Latent Segments and Exfixation Introduction



Download 497.76 Kb.
Page1/5
Date04.06.2017
Size497.76 Kb.
  1   2   3   4   5

Cheryl Zoll Parsing Below the Segment. . .

Latent Segments and Exfixation

Introduction


Chapter 1 demonstrated that invisibility to the syllable constitutes the only automatic consequence of representing a phonological element as a floating feature. Traditionally the realm of floating features embraces exclusively phenomena like those discussed in Chapters 2, 3 and 4, where the features are realized only by docking onto some existing segment. However latent segments also behave exceptionally with respect to parsing, and as argued in Chapter 1, should likewise be represented as floating features. For all floating features the grammar dictates their optimal target. Latent segments differ from prototypical floating features only in that their target is an inserted root node rather than a consonant or vowel present elsewhere in the string.

(1) Exceptional parsing No root node



surface:

full segments

latent segments and
floating features



underlying:

Root

|

features



features

In this chapter I do a case study of the realization of three kinds of latent segments in Yowlumne.1 The language is most interesting in this regard because its floating glottalization may either target an existing segment or dock onto an inserted root node, depending on the context. Yowlumne has in addition two other types of latent segments. In this chapter I will show how the behavior of all three types follows from an analysis parallel to the one developed for the Inor floating features in Chapter 3, and motivate the use of  *STRUC(s), a constraint against excess structure, by demonstrating its role in accounting for asymmetries in the behavior of Yowlumne’s latent consonants and latent vowels. This chapter further illustrates the generality of the approach to subsegments introduced in the earlier chapters.

Yowlumne also provides an opportunity to introduce a third class of affixes into the typology, one not previously analyzed in Optimality Theory. Up until now we have seen edge-bound affixes (2a), and infixes (2b), both segmental and subsegmental. Under the analysis in previous chapters, the differences between the two types of affixes follows from the relative ranking of their precedence constraints with respect to other constraints in the grammar. Yowlumne provides evidence of a third class of affix called an exfix (2c).2 Unlike full infixes, these suffixes are discontinuous in the output because only a portion of the suffix works its way into the base. I show that this behavior likewise follows from the same sort of No-Intervening constraint which governs the edge-bound and fully infixing affixes. No such general account is available under the McCarthy and Prince 1993 coincidence-based conception of affix placement.


(2)








Subsegmental

Segmental



edge-bound affixes

Inor palatalization

Ilokano ag- prefix



full infixation

Inor labialization

Japanese mimetic palatalization



Ilokano um- infix



exfixation

Yowlumne glottalizing suffixes

Hamer -ta suffix



Latent glottals in Yowlumne

How do you identify a subsegment in Yowlumne?


The Yowlumne glottalization process (Newman 1944, Kuroda 1967, Kisseberth 1970, Archangeli 1983, 1984, 1991, Noske 1984, Archangeli and Pulleyblank 1994) is an example of a subsegment difficult to classify as either ghost segment or floating feature. Depending on context it either moves to find a suitable docking site in the base or projects its own root node. (3) provides some examples. Suffix induced glottalization targets the rightmost post-vocalic sonorant in the base (3a). Otherwise, in a biconsonantal root it will manifest itself as a suffix initial glottal stop (3b). Note that in (3a) vowel shortening is unnecessary since [constricted glottis] can dock as a secondary feature on the preceding sonorant, whereas in (3b), a biconsonantal stem with no glottalizable sonorant, vowel length is sacrificed to the parsing of the full glottal.3 In the triconsonantal root in (3c) there is no way to parse the feature since there is no post-vocalic sonorant, nor is there space for a full glottal stop without epenthesizing a vowel, so the glottalization is not expressed. Other glottalizing suffixes are shown in (4).4
(3) Glottalization in Yowlumne

(/)aa durative: (/) = [constricted glottis]

([+G] of Archangeli 1983)

a. glottalize rightmost post-vocalic sonorant

/caaw-(/)aa/ caaw/aa- shout

//iilk-(/)aa/ /el/kaa- sing

b. otherwise full glottal root finally if room

/maax- (/)aa/ max/aa- procure

c. otherwise fail to surface (*CCC)

/hogn- (/)aa/ hognaa- float
(4) Suffixes which induce glottalization (Archangeli 1983: 379)

(/)iixoo

consequent auxiliary

(/)aas

habitual genitive

(/)aa

continuative

(/)in/ay

contemporaneous gerundial

(/)ic

agentive

(/)anaa

desiderative agentive

The absence of epenethesis to rescue the glottal in (3c) sets this potential glottal stop apart from regular segments. As shown in (5), a full segment in the same triconsonantal context does trigger vowel insertion to save itself.


(5)

a. Full segment: Epenthesis into CCC cluster


/wo/y-hin/ *woy. hin wo. /uy. hin sleep (passive aorist)

b. Compare subsegmental glottal: no epenthesis

/hogn- (/)aa-/ hog. naa- *ho. gin. /aa- gloss

How should the glottal be represented?


Chapter 1 demonstrated that irregularity of parsing constitutes the single reliable diagnostic of underlying structure. Therefore I follow Archangeli 1984 in representing the glottal as a floating [constricted glottis] feature (6).
(6) Exceptional parsing No root node




full segments

latent glottal



underlying:

Root

|

features



[constricted glottis]




*STRUC(s):


The first task is to account for the fact that full segments trigger epenthesis to facilitate parsing while the latent glottal does not. The parsing constraints that make the necessary distinctions between floating features and full segments have already been motivated in Chapter 2 (7-8).
(7) Max (Seg) Every segment in Sj has a correspondent in S0

(i) x ((Segment(x)  Sj(x)) y(S0(y)  xRy))

(ii) Assess one mark for each value of x for which (i) is false

(8) Max (Subseg) Every subsegment in Sj has a correspondent in S0

(i) x (Subsegment (x)  Sj(x)) y(S0(y)  xRy))

(ii) Assess one mark for each value of x for which (i) is false

Because only potentially unparseable full segments trigger vowel insertion, some constraint against epenthesis must come between the two parsing constraints. Zoll 1993a proposes *STRUC(s), a constraint which functions to minimize the total number of syllables in a word (9). This is an OT implementation of Selkirk (1981)’s Syllable Minimization Principle. (See also Broselow (1995 fn 19), and Noske (1984)).
(9) *STRUC(s): ‘No syllables’ (Zoll 1993a)

(i) xSyllable(x)

(ii) Assess one mark for each value of x for which (i) is false
Max(Seg) must outrank *Struc() in Yowlumne, because a vowel will be inserted to facilitate parsing of an input segment. In the tableau in (10), (10a) is optimal because it best satisfies Max(Seg). It outdoes (10b) despite the greater number of *Struc() violations because *Struc() sits below Max(Seg) in the hierarchy.

(10) /wo/y-hin/  wo. /uy. hin sleep (passive aorist)






Candidates

Max(Seg)

*Struc(s)

Comments

a.

wo. /uy. hin




***

u is epenthetic

b.

woy. hin

*

**

/ deleted

By the same reasoning, *Struc() must outrank Max (Subseg) since a floating feature deletes rather than trigger epenthesis (11).


(11) Ranking: *STRUC(s) » Max (Subseg)

Rationale: No vowel epenthesis to make room for latent feature as segment


For /hogn-/aa/, in the tableau in (12), for example, no sonorants in the base attract the glottalization since none are post-vocalic.5 Yet unlike the example in (10), the floating [constricted glottis] also fails to materialize as its own segment. This is because the CVX maximal syllable limit in Yowlumne keeps [c.g.] from turning up as a full glottal stop without vowel epenthesis, but epenthesis would lead to a fatal violation of *STRUC(s) (12a). The most harmonic candidate fails to parse [constricted glottis], thereby avoiding the more serious *STRUC(s) violations which would otherwise ensue.

(12) Segment Structure, *STRUC(s) » Max (Subseg) from /hogn, /aa/






Candidates

*STRUC(s)

Max (Subseg)

a.

ho. gIn. /aa

***!




b.

hog. naa

**

*

This results in the hierarchy in (13). The domination of *Struc() by Max(Seg) accounts for the fact that potentially stray consonants can trigger epenthesis. The low ranking of Max (Subseg) captures the fact that floating features will delete rather than prompt life-saving insertion of an additional vowel.


(13) Max(Seg) » *Struc() » Max (Subseg)
There are of course a number of other possible ways to keep a floating feature from manifesting itself as a full segment. A Fill (McCarthy and Prince 1993) or Dep (McCarthy and Prince 1995) constraint (termed No New Root in Zoll 1993a) could proscribe insertion of a new root node , or we could count additional segments rather than syllables as extra structure violations. However only *STRUC(s) correctly predicts an asymmetry that exists between the behavior of floating consonantal and vocalic features (14).6 Namely, since extra consonants can be an onset or coda to an already existing syllable, they will not necessarily violate the constraint when they surface as independent segments. In a language without diphthongs however, an inserted vowel inevitably adds a new syllable and thus violates *Struc(). It follows from this that latent consonants will surface more readily than latent vowels. This prediction is confirmed below in Yowlumne.

(14) Prediction of *Struc(s):

Asymmetry between consonant and vowel epenthesis

(® = inserted root node)



a. inserted V violates *Struc(s): b. inserted C needn't



Affix placement

Exfixation


Candidate evaluation with No-Intervening is straightforward when the affix exists as a contiguous substring within the stem, but cases such as the glottalizing affixes in Yowlumne, where affix integrity may be compromised, require a more articulated statement of No-Intervening. In /el/kaa (15), for example, nothing intervenes between the right edge of the affix and the right edge of the word, but the non-affixal k does follow the glottalized l (which corresponds to the suffix). The surface discontinuity of the suffix constitutes an example of exfixation./ Clearly the constraint requires modification to allow it to handle cases where the affix does not behave as a block.
(15) /el/kaa from //elk, /aa/




Candidates

No-Intervening

comment

a.

/el/kaa]

??

nothing intervenes between the right edge of the affix and the edge

base k intervenes between the glottalized l/ and the edge


Exfixation is not restricted to affixes with a subsegmental component. The data in (16) from Hamer (South Omotic) illustrate a case involving only full segments. Here the initial consonants in the two suffixes -ta and -no undergo metathesis, resulting in the intermingling of segments belonging to base and suffix.


(16) Hamer (South Omotic) Metathesis Lydall (1976: 408-409)

a. isin sorghum isinta small amount of sorghum

rac Rac (clan) ratca Rac man

b. oto calf otono all calves

isin sorghum isinno all sorghum

rac Rac (clan) ranco all Rac



No-Intervening II:


In order to handle cases like these, the No-Intervening constraint must be modified to refer to individual segments in the affix, rather than treat the affix as a single block. The constraint in (17) does this by having both variables range over separate consonants and vowels. Here Base(x) is a shorthand for elements in S0 that correspond to the base and Affix(y) refers to elements in S0 that correspond to the affix. Thus the individual segments that correspond to the affix count as possible values of the second variable, y. This constraint penalizes segments which correspond to the base which intervene between any part of the affix and the right edge of S0.
(17) No-Intervening II:

(i) xy (Base(x)  Affix(y)  x intervenes between y and edge E)

(ii) Assess one mark for every x which falsifies (i)
To illlustrate, consider the evaluation of candidates in Hamer. The precedence constraint for the suffix -ta is given in (18).
(18) No-Intervening(ta;R)

‘nothing intervenes between any part of ta and the right edge of the word’

(i) xy (Base(x)  ta(y) x intervenes between y and edge R)

(ii) Assess one mark for every x which falsifies (i)


In Hamer, only coronal consonants and nasals homo-organic to a following consonant are allowed in the coda (Lydall 1976: 404), forcing exfixation. A descriptively accurate coda condition is given in (19). It sanctions non-coronals only in positions which open up into a vowel.7 This allows codas consisting of the first half of any geminate, a nasal which shares place with the following onset consonant, or any coronal segment.
(19) Coda condition ‘Noncoronal place must open into a vowel’
Following now standard reasoning, the Coda-Condition must outrank the affix constraint, since No-Intervening will be violated in order to contrive well-formed syllables in the output. As shown by the tableau in (20), this ranking selects (20b) as the optimal candidate since it satisfies the coda condition while only minimally violating No-Intervening.
(20) ratca from /rac, ta/




Candidates

Coda-Condition

No-Intervening(ta)




a.

racta]

*!







b.

ratca]




*

c intervenes between t and right edge

c.

traca]




***!

r, a, c intervene between t and right edge

This modified No-Intervening constraint likewise handles the affixation cases discussed in previous chapters. Where there is only one segment which corresponds to the affix, as in Inor or in the Japanese Mimetic palatalization, the two versions of the constraint are obviously the same. Interestingly, this constraint handles cases of full infixation of longer affixes as well. Compare Hamer’s exfixation with the total infixation of um- in Ilokano. (21) shows the constraint that applies.


(21) No-Intervening(um-; L)

(i) xy (Base(x)  um(y) x intervenes between y and edge L)

(ii) Assess one mark for every x which falsifies (i)

The candidates in the tableau now include those with both full (22a) and partial (22b) infixation, both of which minimally violate No-Intervening. The difference between Ilokano and Hamer follows from the nature of the dominating constraint and the shape of the affix. The only way to satisfy No-Coda here is by infixation of the entire affix. (22a) is more harmonic than (22b) and (22c) because it contains the fewest closed syllables.

(22) No-Coda » No-Intervening (um; L) , from {um, kagat}Stem




Candidates

No-Coda

No-Intervening

Comments

  1. 

[kumagat

*

*

full infixation



[ukmagat

**!

*

exfixation



[umkagat

**!




no infixation



Exfixation in Yowlumne


Exfixation in Yowlumne yields to the same sort of account developed for Hamer. The Yowlumne precedence constraint in (23) governs the behavior of all its suffixes so I have not specified the morpheme to which it applies in each example.
(23) No-Intervening(affix; R)

(i) xy (Base(x)  affix(y) x intervenes between y and edge R)

(ii) Assess one mark for every x which falsifies (i)

Because glottalization appears base internally when necessary it must be the case that No-Intervening is outranked by a constraint which will force it to be violated (24). This case is completely parallel to Inor where the dominant constraint is Max (Subseg).


(24) Ranking: Max (Subseg) » No-Intervening (affix;R)

Rationale: Precedence constraint violated by floating glottalization
This is illustrated by the tableau in (25).The glottal cannot surface as an onset to the suffix since this would create an ill-formed trisyllabic cluster. The root final k is not a licit target for glottalization, but the floating [constricted glottis] can dock to the root internal post-vocalic l, as in (25b). This candidate is more harmonic than (25a) where [constricted glottis] is unparsed, because although nonparsing of the floating feature vacuously satisfies No-Intervening, it results in a breach of the higher ranked Max (Subseg).

(25) Max-Feature » No-Intervening from {/elk, [c.g.]aa }






Candidates

Max (Subseg)

No-Intervening




a.

/elkaa]

*!




[c.g.]not parsed

b.

/el/kaa]




*

k intervenes between l/ and right edge

The lack of a sanctioned mooring at the edge of the root thus sends a latent feature sailing inside to look for one, since its need to be parsed exceeds the importance of perfect alignment. But what sets the Yowlumne glottalization apart from prototypical examples of floating features is that under certain circumstances it does show up as an autonomous segment . Where there is no glottalizable sonorant, it will emerge as a full segment if, in the process, it neither displaces a full consonant nor requires vowel epenthesis.

This is illustrated by the tableau in (26). There is no glottalizable (post-vocalic) sonorant in maax so secondary glottalization is impossible. The only way Max (Subseg) can be satisfied then is by the insertion of a full glottal stop (26b), despite the resulting shortening of the template’s long vowel. The need to parse the feature outweighs any cost incurred by shortening.8 Note that the precedence constraint operates here to make sure that the resulting glottal stop surfaces as the onset to the suffix. Therefore the logically possible candidate in (26c), where the suffixal glottal precedes a segment which correponds to the base, will never be optimal in Yowlumne because it results in more violations of the precedence constraint. No-Intervening thus subsumes the place-holding role usually attributed to the root node.
(26) Max (Subseg) » Ident() , No-Intervening(affix, R) from /maax, aa/




Candidates

Max (Subseg)

Ident()

No-Intervening

Comments

a.

maa.xaa]

*!







[c.g] unparsed

b.

max/aa]




*







c.

ma/xaa]




*

*!

x intervenes between suffixal / and right edge

The mixed behavior of the Yowlumne glottal thus follows from the interaction of a hierarchy of general constraints with the latent (rootless) glottal feature (27). The ranking of the constraints is given in (27). The domination of No-Intervening by Max (Subseg) allows mobility of affixal material. High-ranking *STRUC(s) favors deletion over epenthesis as the resolution of potentially triconsonantal clusters where one of the consonants is a latent segment. Finally the effect of the low-ranking precedence constraint is to keep the glottalization as close to the beginning of the affix as possible, subsuming what has been considered the place-keeping function of the root node. In the next two sections I will show how this hierarchy also accounts for the diverse behavior of Yowlumne’s other latent consonants and vowels.


(27) Yowlumne Hierarchy:

Max(Seg) » *STRUC(s) » Max (Subseg) » No-Intervening(root,affix), Ident()




Share with your friends:
  1   2   3   4   5


The database is protected by copyright ©sckool.org 2019
send message

    Main page