THE ALSEA NOUN PHRASE

The structure of the noun phrase (NP) in Alsea, an extinct language of the Oregon coast, is examined with particular attention to the behavior of a clitic occurring in second position within the NP. A presentation of the basic facts includes the following: referential(s) and the deictics, possessive pronouns, third-person po3sessive, the ergative, modifiers, and the dative morpheme. A formal analysis is then provided, although because it is based on incomplete data limited by the contents of the available texts, certail' issues cannot be resolved definitively. The analysis covers Klavens' parameters, the basic structtlre of the noun phrase, and the clitics. It is concluded that dative is lexically marked so that the preposition is forced to move to the end of the clitic group. Contains 11 references. (LB) *******0*************************************************************** Reproductions supplied by EDRS are the best that can be made from the original document. *********************************************************************** DE DEPAATINENT OF EDUCATION Office of EducaNceel Rusaterl Arte Imormement EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC) iiMa document Ass been reProduced as waved horn the person or argent:short omprtatmg it CI Minos cganpes rtaye been mute to Improve reproduction outlay Pants of veer or opinions stated in tots clocumerit do riot necossarry represent officio; OEM 'gagman or pokey As.tairssluslisa THE ALSEA NOUN PHRASE Eugene Buckley "PERMISSION TO REPRODUCE THIS MATERIAL HAS BEEN GRANTED BY TO THE EDUCATIONAL RESOURCES INFORMATION CENTER (ERIC)" This paper examines the structure of the noun phrase in Alsea, an extinct language of the Oregon coast, with particular attention to the behavior of a clitic occurring in second position within the NP. I will first present the basic facts and then consider an appropriate formal analysis. Sirce this analysis is based on incomplete data limited by the contents of the available texts, there are certain issues which cannot be resolved definitively. The basic facts, however. are clear and interesting enough to warrant an attempt at a formal treatment)

The Abet noun phrase has the following basic structure, where It) and IS) are elements which will be explained later: (a) DETERMINER modifier-it) Isi-noun The determiner further has the following composition: (b) DEICTIC-REFERENTIAL In addition, the entire phrase may be preceded by an ergative case marker These elements will be discussed in the sections below.
1,1 The reervaial le and the deictics use: A noun can occur by itself; generally this is in a nonreferential, generic By far the most.common case, however, is to have both a deictic and the referential present.There are three deictic prefixes relevant here, of the forms lal.(kul, and Its, tel.They indicate the position of the noun in physical space but also seem to serve discourse functions as well; since their precise meaning is not important for the present purpose, they will all be glossed DEI.
Following the deictic element is the referential Isl.Together they form a çt determiner which precedes the noun: (2) tanys-x ku-s Occasionally a noun is found without the referential but with a deictic marker.Often these examples are in quoted speech and refer to characters in a story.It is assumed that this omission serves some pragmatic function, since under normal circumstances an object must be referential in order for its location in space to be specified: (4) te-714-axan ku ma.lict .
TR-refuse-CMPL-IsgS DE1 sea.otter 'I refuse (to marry) that sea otter!' (92.13)There are also a few examples of the referential NJ without a deictic.This may at times be a reduction of Iasi, but in some cases such as (s-lahwil 'the world, the earth'.the vowel is nearly always omitted and the entire phrase may be lexicalized.In this particular csse it seems reasonable to omit the deictic since 'the earth' is always in the same place. Deictics can also be added directly to emphatic pronouns, though this use is not tommon since the position of, say, the speaker is generally obvious.In all of the examples above, the referential (s) can be replaced by a first- or second-person possessive pronoun The possessive pronoun and the referential never cooccur, presumably because for a thing ta be possessed it must be referential (i.e. the NP must have a specific referent).In other words, the possessive pronouns are really more specific variants of the referential.
(6 ) ci.s-tuxs DEI-I pIP Ivin-COLL 'our winnings' (26.17) (7) qi lint DEI-2p1P younger .sister'your younger sister ' (22.10)Often the possessive pronoun occurs without any deictic, probably because the indication of the possessor is usually sufficient to identify the referent without specifying its location.The remarks above with regard to the first-and second-person possessive pronouns are largely true of the third-person possessive circumfix The initial element ici appears prefixed to the noun, while the final 41 appears suffixed to the noun.Most frequently there is no deictic: (9) c-hty-jr If a noun indicating the possessor is present, then it generally occurs before the possessed noun (which is marked with the circumfix).Ir this case the referential Is) does appear at the beginning of the.NP, but it seens to refer to the possessor and not the entire NP.This is because the head of the NP is already marked as referential by toi: The correct generalization seems to be that if the possessor precedes the head, then the NP as a whole is not marked with a deictic if the possessor follows, then the head (and therefore the whole NP) can be marked.This restriction may be related to the proposed origin of ici as a deictic and referential determiner.Note the similarity to the possessive pronouns, which also generally occur without a deictic: the same explanation, that they are subtypes of the referential, plays a part with the third-person possessive as With multiple possession the head comes to the right or its complement in each case: Here again the determiner Ikusi seems to refer lust to the dependent noun (s'Crku) and not to either of the possessed nouns.Unmodified nouns (including proper names) have a general requirement for a determiner.
while nouns with Ic...tI are already determined: thus the (kus1 must refer to Suku, and must not refer to the other nouns.Since the referential and the possessive pronoun are in parallel structures in all ways examined so far. it seems reasonable to argue that in ( 12) and ( 15) as well the two are parallel and both modify the immediately adjacent N. not the higher argument NP.

I 4 The ereative
Nouns which serve as the subject of a transitive verb are preceded by the ergative marker km sal This proclitic can occur with or without a deictic, but when it is used with a nonpronominal it (like deictics themselves) always seems to combine with the referential Is) or one of it: possessive substitutes (17) iimxad-u ERG-DE1-lsgP child-PL 'my children' (86.17It may be correct to say that the ergative always occurs with a deictic, since instances of Nal are phonologically ambiguous between the simple ergative on the one hand and the ergative and an elided deictic (al on the other.Not enough is understood of the use of the deictics to give direct contextual evidence for or against the underlying presence of the tal in such cases. The ergative comes at the beginning of the NP and not before the head, though of course ergativity is the property of an argument and thus of the heed.not some dependent 110LIA: In ( 23) the ergative (sal refers to the entire NP, which is the subject of the sentence, while it seems best to say that the referential Es) refers to the immediately following noun, as discussed above.In fact, the underlying structure may be isa-a-si.where the refers to 'earth people' and /za,/, which is procliticized on the entire phrase, refers to the argument NP 'children of the earth people'.

LiaiggiLien
The label 'modifier' given at the beginning of this paper refers to adjectives and numerals and occasionally nouns.The numerals (which seem to be a subclass of adjective) are always followed by the adjectival suffix I-13 when in attributive position.Adjectives used attributively generally also take this suffix but not always; the conditions determining its occurrence are not entirely clear to me.but it appears to be a lexical property of the adjective itself.
In addition to the adjectival It) which is suffixed to the modifier, there is a formative Isi which appears prefixed to the following noun.This Isi can be found even when the [ti is absent.It is homophonous with the referential marker, and may historically be related to it, but since the two can ceoccur, and are found in different places, it seems unreasonable to call them the same morpheme.
Note, however, that a modified noun often occurs without a determiner, a fact which may be a result of this historical origin (more on this in section 3): The last example shows a noun used as modifier; this is much less common than the adjectives exemplified in ( 25) and ( 26).I have glossed the Isi as CSTR for construct, since its function is similar to that of the Semitic consul -t state, which marks a noun as being 'in construct with' a dependent noun.The Alsea morpheme indicates that the head noun is modified by the preceding word -whether adjective or noun.Although in nearly every example this fil occurs before a noun, this may be an accidental result of the fact that all the nouns have just one modifier.The one example I found of a noun with two attributive adjectives shows a somewhat different situation: (28) qiwx-at s-ma7s7qayt-it high-ADJ CSTR-rich-ADJ CSTR-person   'a rich person living in the sky ' (228.27)It seems here that the clitic is) is relatec to the adjective preceding it rather than to the word following it, which can be either a noun or another adjective.This notion receives support from the fact that certain modifiers (i.e.quantifiers) do not occur with this fsl.though they do occur with the referential is) found with deictics: (29) meqamin-t hitsiam many-ADJ person The modifier Is) might thus be argued to be the type of chile described by Kievans (1985) which is syntactically dependent on the previous word but phonologically dependent on the following I will claim later.however.that the construct analysis is correct The modifier of the noun is considered a separate constituent for the placement of a second-position sentential :laic, such as those marking the subject.That is, these clitics occur after the first word or the sentence, even if this involves splitting an argument: See also (30).Examples such as these, as well as the fact that they do not always cooccur.motivate the separation of Cti and Is1 into different formatives, even though both are present due to the preceding modifier.
The possessive prefix fci patterns with the construct Is) in that it is prefixed to the noun: for this reason modifiers precede it.Note that when the prefix (el is present the construct Is) is absent: This situation is parallel to the referential Es) which also cannot directly precede the possessive prefix.
Due to the limitations of the data available for Alsea, there is no good evidence for deciding whether the first and second-person possessive pronouns belong to the same pattern as the construct is) and possessive fa That is, it is impossible to say for sure whether an adjective would precede or follow the possessive pronoun.In his field notes at the Smithsonian, Frachtenberg gives the following constrast (the parse is mine): (36) qan-1-sal sin ta.?
die-INCH-DSTR IsgP father 'My father is dead: (37) qanist-it sin tr.? dead-ADJ I sgP father 'my dead father' The first example, (36).I would translate 'My father died (long ago)', since the distributive suffix /sal/ is used to mark the remote past and the perfect aspect (Buckley 1986).At any rate this is definitely a verb-subject construction.In (37) the situation is not so clear.I am confident that the first word is an adjective, but it is not clear that thir is a noun phrase.Frachtenberg puts a question mark next to these two sentences as though he is not sure of the translations.Since there is no copula in Alsea, the adiective could easily be predicative rather than attributive, in which case (37) would be a sentence meaning 'My father is And.' Thus it is difficult to be certain even relevant to the present discuss'on.
In the published texts, there Is no clear example of modifier and possessive pronoun cooccuring, except for the which is free to float outside the NP anyway: (38) Here the element Iytixwi'si precedes the possessive pronoun.But again, the significance of this fact is unclear since brim/ill is not a typical modifier.it takes no It1 suffix here, and has the internal morphology of a noun.I have also found no examples of ()quail) serving as a modifier of a noun with a deictic-referential determiner, with which (40) and ( 41) could be compared.
One fact argues in the opposite direction, that the possessive pronouns pattern like the referential [4 and should precede modifiers: the transcription that Frachtenberg uses.He writes both the possessive prefix I cl and the construct Is) as part of the word that follows -consistent with the common syntactic patterning described above -but the deictic-referential determiner and the possessive pronouns as separate words, suggesting that perhaps they pattern together as well.Since they are also found with (secondary?) stress as in (17) they are unlikely to be prefixes on the noun: I think they are probably clitics which occur at the beginning of their noun phrase.
It may be that there are other examples in Frachtenberg's notes which bear on this issue, but for the time being I will assume that the possessive pronouns pattern with the referential.
There is a dative morpheme WI, most often indicating rotion towards the noun tallativel.which, in an unmodified noun phrase, occurs directly after the noun.Note that this is the situation whether there is just a noun, or a noun with a determiner: one-ADJ-DAT place 'to one place ' (46.11)In an NP where the head noun is preceded by a possessor NP, the dative comes after the possessor: (49) ku-h am t1-ica-itx2 DEI-your father.DAT-PL POSS-sweathouse-POSS 'to your fathers' sweathouse ' (130.29)Note that in (49) the movement is toward the sweathouse.not 'your fathers', even though the dative comes after the latter element.
There is no example of a modified noun with both a determiner and the dative present.This is not terribly surprising since neither the determiner nor the dative is found very frequently with a modified noun in the first place.Still, due to this gap in the data we cannot be absolutely certain that in an example such as (26) the dative would come after the adjective instead of after :he determiner or even somewhere else.But given the behavior of the dative ia ( 45) and ( 49), where the determiners are ignored for the definition of 'second position', it seems safe to suppose that the dative would in fact come after the adjective in ( 26). 1 will make this assumption in my analysis.the few examples available are with locative adverbial,.There the adverb (ci-mal 'very (much) seems to be ignored for the placement of Its): (30) Mica I cilium ni.sk-ip J ay-iy just very far-DAT go-1NCH 'he vent very far (163.9)But ki.ma1 is also found completely separated from the phrase it modifies.making it unclear where it should be positioned in the structure: (51) ni-sk-Ms hlka caws ay-iy far-DAT just very go-1NCH 'he went very far ' (168.15)The same facts relative to fis! seem true of another adverbial modifier.'a little': (52) zersi-clawzin-jcs clittle above-DAT The moved] a little higher ' (76.35)Since the adverb in many cases occurs rather freely separated from what it seems to modify (similar to the quantifiers), and since these examples are adverbial and not nominal, I will not attempt to incorporate them into my analysis.A phrase such as 'to the very big house' would be necessary to determine with any certainty the behavior of these adverbs.

2.2.Amdira-12.0211d.--1221111211namnahriar.ral=
There is a common clitic fawkcJ 'in.inside' which seems to have several possible domains: S. V. and NP.Although it is clearly attested in only one example, it appears that (awki can behave like fg.$) when it is prepositional (53) zinact-awt one-ADJ-in CSTR-year 'for one year ' (120.26)This clitic shows up in the same position as the dative, and can be analyzed in the same way.3tinityiLs I will now offer a formal analysis ot the data discussed above Unfortunately it cannot be pursued with complete thoroughness because the data necessary to check the validity of certain predictions are simply not 1 1 available in the existing corpus.Still, it should useful to try to formalize what has been described.

It I_ Mavens parameters
In the typology of clitics in Klavans (1985), it is possitie to specify the behavior of a clitic with three parametert PI) Dominance -whether the clitic is positioned relative to the initial or final constituent within its domain; P2) Precedence -whether the clitic is positioned before or after that constituent; end P3) Phonological Liaison -whether the clitic attaches phonologically to the preceding or following word (i.e.whether it is an enclitic or proclitic).
Klavans allows the specification of the domain of the clitic (the constituent to which P1 makes reference), including S. V, and N', but does ?lotexplicitly integrate this option into her set of parameters; the domain should perhaps be thought of as a fourth parameter, different from the others since it is not a binary choice.Among the eight clitic types possible given the three binary parameters, the dative [gill is a Type 3 clitic it encliticizes (P3) after (P2) the initial element of its phrase (P1).Unlike the common second-position clitics which occur in this position under 5, however, the Alsea clitic occurs under NP.
There is a potential problem in this description of the placement of lics).where do the determiners fit in?Recall from ( 44) and (45) that t.,..ey are ignored for the purposes of determining the initial constituent of the NP.If we consider the determiners to be clitics themselves, then they could automatically be ignored by whatever rule places Iksi after the first (nonclitic) constituent of the NP.Note that the determiners would be Type 2 clitics, which procliticize before the first element of the phrase.There is some independent evidence that the determiners are in fact clitics.For example, both the deictics and the referential are found as part of the same stress group as the following full word.The only time stress is marked on a determiner is when it is more than one syllable, in which case it probably reflects the relative stress on the two syllables of the determiner itself -a secondary level of stress which is subordinate to the primary stress of the word to which the determiner clitic attaches.3There is no direct evidence for this stress being secondary, but note that while unusual it is not impossible for a clitic to receive stress (Wanner 1978).The Klavans typology is not completely satisfying as an analysis because it simply describes the position of the clitics without offering a formal explanation of how they got there.I will now propose an analysis which accounts for the behavior of the clitics u well as the overall structure or the noun phrase as described above.Although it is not central to the analysis.I will assume Government-Binding Theory (Chomsky 1981).aure The deictics, the referential Isi, and the possessive pronouns are base- generated in specifier position in the NP.They are created from their component parts, e.g.Nu) and (s), by the morphology.These determiners are lexically specified as proclitics, so that they must combine with a full word at some point in the derivation.4Adjectives are adjoined to the noun that they modify.This allows for the iteration found in example (28).The adjectival It) and construct (s) are generated by a structure-sensitive rule similar to that which generates the English genitive: an adjective or noun adjoined to an N takes the suffix It), and the adjoined-to noun takes the prefix (s): Exception must be allowed for those modifiers (such as quantifiers) which allow the omission of one of the affixes.
There is a alight complication due to the fact that when a noun already has the possessive prefix (c) it does not take the construct Is) or referential Isi in addition.This exclusion Can be explained naturally by treating (c) as carrying the features (*referential, construct), thus making either Is)   redundant.These features are semantically motivated since a possessed noun is necessarily referential, and is in construct with its possessor (whether phonetically realized as a separate NP or not).
Note that under this analysis all nouns, including those consisting internally of an adjuncuon structure, must be marked with the construct prefix' This structure reflects the data given in (28).where the terminal string comes out as A-t s-IA-t s-NI.Since the rightmost adjective-noun combination is itself a noun, we are not required to say that the first construct Is) is prefixed to an djective.
As mentioned earlier, one misfit want to argue for an analysis which equates the referential and construct Is).For example, there is often no referential Is) before a modifier-noun combination, as in (29).so there is some motivation for saying that the Is) preceding the noun is actually the referential.This is problematic, however, because it is possible to have more than one construct Is) when more than one modifier precedes the noun.If this construct is really the referential then the second instance is redundant -it seems in fact more structural (construct) than meaningful (referential). in addition, the familiar deictic-referential combination appears when a deictic is present.e.g. ( 26), and this does not affect the presence of the construct.Thus there would have to be some ad hoc rule stating that a deictic preceding a modifier-noun combination takes a redundant referential suffix -even though such a redundant suffix is not used before the possessive (c).Of course, under my construct analysis it is still necessary to say that sentence.The same is true of the prepositions ir (37).which share membership in PP with the following A or N.
The attachment of the determiner to the following word is straightforward; since the determiner is base-generated in front of its chosen host, it need not even be marked lexically as a proclitic.The preposition (01, on the other hand, must be marked as an enclitic.and this specification must be satisfied at the level of Phonetic Form.The preposition chooses its host word in the normal way: according to maximal shared category membership.At PF. however, it moves to the end of the clitic group to satisfy the feature (*after), to use Klavans' term for the precedence parameter.5 In the illustration below, W refers to a phonological word, the smallest unit to which the prosodic hierarchy can refer.Each C is a clitic group: every lexical category is automatically the head (host) of a C. Higher levels discussed by Hayes will be ignored here.The preposition WI is associated with the following C by the principle of maximal shared categories (58a): At PF the preposition moves from the left side of the clitic group to the right side, not changing its association but merely its position relative to the host (58b).Note that this movement does not result in crossed association lines since the clitic and its host are associated with the same C node, not adjacent ones The Id vowel in the output ( 46) is inserted by a postlexical rule of epenthesis whose domain is the clitic group.
The movement of the clitic is force.1 by the (*sifted feature.If one wants to have free movement at PF, as in the syntax, then the determiners and similar clitics would have to be lexically marked as (*before) to prevent them from undergoing the same movement.If movement if, allowed only when necessary to satisfy feature requirements, then no lexical marking is needed for the simple cases.This same explanation of the behavior of fts) can be applied to the clitic lawg) in (53).
Although this is not an issue in Alsea, Mavens discusses a number of examples of clitics which have divided loyalties: the word which is relevant for the positioning of the clitic (also the one with which it shares the most category membership) is different from the word to which it attaches phonologically.ln other words.Hayes' rule of clitic attachment is violated In Kwakwala.for example, case markers occur before the noun they mark but attach phonologically to whatever word precedes them.I assume that clitics like these (which are fairly uncommon cross-linguistically) are lexically marked as exceptions to the normal method of clitic attachment.For example, the Kwakwala case markers can be specified as always associating with the preceding clitic group.or (*enclitic) in Klavans' phonological liaison  parameter.
clear what role adverbs play in the noun phrase.The dative has not been found together with an adverb in attributive position before a noun: It appears that historically the initial Ici may come from the deictic- referential Rasi, but synchronically this (c1 is simply required in order to mark a noun as possessed by a third person.Interestingly, though, it seems to have retained its referential meaning and is thus parallel to the possessive pronouns: no referential Is; cooccurs with either of them.Thus one finds the simple deictic, including the possible cognate Itoi.attached directly to Id DEI-POSS-head-POSS'his head' (120.25) Compare the following example.