Corps de l’article

In memory of Susanne Mortensen (1956-2006)

Introduction

In a relatively early phase of his long career, the Dutch linguist Christiaan Cornelis Uhlenbeck (1866-1951) became involved in the study of the Eskimo languages. Basing himself on the published descriptive sources available at the time, he wrote a comparative sketch of the inflectional morphology of Eskimo languages and dialects (Uhlenbeck 1907b), as well as several articles concerning the possible long-distance relations between the Eskimo languages and the Uralic, Altaic and Indo-European languages. In the present article, Uhlenbeck’s work on Eskimo will be described and its reception among his contemporaries evaluated. The value of his contributions with respect to the present state of the art will be discussed. It is beyond the scope of this article to systematically evaluate all of Uhlenbeck’s proposed cognates independently from his contemporaries. That would involve a research project requiring expertise both in the field of Eskimo-Aleut linguistics and in Indo-European, Uralic and Altaic historical linguistics. In what follows, I will, after a brief introduction to the Eskimo languages, discuss the four aspects of Uhlenbeck’s involvement with Eskimo: his descriptive work; his contribution to Aleut studies; his comparative Eskimo-Uralic work; and his comparative Eskimo-Indo-European work.

The Eskimo languages

It is presumed that the Inuit have come ultimately from Asia and that their ancestors crossed the Bering Strait somewhere between 6,000 and 8,000 years ago (Damas 1984). Together with the Aleut languages, the Eskimo languages form a genealogical-linguistic stock known as the Eskimo-Aleut language family. The first documented sample of an Eskimo language is the 17-word list of Inuktitut, taken on Baffin Island in 1576 by captain Christopher Hall during Martin Frobisher’s first expedition. Since then, scholars have been wondering where the Inuit and their languages have come from. Over the subsequent centuries, both have been documented and studied extensively and a huge corpus of scientific literature has been built. On the basis of the meticulous study of the Eskimo language in all its varieties, our understanding of both their internal and external historical-genealogical relationships has become very refined. Nowadays the dialectological picture of the Inuit language and its Yupik sister languages has been practically completed and has culminated in the work of modern Eskimologists like Dorais (1990), Fortescue (1983), Fortescue et al. (1994), Jacobson (1998) and others.

The first one to notice the relationship between the markedly distinct Eskimo and Aleut languages was the Danish linguist Rasmus Rask (ms. ± 1820 in Thalbitzer 1916, 1921). The historical linguistic study of Aleut was developed especially by the Norwegian linguist Knut Bergsland (1986, 1989) and nowadays the Eskimo-Aleut linguistic family is an established fact. The question of the relationship between the Eskimo-Aleut and the Old World languages is still not settled, but serious advances have been made especially by Michael Fortescue (1998). Uhlenbeck has played an important, although nowadays somewhat forgotten, role in all these developments, as will be shown in the following sections.

Uhlenbeck’s contributions

In the course of his scholarly life, as sketched in the obituaries by de Josselin de Jong (1952) and Hammerich et al. (1953), Uhlenbeck was an Indo-Europeanist who became deeply involved in the study of the Basque, Amerindian and Eskimo languages, applying the Neogrammarian principles of regular sound change to non-Indo-European languages. Uhlenbeck enriched the study of the Eskimo languages by taking up the issue of their possible relationships to the Uralic and Altaic language families and to the Indo-European languages in a scientific manner. His work on Eskimo can be divided roughly into two periods: the early period, which has seen publications from 1905 to 1907; the late period, with several publications between 1935 and 1950. Except for the odd review (Schmidt 1908), the reception of Uhlenbeck’s work from the early period did not become duly acknowledged until some 20 years later. Only from the mid-1930s, the issues he had raised in the early period started to be discussed to a greater extent by his colleagues, and he participated in the debate with the articles that represent the second period of his work on Eskimo. The fact that Uhlenbeck often wrote in Dutch has probably limited his readership.

Descriptive work on Eskimo

Uhlenbeck’s 1907b comparative morphological sketch of the Eskimo languages is a 76-page monograph that gives an overview of the inflectional and pronominal forms, and the derivational elements of four Eskimo languages, including three Inuit dialects. This sketch represents one of his first publications on Eskimo. Based entirely on published sources by the likes of Kleinschmidt (1851) and others, it forms the descriptive foundation of his comparative work. It contains an extensive coverage of the morphology of the West Greenlandic (Groenlandsch) and Labrador dialects of the Inuit language and the Central Alaskan Yupik language (Alaskisch). The data it contains on the Inuit dialect of the Mackenzie Delta, the Pacific Yupik language of the Alutiiq/Sugpiaq (Kadjakisch)[1] and the Central Siberian Yupik language (Namollo/Eskimo-Tsjoektsjisch)[2] are relatively sparse, due to the lack of comprehensive or reliable sources (Uhlenbeck 1907b: 17, 25).

In the foreword to his sketch of Eskimo morphology, Uhlenbeck explained his non-inclusion of Aleut by the difficulty in establishing its precise position with respect to the Eskimo languages. He nevertheless assumed that there was a genealogical relationship: he did mention similarities between Eskimo and Aleut forms occasionally (Uhlenbeck 1907b: 12, 13, 18, 42, 45, 60 and elsewhere). On certain occasions throughout his sketch, Uhlenbeck reconstructs forms of the proto-language (Oer-Eskimo) on the basis of comparison of Eastern and Western Eskimo forms (e.g., Greenlandic and Central Alaskan Yupik in Uhlenbeck 1907b: 21).

Language-internally, Uhlenbeck (1907b: 18, 54) notes the formal similarity between possessive suffixes on the noun and argument cross-reference suffixes on the verb. He furthermore observes the alternative third person possessive forms and distinguishes them as α-forms (non-reflexive possessive) and β-forms (reflexive possessive). The reflexive forms, as well as the homophonous verbal same-subject forms in subordinate clauses (ibid.: 71-73), have in the Eskimological literature often been called fourth person forms. Finally, Uhlenbeck (1907b:18) regards the personal pronouns as demonstrative and possessive in nature, on the basis of formal characteristics, e.g., uanga ‘I’ is interpreted as ua-nga ‘my here-ness’ because of the similarity with the Mackenzie Eskimo first person possessive suffix -nga. That possessive reading was justly criticised in a review by Schmidt (1907-1908: 334), since the same person suffixes also occur in the nominal participial mood, e.g., tusartunga ‘me hearing’[3]. From a pragmatic viewpoint, it is not so strange to draw a parallel between Eskimo pronouns and demonstratives. As Eskimo is a language with anaphoric person inflection (cross-reference morphology), pronouns are not obligatory in the construction of a grammatical sentence, which makes them available for emphatic purposes. The involvement of a demonstrative root in the etymology of the first person pronoun (but not of the second person) is also assumed in recent comparative work (Fortescue et al. 1994: 383).

In general, throughout his work, Uhlenbeck has wrestled with the concept of ergativity, which he sometimes called the casus transitivus, but for which there was no generally accepted term at that time[4]. In Eskimo the picture is further complicated by the fact that the ergative case marker of the transitive subject is homophonous with the possessor marker in possessive constructions. In his 1936 article, he stated that:

The central enigma of the grammar of Eskimo is the casus transitivus, which adverbally indicates the logical subject of the transitive verb and adnominally has more or less the function of a genitive.

Uhlenbeck 1936: 228, my translation

Uhlenbeck subsequently speculates that understanding the “transitive case” in Eskimo might also cast a light upon ancient Indo-European phenomena[5]. He notes furthermore that many other languages have a transitive case too, without, however, a genitive function. Therefore, Uhlenbeck (1936: 228) suggests that the ergative case in Eskimo would require a more specific term. In this respect he rejects the Danish Eskimologist William Thalbitzer’s term “relative case,” which is nowadays used by many Eskimologists to refer to the ergative case.

Uhlenbeck was not the only linguist to wrestle with the ergative/relative case in Eskimo. Louis Hammerich, a Danish Indo-Europeanist who knew Dutch and who became interested in the Eskimo languages, begins his monograph (1936) on Eskimo person marking with a reference to Uhlenbeck’s (1907b: 18) observation about the homophony of nominal and verbal inflection in Eskimo. This, he says, had inspired him to write his book, and he commends Uhlenbeck for having recognised the true nature of the casus transitivus, although he adds that Uhlenbeck did not get completely to the bottom of it (Hammerich 1936: 11). In his reviews, Uhlenbeck (1937c, 1938) neither rejects nor accepts Hammerich’s critical remark, but only criticises the use of certain terminology (among which the term “fourth person”), and concludes that we are, with Hammerich’s “excellent” book, still “standing at the beginning, not at the end” (Uhlenbeck 1937c: 687). Uhlenbeck was quite correct in stating the latter. Both Thalbitzer and Hammerich, as well as other Eskimologists of the 20th century, have claimed—on the basis of the aforementioned identity in form between possessive and cross-reference marking—that in the Eskimo languages verbs form a semantic subclass of nouns (Hammerich 1936: 220). Some have even attributed this alleged nominal nature of Eskimo verbs to a supposedly more primitive state of the Inuit mind. Recently, this myth was mercilessly taken apart in a splendid article by Jerrold Sadock (1999). Uhlenbeck does not seem to have fallen into the same trap, although his remark that Hammerich’s “nominalistic explanations in the long run may have to be revised in a verbalistic manner” (Uhlenbeck 1937c: 686, and similar remarks elsewhere, e.g., 1936: 227) indicate that he was about to fall in the opposite trap of considering Eskimo nouns as a subclass of verbs.

Comparative perspectives on Aleut

Uhlenbeck had a reputation to lose in comparative Indo-European philology. Even though his research agenda on the possible long-distance relationships of Eskimo was indeed “courageous” (Hammerich 1936: 222), his observations show he always practiced good care in following scientific methods and principles, however exotic the languages he worked on. This is usually acknowledged by his contemporaries (e.g., Thalbitzer 1945: 68). In his first observations about the Aleut language, Uhlenbeck (1906b: 114) notes that it must be related genealogically to Eskimo, but at the same time, being so different from the Eskimo languages, it cannot be considered to be one of them. He concludes that Aleut has not developed from Proto-Eskimo, but that the relationship could be summarised under the denominator “Eskimo-Aleut linguistic stock,” He furthermore observes that the reconstruction of the Proto-Eskimo-Aleut language is a prerequisite for a thorough evaluation of the relationship with the Uralic languages. In the foreword to his comparative sketch of Eskimo, Uhlenbeck (1907b: 3) tries to qualify his previous conclusion about Aleut, and estimates that upon closer investigation, Aleut may turn out to be a highly deviant “dialect” of Western Eskimo. He nevertheless notes that the similarities between Western and Eastern Eskimo are stronger than between Western Eskimo and Aleut.

Uhlenbeck was not the first to speculate about the relationship between Aleut and Eskimo. A century earlier, the Danish linguist Rasmus Rask had already compared Greenlandic and Aleut in a manuscript that was made accessible by Thalbitzer only in 1916. In his articles about this manuscript, Thalbitzer (1916, 1921) also spoke admiringly of Uhlenbeck’s comparative work on Eskimo and cited Uhlenbeck’s above-mentioned remarks on Aleut. Conclusive evidence of a genealogical relationship between Aleut and Eskimo, including phonological and morphological correspondences, was put forward only later by Bergsland (1951), who cites Uhlenbeck a number of times, using his observations. In subsequent work, Bergsland (1986) reconstructed what must have been a “catastrophic” breakdown of Aleut grammar, ultimately caused by a phonological change, which had made it appear so different from Eskimo. Nowadays it is estimated that Aleut and Eskimo began to diverge about 4,000 years ago (Fortescue 1998: 36).

Comparison of Eskimo with Uralic, Altaic and Paleo-Siberian

Already in 1576, on the occasion of what was one of the first documented encounters between Inuit and Europeans, a relationship was presumed to exist between the North American Inuit and the inhabitants of Siberia. During Martin Frobisher’s expedition to find the Northwest Passage to China, sailing master Christopher Hall thought that the Inuit resembled “Tartars” because of their physiognomy (Oswalt 1979: 26). The first one to characterise the Eskimo language as similar to “Tatar” was the German scientist and traveller Adam Olearius (1656: 170), on the basis of a word list taken from Inuit kidnapped on a Danish expedition to Greenland (Thalbitzer 1932: 12). During the following centuries, linguists noticed similarities between Eskimo and the Uralic and Altaic languages, which are spoken in Asia and Europe (e.g., Petitot 1876: lxiv; Wöldike 1746), but the first one to approach the question of a possible linguistic relationship between Eskimo and the Old World languages in a systematic manner was Uhlenbeck (Bergsland 1959: 7; Dorais 1993: 17).

In his first publication concerning Eskimo, Uhlenbeck (1905) discusses the similarities between Eskimo and Uralic and Altaic languages. He observes that the Eskimo languages are similar to those languages because of their almost exclusively suffixing morphology, in which respect they differ from most of the Amerindian languages of North America, only few of which are predominantly suffixing or prefixing and most do both. Not content with a superficial typological impression, Uhlenbeck goes on to review the concrete formal similarities he encountered. As a phonetic parallel between Eskimo and Finno-Ugric, he mentions the lack of original voiced stops (mediae) b, d, g and the presence of voiceless stops (tenues) p, t, k. He furthermore discusses a number of rather striking morphological similarities. Both the form and the function of the Eskimo dual and plural number morphemes, -k and -t respectively, are also attested in Finno-Ugric languages: Greenlandic igdlo[6] ‘house,’ igdluk ‘two houses’ and igdlut ‘houses’ versus Uralic Vogul (Mansi) kval ‘house,’ kvalag ‘two houses’ and kvalet ‘houses.’ The element -t as a plural marker is attested also in Finnish, in Altaic languages and in Chukotko-Kamchatkan Chukchi. In a number of Finno-Ugric languages the plural -t was preserved, while the dual category was lost together with its form. In other Finno-Ugric languages the dual category was lost, while the dual form took over the function of the plural, such as the Hungarian plural marker -k. Note that nominal number is expressed in Eskimo by the same form as person number in cross-reference marking on verbs, and in this respect it is also similar to Hungarian.

Other striking parallels are the Eskimo locative case inflection -ne versus -na in Finno-Ugric languages and -nan in Samoyedic languages; the Eskimo prosecutive case inflection -kut vs. Samoyedic -ut or -un(a); the Eskimo demonstrative prefix ta- vs. Finno-Ugric, Samoyedic and Altaic t-; the Eskimo interrogative element ki- versus Hungarian ki and in other Finno-Ugric languages ku- and similar forms. Even though there are no apparent parallels between personal pronouns in Eskimo and the Old World Arctic languages, there is a stong similarity between Eskimo first person possessive ergative -ma and Uralic and Altaic first person possessive and cross-reference -ma, -mo, -m. Uhlenbeck also mentions Eskimo second person singular possessive and verbal cross-reference -t, which is similar to Finno-Ugric second person possessive and cross-reference -t and -d, as in e.g., Greenlandic kivfat ‘your servant,’ toqupat ‘you killed him’ versus Hungarian házad ‘your house,’ várod ‘you’re waiting (for him)’ (Uhlenbeck does not provide any segmentation of the forms).

In the same article, Uhlenbeck makes some observations on formal variation in person marking between Eskimo dialects, and between Eskimo and Aleut, with regard to which he draws parallels to Altaic languages. For example, the second person singular suffix (both as nominal possessive inflection and as verbal cross-reference marking) is in most dialects of Eskimo -t, which has become -n in Mackenzie Eskimo and Aleut, similar to the Turkish second person singular possessive suffix -ṅ. Uhlenbeck concludes that these and many other inflectional similarities between Eskimo and Asiatic languages cannot be coincidental. He adds that they call for an explanation in terms of a common origin, which is all the more significant in view of the fact that hardly anything in Eskimo grammar could justify any relationship with the Amerindian languages (Uhlenbeck 1905: 764). Nevertheless, Uhlenbeck cautions the reader not to jump to conclusions. He emphasises that the perceived similarities suggest the possibility of a linguistic relationship, but that certainty can be attained only through thorough comparison of sounds and vocabulary (ibid.: 758).

In subsequent articles, Uhlenbeck (1906b, 1907a) makes notice of several other similarities with Uralic languages, such as the Eskimo third person plural possessive ending -i, which functions as a general possessive pluraliser in Hungarian, e.g., WG[7]igdlue ‘his houses’ versus Hungarian házai ‘his houses.’ Another example is the Eskimo suffix -lik ‘provided with,’ as in sákulik ‘armed,’ which resembles the Turkish abstract noun suffix -lik, as in dewelik ‘camel stables,’ the meaning of which Uhlenbeck suggests may have developed from an original possessive meaning. Uhlenbeck (1907a) also made a number of etymological observations that point to an Eskimo-Indo-European connection.

In his later work Uhlenbeck insists that the explanation of the similarities as due either to a single genealogical origin or to borrowing, is too simplistic, and that they more likely are the result of prolonged prehistoric periods of contact between the respective linguistic groups. According to his new insights first laid down in Uhlenbeck (1935a), the Proto-Indo-European (Oer-Indogermaansch) language[8] was really a “mixture” that consisted of two important strata. These strata, or layers, which he termed A and B, predate the well-known division between centum and satem languages (which represents an isogloss, meaning ‘hundred,’ that divides the Indo-European languages according to a combination of specific sound changes). Layer A consists of pronominal elements, verbal roots and their derivations. Layer B includes the numbers, kinship terms, many body-part names, animal and tree names, and many other non-verbal words. The language that Uhlenbeck presumes to lie at the foundation of layer A must have been grammatically “regular,” in a way similar to Uralic and Altaic (ibid.: 135), and he also shows affinities with Eskimo (Uhlenbeck 1935b: 180). He assumes that with regard to the preservation of palatal stops, layer A coincides with the centum languages and the Uralic languages (Uhlenbeck 1935a: 146). The origin of layer B must have been a grammatically irregular language of which Uhlenbeck is not able to provide a modern typological and possibly historically related parallel. It is in this context that Uhlenbeck falls back on his earlier observations on lexical similarities between Eskimo and Indo-European and remarks:

At that time, I did not yet know what to do with my findings, since in those days I was still too much occupied by the old, neogrammarian impression that “genetic relationship” and “borrowing” were completely distinct concepts, and I did not yet suspect to what extent mixing and crossbreeding dominate the history of languages as well as peoples and races.

Uhlenbeck 1935a: 142, my translation

He attributes his broader view, which is also aired in Uhlenbeck (1927, 1936, 1937a, 1937b, 1938, 1941 and elsewhere), partially to the influence of Hugo Schuchardt and Franz Boas, and blames his former neglect of Eskimo-Indo-European comparison to his fascination for the much clearer structural similarities between Eskimo and Uralic. This view explains Uhlenbeck’s suspicion (as hinted at elsewhere, e.g., Uhlenbeck 1906a: 289[9], 1907a: 436, 1935a: 144) that the relationship between Eskimo and Indo-European must be indirect, through Uralic and Altaic.

Only in his 1941 article Oude Aziatische contacten van het Eskimo (‘Old Asiatic contacts of Eskimo’), does Uhlenbeck return to concrete examples of parallels between Eskimo and Uralic, even though the main focus is on Eskimo-Indo-European relationships. On pages 219-222 he presents 15 lexical similarities with Uralic and Altaic languages[10], such as WG atéq ‘name’ versus Tatar āt ‘name,’ Yakutian aata ‘name’; WG imeq ‘water’ versus Chukchi i̯ml ‘water,’ Koryak i̯ml̥ ‘water’; WG kamik ‘boot’ versus Saami (“Lapsch”) kāma ‘boot, shoe’; WG mánik ‘egg’ versus Finnish muna ‘egg,’ Tatar meni ‘seed’; WG niviaq ‘girl’ versus Chukchi ñéwän ‘wife,’ Samoyed nevea ‘woman.’ Subsequently, Uhlenbeck (1941: 24) discusses morphological similarities between Eskimo and some Paleo-Siberian languages (Tsjuktsjisch-Korjakisch-Kamtsjadaalsch ‘Chukchi-Koryak-Itelmen’) and sometimes between Eskimo, Finno-Ugrian and Uralic. He furthermore notices typological similarities with Paleo-Siberian, such as ergative (“the passive interpretation of the transitive verb”), but remarks that this phenomenon is also attested elsewhere in the world (ibid.: 25). As a significant typological difference with Paleo-Siberian (and with Algonquian), he notes the absence of composition in Eskimo (ibid.: 26).

Uhlenbeck’s sensational observations about Eskimo-Indo-European attracted much attention (e.g., Hammerich 1936, Thalbitzer 1945) and, consequently, his Eskimo-Uralic-Altaic observations were not so extensively discussed at the time. After Uhlenbeck’s death, Sauvageot (1953) and Bergsland (1959) developed the Eskimo-Uralic hypothesis further, often referring to Rask and Uhlenbeck. At present better data have become available on other Paleo-Siberian languages besides Eskimo, such as Yukaghir, Nivkh, Koryak, etc. as well as on many Uralic and Altaic languages. This has enabled better reconstructions of established language families, which also improved the basis for long-range comparisons.

One of the recent proposals is the Uralo-Siberian hypothesis (Fortescue 1998). This proposal includes the Uralic, Yukagir, Chukotko-Kamchatkan and the Eskimo-Aleut languages into one language “stock,” or rather, “mesh,” that is, a group of historically connected languages that are either related genealogically or through areal diffusion, which cannot be traced back to a unitary proto-language. Fortescue writes:

[Uhlenbeck] assumed that Uralic and Altaic were related, but in doing so he attempted to distance himself from the traditional genetic versus borrowing dichotomy, envisaging rather an areal model of deep affinity between all these languages as the result of successive episodes of mixing, in a manner more reminiscent of the position expressed in the present book.

Fortescue 1998: 47-48

Lately, Seefloth (2000) has expanded on certain aspects of the Uralo-Siberian hypothesis. Some recent developments in historical linguistics suggest that the Uralo-Siberian stock includes an Indo-Uralic branch that unites the Indo-European-Hittite and Uralic languages in one sub-stock (e.g., Greenberg 2000, Kortlandt 2002). Whatever concept of “proto-language” this view entails, it is a possible explanation for the parallels Uhlenbeck encountered between both Eskimo and Uralic, and Eskimo and Indo-European.

Comparison of Eskimo and Indo-European

Uhlenbeck’s (1907a: 436) first observations about a possible relationship between Eskimo-Aleut and Indo-European (Indogermanisch) are found in the second part of his article Zur Eskimogrammatik. In his discussion of the Eskimo element -na-, which forms part of negative words as (WG) nāgga ‘no,’ he notes that similarity with IE *ne is probably no coincidence, since there are indications that both Eskimo and IE seem to be related to the Uralic languages. In the rest of the paragraph he lists 20 lexical and two pronominal similarities between Eskimo and IE, alphabetically ordered. These early publications on Eskimo were hardly discussed by Uhlenbeck’s contemporaries at the time, whilst he directed his attention to Basque and Amerindian studies.

Apart from sparse observations on Eskimo in the margin of other issues (e.g., Uhlenbeck 1923, 1927), it was not until almost three decades later that Uhlenbeck (1935a) returned to publishing on Indo-European and the possible connection between Eskimo and Uralic. The new ideas about the nature of language relationships that he had developed were discussed in the previous section. In his subsequent article (Uhlenbeck 1935b), which he considers as an appendix to Uhlenbeck (1935a), he presents a revised and much extended overview of formal similarities between Eskimo and Indo-European. He lists 94 lexical similarities which he explains not by a genealogical “proto-relationship” between Eskimo and Indo-European, but by ancient cultural contacts and “language mixing” (Uhlenbeck 1935b: 179-180). In the article he disregards possible grammatical similarities, because those can only be dealt with in relation to the Eskimo-Uralic and Indo-European-Uralic issues. He assumes that his comparison of Eskimo and Indo-European mainly involves the A-complex[11], but does not exclude the possibility that the affinities emerged after the amalgamation of A and B. Before setting off on his 15-page list of (non-accidental) lexical similarities, he notes that these are even more significant in the light of the lack of such similarities between Eskimo and adjacent language families in North America.

Uhlenbeck’s observations were well received by a number of colleagues (e.g., Hammerich 1936, Thalbitzer 1952), although he was not always understood in the way he wanted. On the last page of his treatise on Eskimo inflection, Hammerich (1936: 222) mentions Uhlenbeck’s (1935b) recent “courageous attempt to directly compare Eskimo and Indo-European words,” but mentions that Eskimo-Finno-Ugric relationships might also explain the similarities, and concludes that morphosyntactic comparison should precede lexical comparison. In his 1941 article, Uhlenbeck (1941: 204) rejects Hammerich’s latter criticism. In the same article and in his 1938 review of Hammerich’s book, Uhlenbeck (1938: 18) responds that according to his more recent ideas, the observed similarities between the Eskimo and Old World languages might be explained by “language mixing and acculturation of, let us say, 5,000 years ago.” To this he adds that if Holger Pedersen is right about “Nostratic,” Eskimo and the Uralic and Altaic languages will have to be included in that stock. Uhlenbeck (1938: 18) concludes that with the concept of taalbonden (i.e.Sprachbünde ‘linguistic areas’), “Troubetzkoy and Jakobson have opened broad perspectives.”

At the meeting of the Linguistic Circle of Copenhagen of 25 November 1937, Thalbitzer and the Danish general linguist Louis Hjelmslev discussed Uhlenbeck’s hypotheses. In Hammerich’s (1939) report of this meeting, Thalbitzer mentions both Uhlenbeck’s earlier and newest work, and criticises some of the similarities as “hazardous” because they are based on weak Eskimo data or lack etymological analysis. He admits, however, that many similarities are quite striking and that Uhlenbeck does not claim to have proven any genealogical relationship. In the same report (Hammerich 1939), Hjelmslev focusses on Uhlenbeck’s (1935a) above-mentioned hypothesis on the two strata in Indo-European. Hjelmslev states that it is hardly possible to make a distinction between productive roots and non-productive roots in Proto-Indo-European, and that the distinction merely holds in the historic languages, the data from which have contributed to the reconstruction of the proto-language. Hjelmslev correctly understands Uhlenbeck’s notion of “mixed language” not as a creole language, but as a language “of the same type as English or Albanian, for example,” suggesting a view in terms of areal linguistics. However, Hjelmslev does not accept this notion as an alternative to the classic genealogy-versus-borrowing dichotomy. Hjelmslev can imagine non-genealogical similarities only to be borrowings, even if that cannot be proven.

It is a bit surprising that the German linguist Hans Jensen’s (1936) rather critical review of Uhlenbeck’s Eskimo-Old World hypotheses only concerns his early articles (Uhlenbeck 1905-1907a). Jensen comments on Uhlenbeck’s early 22 similarities one by one, dismissing a number of them dogmatically and without further explanation as “impossible on grounds of specific sound laws,” noting at the end that there can be no relationship between Greenlandic and Indo-European in the “scientific sense” (i.e. a genealogical linguistic relationship), not even when it is claimed by “an authority like Uhlenbeck.” Jensen (1936: 158) concludes indirectly that the similarities are to be considered as worthless. Maybe the review was an old idea that Jensen had saved for a Festschrift without doing much further research (even though it does contain some valuable criticism as well). Uhlenbeck, however, responded quickly, also in a Festschrift (1937a), referring subtly to Jensen’s conclusion by the title Über den Wert eskimoisch-indogermanischer Wortähnlichkeiten (‘About the value of Eskimo-Indo-European lexical similarities’). Uhlenbeck explains that Jensen was misguided in assuming that he would make the mistake of trying to relate Greenlandic directly to Indo-European. He then explains at some length his more recent ideas about long-distance language relationships and the role of language contact. Jensen might not have had the chance to read Uhlenbeck (1935b), but he—had he bothered—could have been familiar with Uhlenbeck (1927: 231) who quotes himself: “I strongly suspect that all larger linguistic families are, after all, products of assimilation and acculturation with subsequent differentiation.” Uhlenbeck ends by saying that the issue of batches of systematic Eskimo-Indo-European lexical similarities should not be wiped off the agenda just like that.

Uhlenbeck’s 1941 article is partially a reaction to earlier reviews of his work, partially a revision of it and partially an extension. After another swipe at Jensen, Uhlenbeck (1941: 211), expresses his satisfaction with Hammerich’s (1939: 9) report of the above-mentioned discussion at the Linguistic Circle of Copenhagen, in which he was called upon to flesh out his comparative Eskimo research. On the subsequent pages Uhlenbeck (1941: 212-219) discusses 26 Eskimo-Indo-European similarities, the majority of which he had proposed previously, defending some, rejecting others. On the following pages (ibid.: 219-222) he discusses 15 Eskimo-Uralic similarities, and in the final pages he evaluates the consequences of the data for the reconstruction of the migration history of the Inuit.

In the introduction to his 1945 article, Uhlenbeck explains that the idea of a distant relationship between Eskimo and Indo-European (as put forward in Uhlenbeck 1907a) had changed fundamentally because of a new understanding of language relationships (as represented in Uhlenbeck 1935b). The new article is an extensive revision of the 1935b and 1941 articles, on the basis of important feedback from colleagues. It contains a list of 101 lexical similarities between Eskimo and Indo-European. A number of entries are of sound symbolic nature, such as WG pátagpā ‘slaps it with the hand’ versus Gr. πατάσσω ‘I hit, making noise,’ or represent words originating in children’s babbling, such as WG amāma ‘breast’ versus Vulg.Lat. mamma ‘breast’; WG mamik ‘edible underside of skin’ versus Skr. māṁsá- ‘meat’; WG atāta ‘father’ versus Lat. atta ‘father’[12]. Uhlenbeck (1945: 135) is aware of the weakness of such similarities as evidence for linguistic relationship, but he includes them anyway, reasoning that the similarities cannot be accidental since he found “a whole series” of them. Thalbitzer (1945: 79) agrees with him in this respect. The majority of Uhlenbeck’s forms, however, are non-iconic, and have a higher potential as evidence. Uhlenbeck (1945:135) is always very careful in his motivation to include a specific entry, and sometimes lists a highly uncertain form because “one cannot know what prospects future research may attain.” A good number of forms show striking similarities, such as WG avdla ‘other’ versus Lat. alius ‘another’; WG qapuk ‘foam’ versus Skr. kapha- ‘slime,’ Persian kaf ‘foam’; WG nu- in words like nutāq ‘new,’ nukaq ‘younger sibling,’ etc. versus Skr. náva-, Lat, novus ‘new’ and Skr. nu, Gr., νύ ‘now’; Asiatic Eskimo kangkole ‘lobster’ versus Lat. cancer, IE *kaṅkro- ‘lobster.’ There are also many forms that are rather doubtful as evidence. Although Uhlenbeck did understand the grammatical structures of Eskimo very well, he was not so well-acquainted with Eskimo etymology and with the lexicalised forms that would be transparent to a trained Eskimologist. Therefore Uhlenbeck’s entries often contain lexicalised morphological structures or disregarded morphological boundaries in complex Eskimo forms that are compared to Indo-European as if they had been monomorphemic.

Uhlenbeck’s 1945 article was already in press when he got hold of Thalbitzer’s 1945 article which contained an extensive detailed discussion of Uhlenbeck’s ideas and discoveries up till 1941. Thalbitzer points out a number of weaknesses, such as the disregarded internal structure of Eskimo forms. A telling example is the semantically empty root pi- (before uvular consonants [pe]) in the Eskimo languages, which is sometimes more or less translatable as ‘thing.’ This root serves as a formative that enables bound derivational morphemes (which often have a lexical meaning) to occur as independent nouns or as verb stems, as in WG: pe-qar-poq, lit. thing-have-(s)he.indicative, ‘(s)he has (something)’; pi-ler-poq, lit. thing-begin-(s)he.indicative, ‘(s)he begins,’ etc. There are also many lexicalised combinations with pi- in Eskimo. Even though Uhlenbeck knew this root and already listed it among the 20 potential cognates with Indo-European in his 1907 article, he first began to recognise its true nature in his 1945 article. And even then he continued to list three forms independently, in which he did not identify the empty root: WG pernarpoq ‘(s)he does it for the first time,’[13] which he contrasts with Lith. pérnai ‘in the previous year’; WG perquvā ‘allows, offers, orders,’ contrasted with IE *perx-, *prex- ‘to ask, request,’ Lat. precor ‘to request’ and WG persoq ‘snow storm,’ contrasted with Skr. pṛṣat- ‘drop’ or pṛçni- ‘spotted’ etc. Thalbitzer (1945: 74-75) justly criticises these proposals for potential cognacy, because the first pair involves the (in WG lexicalised) suffix -rnar- ‘do for the first time’ and the second example involves the productive suffix -rqu- ‘requests, offers,’ along with the semantically empty root pi-. Another example of wrong analysis criticised by Thalbitzer (1945: 70) is from Uhlenbeck’s 1935b article. In the pair WG piuaipoq[14] ‘(s)he is mild, gentle’ versus Lat. pius ‘respectful, mild,’ the Eskimo form contains a negative suffix -it- and means literally ‘(s)he is not treacherous.’ Without this suffix the word, rendered as piuarpoq, means ‘(s)he treacherously steals upon (someone),’ which again is derived from piorpoq ‘(s)he advances quickly (upon game).’ So the meaning of Uhlenbeck’s WG root is actually the opposite of the formally similar Latin one. If these Eskimo and Indo-European forms are really related (either genealogically or through contact), they must have originated from Eskimo, where they are transparent, rather than from Indo-European, where they are not.

Thalbitzer’s article represents the only extensive response to Uhlenbeck’s claims from an Eskimologist’s perspective. Even though not all of Thalbitzers comments and claims are tenable, it contains a relatively thorough evaluation of the similarities noted by Uhlenbeck, and also adds some interesting new ones. The four page appendix includes a survey of all of Uhlenbeck’s parallels from the 1935b and 1941 articles. It was not until 1950 that a response from Uhlenbeck appeared. In this brief article he acknowledges Thalbitzer’s 1945 review, and presents a discussion of 35 Eskimo-Indo-European similarities, sometimes countering Thalbitzer’s arguments, sometimes accepting them. Some of the entries are new. Apart from two subsequent reviews concerning the work of others, this is Uhlenbeck’s last genuine article.

After Uhlenbeck’s death, his work on Eskimo-Indo-European was briefly evaluated by Hammerich (1951) and Thalbitzer (1952). Hammerich gives Uhlenbeck the credit of having raised the Eskimo-Indo-European question and indicates that Thalbitzer and Hammerich himself are favourable with respect to the idea of a relationship, without wanting to define the nature of that relationship. After some interesting observations about the different organisation of Eskimo and Indo-European semantics, Hammerich sets off on an excursion of his own ideas involving case marking and the nominalist hypothesis discussed above. He concludes that more research is necessary before the question as to whether Eskimo is related to Indo-European can be answered.

Thalbitzer (1952) acknowledges a number of striking parallels between Eskimo and Indo-European that were discovered by Uhlenbeck and that may represent possible cognates, such as WG anerpoq ‘(s)he breathes’ versus Skr. ániti ‘breathes’; WG anori ‘wind’ versus Skr. ánila ‘wind’; WG auṷssaq ‘spring, summer’ versus Lat. ver ‘spring’; WG milugpoq ‘(s)he sucks’ versus IE *miluks, Lat. mulgeo ‘I milk’; WG tikeq ‘index finger’ versus IE *deik-, Gr. δeik ‘points at’; WG tutivā ‘(s)he treads on it’ versus IE *(s)teud-, Lat. tundo ‘knock, tread’; WG tunivā ‘(s)he gives it’ versus IE *(s)teup-, Lat. do ‘give’ (Thalbitzer 1952: 53-54). He commends Uhlenbeck for being careful not to jump to conclusions with regard to cognacy or the direction of borrowing.

Uhlenbeck (1937a: 114), however, always remained resourceful in defending the possibility of a relationship, if only in order to “prevent that the […] question regarding ancient relationships between Eskimo and Indo-European be taken off the agenda just like that.” A case in point is the parallel WG ingneq, Asiatic Eskimo eknek[15] ‘fire’ versus Skr. agní-, Lat. ignis ‘fire.’ The etymology of the Eskimo word, however is based on iki- ‘to ignite,’ combined with the (here lexicalised) nominaliser -neq. Uhlenbeck was aware of the transparency of the Eskimo form, and wonders whether the Proto-Indo-European form might have originated from Eskimo or whether it might have had a similar root like eg-. He concludes:

To exclude the similarity ingneq : agní- already in this stage from the Eskimo-Indo-European data, which urgently require serious research, as a case of mere coincidence[16] would not be sensible scepticism, but irresponsible levity.

Uhlenbeck 1945: 138, my translation

After all, a number of striking lexical similarities remain. However, their number is rather modest and not accompanied by strong typological linguistic similarities or by archaeological and genetic data. What then is the significance of these similarities? Are they traces of an old prehistoric relationship? We do not know the answer to these questions. Unlike the Eskimo-Uralic hypothesis, the Eskimo-Indo-European hypothesis went largely out of fashion in the second half of the 20th century. The new developments mentioned at the end of the previous section, however, may in the end confirm Uhlenbeck’s (1935a: 144) suggestion, that a relationship exists through a Uralic intermediary.

Conclusion

Uhlenbeck’s early period of involvement in Eskimo linguistics resulted in a comparative morphological sketch of a number of Eastern and Western Eskimo languages and dialects (1907b), and in three seminal articles on the possible relationships between Eskimo and Uralic languages (1905, 1906b, 1907a) and Indo-European languages (1907a). Throughout his later work on Eskimo, (e.g., 1935, 1941), Uhlenbeck has emphasised that the Neogrammarian dichotomy between “original genetic relationship” and “borrowing” should be put into perspective. It seems unlikely that correspondences between the Eskimo-Aleut, Uralic, Altaic and Indo-European language families should ultimately be traced back to a single proto-language. Of course it is possible that similarities between and within language families reflect a common proto-language which did exist in the prehistoric past and which cannot be proven. It is also possible, however, that such similarities have arisen through an areal-linguistic complex of prolonged periods of contact, assimilation, secondary diversification and re-grouping, which Uhlenbeck regards as a likely scenario for the similarities between Eskimo-Aleut, Uralic, Altaic and Indo-European language families. In this respect, Uhlenbeck’s ideas form a harbinger to modern perspectives in the comparative study of Old and New World Arctic languages, the relationships between which have lately been viewed as a combination of areal and genealogical features, characterised expressively as a “mesh” by Fortescue (1998, 2003).