Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv




Скачать 93.88 Kb.
НазваниеBackground Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv
Дата конвертации12.05.2013
Размер93.88 Kb.
ТипДокументы
15è Colloqui de Gramàtica Generativa

4-6 Abril 2005

Universitat de Barcelona


The allomorphy of Capeverdean object pronominals:

a case for late insertion


Fernanda Pratas Andrés P. Salanova

Universidade Nova de Lisboa Massachusetts Institute of Technology

fpratas@fcsh.unl.pt kaitire@mit.edu

  1. Introduction


The aim of this paper is to describe and explain a striking constraint in Capeverdean1 (henceforth CV), which prevents the cooccurrence of object enclitics with verbs affixed with the past tense marker {-ba}. The relevant data are the following:2


(1) a. N ódja ‘I heard’

b. N odjába ‘I used to hear’

c. N odjá-l ‘I heard him’

d. N odjába el ‘I used to hear him’

e. *N ódjaba-l/odjába-l/odjabá-l

    1. Stress rules



- We argue that the explanation for the ungrammaticality of (1e) is intimately tied to the process of stress shift that occurs in (1). The stress rules of the language would require that in {odja + ba + l} stress be placed on the last syllable. This would violate an independent constraint of the language that requires stress to fall within the stem.


- *odja-bá-l is ruled out since if the stress rule of CV is to be obeyed stress falls on the tense inflection {-ba}, out of the stem. This forces form (1d) to be chosen, where the free-standing form of the object pronoun is used.


    1. Distributed Morphology (DM) and Optimality Theory (OT)



- If our hypothesis is correct, these CV phenomena seem to favour a DM approach (Halle & Marantz 1993, Embick & Noyer 2001), according to which morphemes are bundles of abstract (syntactic-semantic) features provided with phonological features at Vocabulary Insertion (VI); this one operates at a post-syntactic module, and is conditioned by phonological rules and constraints.


- We will also argue, in section 4, that the choice among the several possible results obtained with the different available VIs for the specific position described above can be accounted for by Optimality Theory (OT) (McCarthy 2002) tableaux.
  1. Background



Before we discuss these data in detail, we present some background information on relevant aspects of CV grammar.

    1. Forms for personal pronouns in CV



(2)





Emphatic forms

Free forms

Subject clitics

Object clitics

1sg.

ami

mi

N

-m

2sg (informal)

abo

bo

bu

-bu / -u

2sg (form, masc)

anho

nho

nhu




2sg (form, fem.)

anha

nha







3sg (fem., masc)

ael

el

e

-l

1pl

anos

nos

nu

-nu

2pl

anhos

nhos







3pl

aes

es




-s



The free (non-clitic) forms of pronouns may appear in all case positions (except for genitive, which has an independent set of pronouns), alternating with clitics (of slightly different form). In this paper, we are concerned with the alternation between free forms and object clitics. The paradigms for subject and object clitics are defective: there are no subject clitic forms for 2sg formal fem., 2pl and 3pl, and there are no object clitic forms for 2sg formal and 2pl; in these cases free forms have to be used instead.

    1. Tense in CV



Tense and Aspect in CV is usually expressed by free-standing particles. The only temporal notion expressed through an affix is the past tense with {-ba}. The following table gives a synopsis of the semantics of {-ba} with different verbs and auxiliaries.


(3)



Non-stative verbs

Stative gosta ‘like’

Stative sabe ‘know’

Simple Pres.

E ta odja ‘he sees’

E (ta) gosta ‘he likes’

E (*ta) sabe ‘he knows’

Past Perfect

E odja ‘he saw’

E (ta) gostaba ‘he liked’

E (*ta) sabeba ‘he knew’

Imperfective

E (ta) odjaba

Ex: ‘he used to see’

E (ta) gostaba

Ex: ‘he used to like’

E (*ta) sabeba

Ex: ‘he used to know’

3Pres. Prog.

E sata/ sta ta odja

‘he is seeing’







Past Prog.

E ta staba ta odjaba

‘he was seeing’







Summing up, with non-stative verbs like {odja} ‘see’, post-verbal {-ba} is a marker of imperfective, while with stative verbs like gosta ‘like’ and sabe ‘know’, postverbal {-ba} stands for past (perfective and imperfective). The difference between {gosta} and {sabe}, irrelevant for our present purposes, is that the first allows the morpheme {ta} (which marks present and possibly also imperfective in non-stative verbs), while the second does not. This must depend on aspectual properties of each verb.


The only other verbal affix is the passive marker {-du}, which for independent reasons doesn’t cooccur with {-ba}, and only occurs with object pronominals in passives of double object constructions (DOC’s).


We will assume that what on historical or comparative grounds could be considered a theme vowel (i.e., the final vowel in {odja} ‘see’, {kre} ‘want’ and {xinti} ‘feel’) is part of the stem, as we know no morphological or phonological reasons to think otherwise.


Costa & Pratas (2003) argue that {-ba} is affixed to the verbal stem by lowering rather than by movement of V to T, based on the facts of adverb placement and on the adjacency requirement between V and T. We share this assumption here.

    1. The stress rule of CV



CV displays a simple weight-sensitive stress rule that assigns primary stress to the syllable containing the penultimate mora. We will not say anything about secondary stress in this paper. All coda segments, i.e., /r, l, m, n, s/, count as moraic. This rule accounts for the stress pattern of nearly all words in CV:4


(4) a. káza ‘house’

b. imbés ‘instead of’

c. armúm ‘brother’

d. katxór ‘dog’


A few exceptional cases exist; the last consonant of these has to be marked as extrametrical in the lexicon:


(5) a. ménos ‘less’

b. kinhéntus ‘five hundred’


No special provisos have to be added for assigning stress to morphologically complex words; as can be seen in the examples (6a) and (6b), stems with derivational affixes behave just as we would expect.


(6) a. padjínha ‘weak hashish (pádja)’

b. bistídu ‘dress (from the V bísti)’

c. kárus ‘cars’


The inflectional suffix {-s} ‘plural’, as in (6c), is not very productive. Compare to (7):


(7) a. Kes livru bunitu kusta dinheru txeu.

these-pl book beautiful cost money lot

‘These/The beautiful books are expensive.’

b. Tudu mininu gosta di livrus bunitu

all child like of book- pl beautiful

‘Every child likes beautiful books.’


As it as been proposed in Castro & Pratas (2003), in the Capeverdean DP the plural marker is a singleton and it surfaces on the D head (as a suffix): if there is an overt element on D (7a), it attaches to that element; if there is no overt element on D (7b), there occurs lowering to the head of the complement.

Note that in the DP ‘Tudu mininu’ there is no plural marker, since the presence of the universal quantifier in a pre-DP position seems to determine the {-s} as unnecessary.


In fact: whenever it occurs, this /s/ has to be marked as extrametrical:5


From the behavior of stress in the forms in (1), it can be inferred that the inflectional suffix {-ba} and the clitics are in the domain of the stress assignment rule of CV. This is a crucial part of our story, to which we will return below.


Capeverdian stress assignment


Slightly more formally, the stress rule is the following6, illustrated in (8) with the word {divugádu} ‘lawyer’.


(8) line 2 *

line 1 * *

line 0 (* *) (* *)

d i v u g a d u


Line 0: within a prosodic word, project an * from all syllable nuclei and codas, save for extrametrical segments. Group the ** into feet, beginning from the rightmost pair.

Line 1: project an * from the leftmost * in each foot in Line 0 (i.e., create trochees).

Line 2: project an * from the rightmost * in Line 1.
    1. Expansion of the problem



We now describe a couple of subsidiary aspects of our topic problem, which were more central to previous approaches. The first of these is that in CV there are effectively no clitic clusters:


(9) a. Modi ki el ten kel karu-li?

How rel he have that car-here?

b. * E si pai ki da-l-l.

Is his father rel give.3sg.3sg

‘It was his father who gave it to him’

c. E si pai ki da-l el

Is his father rel give-IO3sg DO3sg

‘It was his father who gave it to him.’


In (9c) there is an optional second pronoun, which stands for the direct object.


In addition to being barred from occurring after verbs that have the affix {-ba}, as shown in (1), cliticization is also impossible after passive verb forms in {-du}. This can be tested with passives of double object constructions (DOCs).


(10) a. * E odjaba-m / * E flaba-m

he saw-imp-1sg / he told-imp-1sg

b. * N dadu-l

I given-3sg

‘I have been given it’

c. N dadu el


Encliticization is also blocked if…


- …one of the few free-standing lexical items allowed postverbally7 intervenes between the verb and the pronominal form; this is exemplified here with the adverb {so} ‘only’ and the preposition {di} ‘of’:


(11) a. N odja so el. / * N odja so-l.

I see only he

‘I saw only him.’

b. N gosta d’el. / * N gosta di-l.

I like of he

‘I like him.’


- …or if the pronominal is coordinated with another DP (no matter the order between them):


(12) a. * Nu odja-l ku Maria (possible when ku has a preposition value – see (13))

we see.3sg and Maria

b. Nu odja-l ael ku Maria

we see.3sg 3sg and Maria

‘We saw him and Maria.’

c. Nu odja Maria ku el / * Nu odja Maria ku-l

we see Maria with 3sg

‘We saw Mary and / with him.’


Note that in (12a) the CV sentence is possible with a secondary predicate, with a PP (the ambiguity here is due to the fact that ku can mean ‘with’ and ‘and’) in right adjunction to VP. See the next examples:


(13) Q: Modi ki nhos ta kume katxupa?

how rel 2pl tma eat katxupa

‘How do you eat katxupa?’

A: ‘Nu ta kume-l ku kudjer.’

1pl tma eat.3sg with spoon

‘We eat it with a spoon.’


To sum up, in previous approaches it was considered a central fact that any material intervening between the verb and the pronominal form would block cliticization. In the approach we defend here, this epiphenomenon has at least two sources, as we will see.

  1. Previous approaches



We now present our arguments against the two previous analyses.

    1. Baptista (2002)



The approach: there is a morphological template where tma’s for imperfective or for passive compete for a slot with person marks. This author also assumes that there is V-to-T movement in this language, being {-ba} affixed in the process, i.e., in the syntactic component of the grammar.


Our point here: aside from the fact that the mentioned competition is a plain stipulation that doesn’t connect this fact with anything else in the grammar of CV, putting together direct and indirect object clitics with tense markers is both intuitively and empirically wrong.

The latter point follows from the two irregular imperfective forms {era} ‘used to be’ and {tinha} ‘used to have’ (the latter alternating freely with regular {tenba}). We analyze these as consisting of an allomorph {-a} for the imperfective, which in turn selects for the allomorphs {er-} and {tinh-} of the verbs {e} and {ten}, respectively.8 We see that there is allomorph selection between the stem and the tma morpheme, while no such allomorphy occurs between stems and person clitics.

In addition to this, person clitics have corresponding freestanding forms that are required in various contexts such as coordination or focalization, while the tma ending {-ba} is exclusively bound.

We conclude that there is no morphological unity in the class of elements that affect stress placement. We will argue below what they share is being in the same phonological domain.

    1. Costa & Pratas (2003)



The approach: the cooccurrence of object clitics with verbs in the imperfective form would not be coherent with the lowering story for the merger of V and T. According to this, inflection gets attached to the verb not by verb movement to T (the relevant functional head in the language), but rather by lowering of T to V, in the morphology component of the grammar, under conditions that include strict adjacency.


Lowering is furthermore constrained in the following way:

In a configuration X * [ Y * Z ], where * represents linear precedence and the brackets represent morphological constituency, lowering of X to Y cannot take place, since it would “break up” a morphological constituent.


The authors argue that this is behind the ungrammaticality of forms such as {flábal}: {fla} + {l} form a morphological constituent, preventing the lowering of {-ba}.


Our point here: to say that V+clitic form a morphological constituent to the exclusion of inflection is as much a stipulation as stating that clitics and tense endings occupy a single morphological slot. Furthermore, it doesn’t explain the fact that the second clitic can’t lean in a double object construction, i.e. *da-l-s.

  1. The morpho-phonological approach





  • We will assume that this is a matter for phonology;

  • the two object forms are syntactically identical, contra our previous work, where we assumed the free-standing forms as DPs, while the clitics were heads;

  • as it is predicted by DM, phonology receives from syntax a morpheme (in the sense of a bundle of abstract features), i.e. a representation where the choice of form is underspecified; in the case of the object pronominals we are dealing with, we will adopt a feature representation (i.e., [ 3sg ], etc.) for convenience;

  • there is no reason to assume that other lexical items are not already inserted at this point;

  • as it is also predicted by DM, the choice between the Vocabulary Items (VI’s) available in each language depends on some rules at a morpho-phonological level (the morphological part would be relevant if there were, for instance, different forms for each morphological Case – Accusative and Dative –, which does not apply in CV); the constraints involved in our problem are, in fact, phonologic.
    1. Formalization



A.

2 constraints involved in the realization of CV object pronominal forms


- the language-wide stress rule

- a constraint against stress falling out of the stem


B.

There are two listed allomorphs for each combination of person and number features (some of these allomorph pairs differ only in being or not being syllabic, but the relation of other members of a pair is not as straightforward)


C.

Effectively, the insertion of the free-standing form of the object pronominal is more costly to the grammar, since it results in the creation of a separate stress domain, which is penalized by the constraint *Structure (the simpler the better). Preliminarily, we put these constraints together in the following OT tableaux:


(14)


odja, [3sg]

Stress rule

Stress stem

*Structure

([ódja]Stl)PrW

*!







([odjá]Stl)PrW










([ódja]St)PrW (él)PrW







*!




odjaba, [3sg]

Stress rule

Stress stem

*Structure

([odjá]Stbal)PrW

*!







([odja]Stbál)PrW




*!




([odjá]Stba)PrW (él)PrW







*



There is one situation which our proposal so far doesn’t deal with adequately: if a pronominal form follows a word that is not a verb, what prevents it from encliticizing? That is, what prevents the direct object pronominal from encliticizing to an indirect object DP in (15)?


(15) a. E da Maria el

He give Maria it

‘He gave it to Maria.’

b. * E da Maria-l


To deal with these cases (and partially with the cases introduced in (11) above), we need to introduce a constraint that prevents pronominals to encliticize to anything but the verb; or, more generally, that a pronominal can only encliticize to something whose syntactic projection it is within.


(16)


da Maria [3sg]

Stress r.

Stress stem

[ [ X ] cl ]XP

*Structure

(da) PrW (Maria) PrW (el)PrW










*

(da) PrW (Marial) PrW







*!






This constraint might sound stipulative, but it is parallel to another well known case of prosodically-conditioned allomorphy, that of the feminine definite article in Spanish.

As it is known, the definite article {la} has the allomorph {el} before words that begin in stressed /a/; this is clearly prompted by hiatus avoidance.


(17) a. agúja (fem.): la aguja, las agujas

b. águila (fem.): el águila, las águilas

c. ábaco (masc.): el ábaco, los ábacos


This allomorphy of the definite article nevertheless doesn’t take place before an adjective, even if a hiatus would result:


(18) la álta águila / * el alta águila


Whatever the correct analysis of this allomorphy is, it is clear that, on the one hand, it’s prompted by phonology, and, on the other, it is sensitive to syntactic information much in the way that CV cliticization is.

    1. The underlying representation of clitics



A more serious challenge to the prosodic approach comes from the fact that a sequence of two clitics is not permitted even when they could be syllabified as a single mora, i.e., as in the following examples:


(19) a. * Da-l-u, *Prizenta-s-u


b. N da-u el.

I give.2sg 3sg (IO , DO)

’I gave you it.’ / ‘I gave it to you.’


c. N prizenta-s bo.

I introduce.3pl 2sg (IO , DO)

‘I introduced them you.’ / ‘I introduced you to them.’


A not altogether implausible solution to this problem could follow if the underlying form of clitics is made to contain a mora.9 Forms such as those in (19a), whatever their syllabification, would get stress outside of the stem, and thus be ruled out.
    1. Remaining problems



One of the problems left for further research is tied with the base order in DOCs. From the examples above, and also for the one below, it seems that the base order in CV is V IO DO:


(20) Q: Si bu teneba dinheru txeu kusé k’-u ta fazeba ku el?

if 2sg have.tma money lot what rel.2sg tma do with it

‘If you had a lot of money what would you do with it?’

A: N ta daba nha mai el. / ??? N ta daba el nha mai.

1sg tma give.tma my mother it

‘I would give it to my mother.’


This could be an obstacle to our proposal, since it would be difficult to explain how sentences like N da-l Maria ‘I gave.3sg Maria’ (DO, IO) would form: how would the clitic (DO, in this case) appear in a pre-IO position. But it is our intuition that sentences like this one also involve some kind of focalization, leaving the new information on the right. This said, N da-l Maria must be the answer to some question like: “What did you do to the book?’, and the non-marked form would be N da Maria el.

For the moment, we can also provide the contrast between (20), above, and (21):


(21) Q: Si bu teneba dinheru txeu (pa) kenha ki bu ta daba el?

if 2sg have.tma money lot (to) who rel 2sg tma give.tma it

‘If you had a lot of money you would give it to whom?’

A: N ta daba el nha mai.

1sg tma give.tma it my mother

‘I would give it to my mother.’


Here the IO appears on the right edge of the sentence, in spite of the enclitic for the DO being barred (for the reasons we have just discussed – the {-ba}, which in the Answer, under a compositional interpretation of the sentence, stands for a conditional, prevents the enclitic form and the free standing form is chosen).

  1. Conclusions



We have analyzed encliticization in CV as a case of prosodically conditioned allomorphy. Enclitic forms and free forms are allomorphs that are chosen following the stress rules of the language. To fully analyze the Capeverdean clitic cluster problem we have to choose between the best of three sets of stipulations:

  1. There are no clitic clusters in the language, and the morphemes {-ba} and {-du} are clitics (Baptista 2002).

  2. V+clitic forms a morphological constituent that disallows the lowering of {-ba}, and, in addition, clitics cannot lean on clitics (Costa & Pratas 2003).

  3. Stress cannot fall outside of the verb stem, and clitics can only lean on a prosodic word that contains the verb (the present analysis).


We believe to have provided some evidence for a DM approach, particularly by showing that this “choice” between allomorphs is phonologically motivated, and thus a case for late insertion.


References

Baptista, Marlyse 2002. The Syntax of Cape Verdean Creole, the Sotavento varieties. Amsterdam/ Philadelphia: John Benjamins.

Castro, Ana & Fernanda Pratas. 2003. Capeverdean DP-internal number agreement, additional arguments for a Distributed Morphology approach. Lisbon Workshop on Agreement, Universidade Nova de Lisboa, July.

Costa, João & Fernanda Pratas. 2003. Capeverdean Creole, some parametric values. ACBLPE meeting, Universidade de La Coruña, June 2003.

Halle, Morris & William Idsardi. 1993. General properties of stress and metrical structure. In John Goldsmith (ed) A Handbook of Phonological Theory. Oxford, Blackwell Publishers:…….

Embick, David & Rolf Noyer. 2001. Movement Operations After Syntax. Linguistic Inquiry 32:555-595.

Halle, Morris & Alec Marantz. 1993. Distributed Morphology and the Pieces of Inflection. In Kenneth Hale and Samuel J. Keyser (eds) The View From Building 20. Cambridge, MA, MIT Press:111-176.

McCarthy, John J. 2002.A Thematic Guide to Optimality Theory. New York: Cambridge University Press.

Ximenes, Cristina. 2004.

1 A Portuguese based Creole language, the native language of all the 400.000 Cape Verde inhabitants.

2 Our transcription of the examples follows the official Capeverdean orthography, except in that word stress, since it is relevant to our purposes, is always indicated.

3 In some Capeverdean varieties there is no sata form (Baptista 2002 does not even mention it), but only sta. In the inland of Santiago, however, there is an interesting distinction between both forms, but this is out of the scope of this paper.

4 Many details could be added, v.g., about the prosodification of rising vs. falling diphthongs, the stress pattern of compounds, etc. We omit these, since they are beside the point here.

5 It would also be possible to analyze the special behavior of the plural as following from it belonging to a putative affixal class 2, which wouldn’t affect stress. It has been shown by Ximenes (2004) that the plural morpheme in Brazilian Portuguese is outside of the domain where coda /s/ is a trigger for diphthongization:

(i) paz [pajs] ‘peace’

pá+s [pas] ‘shovels’

6 The rule here is given loosely within the formalism of Halle & Idsardi (1993) for concreteness. Later in the paper we subsume the stress rule under a single OT constraint; since there’s nothing crucially derivational in CV stress, the statement of the stress rule made above is easily convertible to any constraint-based one.

7 According to our informants, there seem to be requirements of strict adjacency between the verb and the arguments:

(ii) (Tudu dia) Djon (*tudu dia) ta odja (*tudu dia) tilibison (tudu dia) na ora di djanta.

(Always) Djon (*always) tma see (*always) tv (always) in time of dinner.

‘Djon always watches tv at dinner.’

8 Alternatively, one could say that there is suppletion for the V+T. Our point in this section isn’t altered if this alternative analysis is adopted, but our analysis of stress shift and the restriction on clitics depends on a stem being separated from inflectional endings.

9 We leave this as a plain stipulation here, since our knowledge of the language is still insufficient to make this follow from other facts (i.e., a general condition on allomorphy). Also, we are aware that there is overlap between the work that’s done by positing clitics to be inherently moraic and the stress rule that we propose for CV, yet if we have reasons to doubt that coda /s/ is moraic, the overlap is significantly curtailed.




Добавить в свой блог или на сайт

Похожие:

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconPersonal data date and place of birth: 7/1/39, Springfield, Massachusetts Citizenship: usa educational background

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconWe discuss in detail all of the required aerodynamic and aerothermal information that is needed in order to launch a 200g, 1kg, and 5kg payload into low earth

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconThis document provides an overview of baseline biological information relevant to risk assessment of genetically modified forms of the species that may be

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconThis document provides an overview of baseline biological information relevant to risk assessment of genetically modified forms of the species that may be

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconThis document provides an overview of baseline biological information relevant to risk assessment of genetically modified (GM) forms of the species that may be

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconBackground information

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconBackground Information

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconBackground information

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconA. Background Information

Background Before we discuss these data in detail, we present some background information on relevant aspects of cv grammar. Forms for personal pronouns in cv iconBackground information


Разместите кнопку на своём сайте:
lib.convdocs.org


База данных защищена авторским правом ©lib.convdocs.org 2012
обратиться к администрации
lib.convdocs.org
Главная страница