Sarah G. Thomason
The Pacific Northwest region of North America is one of the world’s most famous linguistic areas, but the histories of the numerous structural features that are widely shared in the area are poorly understood. The primary goal of this chapter is to show which areal features can be traced historically and which areal features cannot be, and to explain what the difficulties are for the latter set of features. Section 1 sets the stage by characterising the concept ‘linguistic area’ and outlining the four possible historical explanations for shared structural features in this and other linguistic areas. Section 2 discusses the geographical range of the Pacific Northwest linguistic area and the languages that belong to it. Section 3 focuses on shared features that are more likely to be traceable historically – namely, features that have limited range in the area and that are presumably relatively recent innovations – and Section 4 addresses the difficulties that arise in efforts to provide historical explanations for the ancient area-wide features. Section 5 is a brief conclusion that points to possibilities for more satisfactory historical explanations in the future.
First, a definition: “a linguistic area is a geographical region containing a group of three or more (usually more) languages that share some structural features as a result of contact rather than as a result of accident or inheritance from a common ancestor” (Thomason 2001: 99). Some scholars would argue for modifying parts of this definition and/or for additional criteria to identify a linguistic area. One recurring proposal, for instance, is that “the existence of a language area depends crucially on the presence of genetically unrelated (or only distantly related) languages” (Tosco 2000: 335). This proposed requisite is unappealing; the Comparative Method certainly makes it possible to distinguish shared inheritances from diffused features in many or most cases, and in such cases it will still be quite possible to establish the existence of a linguistic area even if the languages are fairly closely related. It is certainly easier, as Mithun has noted, to establish that shared features are due to contact when the languages in contact are unrelated or distantly related (1999: 314). Surely, however, ease of historical analysis should not be criterial for identifying a linguistic area. According to another proposal, there needn’t be several structural features in a linguistic area – even a single feature could be sufficient (Masica 1976: 172; see discussion in Campbell et al. 1986). This proposal also seems unpromising, mainly because a multi-language contact situation that produced just one single diffused structural feature is so implausible: any contact situation that is intense enough for one structural feature to be transferred from one language to one or more others is intense enough for more than one feature to be transferred. Claiming a linguistic area on the basis of a single shared structural feature is therefore unjustified on both methodological and theoretical grounds (see Thomason 2000 for further discussion of this point). But for present purposes we needn’t worry about modifications or additions to the definition above because, as we will see below, the Pacific Northwest clearly satisfies everyone’s favourite criteria for a linguistic area.
There are four possible historical sources of shared features in the languages of a linguistic area. First, they could be inherited from a remote common ancestor, even if the languages are not known to be related at all.
Second, they could be accidentally, or ‘accidentally’, shared – that is, some of the shared features could be due to sheer accident or to the operation of universal structural tendencies. To take one obvious example, the presence of a phoneme /t/ in all the relevant languages would not require a historical explanation of any kind, because the vast majority of the world’s languages have such a phoneme. Similarly, a sharp distinction between nouns and verbs would be assumed to be due to universal structural tendencies, since such a distinction is universal (or nearly so) in the languages of the world.
Third, shared features could be due to spread from one of the languages or language groups in the area to other language(s) or language group(s). Diffused features in this category could either be old in the source language and its relatives (if any) or innovative in the source language.
Fourth, shared features could arise through a process of ‘negotiation’ (Thomason 2001: 142–146). This would happen when speakers of one language who have limited competence in another misperceive a feature in that other language and produce a structure that is innovative with respect to both the innovator’s own language and the other language. This is what Lindstedt (2000: 231) has called “contact-induced change by mutual reinforcement”, with convincing examples from the Balkan linguistic area.
Three of these four possible historical sources of shared features – all but the second one – signal a historical connection between the languages; and of these three, the last two involve contact-induced change. As we will see below, it is not always possible to determine whether a shared feature results from inheritance or diffusion, or whether a shared non-inherited feature results from diffusion or negotiation.
Even a cursory glance at the literature on the Pacific Northwest linguistic area will show that there is no complete consensus on its geographical boundaries, its component languages, or even its name. One common alternative name is the Northwest Coast linguistic area; some scholars divide the area into several subareas, most prominently the Northern Northwest Coast area. The core region comprises Oregon, Washington (state), and neighbouring parts of British Columbia. Some experts expand the territory north to Alaska, south to the Oregon/ California border, and east from the Northwest Coast geographic and culture area to the Plateau culture area, including northern Idaho and western Montana. In this chapter I will focus on the core region plus northern Idaho and western Montana; my reason for including more eastern regions is that Salishan, one of the three core language families of the area, has eastern outliers in those states and in eastern British Columbia as well.
The three core language families of the Pacific Northwest linguistic area are Salishan (23 languages, most of them on or near the Pacific coast), Wakashan (6–8 languages, depending on who’s counting, all on the Pacific coast), and Chimakuan (2 languages, both coastal). These three language families cannot be, or at least have not been, shown to be related to each other. Campbell discusses early proposals for a Mosan super-family comprising the three families (Frachtenberg 1920: 295, Sapir 1929 [1949], and Swadesh 1953a, 1953b), concluding that “[s]ubsequent research has called this classification into question and it is now largely abandoned” (Campbell 1997: 288). In Greenberg’s well-known classification they form one subbranch of his proposed Amerind language family (1987), but serious questions about his methodology have led historical linguists to reject his classification almost unanimously. The lack of evidence for connecting the three families genetically does not, of course, mean that they are 垂related. Nor is it safe to assume that a hypothesis of relatedness cannot be tested. So far no systematic evidence has been presented to support such a hypothesis, but future work could in principle turn up probative evidence, although the families would have to be so distantly related that the chances of finding adequate numbers of systematic correspondences are likely to be slim.
Other nearby language families are peripheral to the linguistic area, both geographically and linguistically: Tsimshian to the north, Chinookan and Sahaptian to the east, and the isolate Kutenai, also to the east. Kinkade (1997) also includes, as peripheral to the Pacific Northwest linguistic area, a number of indigenous languages of Oregon: Cayuse in eastern Oregon and, in western Oregon, Alsea, Siuslaw, Coosan, Takelma, Molalla, and Oregon Athabaskan. The rest of the large Athabaskan family, with member languages in and near the linguistic area, does not share most of the characteristic areal features, and I will not discuss it further. I will also omit discussion of the even more peripheral coastal languages to the north: Nisga’a, Coast Tsimshian, Tlingit, Haida, and Central Alaskan Yup’ik.
The three core families share the most areal features, of course. Because the Salishan family is such an integral part of the linguistic area, and because the easternmost languages (the ones in Idaho, Montana, and interior British Columbia) share all of the area-wide features with languages of the other two core families, it makes sense to treat the interior Salishan languages as belonging to the linguistic area.
The peripheral languages and language families share a smaller number of the areal features, in some instances no doubt because of diffusion. This fact highlights a central problem for those who believe that a linguistic area can and should be sharply delineated: here as in other linguistic areas, the most intense contacts and therefore most of the contact phenomena linked the core language families. But this does not mean that contacts were nonexistent between languages of the core families and speakers of peripheral languages, or that areal features were somehow blocked entirely from diffusing to languages outside the core area. In other words, the boundaries of a linguistic area are inherently fuzzy: some languages are clearly in, some languages are clearly out, and some languages have at least a few of the characteristic areal features under circumstances that suggest diffusion. In the following sections I will not consider the peripheral languages in discussing historical sources of the areal features, because I know of no evidence that suggests diffusion of Pacific Northwest areal features from any of the peripheral languages to any of the core languages. This does not, of course, mean that there was no such diffusion; historical reconstruction of proto-languages for most language families in this part of the world has barely begun, and without comparative reconstruction we cannot know what is old and what is new in a given language.
The age of the Pacific Northwest linguistic area turns out to be impossible to determine. As we will see in Section 4, some of the areal features are ancient in all three of the core families. That means that if those ancient features are shared due to language contact, the linguistic area must have been in existence by at least 4,000 years BP, the very rough time depth estimate for the Salishan family (M. Dale Kinkade, personal communication 1998, based on perceived comparability to solidly estimated time depths for relevant Indo-European branches; see also Holman et al. n.d.: 33, which estimates the time depth at 3,827 years bp). There are also much more recent areal features, and recent and current close contacts that could lead to contact-induced change.
The Pacific Northwest area also includes shared non-linguistic aspects of indigenous culture. In ethnobotany, for instance, one can point to shared aspects of the harvesting and preparation of food plants, uses of medicinal plants, and plants in trade and exchange (Turner 1997). In verbal art there are shared features in sound imagery and stylised characters’ speech (diminutive symbolism, special voice qualities, etc.) and in conventions about such things as the proper season or time for myth recitals and the proper posture for listeners (Seaburg 1997). Although it would clearly be useful to analyse these and other shared cultural features historically in conjunction with an analysis of the Pacific Northwest as a linguistic area, I will focus in this chapter solely on shared linguistic features.
It is impossible, at this late date, to get a clear picture of the nature and extent of language contacts in the Pacific Northwest several thousand years ago, but the indirect evidence for extensive intertribal contact, including much multilingualism, is strong. The original homelands of the three core families are now generally considered to be in the central coastal area of what is now Washington and neighbouring British Columbia (Kinkade 1990): Proto-Salishan along the coast south of the Fraser River (near the present-day Canadian/US border) and probably north of the Skagit River (Washington); Proto-Wakashan on Vancouver Island and neighbouring mainland coastal areas; and Chimakuan on the Olympic Peninsula in Washington. Speakers of Wakashan and Salishan languages are known to have expanded their territories, and their languages must have replaced the languages previously spoken in the expansion areas. For instance, Kwakiutl (Wakashan) moved “southward during the nineteenth century at the expense of the Comox and Sliammon” (both Salishan) (Kinkade 1990: 205); and at an older period Salishan languages expanded southward and to the interior from the original homeland, at the expense of (among other languages) Chimakuan (Kinkade 1990). These expansions must have involved linguistic contacts, and language shift by indigenous populations to the languages of the intruders could well have left traces in the newly arrived languages. Direct linguistic evidence for old as well as more recent contacts is seen in the presence of loanwords that are widely shared in the area, especially terms for flora and fauna. Known areal cultural institutions such as slavery and intermarriage, as well as ubiquitous trade relations, also speak to extensive language contacts (Thomason 1983: 862). At the time of initial contact with Westerners there was extensive multilingualism throughout the area.
Many of the Pacific Northwest areal features have area-wide distribution; many others, however, are limited in their spread within the area. As I will argue below, the shared limitedrange features appear to be more recent developments than the area-wide features, and our chances of discovering the histories of these more recent features are better than for the older shared features. We will therefore begin the historical discussion with the limited-range features.
One example is a common sound change from plain (non-labialised) velars to alveopalatals, as in *k *k’ *x > č, č’ š; this happened in a range of contiguous languages from the west coast of Vancouver Island to the mouth of the Columbia River, including most Salishan languages and some Wakashan languages, as well as Chemakum (but not Quileute) (Campbell 1985: 45; Campbell 1997: 333). Other examples, each of which occurs in a few languages spread across all three core families, are a change from nasal consonants *m *n to voiced oral stops b d and the development of tones or pitch accents.
Determining the origin and spread of these limited-range features is sometimes easy. For instance, most Salishan languages have elaborate consonant clusters in both onset and coda position in the syllable, as in these words from Salish-Pend d’Oreille (the easternmost Salishan language, spoken in western Montana, previously known as Flathead): sxwčšt’sqá ‘a person whose job it is to take care of livestock’ and ta qesm’l’mél’šstmstxw! ‘Don’t play with it!’ This was presumably a feature of Proto-Salish. But one Coast Salishan language, Comox, has no syllable-initial consonant clusters at all, thanks to diffusion from neighbouring Wakashan languages with which Comox speakers have been in close contact. Another example, reported in Kinkade (1997), is an alternation that developed in certain environments through hardening and devoicing of *w and *y to kw and č, respectively, in Makah (Wakashan), Klallam (Salishan), and Chemakum (Chimakuan), all spoken on the northern edge of the Olympic Peninsula. This is a regular conditioned sound change in Klallam (shared in part by several other Salishan languages); the alternation then spread by borrowing to Makah and Chemakum.
Another easy case is the borrowing of a non-glottalised lateral affricate /λ/ by Quileute (Chimakuan) from Wakashan: this phoneme came into Quileute with Wakashan loanwords but then spread to native Quileute words as well when sequences of /t/ and /ł/ coalesced into a unitary affricate (Jacobsen 1979a: 795, citing Powell 1975).
Sometimes tracing the path of shared features is not so easy. Consider the lexicon, for example. Many words, especially words for flora, fauna, trade items, and other vocabulary with particular cultural significance, have spread differentially within the area. Sometimes there are clues to their provenance, but often there aren’t. Many languages in the area have a word kapi ‘coffee’, for instance. This must be a loanword that postdates the first contacts with Whites, because coffee was introduced by Whites. Its recent origin is also confirmed in some of the languages by the fact that it starts with k; in Salish-Pend d’Oreille, for example, it is one of only three words in the language that have a phoneme /k/, because all the other non- labialised velar obstruents changed to alveopalatals (the other words with /k/ are also loanwords). Further, the word must ultimately come from French rather than from English, because the second-syllable stress fits French and not English (and Salish-Pend d’Oreille has free stress, so the stress surely matches that of the source word). But any given language could have borrowed it either directly from French café or from French via Chinook Jargon, the pidgin language that was widely spoken throughout the coastal area and as far east as Idaho. The form would be the same in either case, because f is not a common phoneme in this region and is regularly replaced by p in loanwords – including French loanwords in Chinook Jargon, where the word for ‘coffee’ is also kapí. Moreover, the word could have entered a language from another Native language rather than directly from either French or Chinook Jargon. Other loanwords have followed similarly crooked paths, and although some paths can be traced by means of phonological clues, others cannot be. In such cases, without knowledge of the external circumstances (such as where a trade item originated), the source and differential spread of a loanword cannot be determined.
In many cases it is impossible to discover the history of a feature with limited spread in the linguistic area. One example is the five so-called nasalless languages, in which *m, *n have changed to b, d : Quileute (Chimakuan), the Salishan languages Twana and Lushootseed, and the Wakashan languages Nitinaht (a.k.a. Ditidaht) and Makah. This feature was certainly not inherited from a single remote ancestor, because the other languages in all three families have m, n and not b, d. The voiced oral stops are definitely innovations in the handful of nasalless languages. It is conceivable that the innovations arose independently in all the nasalless languages, i.e., that they are accidentally similar in this respect. This possibility arises primarily because the nasalless languages are not all contiguous – rather, they cluster in two non-contiguous areas, one on Puget Sound (Lushootseed and Twana) and the other in neighbouring areas on the outer coast (Quileute, Nitinaht, and Makah) (Kinkade 1997). But nasal consonants are virtually universal in the world’s languages, so the loss of all nasals is a typologically rare feature; this makes independent innovation less appealing. A more likely possibility is that the change to voiced oral stops occurred in one of the languages with subsequent spread to the others, though in that case one would like to know how the innovation skipped over intervening languages. Kinkade (1985) attributes the innovation to a phenomenon found in almost all the coastal languages from Comox to Lower Chinook in which nasality is phonetically weak and/or there are alternations between nasal stops and voiced oral stops. For him, presumably, it would therefore be reasonable to hypothesise that the change happened in one language in each of the two different areas and then spread to the other language(s) in the same area. But we have no way of testing Kinkade’s hypothesis, or any other hypothesis, concerning the innovation and spread of the nasal-to-oral change.
An even more puzzling example is the case of pharyngeal consonants in the linguistic area. Several languages, distributed over three unconnected parts of the Pacific Northwest and over two of the three core language families, have pharyngeal consonants (the account here is drawn primarily from Kinkade [1997]). In Nootka and Nitinaht (Wakashan) and northern Haida (a probable isolate, southern Alaska), the uvular ejective q’ changed to a pharyngealised glottal stop and the uvular fricative changed to a pharyngeal fricative ḥ. The third region with pharyngeals is Interior Salish territory; all seven languages in this branch of the Salishan family have pharyngeal consonants, but unlike the Nootka-Nitinaht and Haida pharyngeals, the Salishan consonants are sonorants, not fricatives. Salish-Pend d’Oreille, for example, has four pharyngeal consonant phonemes: /ʕ,ʕw,ʕ’,ʕw’/. As in the other Interior Salishan languages, these consonants pattern with the other resonant consonants, and they cannot be shown to have developed from any other kinds of sounds; in fact, they must be reconstructed for Proto-Salish. The historical puzzle, then, is that on the one hand, the Wakashan/Haida and Interior Salishan pharyngeals are so different phonetically and phonologically that it is hard to see how they could be historically connected; but on the other hand, pharyngeals are so very rare cross-linguistically that it is rather difficult to swallow the conclusion that such a rare consonant type just happened to emerge independently in three different parts of the Pacific Northwest. Moreover, none of the three regions with pharyngeals is close to any of the others, which would make a diffusion scenario difficult to imagine even if the consonants themselves were not so different. The puzzle remains, as it does with quite a few of the other limited-range shared features in the Pacific Northwest.
The situation is far worse when we try to discover the histories of the many shared area-wide structural features. To set the stage for the historical discussion, we will begin with a non-exhaustive list of features that occur in all three of the core language families and, in some cases, in peripheral languages as well. First, some characteristic phonological features: very large consonant inventories, typically containing a full series of ejective stops and affricates, lateral obstruents (usually only λ’ and ł), labialised dorsal obstruents and an opposition between velar and uvular obstruents (e.g. /kw, kw’, qw, qw’/), and often glottalised resonant consonants; small vowel inventories, typically just four vowels.
Second, here are some characteristic morphosyntactic features: poly synthetic word structure, e.g. Salish-Pend d’Oreille qwo č-taxwl-m-nt-cút-m-nt-m ‘someone came up to me’ (literally: ‘me to-START-derived.transitive-transitive-reflexive-derived.transitive-transitive-indefinite.agent’); a weak lexical noun/verb distinction (according to Bach [1997], the existence of the distinction has been debated for Wakashan, Chimakuan, and some Salishan languages, though most specialists now agree that the distinction exists); many suffixes, relatively few prefixes; many lexical suffixes, with concrete meanings such as ‘hand’, ‘water’, and ‘lodging’; several reduplication processes, including an optional distributive plural, e.g. Salish-Pend d’Oreille qe c’uw ‘we’re gone’ (we left in a group) vs. qe č’uč’úw ‘we’re gone’ (we left one at a time); verb-initial basic word order; sentence-initial negation; numeral classifiers, e.g. Salish-Pend d’Oreille č-t’áq’n ‘six people’ vs. t’áq’n ‘six of anything else’; lexically paired singular and plural verb stems, e.g. Salish-Pend d’Oreille č’n n?λxw ‘I went in’ vs. qe npilš ‘we went in’ or cqnten ‘I put it down’ vs. slnten ‘I put them down’; a yes/no question particle; an imperative construction that translates roughly as ‘It would be good if you did X’; and aspect as a more important verbal category than tense.
These features are of course not expressed identically in all the languages. Chemakum, for instance, has numeral classifiers for counting objects, persons, canoes, and several other categories, whereas Salish-Pend d’Oreille distinguishes only between counting people and counting other things. Nor are the area-wide shared features all confined to the three core families. To take just two examples, a reduplicative plural is found in Tsimshian as well as in Wakashan, Chimakuan, and Salishan (Sapir 1929 [1949], cited in Campbell 1985: 47), and Kutenai has some lexical suffixes (Kinkade 1997).
The question is, where do these shared area-wide features come from? The answer, in every single case, is that we do not know and most likely will never know. The problem is that almost all of these features must be reconstructed for all three of the core proto-languages – Proto-Salish, Proto-Wakashan, Proto-Chimakuan. The sole exception may be the numeral classifiers, whose status is complex. Even in Salishan languages, whose histories have been most intensively studied, there is so much variation in numeral classifier formations that reconstruction is doubtful. Newman (1976: 240) observes, for instance, that a few Coast Salish and Interior Salish languages have special prefixes which are “used with numerals for counting human beings,” but that these cannot be reconstructed for Proto-Salish. Newman (1976: 240) goes on to note that at least two languages peripheral to the Pacific Northwest linguistic area also have numeral classifiers: Haida apparently has the same distinction as in Salish-Pend d’Oreille, and Tsimshian has “a more complex system of numeral classifiers, in which human beings are distinguished from several other categories of objects.” Nevertheless, in spite of the messy numeral classifier picture, many of the formations involve reduplication as a category marker, and it still seems possible that numeral classifiers will eventually be reconstructed for one or more of the core families. The remaining features, in any case, must be reconstructed for all three proto-languages, which means that they are very old.
This in turn means that these ancient features could be inheritances from a common remote ancestor, or they could be ‘accidentally’ shared (and perhaps preserved by areal pressures), or they could be the result of innovations within the linguistic area – not all at once in a single language, but rather, quite possibly, in individual innovations in various parts of the area – and then diffused throughout an ancient linguistic area, namely, an area in which Proto-Salish, Proto-Wakashan and Proto-Chimakuan were all in intimate contact. (’Negotiation’ is also a possibility here, but discovering its existence is unlikely to be feasible, so I ignore it in the following discussion.)
Unfortunately, there is no evidence for or against any of these three hypotheses. Accident, with or without the influence of universal structural tendencies, seems rather unlikely for some of the features, however – notably the ones that are highly marked cross-linguistically, i.e. rare in languages of the world. A voiceless lateral fricative, for instance, occurs in (almost?) all of the languages in this area but only in about 9 per cent of the languages of the world (Poser 1997); even ejective stops and affricates, which are reasonably common cross- linguistically, occur only in about 16 per cent of languages (Poser 1997). Other cross- linguistically rare area-wide features are the lateral affricate, labialised dorsal consonants, distinctively uvular consonants, polysynthetic word structure, a weak noun/verb distinction, lexical suffixes as a special suffix type, and numeral classifiers. Some of the other widely shared features are also relatively uncommon in languages of the world. Therefore, although we cannot rule out accident or the operation of linguistic universals, that seems by far the least promising of the available historical explanations.
The criteria for establishing that any two or all three of the core families are genetically related have not been met, so at present the first hypothesis (inheritance from a remote common ancestor) is untestable. This situation could certainly change in the future; the Comparative Method has not yet been fully applied to any of the three core families, but once we do have substantial reconstructions for the three proto-languages, we can compare the reconstructed proto-languages in an effort to find the kinds of systematic correspondences that would permit the establishment of genetic relationship. It is difficult to be very optimistic about the prospects for success, however, because a proto-language that gave rise to two or all three of the core families must have been spoken so long ago that the necessary correspondences are liable to have decayed beyond recognition.
Two of the four criteria for proving that contact-induced change has occurred also cannot be satisfied. The four requisites (see Thomason 2001: 93–94 for discussion of these criteria) are: (1) prove the existence of contact; (2) find shared features; (3) prove that the proposed source language (A) had the shared features before contact, i.e. prove that Proto-A has not innovated; (4) prove that the proposed receiving language (B) has innovated, i.e. prove that Proto-B did not have the shared features. In the case of the Pacific Northwest, requisites (1) and (2) are satisfied: we can establish the fact of contact, and we can certainly find many shared features across the whole linguistic area. But requisites (3) and (4) are not satisfied, since all or almost all the shared area-wide features are found in all three proto-languages. Even if we could identify a proposed source proto-language and a proposed receiving proto- language for one or more of these features, we could not prove that Proto-A had the features before coming into contact with Proto-B, and we could not prove that Proto-B did not have the features before coming into contact with Proto-A. Establishing that any of the three proto- languages acquired any of the shared area-wide features by diffusion from another proto-language is therefore hopeless, with the possible (but not probable) exception of the numeral classifiers.
This leaves us with a completely indeterminate historical picture, as far as the shared area- wide features are concerned. Historical linguists are all too familiar with indeterminacies, of course, but this situation is unusual in the extreme age of the shared features. (As far as I know, no time estimates have been proposed for Proto-Wakashan and Proto-Chimakuan. But since the break-up of Proto-Salishan into separate branches appears to have happened about 4,000 years ago, as noted in Section 2 above, an ancient linguistic area cannot have developed later than that; and if some or all of the shared features did result from shared genetic inheritance in two or all three of the families, that would push the dating of the shared features back several thousand more years.) We can surmise that contact is responsible in large part for the current status of the features, if only because the ongoing close contacts helped to preserve the ancient shared features. But we can’t prove it, any more than we can prove that two or more of the three core families are genetically related.
Systematic research (unlike that of Edward Sapir and Joseph Greenberg) to try to establish genetic relationships of two or all three of the core language families could prove fruitful, and could help to explain the area-wide shared features historically, if and only if adequate reconstructions of the three core proto-languages can be carried out. Further research into the specifics of tribal interactions, e.g. trade relationships, might well turn up evidence that would help to identify the historical routes by which some of the shared limited-range areal features diffused from one language into another. Further reconstruction of the core families – there is a fair amount now for Salishan (see e.g. Thompson 1979; Thompson and Kinkade 1990; Kroeber 1999), less for Wakashan (see e.g. Jacobsen 1979b), and I believe still less for Chimakuan (see e.g. Jacobsen 1979a; Powell 1993) - should also help with efforts to discover the historical developments of some of the limited-range features, by providing evidence as to whether language A or language B has innovated in a given instance.
But for the area-wide shared features, unless the language families can be shown to be related, the situation is hopeless. Without (necessarily distant) relatives outside the Pacific Northwest, we will never know whether those features are inherited from a common remote ancestor or innovated and spread within an ancient linguistic area comprising at least Proto-Salish, Proto-Wakashan and Proto-Chimakuan. As argued above, the only historical explanation we can tentatively rule out is accident (including universal structural tendencies), because it is so very unlikely that a whole range of universally marked structural features – and even unmarked features, for that matter – would just happen to arise independently in three proto-languages in the same geographical region. And finally, regardless of which of the two plausible historical explanations is true, it is reasonable to hypothesise that the ancient area-wide shared features were reinforced by contact up to the present.
Jacobsen, William H., Jr. 1979a. Chimakuan comparative studies. In Campbell and Mithun (eds), 792–802.
——1979b. Wakashan comparative studies. In Campbell and Mithun (eds), 766–791.
Kroeber, Paul D. 1999. The Salish language family: reconstructing syntax. Lincoln: University of Nebraska Press.
Thompson, Laurence C. 1979. Salishan and the Northwest. In Campbell and Mithun (eds), 692–765.
Thompson, Laurence C. and M. Dale Kinkade. 1990. Languages. In Wayne Suttles (ed.) Northwest Coast, Volume 7 of the Handbook of North American Indians. Washington, DC: Smithsonian Institution, 30–51.
Bach, Emmon. 1997. Areal grammatical features in Pacific Northwest languages. Paper presented at the symposium ‘The Pacific Northwest as a Linguistic and Cultural Area’, Annual Meeting of the American Association for the Advancement of Science, Seattle.
Campbell, Lyle. 1985. Areal linguistics and its implications for historical linguistic theory. In Jacek Fisiak (ed.) Proceedings of the Sixth International Conference on Historical Linguistics. Amsterdam/ Philadelphia: John Benjamins, 25–56.
——1997. American Indian languages: the historical linguistics of Native America. New York: Oxford University Press.
Campbell, Lyle, Terrence Kaufman and Thomas C. Smith-Stark. 1986. Meso-America as a linguistic area. Language 62: 530–570.
Campbell, Lyle and Marianne Mithun (eds) 1979. The languages of Native America: historical and comparative assessment. Austin: University of Texas Press.
Frachtenberg, Leo J. 1920. Abnormal types of speech in Quileute. International Journal of American Linguistics 1: 175–182.
Greenberg, Joseph H. 1987. Language in the Americas. Stanford: Stanford University Press.
Holman, Eric C., Cecil H. Brown, Soren Wichmann, André Müler, Viveka Velupillai, Harald Hammarströ, Sebastian Sauppe, Hagen Jung, Dik Bakker, Pamela Brown, Oleg Belyaev, Matthias Urban, Robert Mailhammer, Johann-Mattis List and Dmitry Egorov. n.d. Automated dating of the world’s language families based on lexical similarity. [Available at: www.staff.eva.mpg.de/wichmann/AutomatedDatingFinal.pdf; Accessed 29 August 2013].
Jacobsen, William H., Jr. 1979a. Chimakuan comparative studies. In Campbell and Mithun (eds), 792–802.
——1979b. Wakashan comparative studies. In Campbell and Mithun (eds), 766–791.
Kinkade, M. Dale. 1985. More on nasal loss on the Northwest Coast. International Journal of American Linguistics 51: 478–480.
——1990. Prehistory of Salishan languages. In Papers for the 25th International Conference on Salish and Neighbouring Languages. Vancouver: University of British Columbia, 197–208.
——1997. The emergence of shared features in languages of the Pacific Northwest. Paper presented at the symposium ‘The Pacific Northwest as a Linguistic and Cultural Area’, American Association for the Advancement of Science Annual Meeting, Seattle.
Kroeber, Paul D. 1999. The Salish language family: reconstructing syntax. Lincoln: University of Nebraska Press.
Lindstedt, Jouko. 2000. Linguistic Balkanization: contact-induced change by mutual reinforcement. In Dicky Gilbers, John Nerbonne and Jos Schaeken (eds) Languages in contact. Amsterdam: Rodopi, 231–246.
Masica, Colin P. 1976. Defining a linguistic area: South Asia. Chicago: University of Chicago Press.
Mithun, Marianne. 1999. The languages of Native North America. Cambridge: Cambridge University Press.
Newman, Stanley. 1976. Salish and Bella Coola prefixes.International Journal of American Linguistics 42: 228–242.
Poser, William J. 1997. Areal phonological features in Pacific Northwest languages. Paper presented at the symposium ‘The Pacific Northwest as a Linguistic and Cultural Area’, Annual Meeting of the American Association for the Advancement of Science, Seattle.
Powell, James V. 1993. Chimakuan and Wakashan – the case for remote common origin: another look at suggestive sound correspondences. In Anthony Mattina and Timothy Montler (eds) American Indian linguistics and ethnography in honor of Laurence C. Thompson. Missoula: UMOPL- Linguistics Laboratory, University of Montana, 451–470.
Powell, Jay. 1975. Quileute language: Book 1. La Push, WA: Quileute Tribe.
Sapir, Edward. 1949 [1929]. Central and North American languages. In David G. Madelbaum (ed.) Selected writings of Edward Sapir in language, culture, and personality. Berkeley: University of California Press, 169–178.
Seaburg, William. 1997. Areal features of style in Pacific Northwest folklore. Paper presented at the symposium ‘The Pacific Northwest as a Linguistic and Cultural Area’, Annual Meeting of the American Association for the Advancement of Science, Seattle.
Swadesh, Morris. 1953a. Mosan I: a problem of remote common origin. International Journal of American Linguistics 19: 26–44.
——1953b. Mosan II: comparative vocabulary. International Journal of American Linguistics 19: 223–236.
Thomason, Sarah G. 1983. Chinook Jargon in areal and historical context. Language 59: 820–870.
——2000. Linguistic areas and language history. In Dicky Gilbers, John Nerbonne and Jos Schaeken (eds) Languages in contact. Amsterdam: Rodopi, 311–327.
——2001. Language contact: an introduction. Edinburgh and Washington, D.C.: Edinburgh University Press and Georgetown University Press.
Thompson, Laurence C. 1979. Salishan and the Northwest. In Campbell and Mithun, (eds), 692–765.
Thompson, Laurence C. and M. Dale Kinkade. 1990. Languages. In Wayne Suttles (ed.) Northwest Coast, Volume 7 of the Handbook of North American Indians. Washington, DC: Smithsonian Institution, 30–51.
Tosco, Mauro. 2000. Is there an Ethiopian language area? Anthropological Linguistics 42: 329–365.
Turner, Nancy J. 1997. Patterns in Pacific Northwest ethnobotany. Paper presented at the symposium ‘The Pacific Northwest as a Linguistic and Cultural Area’, Annual Meeting of the American Association for the Advancement of Science, Seattle.