Table 2. UT phonemic inventory a. Bergsland (1994:xvii), with minor simplifications, e.g. dialectal elaborations are not noted. Under- lined phonemes are found in Russian or English loanwords; orthographic conventions are given in brackets. Table 3. PE and UT regular sound correspondences Nevertheless, there are still many questions and uncertainties in the recon- struction of PEA, some of which have been explicitly acknowledged and others of which have not. The following descriptions, given as (i)-(iv), give overviews of fable 4. Proposed distribution of PEA, PE, and PUT fricatives A detailed examination and exp lanation of the less well-understood aspects. historical reconstruction in EA is beyond the scope of this work, as are studies | other aspects of historical reconstruction, including reconstruction of the gran mar (Bergsland 1989, 1997b; Fortescue 1998, 2002, inter alia, Berge 2016b); so tc is searching for connections between Eskimo-Aleut and other language familie most famously Indo-European (Uhlenbeck 1935), Wakashan (Swadesh 1962; Hol 2004; Fortescue forthcoming), Chukotko-Kamchatkan (Swadesh 1962; Fortesct 1998) and Uralo-Altaic (Fortescue 1998). In the discussion that follows, I focus ¢ the problems of assumed cognates, here regarding sound changes and t and do not focus on the issues I have rais« he reconstruction of PEA. The high degree of lexical divergence in UT paired with the low level of bo rowings from other languages has ancient split. But findings in other been one of the main reasons for positing < fields suggest a more recent EA split; lexic studies suggest more borrowing than previously assumed (Berge 2016a, 201 forthcoming), and some words thought to be good cognates refer to technol gies that postdate the EA split and therefore cannot have been directly inherited ilictrate thece nninte with twa evamnlec While percentages for groups 2-3 are unsurprising, the percentage of ques- tionable ‘exact cognates’ is noteworthy, as many of these cognates have been con- sidered among the most illustrative of an old EA relationship. Cognates with no phonological or semantic differentiation seem improbable over the course of 3000-4000 years. The lack of phonological variation is striking when we con- sider the typologically unusual nature of the “regular” sound correspondences as against the otherwise complex morphophonological rules in both UT and Eskimo languages resulting from their extreme polysynthesis. Lack of semantic variation is especially noteworthy given the different geographical and climatic environ- ments of the different groups, the different technological advances, and the differ- ent experiences the respective groups were subject to during this extended period of time. In fact, this category is complex and probably includes both inherited and borrowed words. Thus, despite a lack of phonological variation, the semantic vari- ation in group 4.1.1 suggests long independent development, as does the frequency of use and commonality of the words in group 4.1.2 and they may reflect an old stratum of EA; group 4.1.3, however, may not.
