An Overlooked Key Element in North Wakashan Morphophonemics

Investigated is a morphophonemic entity that manifests itself in a rich variety of ways including doubling or glottalization of a resonant; the phonemes /a/, /i/, /u/; the morpheme boundary marker /ə/ and, in the variety of North Wakashan made famous by Franz Boas, an opposition type unknown in the other varieties. It is a byproduct of some of the complex processes triggered by attaching certain types of suffix to a stem but occurs also as a constituent of morphemes and as a device to turn bound roots into free forms. It has never before been identified and is a warning that we are nowhere near understanding the history of North Wakashan, let alone its possible relationships to neighbouring languages. No North Wakashan expertise is necessary to follow the discussion which is preceded by essential comparative phonemic and morphophonemic information.

Mudge. These varieties have been called different languages, including by this writer, but their similarity would also justify calling them major dialects of one North Wakashan language. To avoid the issue this paper will call them isolects. Each of the four isolects was at one time a group of regional variants and some of this diversity persists to this day, particularly in the case of Haisla and Kwakwala. Indeed the native terms of which "Haisla" and "Kwakwala" are anglicizations once did not refer to whole isolects but only to particular local variants. Pace junctures .
(1) A morphophonemic process involving an entity with many possible realizations that is discussed extensively in this paper.
(2) Reduplication of the wordinitial phoneme. 2 It should be noted that KW examples of /i̓ / or /u̓ / plus vocalic element are rare compared to those of /i/ and /u/ plus same, and they are plural forms such as the mentioned /mi̓ ṃgʷat/ and the following two words, KW /hi̓ inux̌ʷ/ 'the one who always does that' and KW /ǧin̓ i̓ isḷa/ 'salmon with roe in body'. 3 The /i̓ i/ in these two items is correlated with a stem ending in a palatal resonant adjoining a suffix beginning in one, an environment that is known to cause morphophonemic irregularities in all four isolects and does not involve the morphophonemic entity this paper is about.

Tones and Accent
A HE vocalic plain resonant that is neither preceded nor followed by another one has either high tone indicated by an acute, or low tone indicated by its absence, cf. HE /k ̓ ʷás/ 'mussel' versus HE /k ̓ ʷas/ 'sit on the ground outside'. The other isolects have accent instead. In HA it is free and to be indicated whenever not automatic, in this paper by the prime symbol before the accent bearer.
Thus, HA /ƛ'ix̌əzu/ 'board for beating time' but HA /c̓ xa/ 'to singe'. In OO/KW the accent is not written in the vast number of cases it is predictable with the simple rule stated below. Vocalic glottalized resonants never have HE high tone or distinctive HA/OO accent. They don't as a rule have distinctive accent in KW either but there are numerous exceptions, all explicable morphophonemically, e.g. KW /h'i̓ sta/ 'to go right around' in which /i̓ / stems from telescoping of /ii̓ /.
A series of vocalic plain resonants has only one tone or accent; the placement of the tone or accent in the series is not distinctive. morphophonemically but the rules are too many to be explained in this paper and the reader has to be referred to Rath (1985). It is useful to remember at least this. After consecutive vocalic plain resonants or a series of plain resonants with vocalic and consonantal alternating, no more high tones are possible. Allowing for this rule which overrides even the lexical high tone of some suffixes, any series of vocalic plain resonants takes one high tone while two plain vocalic ones separated by a consonantal one have high tone on both vocalic ones. Seeming exceptions to this rule have a morphophonemic cause. For example, HE /xálá/ 'come!' exhibits the default tone pattern whereas the high-low pattern of HE /p̓ ála/ 'to work, worker, etc.' stems from its /á/ being two identical plain resonants that have telescoped into one phonetically but nevertheless still block further high tones.

Gemination
Only OO and HA allow geminate vocalic plain resonants. They are phonetically long in duration.

Reduplication Boundary, Junctures
The reduplication boundary /:/ occurs in e.g. KW /łx̌:łx̌a/ (frequentative of KW /łx̌a/ 'loose or string-like things are somewhere') and signals that the following element is to be pronounced like word-initially; accordingly not only initial /ł/ but also its repeat is separated by a murmur vowel

Redundancy of Resonants' Consonantal-Vocalic Distinction
HE/OO/HA vocalic resonants are in complementary distribution with their consonantal counterparts; therefore the pairs /m, ṃ/, /m ̓ , ṃ ̓ /, /n, ṇ/, /n̓ , ṇ̓ / /y, i/, /y̓ , i̓ /, /w, u/, /w ̓ , u̓ /, /h, a/, /h ̓ , a̓ / can be collapsed into single phonemes while leaving it to rules rather than typography to determine when to pronounce consonantally or vocalically. In this paper these abstract single phonemes will only be used in morphophonemic transcriptions of the kind introduced in section 3. For lack of suitable symbols for resonants that can be consonantal or vocalic the consonantal symbols will represent them. As for the rules determining when a resonant is consonantal or vocalic, those for HA are in Lincoln & Rath (1986:4-8) and those for HE in Rath (1981:61-62). The OO rules are like those for HE except that OO does not allow 3 consecutive vocalic plain resonants with /a/ in the middle (see also the beginning of 2.4). The collapse into single phonemes is difficult for KW due to the split in two of the vocalic palatal and rounded labial resonants. However, these KW idiosyncrasies result from morphophonemic processes and are not part of morphemes' sound constituents. The consonantal-vocalic distinction can therefore be dropped from KW morphophonemic notations too as will be shown in what follows.

Morphophonemic Essentials
North Wakashan roots and suffixes (there are no prefixes) exhibit a great deal of allomorphic variation. The root allomorphs in the following three data sets illustrate variation associated with a morpheme boundary (indicated by a hyphen). There is also a type of variation associated with reduplication of all or part of a morpheme but it is of no concern in this paper.  Rath (1980:36-45), pages written in response to Vink (1977).

HE
A morpheme's base form can become modified by environment-triggered rules. For example, in HE/OO only, a plain plosive that lands morphophonemically before any other kind of obstruent or word-finally, will turn into its fortis counterpart. That's why in rows 4 and 6 the transitivity marker {-d} becomes /t/ in HE/OO. An all-North Wakashan rule is that a fortis plosive landing directly before an obstruent will turn into its homorganic fricative (with special provisions for |p| and |t| that need not concern us here and exemption for reduplicating initial fortis plosives). A good example is |qʷ| > /x̌ʷ/ in KW {dwqʷ-bh} > /dux̌ʷba/ (glossed by the sources as 'to see end, to see point'). Not a good example is |qʷ| > /x̌ʷ/ in the items on row no. 4 because in HA/KW directly before /s/ the opposition is neutralized between a non-initial velar or uvular fortis plosive and its homorganic fricative, see Lincoln & Rath (1986:11). In HE/OO the opposition is not neutralized but there is another complication, the low tone of HE /u/ and the glottalization of OO /u̓ /. It forces recognizing that the row 4 items contain a suffix {-sw} in HA/KW but {-'sw} in HE/OO. The element |'| is assumed to glottalize the plain resonant |w| in {dwqʷ-} and, generally, stems (not necessarily just roots) ending in a fortis plosive preceded by a plain resonant that in its turn is preceded by an obstruent, glottalized resonant or word-initial plain resonant. |'| has no diagnosable effect on stems ending differently. Thus, HE/OO {dwqʷ-'sw-d} > {dw ̓ x̌ʷ-sw-d} > HE /dux̌ʷsút/ OO /du̓ x̌ʷsut/. The step from {dw ̓ x̌ʷ-sw-d} to HE /dux̌ʷsút/ is governed by the tone rules which have to be taken for granted in this paper, see 2.5. This |'| is just one of a set of environment-modifying morphophonemic elements that will be referred to as adaptives in this paper. The set is listed below and will give the reader an inkling of the complexity of North Wakashan morphophonemics. The following abbreviations are used. P pl , P ̓ , P fo mean the 3 kinds of plosive. F is a fricative, O any obstruent, C any obstruent, glottalized resonant or root-initial plain resonant, R a plain resonant and R ̓ a glottalized one. The description of their effect is broad; for example, ignored are constraints on the obstruents following HE/OO |'| and |‼| and the fact that some adaptives are suffix-like and can occur word-finally.

Isolect
Adaptive Diagnosable Effect Attested before |k| and |s|. Imparts high tone to stem-final R. If stem does not end in R but |↔k| or |↔s| is followed by R, the high tone goes to the latter R. |↔| is overridden by the general rules mandating or prohibiting high tone, see 2.5.
Special case: Ch-=h and Cy-=y behave irregularly in HE/OO/HA.
Special case: if RR is Rh, |=| may also be ineffective while in KW there is a third option, Rh-= > R ̓ - Illustrating all adaptives of the chart is desirable but not the aim of this short paper the remainder of which is about |∞|. First we shall look at it as byproduct of other adaptives and next at its occurrence as lexical component of morphemes.

The Effect of |%| and |=| on a Fricative
|%| and |=| coalesce with |s, ł, x, xʷ, x̌ʷ| into |y∞, l∞, n∞, w∞, w∞|, respectively, and note the identical effect on |xʷ| and |x̌ʷ|. As demonstrated below, |∞| can become a replica of the plain resonant before it, turn it into a glottalized resonant, turn itself into |w| and eventually /u/ (if a rounded obstruent follows), turn into |y|, turn into /ə/ (in HA and KW only) or, if followed by a plain resonant in vocalic position (as when adjoining an obstruent or the word end), become It is unknown if the above special cases triggered a search for more examples when first recorded. Perhaps alternation between e.g. KW /kʷṇu̓ gʷa̓ ł/ and /kʷṇw ̓ əga̓ ł/ or between KW */hιḷ ̓ s/, */hἰḷs/ and /hiḷ ̓ s/ was once common but we may never know.

|=| and |+| after a Single Plain Resonant
|=| and |+| usually both have the effect of glottalizing a single stem-final plain resonant, "single" meaning "not preceded by another non-initial plain resonant." Compare (1) .)). See also example 13 in 4.1. Lengthening and postglottalization of a vocalic element are just two of the many possible effects of |∞| in North Wakashan but it is interesting that alternation of these two phenomena in particular seems to be wide-spread geographically, allowing for the possibility of the glottalization coalescing with the next phoneme. Jacobsen (1968) reports the alternation for South Wakashan.

The Heterogeneity of the effects of |%|, |=|, |+|
One of his examples is the word for 'picking berries'. Using his transcriptions which employ /·/ to denote length, one isolect has /ča·yax/ but another /čay̓ ax/. In the latter the length feature of /a·/ has been replaced with a glottal stop that becomes the preglottalization of /y̓ / rather than the postglottalization of a short vowel */a̓ /. Another example is the length feature in /t ̓ i·las/ 'wet spot on ground'. The length is absent in e.g. /t ̓ ilis/ 'wet spot on beach'. The cause of its presence in the root allomorph /t ̓ i·l-/ is an adaptive in the suffix that is known to glottalize non-glottalized plosives, compare /t ̓ iqʷis/ 'to sit on the beach' and /t ̓ iq̓ ʷas/ 'to sit on the ground'.

Suffix-Final |∞|
Suffixes, like roots, can end in two plain resonants plus |∞|. An example is {-yw∞} 'on forehead'  There is no KW example in the database for (2b) and (3c) but this doesn't mean they don't exist.
KW forms with plural reduplication are not abundant in the database and systematic KW plural form elicitation would be a worthwhile project. What is certain as of this writing is that there is no KW example in which reduplication of a root-initial plain or glottalized resonant gives rise to a phonemic word with /ι, ἰ, υ, ὐ/ before a vocalic resonant. We find KW /niṇłḷa/, for example, and not */nιṇłḷa/. It is not clear why reduplication blocks the use of /ι, ἰ, υ, ὐ/. Perhaps it is due to the association of the augmented root with the simplex one and the pronunciation of its |y| or |w|.
A case, in other words, of analogy trumping morphophonemic rule.

Inconsistent KW Prohibition on /a/ before Another Vocalic Resonant
In HE/OO/HA /a/ can be followed by another vocalic plain resonant (not a glottalized one, see low tone in the HE item and the doubling (= lengthening) of /a/ in the OO item point to a structure {ghhlh-} and how this analyzes is an unsolved problem that however need not concern us here.
Example 10. KW /qast'ai/ is the vocative form of /qast/, an address used for male friend of male.

The Inevitability of the KW Vowels
Some of the cases of /a/ = [a] followed by a vocalic element could be of foreign origin, for example KW /haida/. Or they could be dialect borrowings. Boas (1947:205) reports that the KW he describes is the dialect "spoken in the area from Fort Rupert south, and on the small islands at the entrance to Johnson Strait" and that it "has encroached considerably upon the other dialects." Among these other dialects is that of "the most northern part of Vancouver Island as far south as