggested by Syrjänen et al. (2013) based on the same data as here, but analysed with a different algorithm. Various views about the degree of polytomous branching in Uralic languages have been suggested, varying from highly polytomous (Häkkinen, 1983; Salminen, 1999) to strictly binary (Korhonen, 1981) trees, along with different intermediate forms (Kulonen, 2002; Michalove, 2002). This ambiguous nature of the intermediate branchings can also be seen in our results: they show the lowest posterior probability values proposing an uncertain pattern of binary branchings.
Our results suggest a rather ancient divergence for the Ob-Ugric languages (ca. 1900 YBP), as our estimate is close to the upper limit of what previous linguistic studies have estimated, namely between 2000 and 1000 YBP with an average of 1550 YBP (Décsy, 1965; Kálmán, 1988; Honti, 1998; Keresztes, 1998) (Fig. 3). The divergence of Finnic languages into northern (Finnish, Karelian, Veps) and southern (Estonian, Livonian) groups occurred around 1200 YBP according to our results. Our estimate places this divergence some 500 years later than has been obtained on average in earlier research (1700 YBP; Kettunen & Vaula, 1943; Décsy, 1965; Hajdú, 1975; Korhonen, 1981; Janhunen, 2009) (Fig. 3). Similarly, the average divergence time of Skolt Saami from North and Ume Saami is also ca. 500 years more recent in our results (ca. 800 YBP) than suggested by previous scholars (1250 YBP, Korhonen, 1981, 1988; Lehtiranta & Seurujärvi-Kari, 1991) (Fig. 3).
The overlapping HPD values and broad HPD ranges, especially those close to the root of the tree, are partly due to the relatively recent dates of the calibration points used in the analyses. Rather recent divergences were chosen as calibration points as these time estimates were more reliable than the estimates for earlier divergences. Another reason for the overlapping of HPDs lies in the short intermediate branches of the tree (Fig. 2).
Comparisons with abiotic and ‘biotic’ changes
The first divergence of the Uralic language tree occurred ca. 5300 YBP according to our results. Based on earlier divergence time estimates, Carpelan & Parpola (2001) connected Proto-Uralic to the Lyalovo culture (7000–5650 YBP). Our results, however, also allow the possibility that the initial divergence of Proto-Uralic happened during Volosovo culture which followed the Lyalovo culture. At the time of Lyalovo culture period, the climate was relatively warm (Holocene thermal maximum ca. 7500–5000 YBP; Kremenetski et al., 1997; Davis et al., 2003; Väliranta et al., 2003; Heikkilä & Seppä, 2010). A moderate increase in temperature leading to the Holocene thermal maximum may have induced a rise in population size (Tallavaara & Seppä, 2011) and led to migrations when the carrying capacity of the area was approached or reached (Lidicker, 1962; Zubrow, 1971). Thus, it is possible that increasing population size may have been indirectly associated with cultural alterations (from Lyalovo culture to Volosovo) which further could have had an impact on the diversification of Proto-Uralic. This idea of possible migratory movements among the Proto-Uralic speakers is also supported by suggestions that the Volga-Oka area – the area of the earliest ceramic finds of this part of Europe and the hypothesized speaking area of Proto-Uralic (Toivonen, 1953; Salminen, 1999; Häkkinen, 2009) – continuously produced a demographic surplus beginning in the Mesolithic (ca. 10 000 YBP). This caused migrations especially towards the north and the northwest (Carpelan & Parpola, 2001). Thus, it can be suggested that a significant underlying cause for the divergence of Proto-Uralic was an ongoing climate change, which led to migrations due to a high population density, which in turn induced cultural changes.
The divergences of Finno-Ugric, Finno-Permian, Ugric, Finno-Volgaic and Finno-Saami occurred ca. 3900–2500 YBP. Carpelan & Parpola (2001) connect the Proto-Finno-Ugric phase to Volosovo culture (5650–3900 YBP; Fig. 3), but also suggest that during this period a cultural border was formed between the Proto-Permian, Proto-Volgaic and Proto-Ugric linguistic communities. Our results also connect the Proto-Finno-Ugric phase to Volosovo culture but instead of showing further divergences during Volosovo these divergences occur rapidly after the end of Volosovo period (Fig. 3). Therefore, we argue on the basis of our results, suggested also by archaeological evidence, that the Proto-Finno-Ugric started diverging into several different clades during Volosovo period after which several divergences occurred during a rather short period. Moreover, the divergences of Finno-Ugric, Finno-Permian, Ugric, Finno-Volgaic and Finno-Saami coincide with a period of cooling after the Holocene thermal maximum (for the variation in temperature, see Kremenetski et al., 1997; Davis et al., 2003; Väliranta et al., 2003; Heikkilä & Seppä, 2010, Fig. 2). In hunter–gatherer populations, declining temperature can lead to lowered carrying capacity through lower food abundance (Tallavaara & Seppä, 2011), which forces inflicting populations to either gradual extinction or migrations (Lidicker, 1962; Zubrow, 1971). Thus again, the ultimate cause for these language divergences may have been a change in temperature which led to migrations, this time through lowered carrying capacity, which again in turn induced cultural changes. It seems reasonable to conclude that it is the change in temperature which matters, not the direction of the change. The study of causal linkages from changes in climate to changes in population size and migrations through different pathways remains out of scope for this study, but for potential pathways of causal linkages in preindustrial agricultural societies, see Zhang et al. (2011).
According to our results, the diversification of Ob-Ugric into different languages occurred ca. 1900 YBP. The divergence of Samoyed has been dated to ca. 2000 YBP on the basis of alleged contacts with Turkic tribes (Hajdú, 1975; Korhonen & Kulonen, 1991; Janhunen, 1998). As the assumed living areas of the speakers of Proto-Ob-Ugric and Proto-Samoyed were rather close to each other, the migrations of the Turkic tribes affecting the Samoyeds may also have affected the Ob-Ugric speakers, further impacting on the divergence of Khanty and Mansi. These migrations also co-occured with declining temperatures around 2000 YBP (note a difference to Fig. 2 in which the temperature increases around 2000 YBP due to the location on the other side or the Ural Mountains), which changed the vegetation from steppe to forest on the eastern side of the Urals (Zakh et al., 2010), and which most likely had an impact on the local human populations as well. In sum, we consider it possible that the migrations of Turkic tribes affected both Samoyed and Ob-Ugric speakers, while the subsequent migrations of the speakers of Khanty and Mansi could have been intensified by changes in their living environment and its carrying capacity.
Our results suggest that the Finnic languages diverged into northern (Finnish, Karelian, Veps) and southern (Estonian, Livonian) groups around 1200 YBP. It is assumed that the temperature was slightly warmer again after the gradual cooling and the short cold period (Fig. 2). The density of human settlements is known to have increased in all coastal Baltic areas between 1600 and 1200 YBP (Meinander, 2006). This increase in population density was probably induced by the intensification of agriculture around 2500 YBP in the area what is now Latvia (Heikkilä & Seppä, 2010). Correlation between temperature and population size is weaker in agricultural than in hunter–gatherer societies (Tallavaara & Seppä, 2011). Instead, population growth is high soon after transition to agriculture (Bocquet-Appel, 2002; Bandy, 2005), which could explain why the density of human settlements increased despite only a slight rise in temperature. Migrations across the Gulf of Finland had been occurring reciprocally and irregularly for thousands of years (Miettinen, 1996), and it can be suggested that the increase in population density in the southern coast of the Gulf of Finland would have led to one of these migratory waves to the north. In effect, similar archaeological finds from the northern and southern coasts of the Gulf of Finland dated to 2000–1600 YBP indicate gradual northward migrations across the gulf to the current area of Finland (Kivikoski, 1961). This migration period precedes our divergence time estimate for the Finnic languages. Nevertheless, it could have been the starting point for the process of language divergence which was possibly intensified by different economies on the opposite sides of the Gulf of Finland (agriculture in Estonia; slash-and-burn agriculture, fishing and hunting in Finland) and by the political split in early mediaeval times (ca. 900–800 YBP; Carpelan & Parpola, 2001). Thus, we suggest that the division of the Finnic languages into southern and northern groups was at least partly driven by agriculture-induced increase in population density, which led to migration from the southern coast of the Gulf of Finland and was further deepened by nascent cultural differences.
The divergence of Finnish from Karelian and Veps occurred around 800 YBP (Fig. 2). This happened soon after the time when East Slavic tribes arrived in the Baltic area (Bjørnflaten, 2006). These new arrivals introduced many loanwords, especially into the eastern Finnic languages, and brought the Karelians and Vepsians under Eastern Slavic cultural influence and the Greek Catholic Church around 900 YBP (Hajdú, 1975; Laakso, 1991). At that time, the speakers of Finnish on the other hand were brought into close contact with Sweden and the Roman Catholic Church (Hajdú, 1975; Meinander, 2006). Therefore, it can be suggested that this divergence occurred mainly due to cultural reasons.
The influence of Novgorod (the centre of the Eastern Slavic culture) extended also to northern areas (Hajdú, 1975), coinciding with our results on the divergence time of Skolt Saami from North and Ume Saami (ca. 800 YBP). Reindeer herding, which is still practised by the Saami tribes, may have also had an influence on the diversification of the Saami languages. Kortesalmi (2008) proposed that reindeer herding and nomadism existed around the area of Ume Saami already around 1200–800 YBP. From there, it moved towards the north reaching the North Saami area in 400 YBP at the latest (Kortesalmi, 2008) but without ever fully reaching the more Eastern Saami areas (to which Skolt Saami belongs to). As the availability of draught animals facilitates long-distance travel, it is possible that reindeer herding in the western areas strengthened the connection between Ume and North Saami. This economical border separating the areas with and without reindeer herding and most recently the state border in the east may have encouraged closer contacts between Ume and North Saami leaving Skolt Saami to develop more in the eastern sphere of influence. This probably affected the divergence of languages as well.
We show here that both ‘biotic’ and abiotic processes are involved in language diversification and that this may happen primarily through climatic changes triggering secondary forces, such as cultural alterations. The connection between cultural changes (‘biotic’ processes) and the diversification of languages can also be seen in the Indo-European language family where major processes of diversification within the Germanic and Romance subgroups of Indo-European co-occur with the significant cultural changes known as the Migration period (Gray & Atkinson, 2003; Büntgen et al., 2011). These involved, besides migratory movements, the disintegration of Roman administrative structures in many areas (Gray & Atkinson, 2003; Halsall, 2007; Büntgen et al., 2011), and which also interestingly coincided with exceptional variability in climate (Büntgen et al., 2011).