Tomohiro Yokoyama University of Toronto Features - NELS 45

Tomohiro Yokoyama
University of Toronto
Features Wearing Two Hats: Derivation of Object-Marked Verbs in Kinyarwanda
There have been several proposals regarding how an apparent reverse ordering of different syntactic
elements can be derived (cf. Massam 2010, Rachowski & Travis 2000, Johns 2008, Herd 2003, and others).
Kinyarwanda (Bantu) is one of the languages that show such mirror ordering of items. Kinyarwanda object
markers (or incorporated pronouns) appear in an order completely opposite to the basic order of full DP
objects, as shown in (1).
a. Umugore
she-PRES-give-APP - FV man(BEN)
‘The woman is giving food to the dog for the man.’
b. Umugore
she-PRES-it(DO)-it(IO)-him(BEN)-give-APP - FV
‘The woman is giving it to it for him.’
(Kimenyi, quoted by Dryer 1983: 137)
The sentence in (1a) is a ditransitive construction with an applied beneficiary argument. The three object
DPs in (1a) are placed post-verbally in the order: Ben[eficiary], I[ndirect] O[bject], D[irect] O[bject]. The
object markers prefixed on the verb in (1b) are in the inverse order: DO, IO, Ben.
Proposal: In this paper, I propose that while feature valuation motivates internal/external Merge, as
has previously been claimed, checking of uninterpretable features is a distinct operation which triggers
head movement. I further propose that incorporation of clitics takes place at the phase edge (i.e. v* or
Voice) after all the clitics have been merged in their respective argument positions. Consequently, the
surface order of incorporated pronouns can be derived by head movement without having to strengthen the
Morphological Component or implement PF rules. I treat the object markers as clitics that are defective
elements (Cardinaletti & Starke 1999, D´echaˆıne & Wiltschko 2002), namely Dmin/max , which lack internal
structure and a Case feature (Roberts 2010).
Assumptions: Features (F) are assumed to have two separate components: interpretability and values
(Pesetsky&Torrego 2006, Boˇskovi´c 2009, and Wurmbrand 2014). Therefore, there are four possible types
of F: uninterpretable F that are unvalued (uF: ), interpretable F that are unvalued (iF: ), uninterpretable
F that are valued (uF:val), and interpretable F that are valued (iF:val). I follow Wurmbrand’s (2014) Merge
Condition and assume that Merge is motivated by downward feature valuation. I follow the traditional view
and assume that head adjunction is to the left of the host. I adopt Low Applicative Phrase for ditransitives, High Applicative Phrase for benefactives (Pylkk¨anen 2002), and Voice Phrase as an agent introducer
(Kratzer 1996). The assumed partial structure is as follows:
[VoiceP Agt [Voice [ApplHP Ben [ApplH [VP V [ApplLP IO [ApplL DO]]]]]]]
Mechanics: Feature valuation has been claimed to motivate internal/external Merge, and here I claim
that uninterpretability of features triggers head movement and that uninterpretable features are checked
after they head-adjoin to the host with corresponding interpretable features. All the functional heads (Voice,
ApplH , ApplL ) bear an uninterpretable unvalued V-features (uV: ) and interpretable unvalued ϕ-features
(iϕ: ). The verb V possesses an interpretable valued V-feature (iV:val) and an interpretable unvalued
ϕ-feature (iϕ: ). Full DP arguments carry an interpretable valued ϕ-feature (iϕ:val) whereas clitics or
Dmin/max have an uninterpretable valued ϕ-feature (uϕ:val). The uninterpretability of V-features cause the
verbal elements (Voice, ApplH , ApplL , V) to go through cyclic head movement to the phase edge first.
Subsequently, the uninterpretability of ϕ-features trigger non-cyclic head movement of clitics. The closest,
thus structurally highest, clitic (Ben) will move first, and the farthest clitic (DO) will move last, resulting in
the correct order of prefixes. In order to account for the non-cyclic head movement of clitics, I propose the
following conditions.
Tomohiro Yokoyama
University of Toronto
Only one type of feature is active and only one feature gets valued and optionally checked at
the time of Merge1 , and other features are inert
All instances of head movement conform to anti-locality (Grohmann 2003)
All the uninterpretable features get checked at the phase edge (Voice or v*)
The condition in (2a) essentially means that one feature can only trigger one instance of Merge and one
instance of head movement; therefore, a functional head cannot probe a verbal root and a clitic at the same
time. The condition in (2b) prohibits head movement within a phrase, which results in successive head
movement of functional elements before incorporation of clitics. The third condition (2c) makes it possible
for the complex head at the phase edge to collect clitics in the desired order. The resulting complex head2
for (1b) is shown below:
[DDO [DIO [DBEN [[[ApplL - V] -ApplH ] -Voice]]]]
uV : val
uV : val
uV : val
iV : val
{uϕ:val1} {uϕ:val2} {uϕ:val3}
iϕ : val1
iϕ : val3
iϕ :
iϕ : val2
Consequences: Not only does the proposed analysis generate the correct order of morphemes, it can easily
accomodate the free order of non-human object markers in a double object construction in Kinyarwanda
with slight modification. It can also account for the variation in object marking among Bantu languages.
While Kinyarwanda lacks independent object pronouns, Zulu has incorporated object markers as well as
independent object pronouns that stay in their argument position. The difference between the two types of
pronouns is attributable to the interpretability of features (OM:uϕ:val vs. PRO:iϕ:val). Moreover, Bantu
languages such as Kinyarwanda allow multiple simultaneous object markers whereas others such as Swahili
permit only one object marker on a verb. This difference can be ascribed to the property of the Voice head.
The Voice head filters out all the subordinate ϕ-features in the complex head except for one in Swahili, but
in Kinyarwanda, the Voice head is not a filter and cliticization is unrestricted.
There are theoretical advantages to the above system as well. We can make a clear distinction between
head movement and phrasal movement without having to stipulate special properties of syntactic elements.
Since features play a double role (valuation and checking) in the system, there need not be extra features
that motivate head movement; therefore, the derivation is more economical.
Conclusion: By differentiating feature checking from feature valuation and taking feature checking as
motivation for head movement, we are able to account for the surface order of incorporated pronouns in
Kinyarwanda without resorting to post-syntactic operations. This paper may also shed light on object marking systems in other Bantu languages as well as word-formation in general.
REFERENCES Boˇskovi´c. (2009) “Unifying First and Last Conjunct Agreement.” NLLT 27.3:455-496. Cardinaletti&Starke. (2004) “The typology of structural deficiency: On the three grammatical classes.” Clitics
in the Languages of Europe: 145-233. Mouton de Gruyter. D´echaˆıne&Wiltschko. (2002) “Decomposing
pronouns.” LI 33: 409-442. Dryer. (1983) “Indirect objects in Kinyarwanda revisited” Studies in Relational Grammar: 129-40. U of Chicago Press. Grohmann. (2003) “Successive cyclicity under (anti-)local
considerations.” Syntax 6:260312. Herd. (2003) “Deriving prosodic inversion: Clitics, cyclicity and the
organization of post-syntactic interfaces.” In Proceedings of the Fourth Annual Meeting of the Niagara Linguistic Society: TWPL21.U of Toronto. Johns. (2008) “Restricting noun incorporation: Root movement.”
NLLT 25: 535-576. Kratzer. (1996) “Severing the external argument from its verb.” Phrase structure and the
lexicon. Kluwer Academic Publishers. Massam. (2010) “Deriving inverse order: The issue of arguments.”
Autronesian and Theoretical Linguistics: 271-296. John Benjamins. Pylkk¨anen. (2002) Introducing arguments.
Ph.D. dissertation, MIT. Rackowski&Travis. (2000) “V-initial languages: X or XP movement and adverb
placement.” The Syntax of Verb Initial Languages: 117-142. OUP. Roberts. (2010) Agreement and Head Movement: Clitics, Incorporation, and Defective Goals. MIT Press. Wurmbrand. (2014) “The Merge Condition: A
syntactic approach to selection.” Minimalism and Beyond: Radicalizing the interfacesed. John Benjamins.
Valuation and checking must apply to the same feature
Features with underlined values represent unvalued features that were valued through Merge operations. Those with “u”
represent uninterpretable features that have been checked through head movement.