Академический Документы
Профессиональный Документы
Культура Документы
Most of the facts and arguments are left out in the current study. Those details and supporting
arguments can be perused in Yen & Billings (2011). Only the data exemplified here are cited.1
Usually, if one clitic pronoun precedes another in Mantauran, its final vowel is deleted. If
only a single such pronoun appears, there is no final deletion. To illustrate the phenomenon,
combinations involving three distinct phonological shapes are considered in turnbelow in (1),
(5), and (8). The order of clitic pronouns is invariably with the subject first, where the subject
bears any of NOM, GEN, orrarelyDFLT case, whereas the latter clitic pronoun invariably bears
only DFLT case. Mantauran is unique within Austronesianto our knowledgein attesting
paradigms of bound personal pronouns in three morphological cases (Huang et al. 1999:167).
To begin, if the two clitic pronouns underlying shapes are /=()CV =i/ (namely, with
the preceding pronoun ending in a vowel), the pronoun-final vowel is deleted: [=()C_ =i], as
(1b) shows. (The DFLT-case pronouns always begin with /i/.) Namely, /=ni/ in (1b) surfaces
without its final vowel, as the underscorea convention used in Zeitoun (2007a:38)shows.2
1
We are pleased to contribute to the current collection, honoring Alan Prince, who was outside reader of Billings
(1995). More recently, the second author supervised the first authors masters thesis (Yen 2012). A previous
version of this study was also presented (by Yen) at the 11th National Conference on Linguistics, held at National
Hsinchu University of Education, Taiwan, in 2010. The following are also acknowledged for their help at various
stages in its genesis: Lauren Brother, Bill Davis, Jane Grimshaw, Hui-chuan J. Huang, Amy Pei-jung Lee, Lilian Liying Li, Bill Palmer, Valerie Rushanan, Joseph Sabbagh, and Adam Werle. Nonetheless, only the two authors are
responsible for any errors that remain herein. Also, these abbreviations are used: C consonant, CAUS causative,
DFLT default (case), DYN dynamic, EXCL exclusive, FIN finite, GEN genitive, IMPRS impersonal, INCL inclusive,
INV invisible, NEG negation, NFIN nonfinite, NMLZ nominalization, NOM nominative, PL plural, SBJV subjunctive,
SG singular, STAT stative, V vowel, and VIS visible. A preceding equals sign indicates that a morpheme is enclitic.
The numbered examples show glosses, after Zeitoun (2007a), modified slightly to follow the Leipzig Glossing
Rules. All transcription has been converted without further comment using the International Phonetic Alphabet.
2
There are also clitic pronouns that end in two vowels: /=ao/ =NOM.1SG, /=nai/ =NOM/GEN.EXCL1PL, /=mao/
=NOM/GEN.IMPRS, and /=ima/ =DFLT.IMPRS. Incidentally, the two IMPRS forms function exactly like most
personal pronouns with regard to V-deletion. The final vowel of /=mao/ is deleted immediately before a DFLT
pronoun (Zeitoun 2007a:455, 518). One clitic pronoun even ends in three vowels: /=ia/ =DFLT.1SG. If this form as
the subject (rarely) precedes another DFLT-case pronoun, its final // is deleted (Zeitoun 2007a:361, 401). Another
constraint is introduced in Yen & Billings (2011:178) to ensure deletion of only the final vowel. The G subscripts in
(1a) and (8ab) denote cross-referencing (or doubling) of the GEN clitic pronoun, not central to the
current discussion. (Its also possible for NOM and DFLT pronouns to be cross-referenced in this way, not shown.)
(1) a. lo ma-oipi
=niG
[inai vaha
=nai
oponoho]G
if STAT.SBJV-disappear =GEN.VIS3SG [this
language =GEN.EXCL1PL Mantauran
If our language disappears
[Zeitoun 2007a:56]
b. lo pa-kan
=n_
=in
taii
if CAUS-DYN.NFIN;eat =GEN.VIS3SG =DFLT.VIS3SG good
If {she/he}i feeds {her/him}j well, []. [Zeitoun 2007a:78, cited in Yen & Billings 2011:175]
If the cluster-initial pronoun ends in a vowel, the occurrence of this final vowel is not allowed. In
such an environment, this pronouns final vowel is deleted. We propose the following constraint.
(2)
*V]cl[cl.DFLT: A DFLT pronoun does not immediately follow a vowel-final subject pronoun.
[Immediately preceding a DFLT clitic, a subject pronoun (i) ending in a consonant incurs
no violation, as in (4b), (7bc) and (10b) below; (ii) ending in exactly one vowel incurs
one violation, as (4a), (7a) and (10a, c, d) below show; (iii) ending in two or more vowels
(not exemplified in this study but see fn. 2 above) incurs an equal number of violations.]
If V-deletion occurs, then *V]cl[cl.DFLT is satisfied, but MAX-IO, defined as follows, is violated.
(3)
MAX-IO: Every segment in the input has a correspondent in the output. (No
phonological deletion.)
[McCarthy & Prince 1995:264]
In this study, we combinesimilarly to Oda (2005, citing theretofore unpublished work by John McCarthy)the
properties of both data tableaux (the kind instituted in Prince & Smolensky 1993/2004) and comparative tableaux
(introduced in Prince 2003, also discussed in McCarthy 2002:3334). Common to both tableau types is (i) the
arrangement of constraint names along the top, (ii) output formsalso known as candidatesalong the left-hand
column, and (iii) the input indicated in the upper-left cell. The optimal candidate is listed with a preceding pointing
finger (). As in data tableaux, the left side of any other cell in the tableau shows the number of violations of the
constraint named above it by the candidate to its left. In addition, however, as in comparative tableaux, the right side
of the same cells (only in non-optimum rows) shows how the given candidate fares compared to the optimum with
regard to the same constraint using W (indicating that the optimum wins) and L (standing for optimum loses).
(4)
Thus far, we have formalized the heart of V-deletion: *V]cl[cl.DFLT disallows a pronouns
final vowel if this pronoun is immediately followed by a DFLT pronoun. If there is no DFLT
pronoun immediately afterward (not shown in any tableau here), then MAX-IO prohibits deletion.
The next shape is /=C =i/ (viz., with the preceding pronoun ending in a consonant):
(5) a. o-kaka =moo
ana olai
DYN.FIN-beat =NOM.2SG that child
You beat that child.
b. o-kaka =mi
=in
DYN.FIN-beat =NOM.2SG =DFLT.VIS3SG
You beat {her/him}.
[Zeitoun 2007a:286]
[Zeitoun 2007a:286]
Unlike in (1), where deletion of the subject pronouns final vowel occurs if a DFLT-case pronoun
immediately follows, in (5) we see a less straightforward situation. Namely, (5b) is not merely
the result of deleting the final vowel of the clitic pronoun in (5a). Rather, this is an instance of
lexical suppletion, where the selection owes to the following environment. The marked NOM.2SG
allomorph /=mi/ is employed in lieu of final-vowel deletion. Nor is the NOM.2SG form in (5b)
the result of final-V deletion plus some change in the internal vowels quality. There is no
synchronic phonological process whereby the sequence [o] alternates with [i]. Thus, there are
multiple NOM.2SG underlying forms in the lexicon of Mantauran: /=moo/ and /=mi/. The choice
of allomorphs is determined by whether there is a following DFLT pronoun. Additionally, due to
this suppletion, mere deletion of the final vowel of /=moo/ is not attested (i.e., *[=mo_ =i]).
Because there are multiple formsnamely, /=mi/ and /=moo/stored in the lexicon,
we propose the following constraint to determine which underlying form the output selects.
(6)
Entailed in this approach is the notion that the /=mi/ variant is identified somehow in its lexical
entry as the marked member of the set of NOM.2SG underlying forms. In most situations (namely,
if theres no DFLT pronoun immediately following the NOM.2SG clitic pronoun), then this *MKD
constraint is what allows /=moo/ to be selected instead of the marked NOM.2SG variant /=mi/.
We now draw a distinction between the input (in the upper-left cell of each tableau) and
the underlying form (shown between slashes on each candidate row). The former is comprised of
features, presumably those that the syntax utilizes; the latter, the forms selected from the lexicon
after spelling out to the morphological component. The same input is used throughout a tableau,
whereas in the same tableau there can be more than one underlying form. With this new
assumption, we also modify McCarthy & Princes definition of MAX-IO above in (3) just slightly
(1995:264): Every [underlying] segment [] has a correspondent in the output.
At this point, *MKD is still unranked relative to the other three constraints. In (4) above,
our only tableau so far, neither candidate violates this constraint. A tableau with the NOM.2SG
pronoun immediately before a DFLT pronoun allows us to fill in this constraints ranking:
(7)
thereof); the selection is clearly phonological. The marked GEN.VIS3SG allomorph is chosen only
if the host ends in a particular class of sound shapes: /a/ preceded by a velar stop.4
Now, if the selection were by strictly phonological means, one might expect every
/n/-initial pronoun to undergo deletion. However, neither /=nomi/ =NOM/GEN.2PL nor /=nai/
=NOM/GEN.EXCL1PL undergoes such deletion of /n/ in the same environment: [-ka =nomi]
-NEG =GEN.2PL and [-a =nai] -already =NOM.EXCL1PL (Zeitoun 2007a:343, 287). In
order for such a strictly phonological rule to work, it would have to define the sequence of velar
stop plus /a/, followed by a morpheme boundary and /n/ plus /i/. That is, only the sequences
/()ka =ni/ and /()a =ni/ would surface as [()ka =_i] and [()a =_i], respectively. There
would be no phonological motivation for such a rule. This would also be an absurd rule,
selecting a specific morphemethe GEN.VIS3SG pronounby identifying both of its sounds.
Rather than such a phonological rule, we again propose phonologically conditioned lexical
suppletion. The marked allomorph /=i/ combines with a limited phonological shape of hosts.
As for the suppletion of /=ni/ and /=i/ in pronominal combinations, if the subject pronoun
preceding the DFLT-case pronoun is /=ni/, then its final vowel /i/ is deleted, as in (1b) above.
However, if the cluster-initial pronoun is the /=i/ allomorph, then there is no deletion, as in (8b).
That is, if the /=ni/ variant of the GEN.VIS3SG pronoun precedes a DFLT pronoun, it surfaces as
[=n]. Hence, the underlying form /=ni/ has two phonologically conditioned allomorphs: /=ni/
and /=i/. However, the /=i/ variant preceding a DFLT pronoun does not undergo vowel deletion.
Nor is there coalescence in (8), where /i= i/ surfaces as [i]. Mantauran is quantity-sensitive.
In opposition to *MKD, introduced above in (6), which inhibits /=i/, we propose the next
constraint to select the marked /=i/ variant preceded by hosts of a certain phonological shape:
(9)
/Ka =i/: Use the GEN.VIS3SG lexical form /=i/ after a host ending in a velar stop plus /a/.
Our last tableau, in (10), demonstrates how the constraint in (9) works, as well as the notion of
multiple potential optima, a feature not commonly found in the Optimality-theoretic literature.5
(10)
Exactly two recalcitrant examples remain: (i) [ma-aii-a =n_ =i] STAT.FIN-heal-already =GEN.VIS3SG
=DFLT.INV3SG and (ii) [o-tipitipi-ka =n_ =ia] DYN.FIN-beat-NEG =GEN.VIS3SG =DFLT.1SG (Zeitoun 2007a:98,
294). We have not found any instances of a host ending in a velar stop plus /a/ followed by untruncated [=ni] (i.e.,
without a following DFLT pronoun). Still, these examples constitute exceptions to the requirement that hosts ending
in a velar stop plus /a/ select the lexical allomorph /=i/. Currently, these examples are in complementary distribution
with /=i =i/ data as in (8b). We refrain from postulating an analysis of these data in (i) and (ii) without having
established the type of variation going on: intra-speaker or -dialect (perhaps even inter-dialect influence).
5
Following (Prince 2003), we use line-crossing to show losers; e.g., (10ab) each harmonically bound
(10c). The broken arrow () indicates a potential optimumi.e., homophonous to (10d)that is a loser. In addition,
the candidates /-ka =i =in/ *[-ka =<n>i =in] and /-ka =i =in/ *[-ka =<n>_ =in], not listed in this tableau,
would each violate undominated DEP, the constraint prohibiting epenthesis (McCarthy & Prince 1995:264).
Thus, combined with the constraint rankings from the previous tableaux (specifically,
*V]cl[cl.DFLT MAX-IO *MKD), we also now know from (10d~b) that /Ka =i/ *V]cl[cl.DFLT.
We therefore also know that the final ranking is /Ka =i/ *V]cl[cl.DFLT MAX-IO *MKD.
Additional candidatesnot shown here, but see Yen & Billings (2011)involve deletion
of the only segment of /=i/ or of both segments in /=ni/. An additional constraint (crucially
dominating *V]cl[cl.DFLT), which requires morphemes to receive some surface phonological
manifestation but does not specify what it should be (Kurisu 2001:55), is harnessed for them.
In (10cd) we see two candidates with outputs that are phonetically identical. As such,
both of these are attestedi.e., are potential optima. Thus, whereas there can be more than one
attested form in a tableau, only one optimum is possible. (In most Optimality-theoretic work, the
two notions amount to the same thing.) Eliminating (10c) as a loser solves this dilemma.
In this short study we have summarized a fuller paper (Yen & Billings 2011), presenting
the key facts about how detecting losers can be the key to determining the true optimum. In so
doing, we have also clarified the distinction between input features and underlying forms. We
have looked at just one of the ways in Mantauran that vowels across morpheme boundaries are
restricted somehow. Deletion of junctural vowels is found not only between clitic pronouns but
also at affixal boundaries (Zeitoun 1997:343, 2007a:2829, 381), though not investigated here.
References
Billings, Loren A. 1995. Approximation in Russian and the single-word constraint. Ph.D.
dissertation, Princeton University.
Huang, Lillian M., Elizabeth Zeitoun, Marie M. Yeh, Anna H. Chang, & Joy J. Wu. 1999. A
typological overview of pronominal systems of some Formosan languages. In H.
Samuel Wang, Feng-fu Tsao, & Chin-fa Lien (eds.), Selected papers from the fifth
International Conference on Chinese Linguistics. Taipei: Crane, 1.165198.
Kurisu, Kazutaka. 2001. The phonology of morpheme realization. Ph.D. dissertation, University
of California, Santa Cruz.
Li, Paul Jen-kuei. 1977. The internal relationships in Rukai. Bulletin of the Institute of History
and Philology [Taipei: Academia Sinica] 48(1):192 (republished in Li 2004:1.559624).
Li, Paul Jen-kuei. 2004. Selected papers on Formosan languages (Language and Linguistics
Monograph Series C3). Taipei: Institute of Linguistics, Academia Sinica.
McCarthy, John J. 1996. Remarks on phonological opacity in Optimality Theory. In Jacqueline
Lecarme, Jean Lowenstamm, & Ur Schlonsky (eds.), Studies in Afroasiatic grammar:
Papers from the second Conference on Afroasiatic Languages, Sophia Antipolis, 1994.
The Hague: Holland Academic Graphics, 215243.
McCarthy, John J. 2002. A thematic guide to Optimality Theory. Cambridge, MA: Cambridge
University Press.
McCarthy, John J., & Alan Prince. 1995. Faithfulness and reduplicative identity. In Jill
N. Beckman, Laura Walsh Dickey, & Suzanne Urbanczyk (eds.), Papers in Optimality
Theory (University of Massachusetts Occasional Papers 18). Amherst: Graduate
Linguistic Student Association, 249384.
Oda, Kenji. 2005. The accentuation patterns of nominal compounds in Japanese: A preliminary
study. Toronto Working Papers in Linguistics 24. 203231.
Prince, Alan. 2003. Arguing optimality. In Andries Coetzee, Angela Carpenter, & Paul de
Lacy (eds.), Papers in Optimality Theory II (University of Massachusetts Occasional
Papers 26). Amherst: Graduate Linguistic Student Association, 269304.
Prince, Alan, & Paul Smolensky. 1993/2004. Optimality Theory: Constraint interaction in
generative grammar. Technical report RuCCSTR-2, Cognitive Science Center, Rutgers
University, Piscataway, NJ; Technical Report CU-CS-696-93, Department of Computer
Science, University of Colorado, Boulder (republished Malden, MA: Blackwell).
Yen, Doris Ching-jung. 2012. Clitics in Kavalan (East Formosan, Austronesian). M.A. thesis,
National Chi Nan University, Puli, Nantou County, Taiwan.
Yen, Doris Ching-jung, & Loren Billings. 2011. Sequences of pronominal clitics in Mantauran
Rukai: V-deletion and suppletion. In Lauren Eby Clemens, Gregory Scontras, & Maria
Polinsky (eds.), Proceedings of the eighteenth meeting of the Austronesian Formal
Linguistics Association (AFLA), Harvard University, Cambridge (MA), March 46, 2011.
London, Ontario, Canada: Linguistics Program, University of Western Ontario. 168182
<westernlinguistics.ca/afla/proceedings/afla18/>.
Zeitoun, Elizabeth. 1997. The pronominal system of Mantauran (Rukai). Oceanic Linguistics
36(2):312346.
Zeitoun, Elizabeth. 2007a. A grammar of Mantauran (Rukai) (Language and Linguistics
Monograph Series A4-2). Taipei: Institute of Linguistics, Academia Sinica.
Zeitoun, Elizabeth. 2007b. La rduplication en rukai mantauran. In Alexis Michaud & Aliyah
Morgenstern (eds.), La rduplication (Faits de Langues 29). Paris: Ophrys, 3747.