Corps de l’article

Introduction

Utkuhiksalik[1] is a sub-dialect of Natsilik[2] within the Western Canadian Inuit (WCI) dialect continuum (Dorais 1990: 17, 41). Unlike many Western dialects, Utkuhiksalik contrasts /h/ and /s/, /j/ and /ř/. The aim of this paper is to show how these contrasts arose by describing the relevant sound changes from Proto-Eskimoan (PE) or Proto-Inuktitut (PI) to Utkuhiksalik. These sound changes involve PE *ð, PE *c (PI *s), and PE *j. In this paper, we first describe the historical relationships between the Utkuhiksalingmiut and related groups. We then overview the historical development of the Utkuhiksalik consonant inventory, and continue with more detailed descriptions of the historical developments in following sections[3].

Overview of dialect relations[4]

The traditional territory of the Utkuhiksalingmiut (‘the people of the place where there is soapstone’) lay along the Back River, between Chantrey Inlet and Franklin Lake. Utkuhiksalik speakers also lived in the interior, near a place called Kitikkat (‘eskers’), which was southeast of Chantrey Inlet and towards Baker Lake.

Most sources, including current Utkuhiksalik traditions, agree that the Utkuhiksalingmiut moved into the Back River area from the west coast of the Adelaide Peninsula (Ukřulik, ‘having bearded seals’) where they had probably lived adjacent to the Iluiliqmiut. (The present name of the Adelaide Peninsula is Iluiliq.) From the mouth of the river, Chantrey Inlet, they spread westward along the river as far as Lakes Garry and Pelly, and southeastward towards Baker Lake. When exactly these movements occurred is not known. A Chantrey Inlet elder told Briggs that it was in the late nineteenth or early twentieth century, but Back (1836), for example, found habitations all along the river in 1833. Those who moved to Garry and Pelly Lakes, the Hanningařuqmiut (‘the people of the place that lies across’) and the Ualijaqɬit (‘the most western people’) spoke dialects closely related to that of the Utkuhiksalik proper, who remained in Chantrey Inlet (Utkuhiksalik).

The traditional home of the Natsilik was to the north of the Utkuhiksalik-speaking area—King William Island and Boothia Peninsula, later concentrated in the settlements of Gjoa Haven, Spence Bay, and Pelly Bay. Utkuhiksalik speakers currently residing in Gjoa Haven (and some currently living in Baker Lake) consider the speech variety originally spoken in Chantrey Inlet to be “core” Utkuhiksalik. They consider the varieties spoken by the Ualijaqɬit and Hanningařuqmiut to be more or less borderline. They also sometimes classify Iluiliq as being Natsilik.

Historical and dialectal relations between the above-mentioned groups were greatly complicated by famine-induced resettlement and other factors. In addition, whether or not one is currently designated as an Utkuhiksalingmiutaq can also depend on fine-grained considerations which are not necessarily linguistic, involving, for example, a person’s history of residence and social affiliation.

Historical development of the Utkuhiksalik phonemic inventory

We now overview the main developments between the PE and Utkuhiksalik consonant inventories. Example 1 illustrates the PE consonant inventory (phonetic equivalents are provided where these are not identical to the spelling[5]).

The PE consonant inventory is similar to the Utkuhiksalik inventory, except for the following changes: PE *c became /h/ and /s/ (as well /ts/ and /tj/) in Utkuhiksalik (PE *c had previously become *s in PI; Fortescue et al. 1994: xvi, xiv). PE *ð became /ř/, /s/, and /j/. /ř/ is a voiced rhotic fricative; it also has the non-rhotic allophones [ð] and [ž]. Finally, PE *j became /j/, /s/, /ts/, or /i/, and also deleted.

Example 2 illustrates the present-day Utkuhiksalik consonant inventory arising from these sound changes.

A few comments are in order. First, we classify /v/ as a sonorant on the grounds that it patterns like /ř/: both /v/ and/ /ř/ can be followed by /v/, /g/, or /r/, whereas obstruents cannot be. Second, while /s/ is a phoneme, it has a limited distribution, occurring only after the stops /p, t, k, q/. Third, we assume that the laryngeals /h/ and /ʔ/ are not true consonants phonemically (Chomsky and Halle 1968: 303); however, they are included in the ‘obstruent’ section of the table for convenience.

Several phonological environments influenced the sound changes from PE to Utkuhiksalik phonemes, as shown in example 3.

The relevant environments were intervocalic, preconsonantal, and postconsonantal.[8] The development of PE *ð, *c and *j was also influenced by the labial, coronal, or dorsal place of articulation of any preceding consonant. Finally, the development of PE *ð was influenced by the voiced or voiceless status of a preceding proto-dorsal consonant. We overview the major developments involving PE *ð in example 4, PE *c in example 5, and PE *j in example 6.

As shown in example 4, PE *ð became /ř/ (and sometimes /j/) except after originally voiceless dorsals (*k and *q), where it became /s/[9].

Example 5 demonstrates that PE *c became /h/ (also the variant [hj]), /ts/ or /tj/ between vowels; PE *c became /s/ after labial and coronal consonants, and either /h/ or /s/ after dorsal consonants.

As shown in example 6, PE *j became /j/ and /ts/ between vowels, but had other reflexes after consonants.

PE *j either remained as /j/ or vocalized to /i/ after proto-labials. PE *j became /s/ after proto-coronals. PE *j either became /s/ or vocalized to /i/ after originally voiceless *k (and possibly after *q, since both *k and *q are dorsal plosives). PE *j became a fricative, either /g/ or /h/ (phonetically, [χ], in the Utkuhiksalik /qh/ cluster) after originally voiced *g and *r respectively; it also vocalized to /i/ after *r (and possibly after *g, since both *g and *r are dorsal fricatives); finally, it also became /g/ and /r/ in postbases.

All of the changes described above will be discussed in greater detail. We begin by describing historical developments in intervocalic position.

Intervocalic *ð, *c, and *j

PE *ð often became /ř/ between vowels (example 7).[12] ([ð] remains as one of the allophones of /ř/ in Utkuhiksalik.)

A parallel development occurred in Natsilik, where PE *ð became a fricative sound similar to [ž], but rhotic (Fortescue et al. 1994: xvi).

Intervocalic PE *j often remained as /j/ in Utkuhiksalik (Example 8), but it also became /ts/ alternating with /j/ (Example 9).

A comparison of examples 7 and 8 illustrates that Utkuhiksalik maintains a contrast between /ř/ and /j/; these sounds were conflated in many dialects of Inuktitut. Geminate /řř/ also occurs in Utkuhiksalik (Example 10).

In contrast, the existence of a geminate counterpart to /j/ is doubtful; see the section on “*j-initial clusters” for further details.

PE *c (PI *s) developed into /h/ and [hj] between vowels (Example 11), as it did in Natsilik (Fortescue et al. 1994: xvi, xiv).

Some words are pronounced exclusively with [hj] (Example 11.b), while other words have [h] or [hj] in free variation (Examples 11.c, d); there is also an instance of [hj] in free variation with [hi] (Example 11.e).

Other reflexes of PE *c in intervocalic position were /ts/ and /h/ from the same proto-base (see Example 12).

As shown in example 12.a, some proto-bases with intervocalic *c split into two closely-related bases, one containing /h/ and one containing /ts/. In other cases (Examples 12.b-d), PE bases with intervocalic *c became a single base with two allomorphs, one with /h/ and one with /ts/.

Finally, at least one instance of PE intervocalic *c became /tj/ in Utkuhiksalik (Example 13).

The Utkuhiksalik /tj/ cluster shown in example 13 is pronounced somewhat like [tž]; the second consonant is not rhoticized, and the cluster is therefore distinct in pronunciation from /tř/. Furthermore, the pronunciation of /tj/ is distinct from that of a single, lightly-palatalized /t/ [tj] which is found in sequences such as /ti/ [tji][23].

The changes affecting PE *ð, *c, and *j between vowels are summarized in example 14.

In many cases, PE *ð became /ř/, while *c became /h/ (as well as [hj], /ts/, and /tj/) and *j remained /j/ between vowels.

More reflexes of *ð, *c, and *j occurred in consonant clusters than in intervocalic position. We describe clusters that began with *ð or *j below. (Fortescue et al. 1994 list no *c-initial clusters for PE.) Clusters that ended with *ð, *c, or *j are discussed later on.

Clusters beginning with *ð or *j

*ð-initial clusters (as well as another cluster leading to Utkuhiksalik /řg/) and *j-initial clusters are discussed in the next sections.

*ð-initial clusters and other proto-clusters leading to Utkuhiksalik /řg/

PE *ðg became /řg/ in Utkuhiksalik (as can be seen in example 15)[24].

Thus, PE *ð became /ř/ when at the beginning of a cluster. Another source for /řg/ was PE *ɬg (Example 16).

*j-initial clusters

Historically, PE had a contrast between *j and geminate *jj—the latter representing the only type of *j-initial cluster listed in Fortescue et al. (1994). Geminate *jj tended to become /ts/ in Utkuhiksalik, instead of /jj/ (Example 17).

It is possible that the historical contrast between PE *j and *jj was preserved as a contrast between /j/ and /ts/ in Utkuhiksalik. Alternatively, the PE contrast might have been lost in Utkuhiksalik. We leave this matter aside for future research.

There is little evidence for a geminate /jj/ in synchronic Utkuhiksalik: Briggs’s database of several thousand bases lists only one possible /jj/ geminate (Example 18).

The possible /jj/ geminate shown in example 18 is not rhoticized, but is very fricative-like. An Utkuhiksalik consultant rejects the rhotic pronunciation for this geminate in favour of a long [žž]-like sound.

The antecedents of the possible /jj/ geminate in example 18 are shown in example 19 below.

The /aa-ŋa-jjaq-/ or /aa-ŋa-jaa(q)-/ forms in example 18 developed from the proto-morphemes of PE *aga- (Example 19.a), *-ŋa-1 (Example 19.b), and *-tjag- (Example 19.c). (Thus, the possible geminate /jj/ in example 18 likely descends from PE *tj.) Other developments from PE *tj are discussed in example 32.

Historically, then, PE *jj became /ts/ in Utkuhiksalik instead of /jj/, eliminating one possible source of j-initial clusters in Utkuhiksalik. Another possible source of j-initial clusters, PE *tj (Example 19), possibly developed into /jj/ in one instance, but also became either /h/ or /j/ (Example 32). Consequently, there are few, if any, j-initial clusters in Utkuhiksalik.

Our overall conclusions about PE clusters beginning with *ð or *j are as follows. First, *c and *j did not occur as the first member of a consonant cluster in PE (except for proto-*jj, which became /ts/ or /tj/). However, *ð existed as the first member of a consonant cluster in PE; this consonant (and also *ɬ) became /ř/ in cluster-initial position, resulting in a relatively uncommon /řg/ [řɣ] cluster in Utkuhiksalik. An /řv/ cluster, pronounced [řv] or [př], also occurs in Utkuhiksalik postbases (Example 42).

Clusters ending with *ð, *c, and *j

In this section, we describe how clusters ending with PE *ð, *c, or *j developed in Utkuhiksalik. The labial (p, v, m), coronal (t), or dorsal (k, q, g, r) place of articulation of the first consonant of the cluster was relevant in such cases, and so our description is organized according to the labial, coronal, or dorsal place of articulation of the first consonant of the cluster.

Proto-clusters with labials

First, we discuss proto-clusters consisting of a labial phoneme (*m, *p, *v) followed by *ð, *c, or *j.

Labials followed by *ð

The development of clusters of proto-labials followed by *ð is overviewed in examples 20-24.

In at least one case, PE *vð became a /ps/ cluster with a rhoticized /s/ (Example 20).

The /ps/ cluster in 20 consistently has a distinctly rhotic quality, and is pronounced as [pšɹ̥] or [ɸšɹ̥]; the latter pronunciation is similar to the NAI < taafšuma > pronunciation listed in 20. [28][29]

Other PE *vð clusters became /vj/, pronounced as [bj, vj, βj] (Example 21).

Finally, other PE *vð clusters became /př/ in Utkuhiksalik (Example 22).

However, Utkuhiksalik /př/ clusters also arose from the metathesis of clusters beginning with a proto-liquid (/l/ or /ɬ/) followed by proto-*p or *v (Example 23), as well as from the metathesis of *ðv clusters (Example 24).[30]

In summary, PE *vð became /ps/, /vj/ and /př/; /př/ also derived from other PE sources, including *lp, *ɬv, and *ðv.

Labials followed by *c

Proto-clusters containing a labial (*p, *m) followed by proto-*c became /ps/ in Utkuhiksalik (Example 25).

In /ps/ clusters, the /p/ is pronounced either as [p], [ɸ] or [f], and the /s/, as [s] or [sj]. The [s] is not fully palatalized to [š] in /ps/ clusters; in contrast, /s/ sounds more palatal in other types of cluster, described later.

Labials followed by *j

Several different reflexes are attested for clusters consisting of a labial consonant followed by PE *j (Example 26).

First, *pj became /ʔj/ (Example 26.a). (This is also an instance of a general change reported for Natsilik, in which PE clusters of a stop followed by a sonorant became clusters of a glottal stop /ʔ/ followed by a sonorant; Fortescue 1983: 12; 20). Second, in some *mj and *vj clusters, *j vocalized to /i/ (Example 26.b-d).

In summary, the main outcomes of PE clusters of a labial followed by *j include /ʔj/, /mi/ and /vi/. These outcomes also reflect a general constraint against clusters of a true consonant followed by /j/ in Utkuhiksalik. (Recall that neither /ʔ/ or /v/ are true consonants; see example 2 for further discussion.)

Example 27 summarizes the main reflexes of PE clusters of a labial consonant followed by *ð, *c, or *j.

Proto-clusters with coronals

In this section, we describe the development of PE clusters beginning with a coronal consonant (*t or *c) followed by *ð, *c, or *j.

Coronals followed by *ð

Clusters originating from PE *tð are shown in example 28.

Some PE *tð clusters became /ʔř/ (Example 28.a).[31] (The reflex in example 28.a is also an instance of the general change reported for Natsilik, in which PE clusters of a stop followed by a sonorant became clusters of a glottal stop /ʔ/ followed by a sonorant (Fortescue 1983: 12, 20). PE *tð also became /řř/ in some instances (Example 28.b).

Coronals followed by *c and *j

PE clusters of a coronal followed by *c or *j included *tc, *cc, *jj, and *tj. The distribution of PE *tc, *cc, and *jj largely overlapped with the distribution of *tt; however, while *tc, *cc and *jj occurred before any PE vowel, *tt did not occur before /i/ (Fortescue et al. 1994: xvi). These observations suggest an historical contrast, neutralized before /i/, between *tt and the other coronal clusters.

PE *tc, *cc, and *jj became /ts/ in Utkuhiksalik[32], while *tt remained unchanged. Consequently, Utkuhiksalik now has a contrast between /tt/ and /ts/, which is analogous to the PE contrast. Utkuhiksalik /tt/ is pronounced with very little, if any, assibilation[33], while /ts/ is heavily assibilated. While it is clear that /tt/ and /ts/ contrast before /a/ and /u/ in Utkuhiksalik, it is unclear whether they contrast before /i/. We set this matter aside for future research.

Examples 29 and 30 illustrate the historical antecedents of the present-day contrast between /tt/ and /ts/ in Utkuhiksalik. Proto-*tt (and also *lt) became /tt/ before /a/ (Examples 29.a, b), while proto-*tt (and also *lt) became /tt/ before /u/ (Examples 29.c-e) in Utkuhiksalik.

In contrast, proto-*tc (Example 30.a), *cc (Examples 30.b, c) and *jj (Examples 30.d, e) became /ts/ in Utkuhiksalik.

However, it is possible that one instance of PE *jj became /tj/ rather than /ts/ (Example 31).

The WCI reflexes in example 31 suggest that the older PE form, not provided by Fortescue et al. (1994), may have contained a geminate *jj. If so, then one instance of PE *jj may have become /tj/, rather than /ts/, in Utkuhiksalik. Alternatively, however, Utkuhiksalik /tj/ could have derived from PE *c (see example 13) or, plausibly, from *tj, although no examples are attested for the latter.

Finally, PE *tj developed the reflexes shown in example 32.

PE *tj developed into the single consonants /h/ (Examples 32.a, b) and /j/ (Example 32.c) in Utkuhiksalik.

In summary, proto-*tc, *cc, and *jj became /ts/ (the latter possibly also developed into /tj/); proto-*tt remained unchanged; and proto-*tj became /h/ or /j/.[36]

The reflexes of PE clusters containing a coronal followed by*ð, *c, or *j are summarized in example 33.

Proto-clusters with dorsals

We now describe the development of clusters containing a proto-dorsal consonant (*k, *q, *g, *r) followed by *ð, *c, or *j.

Dorsals followed by *ð

In clusters of a PE dorsal followed by *ð, *ð displays two reflexes, depending on the original voicing of the first member of the cluster: *ð became /s/ after originally voiceless dorsal consonants and /ř/ after originally voiced dorsal consonants. A similar development occurred in Alaskan Inupiaq (Dorais 1986: 37). Moreover, both in Alaskan Inupiaq (MacLean 1980) and, apparently, in Natsilik (Hitch 1994: 4-6) /ř/ still regularly devoices to (retroflex) [ṣ] after /k, q/. Examples of the Utkuhiksalik developments are provided in examples 34-36.

Proto-*ð became /s/ after voiceless proto-dorsal consonants (Example 34).

In /ks/ and /qs/ clusters, the plosives /k/ and /q/ are realized as [x] and [χ] respectively, through manner assimilation, and the /s/ is pronounced as [s], [š], or [sj].

Proto-*ð became /ř/ after the originally voiced dorsals *g (Example 35) and *r (Example 36). Both *g and *r then devoiced and became /k/ and /q/ respectively.

/kř/ is pronounced as [gř] and /qř/ is pronounced as [q0ř]; the first member of the [gř] cluster is lenis and at least partly voiced; the /qř/ [q0ř] cluster has a lenis release between the plosive and the /ř/.

Dorsals followed by *c

PE *c became either /s/ or /h/ after proto-dorsals. The /s/ reflex is a more conservative one, retained from PI (recall that PE *c became *s in PI; Fortescue et al. 1994: xvi, xiv). In contrast, the subsequent change from PI *s to /h/ is an innovation shared by most WCI dialects, and is also partly shared by Utkuhiksalik.

Proto-dorsals followed by *c became /kh/ [xx, x̯x̯ɕ] or /ks/ [xs], and /qh/ [χχ], or /qs/ [χs] in Utkuhiksalik (Example 37).

As shown in examples 37.a-c, while /kh/ clusters are often pronounced as [xx], they sometimes have an ‘s-like’ timbre, which we transcribe as a fronted velar fricative [x̯x̯ɕ]. The possible historical significance of s-like pronunciations is discussed below.

Dorsals followed by *j

Clusters of a PE dorsal followed by *j underwent several changes in Utkuhiksalik (Example 38).

PE *j was vocalized to /i/ after *k and *r (Examples 38.a-c). However, it also became /s/ after *k (Example 38.d). PE *j also became a fricative after *q and *g (Examples 38.e, f), and sometimes deleted (Example 38.g). These diverse historical changes resulted in a lack of clusters containing a dorsal consonant followed by /j/ in Utkuhiksalik.

Examples such as 38.e are particularly interesting: in the same elicitation session, one of the consultants pronounced the /qh/ [χχ] cluster with different degrees of s-like timbre (i.e. [χχ] versus a more fronted [χ̯χ̯]).[38] Examples such as 37.a-c ([x] versus [x̯]) and 38.e ([χ] versus [χ̯]) suggest an intermediate stage in the historical change from PE *c to PI *s to WCI /h/, a stage in which PI */kh/ was pronounced as [x̯x̯] and PI */qh/, as [χ̯χ̯]. The intermediate stage would have resulted from place assimilation (that is, from PI *s becoming more like /k, q/ for place of articulation). Utkuhiksalik preserves this intermediate stage in some isolated pronunciations. In summary, PE *j became /s/ (or an s-like sound), /h/, or /i/ after dorsals. It also deleted.

Example 39 summarizes the development of PE *ð, *c, and *j when these occurred as the second consonant in a cluster.

Example 39 illustrates that the place of articulation influenced the preceding consonant in the historical development of *ð, *c, and *j, as did the voiceless or voiced status of preceding dorsals.

Up to this point, we have described the historical development of *ð, *c, and *j in bases, but not in postbases. We have chosen this strategy because historical reconstruction is typically more straightforward for bases than for postbases. In our final descriptive section, we sketch out some basic trends in the historical development of *ð, *c, and *j in postbases.

Postbases containing *ð, *c, and *j

The historical development of Inuit and Yupik postbases is complex: for example, “[…] different affixes with the same original shape may end up with different morphophonemic behaviour. Even more commonly, the same historical affix may end up with different properties (and allomorphs) in different [Inuit and Yupik] languages and dialects or one affix may split both semantically and morphophonemically in the same language” (Fortescue 1992: 18; emphasis added). Consequently, modern Inuit and Yupik languages and dialects often have several postbases descended from a single proto-form, with at least one of the modern postbases being fossilized or less productive. For example, PE *nər ‘one that is more or most’ is the proto-form for two WCI forms: (1) productive -niχχaq- ‘the one that - most, more’ and (2) unproductive, non-truncating -niq- ‘the one that - most, more’ (Fortescue et al. 1994: 414).

We do not yet have enough information to make claims about the productivity of Utkuhiksalik postbases. Moreover, without this information, we cannot postulate any intermediate stages in the development of PE *ð, *c, and *j in postbases. In the interim, however, we provide examples of how PE *ð, *c, and *j developed in Utkuhiksalik postbases, making generalizations where possible.[39]

*ð in postbases

The historical development of *ð in postbases is exemplified in 40-44. PE *ð developed into /ř/ and /s/ in postbases (Example 40).

Postbase-initial *ð also became /ř/ alternating with /t/ in some postbases and in some inflectional affixes (Example 41).

However, as noted in the Fortescue et al. (1994) glosses in example 41, the allomorphs beginning with /t/ likely arose from PE *t, rather than *ð. The examples in 41 thus only provide instances of PE *ð becoming /ř/ in Utkuhiksalik.

Metathesis sometimes affected postbases beginning with PE *ð as demonstrated in example 42.

As shown in example 42, PE *ðvig ‘place or time of’ has three reflexes in Utkuhiksalik: unmetathesized [řvik], metathesized [vřik], and the form [vik], in which /ř/ has been deleted. The developments shown in examples 40-42 are consistent with the historical development of *ð in bases (which was summarized in example 4). 4). As in bases, *ð became /s/ after originally voiceless dorsal consonants (Example 40.b), but became /ř/ elsewhere (Examples 40.a, 41, 42).

In contrast, examples 43-44 illustrate other developments of PE *ð in postbases. As shown in example 43, some postbases originally beginning with PE *ð now have two allomorphs beginning with either /h/ or /s/.

Moreover, as shown in example 44, postbase-initial *ð became /g/ [ɣ] alternating with /r/ [ʁ] and /t/ in some cases.

In postbases, then, PE *ð sometimes developed in a manner consistent with the historical development of *ð in bases: *ð became /ř/ or /s/. However, additional reflexes such as /h/, /g/, and /r/ are also attested. These reflexes could be due to morphological restructuring on analogy with other postbases (e.g., Example 46), an idea which we leave for future research.

*c in postbases

The historical development of *c in postbases is exemplified in example 45.

As shown in example 45, PE *c became either /h/ or /s/ in postbases. These historical developments are consistent with the development of *c in bases (which was summarized in example 5). As in bases, *c became /h/ between vowels (Examples 45.a, c); *c became /h/ (Examples 45.a-c) or /s/ (Example 45.a) after /q/; and *c became /s/ after /t/ (Examples 45.d, e) or *m (Example 45.f). The reflexes of PE *c in postbases thus result from regular historical sound changes.

*j in postbases

The historical development of *j in postbases is demonstrated in examples 46-48. Example 46 illustrates that in some postbases, initial PE *j developed into morphemes with allomorphs beginning with / j, s, g, r / [j, s, ɣ, ʁ].

Example 47 illustrates that in some postbases, PE *j became /h/ or /s/.

Since both /-huq-/ and /-suq-/ can occur after bases ending with /q/, it is possible that one of the postbases is a more archaic or less productive form.

Finally, example 48 illustrates a postbase in which PE *j became [j] or [i][44].

Example 48 differs from the earlier examples in that PE *j occurred after a consonant within the same morpheme in example 48, whereas *j was postbase-initial in examples 46 and 47.

In summary, as in bases (see example 17) all of the changes affecting *j-initial postbases resulted in the elimination of proto-sequences of a true consonant followed by /j/. (We assume that /v/ is a sonorant glide comparable to /w/; see example 2.)

Example 49 summarizes the historical development of PE *ð, *c, and *j in postbases.

Example 49 illustrates that the main difference between the historical development of PE *ð, *c, and *j in bases versus postbases is that the number of historical reflexes tended to multiply when PE *ð, *c, and *j occurred postbase-initially. An interesting matter for future research, partly addressed by Fortescue (1992), is why the boundary between the base and the postbase is historically different from other boundaries within the Inuit word. Following Fortescue (1992), we have speculated that both archaisms and more modern forms coexist, and that the modern forms have arisen both by regular sound changes and through morphological restructuring or analogy.

Conclusion

So far, we have described how PE *ð, *c, and *j developed in Utkuhiksalik bases and postbases. In this final section, we outline the characteristics that Utkuhiksalik currently displays as a result of these historical developments. The present-day Utkuhiksalik system of contrasts resulting from the developments of PE *ð, *c, and *j is summarized in example 50.

Example 50 illustrates several striking characteristics of the Utkuhiksalik sound system: first, /ř/ (from PE *ð), /h/ (from PE *c), and /j/ (from PE *j) all occur between vowels in Utkuhiksalik; Utkuhiksalik is like Natsilik (Dorais 1990: 39) and Uummarmiut (Lowe 1985: 276) in maintaining a contrast between /ř/ and /j/. Second, the phonemes /ř/ (from PE *ð), and /s/ (from PE *ð, *c, and *j) both occur after the plosives /p, t, k, q/ in Utkuhiksalik, while /h/ (from PE *c) also occurs after /k, q/. Natsilik might share with Utkuhiksalik the retention of /s/ after some consonants; however, Dorais (1990: 41) reports that Natsilik tends to have /h/ in positions where most ECI words have /s/. Finally, Utkuhiksalik does not have /j/ after true consonants; /j/ occurs after vowels, and after the non-consonantal sounds /ʔ/ and /v/. It is not clear how Utkuhiksalik compares with other WCI dialects in this respect. In conclusion, a matter for future research is the extent to which Utkuhiksalik differs from, or is similar to, other dialects in the Natsilik grouping in terms of the characteristics described in this paper.