Western Conference On Linguistics WECOL 91

visibility

…

description

347 pages

link

1 file

Women provided me with financial support which allowed me to devote my full attention to writing my dissertation. The inadequacies in this paper are my own responsibility.

Figures (51)

The combinations of flexion and extension of these joints yield four logically possible configurations of the hand. First, the fingers can be "extended", as shown in (4), in which there is no flexion at MCP, PIP or DIP joints, fingers is the knuckle, or metacarpophalangeal (MCF) joint, shown in (3a). The second joint on the I, M, R and P is the proximal interphalangeal (PIP) joint, shown in (3b). Finally, the joint closest to the fingertip of the I, M, R and F is the distal interphalangeal {DIP} joint, shown in (3c). The PIP and DIP joints generally function as a unit (Brand, 1985).

{picture from Tubiana, 1981; my labelling of extensors)

Af® (the head of AfP) is a positien to which a verb has to move in order to take (part of) its verbal morphology (infinitival. participial and progressive suffixes or V-class suffix -the so- called thematic vowel in Romance),”

In addition to the rhymes listed above. Yanggu has the syllabic retroflex lateral { and the syllabic nasal m. bringing the total of rhymes to thirty- nine. 1and ¢ are apical vowels. with the latter being retroflexed. Nove that “is useu to denote both the back mid vowel and the voiced velar fricative. As usaal, a tilde denotes nasality. All vowels are nasalised the svinbot + ins at esting phonuiog ta Mandarin Chinese dialect. has inter jeal behavior which bears on issues of prosodic licensing and the struct eof faterads. In tidy article T exauone the data and slow that the ssHable in Yangau leenses oue articulator each fur the onset and the nat be placed below the coronal node, TIT, ated ble feature ‘latera

a bisvlabie word from a monoss]- This is Type ive aliixation aiso general HWothe stem contains a pre-vocalic glide ¢ or y.

The grammaticality of the Psych construction (24) below shows conclusively that the dative (inversion) nominals can float quantifiers suggesting, rather unequivocally, that these nominals are not initial 1s, in fact they are final 1s. In Kashmiri ONLY final 1s float quantifiers. The grammaticality (23a) suggests that 1s (subjects) can float quantifiers while the ungrammaticality of (23b &c) suggests that 3s (indirect objects) and 2s (direct objects) canNOT float quantifiers, respectively. (23d), a Passive construction, shows that a final 1 can float a quantifier whereas (23e) shows that an initial 1 is unable to float quantifier.

advances to 1 placing the Theme, an initial 2, en chomage. This proposal of advancement is represented in (33). Under the proposal outlined above all the behavioral properties of dative subjects in Kashmiri, such as Quantifier floating, Control PRO, $-S-R and ECM can be stated in terms of its final 1-hood. The quirky “dative” on the Experiencer subject can be accounted for by following the proposals in Gerdts & Youn (1989). They make the distinction between I-Case [=inherent Case in GB theory] which is selected on the basis of semantic role of the nominal and licensed in initial structure, and $-Case [=structural Case in GB] which is the grammatical Case and licensed in the final structure.

Given this generalization, there is no need to stipulate that the specifier of CP is on the left - all possible permutations of the CP projection are instantiated. Furthermore, the location of the specifier and the level at which wh-movement takes place are intimately tied together. Any language which has the specifier of CP position on the left will have overt wh-movement (and vice-versa), while any language in which the specifier is on the right will have LF wh-movement (and vice-versa). This then also accounts for the absence of overt. specifiers on the right.

Figure 1: The output of the parser for the sentence “What did the man say the woman hit?”: a gap-locating process.

Figure 2: The output of the parser for the sentence “Ni shuo ni kanjian le shenme?”: a gap-creating process.

In the initial structure assigned to such clauses by Kimenyi both the Locative and the instrumental of initial obliques and both advance to object. A priori, either instrumenta advancement could be earlier than locative advancement, as represented in (49a), or vice versa as represented in (49b). In fact, Kimenyi supplies the data that allows us to choose between these analyses. As predicted by (49a) but not (49b), only the Locative nominal shows the properties of final object in multiple applicatives. So, for example, the locative can be the subject of a passive, as in (50), but the instrumental and the initial object cannot be, as (*51) and (*52) show.

4NCTETSSEIANgsy, EME GASLTINCCION Dercween “accusative and "dative" verbs with regard to clitic doubling is lost when the "argument" of the verb is a pronoun. As shown in (12), the accusative assigning ‘ver’ and the dative asSigning ‘pegar’ must surface with a clitic when they take pronominal objects.

Let us now consider the effect of cyclic palatalization on the voiceless velar stop. This segment has a fronted counterpart in the UR, as demonstrated in (8): Now, if fronted velars were indeed palatalized counterparts of plain velars, then, given that cyclic palatalization is structure preserving, we would expect it to derive fronted velars. This effect is observed commonly across languages: if the segment inventory of a language contains a palatalized counterpart (e.g., Ati of the segment targeted for palataliza- tion (A/), the result of palatalization is the palatalized counterpart (/ti/) (see Gorecka (1991) for further discussion). Contrary to this expectation, in Russian, velars turn into palato-alveolars under cyclic paiatalization:

Let us now consider the representations which a palato-alveolar, a front vowel, a plain velar, and a fronted velar receive in this model:

It turns out that the voicing may be triggered by material from any level of the morphology, as seen in (9) and (10), where it is induced by the epenthetic a, added at level 5 according to Hargus, or by a vowel from a disjunct prefix, also added at level 5. Another rule which leads to the same problem is that of Voicing Assimila- tion. The voicing of stem-initial fricatives in verbs is predictable on the basis — of the preceding segment: they agree. This can be seen in contrasts such as in (7) and (8), where the initial fricative of the stem appears unvoiced after an unvoiced segment and voiced after a voiced segment.

The solution to this proposed by Hargus is again to retain level 1 brackets (that is, make them exceptions to the BEC) and formulate the rules to make reference to these brackets, as indicated in the formulation of VA given in (13). Voicing spreads onto a fricative following a retained stem bracket.

These rules must satisfy their structural descriptions entirely within the conjunct prefixes, supporting the claim that these domains constitute mor- phological constituents. This is illustrated by the case of Conjugation Schwa Deletion. Simplifying a bit, the conjugation morphemes (position 10) are re- alized as C when they are final in the conjunct domain and are preceded by other conjunct prefixes. Otherwise, they surface as Ca. This alternation is illustrated in (18) and (19).8 Hargus’s version of the rule makes reference to level 1 brackets, but we can treat this as an alternation which is sensitive to the edge of the conjunct constituent, rather than brackets around the stem, as in (21). The Conjugation schwa alternation and the other rules in (19) confirm the first prediction.

However, this causes problems for the description of several alternations in the surface form of the subject markers and the conjugation morphemes. The surface form of the s-conjugation morpheme and of the subject markers, when they are the final morpheme in the conjunct domain, is sensitive to the identity of the following classifier or to its absence; the alternations in question involve the rules of S- Voicing and Aspiration (p. 90), details of the application of the Conjugation schwa Deletion rules (pp. 142-51) and allomorphy of the subject markers (p. 89). I will exemplify only the alternation Hargus dubs s Voicing; the other alternations raise similar issues. There is one set of affixes, namely the position 13 classifiers, which are some- what ambiguous with respect to whether they count as part of the verb stem or as part of the conjunct domain. On the basis of two rules (D-effect and Palatalization), Hargus assigned them to level 1; that is, they attach to the stem more tightly than any other prefix. The fact that they trigger the D- effect and Palatalization, rules which only apply between the classifiers and verb stems, is explained by assigning both the morphemes and the rules tc level 1. One way to interpret this, under the compound analysis, is to treat the classifiers as part of the stem constituent, rather than the conjunct con- stituent. That is, an example such as (27) should receive the analysis in (28).

As seen in the preceding examples, the conjugation prefix (position 10) appears as {z] immediately before the stem, but [s] before the d-classifier. That the underlying representation of the s-conjugation morpheme has an /s/ rather than a /z/ is indicated by the fact that it surfaces as [‘sa] when not the final conjunct morpheme - see (21). Given the stem-core analysis, the rule (be it phonological or allomorphic) which accomplishes the voicing would have to be able to make referetice to the stem, or be able to identify the classifier prefixes, once again requiring retention of level 1 brackets. Hargus’s formulation is given in (31).

Templatic description of the Sekani verb (modified from Hargus 1988)

Now, a baffling counterexample occurs in (19), which is fully grammatical with ezafe preceding just the kind of element it normally cannot accompany; in fact, (19) is identical to (17a) -- except for the absence of the adjective jaleb, "interesting""

Two types of evidence seem to suggest that these suffixes serve the same function as full NPs, First, the suffixes occur in the same string positions as nominals with the same grammatical function - immediately after a lexical head. This can be seen by comparing (2) - (7) with (8) - (13) below, which have nominal arguments, The distribution of these person markers is illustrated in (2) through (7) using th Ist person singular suffix /-y'/. Attached to a noun, as in (2), the suffix indicates the possessor. Attached 1o a preposition, it may realise an oblique argument of a verb, as in (3), or the object of a preposition, as in (4). Attached to a verb it may indicate either a subject, as in (5) - (6), or an object, as in (7). 4

It is useful to first consider the prediction with respect to long-distance extraction in analyses like Sproat’s according to which the verb has moved to C. Suppose the finite verb is in C as shown in the representation in (18): The movement to the SpecCP of the embedded clause is again not at issue. However, if the verb is in C, then the movement from the embedded SpecCP to the matrix SpecCP crosses only one bounding domain, namely, the embedded CP. The matrix VP, and IP are not bounding because the heads of these projections are empty. Thus, under the view that verbs move to C in Welsh, the derivation in (18) is fooele, an incorrect result, It thus appears that the finite verb is in INFL rather inl. asiiaaaiieadcsicimemmninin: — tuuleniiadealici:tnisitaie: aehaanliaal ieee, “Tiel. abled” J Suppose CP is (potentially) bounding in Welsh (cf. Law (1991) for a discussion of how this can be derived in a principled way). Since the heads of the CPs in (17) are not empty, these projections are indeed bounding domains according to (15). Now, if we indeed have reason to think that the finite verb in (17) is in INFL rather than in C (as I have been assuming), then the IPs in these examples would be bounding. The movement from the embedded SpecCP to the matrix SpecCP would then constitute a subjacency violation.

Such configurations are ruled out by the No Crossing Constraint (‘NCC’) Goldsmith 1976). The conclusion is that the epenthesis rule is prevented from The basis of the current AP account of formal integrity originates with ar »bservation of Kaye (personal communication cited by Halle and Vergnauc |980) and independently Kenstowicz, Bader and Benkeddache (1982). Wher ome segment /i/ is inserted into the true geminate of (lla), the geminate -annot house the epenthetic vowel without resulting in crossing associatior ines (11b,c).

Recall that the structure indicated by the tag [7] does not ‘happen’ be- fore or after that indicated by [7} The structures merely encode information characterising the slots which they are associated to. Only these skeletal slots are in sequence. By contrast, in AP F, does indeed occur before F2 and the NCC is required to control the patterns of this temporal association. In Au- tosegmental Phonology, (20b) is ill-formed because it includes as a sub-part the representation in (21a). My view is rather that (20a) is ill-formed because it includes the component (21b). We can compare the effects of the NCC and the Sharing Constraint while constrasting their methods by referring to (20).4

Finally note that because AVP is declarative, a nonmonotonic operation like mitosis is impossible. It would be removing information, and derivations in AVP consist solely of the addition of information.®

Figure 4: FO and underlying tones of Venezuelan unmarked yes/no question.

Figure 5: FO and underlying tones of Puerto Rican negative question

Figure 6: Configuration and underlying tones of unmarked Argentinian yes/no questions In Figures 6 to 9 we represent the basic configurations of the typical contours that are derived from the representations above:

According to Sagey, the Labial articulator node is activated in both Labial consonants and rounded vowels. So, if OCP is effective in primary feature, namely *Labial-Labial, it seems that we will get case (Sa) as follows.

In fact, case (5a) is possible. In Cantonese, there are many such acceptable examples as in (5b). These data can be employed to support three different hypotheses. First, they may support the claim of C-place node vs. V-place node by Clements (1989) because the different phonological behaviours of consonantal labial and vowel labial to OCP prove that consonanta! labial and vowel labial belong to different place node. Second, they may support the claim of C-plane vs. V-plane proposed by McCarthy (1981, 1989) because belonging to different planes, /p/ and /u/ will not violate OCP. Third, they may support the claim about the difference between labial feature and round feature proposed by Sagey (1987). The Labial articulator node is activated in both consonants and vowels, but that only the latter are also specified [+round]. Note that Cantonese has both front and back round feature *round-round, instead of Labial feature *Lab-Lab.

But, in Cantonese, there is a second kind of constraint which holds between the onset and the nucleus. Labial consonants may precede back rounded vowels, but not front rounded vowels, as shown in (6a) and (6b) respectively.

The difference between back and front rounded vowels in (6) is interesting. One hypothesis may be proposed that only front rounded vowels are assigned for Labial feature and are therefore identifiable as Labial and subject to the prohibi- tion. This hypothesis, however, is not tenable. That is because in Cantonese there is a third kind of constraint which holds between the nucleus and coda. In other words, a rounded vowel (either front or back) followed by a labial consonant is completely impossible, as shown in (7a) and (7b) respectively:

As seen in (7a) and (7b), both front and back rounded vowels must be identifiable as Labial between nucleus and coda. Thus it can be seen that (6a) and (6b) is just a pair of counterexamples. To explain this front round vowel vs. back round vowel phenomenon, Yip (1988) suggests to make use of the kind of syllable-structure building rules in which only front rounded vowels are underly- ingly marked Labial. When syllable structure is built, first the prevocalic con- sonant is marked as an onset, then redundant values of [round] are assigned, including [+round] for the nonlow back vowels. Assigning a value of a terminal feature like [+round] automatically entails assigning the superordinate articulator node, in this case Labial. Then postvocalic consonants are marked as codas. The order is shown in (8):

Although Yip’s hypothesis can provide for us the explanation to the problem of the back rounded vowel in case (6) and (7), it fails to explain why OCP is not effective case (5), but must be effective in secondary articulatory kw. The feature geometry of secondary articulatory kw can be presented in several different ways, such as (10a) by Sagey (1986) and (10b) by Clements (1989). It should be pointed out that it is very important in the hypothesis (8) pro- yosed by Yip that linear order in phonological representation, redundancy rule for +round] must be applied after creating onsets and before creating codas. It is nly in this way that the contradiction between (6) and (7) can be solved, as seen n (9):

If the derivation in (12) is correct, we will get the result in (14a); if the derivation of (13) is correct, we will get the result in (14b) instead. In Cantonese however, we have got neither (14a) nor (14b) but (14c).

Since the trigger element of both case (7) and (11) is the secondary feature, Redundancy rule must be applied after creating codas, and OCP should be effec- tive after rule application. Thus the back rounded vowels will have OCP opera- tion, as illustrated in (17): My new hypothesis in (15) can be used to explain all the above mentioned data. Since the trigger element of both case (5) and (6) is the primary feature, Redundancy rule must be applied after creating onsets and before creating codas, and OCP should be effective before rule application. Thus the back rounded vowels will not affect OCP, as seen in (16):

ES ee ee Ae el SEN a ee te If the English possessive in the Spec of DP may get Case from AGR in D and the Chinese possessive in the Spec of DP may get Case from determiner in D, the natural hypothesis is to assume that the determiner in Chinese noun phrases shares the same functional role with AGR in English noun phrases; i.e. they may act as the SUBJECT of noun phrases in the two respective languages. If this line of reasoning is plausible,then the specificity ee ms ee nn. ae See wa ee OE ee ee ee Se a a, ee een. ee 2) Lisi-de san ben shu Lisi's three CL book ‘Lisi's three books'