handout pdf

Varieties of Coordinated Questions in Twi
ACAL 44, Georgetown University, 2013
Bradley Larson
University of Maryland
[email protected]
1 Basic Goals
Are examples (1) and (2) derived in the similar ways?
(1)
What and when did Dana eat?
(2)
What did Dana eat and when?
Giannakidou and Merchant 1998 would say: yes, via sluicing:
(3)
What did Dana eat and when did Dana eat?
(4)
What did Dana eat and when did Dana eat?
Larson 2012 would say: no, only (2) involves sluicing.
(5)
[What] and [when did Dana eat]?
(6)
What did Dana eat and when did Dana eat?
In this paper I look to Twi to support this second view.
I
Twi verbal morphology can be sensitive to transitivity.
II
Twi employs different coordinators in different environments.
III
Twi appears to lack the ability to sluice.
1 2 Bigger picture
Coordinated-wh constructions like (1) are perplexing in languages, like English and Twi,
that only allow single wh-fronting.
(7)
a.
*What when did Dana eat?
b.
*dɛn brebɛn na Kofi dii?
what when C Kofi ate
‘what did Kofi eat when?’
In such languages, it is not the case that the wh-words could have moved independently.
(8)
a.
*Whati and whenj did Dana eat ti tj?
b.
*dɛni ne brebɛnj na Kofi dii ti tj?
what and when C Kofi ate?
‘What and when did Kofi eat?’
Nor is it the case that the wh-words could have moved up en masse as a constituent:
(9)
a.
*[What and when]i did Dana eat ti?
b.
*[dɛn ne brebɛn]i na Kofi dii ti?
There are two ways of resolving this paradox:
One means is that of ellipsis (see Browne 1972, Bánréti 1992, Giannakidou and
Merchant 1998, and Whitman 2002)
2 ‘What and to whom did you give?’
b. Mit és hol javítottál meg?
what and where repaired.indef.2sg pv
‘What and where did you repair?
There is actually two full clauses, thus two wh-movements is fine
This suggests that both wh-words are actually undergoing movement
(10)
(34)
a.
[Whati did Dana eat ti] and [whenj did Dana eat tj]?
Whati and whenj Igor repairs ti tj ?
b.
[dɛni na Kofi dii ti] ne [brebɛnj na Kofi dii tj]?
But in Germanic, we seem to have this sort of pattern
Another means is that of differing
dependency-types.
Verb-Type:
eat
Verb-Type: fix
Wh-order: Arg first
Good
Bad
-Larson
2012
(also
Larson,
Lewis,
and
Kush
2012
and
Kush, Lewis, and Larson,
Wh-order: adj first
Good
Good
2013) posits that coordinated-wh questions are derived via long distance, overt whdependencies that are not derived via movement.
This is best accounted for with an analysis like the following:
(35)
(11)
CP
CP
whati
&P
C
&
CP
when
C’
C
TP
Ivan
T’
T
VP
eat-xi
t
--
This option is permitted under current Minimalist (Chomsky 1995) assumptions.
--
It entails that displacement not be monolithically driven by movement (or lexical
features in transformation-less theories)
"!
It relies on the possibility of a null variable down low to be bound (see Bresnan
Given this structure, the shared material ‘the book’ is c-commanded by elements
1978 and Johnson 2001).
!
--
from both conjuncts. This can easily tested empirically. Elements from both conjuncts should show the effects of c-commanding the shared material. This turns out
not to befrom
the case.
This differs
(12) in which there is coordination of two full clauses plus ellipsis.
Shown in this section, only elements from the second conjunct show c-command
effects while elements in the first conjunct do not.
(12)
[When did Dana eat] and [what did Dana eat]?
12
3 blame the lack of a clear confirmation of the theoretical prediction on an interfering parsing bias.
Another prediction that this approach makes concerns the directionality of the inchoate
conjunct. All of the instances that have been discussed are ones in which it precedes the more
fully fledge conjunct. If the conjuncts were transposed the result would be as in (50):
The sluiced wh-word could not possibly bind a variable in the first conjunct:
CP
(50)!
(13)
CP
when
&P
C’
C
&
TP
Ivan
CP
whati
C
T’
T
VP
eat-xi
t
In the above representation, the wh-word in the second conjunct is buried too deeply to
In this paper I argue that evidence from Twi supports this latter view over the former one.
bind into the main conjunct. The theory proposed here thus cannot predict that these sentences
are derived inGiven
the same
way as the other order. Sentences of this sort are acceptable:
this structure, the shared material ‘the book’ is c-commanded by elements
3 Coordinated
Questions in Twi
from both conjuncts. This can easily tested empirically.
Elements from both con-
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
juncts
show the argument
effects of c-commanding
the adjunct
shared material.
This turns
outa mix of both in
It is possible
toshould
coordinate
wh-words,
wh-words,
and
to be the case.
Twi: here, not
"!Again
when the parenthetical nature is highlighted, the sentence that would otherwise be bad, is improved:
Shown in this section, only elements from the second conjunct show c-command
(i)
(14)
Whateffects
(and while
whenelements
if I’m allowed
to ask)
did Ivy
fix?!
in the first
conjunct
do not.
ɛhe
ne brebɛn na Kofi didii?
where and when C Kofi ate
12
‘Where and when did Kofi eat?’
(15)
hena ne dɛn na Kofi bɔɔ?
who and what C Kofi hit
‘Who and what did Kofi hit?’
(16)
dɛni ne brebɛnj na Kofi dii?
what and when C Kofi ate?
‘What and when did Kofi eat?’
4 Why might we suspect that there is no ellipsis in (16)?
3.1 Transitivity distinctions in Twi
In Twi, certain verbs have different instantiations depending on whether they are used
transitively or intransitively (Ofori 2006).
(17)
a
o-dii nam
b.
o-didii
3sg-ate fish
3sg-ate
‘he ate fish’
‘He ate’
This holds for wh-constructions as well:
(18)
dɛn na Kofi dii?
(19)
brebɛn na Kofi didii?
what C Kofi ate?
when C Kofi ate?
‘what did Kofi eat?’
‘When did Kofi eat?
In coordinated-wh constructions in which the argument wh-word is the leftward one, we
find that only the transitive version of the verb is possible:
(20)
dɛn ne brebɛn na Kofi (*di)dii?
what and when C Kofi ate?
‘What and when did Kofi eat?’
This is unexpected under an ellipsis account. Under such an account the second conjunct
ought to look like that in (19)
5 (21)
*dɛn na Kofi dii ne brebɛn na Kofi didii?
what C Kofi ate and when
C Kofi ate?
‘What and when did Kofi eat?’
Under the differing dependency-type approach, this result is expected as there is a
semantic (though not syntactic) dependency between the argument wh-word and the verb.
When the argument wh-word comes second, it too effects the transitive version of the
verb. This suggests that the argument dependency conditions the change.
(22)
brebɛn ne dɛn
when
na Kofi (*di)dii?
and what C Kofi ate?
‘What and when did Kofi eat?’
3.2 Versions of ‘and’
In Twi there are at least two versions of ‘and’. One for clausal coordination, and one for
coordination of things smaller than clauses (Kobele and Torrence 2004):
Clausal ‘and’: na
(23)
me-huu Kofi na wo-bɔɔ Yaa
1sg-saw Kofi and 2sg-hit Yaa
‘I saw Kofi and you hit Yaa.’
Smaller ‘and’: ne
(24)
me-huu Kofi ne Ama
1sg-saw Kofi and Ama
‘I saw Kofi and Ama.
6 When coordinating full wh-questions, the clausal coordinator na is obligatory:
(25)
dɛn na Kofi dii na/*ne brebɛn na Ama didii?
what C Kofi ate and
when
C Ama ate?
‘What did Kofi eat and when did Ama eat?
But in coordinated-wh constructions, ne obligatory:
(26)
dɛn ne/*na brebɛn na Kofi (*di)dii?
what and
when C Kofi ate?
‘What and when did Kofi eat?’
This again militates against an account in which two whole clauses are coordinated.
Under the account proposed by Larson, we are dealing with coordination of a clause with
a ‘smaller’ constituent and this distinction is expected.
3.3 Sluicing in Twi
It is conceivable that the above data is what it is due to conditions on the use of na in
conjunction with ellipsis. The use of na necessary when coordinating full clauses so long
as there has been no ellipsis.
Yet Twi seems to not have ellipsis constructions generally:
(27)
*Kofi bɔɔ obi
na me-nim hena
Kofi hit someone and 1sg-know who
‘Kofi hit someone and I know who’
Instead, a non-interrogative pronoun must be used:
7 (28)
Kofi bɔɔ obi
na me-nim no
Kofi hit someone and 1sg-know him/her
‘Kofi hit someone and I know him/her.’
This makes an ellipsis-based analysis of coordinated-wh questions less plausible.
--
Recall that sentences like the English (29) were assumed to be derived via ellipsis
(29)
What did Dana eat and when?
--
If such is the case, then this type of sentence should not be possible in Twi. And
they are not:
(30)
*dɛn na Kofi dii na brebɛn?
what C Kofi ate and when
‘What did Kofi eat and when?’
--
Further, note that Twi optionally leaves its wh-words in situ (from Kobele and
Torrence 2004):
(31)
dɛn na wo-noae?
--
(32)
wo-noaa
dɛn?
what C 2sg-cooked
2sg-cooked what
‘What did you cook?’
‘What did you cook?
In these cases, it is sometimes possible to append a coordinated bare wh-word to
the in situ wh-word:
8 (33)
dɛn?
(32) Kofi
!ta bɔɔ
je hena
i ne
kada
Ivana popravila?
What CL and when Ivana repair
Kofi
hit and
whowhen
and what
‘What
(Serbo-Croatian)
‘Who and
what did Kofi
hit?
This extends
to multiple
wh-fronting
languages generally
(33) a. Mit és kinek adtál?
(Hungarian from Liptak)
what that
and the
whonon-clausal
gave.indef.2sg
Note however
coordinator is used here (obligatorily so). This
andistomere
whom
you give?’ going on. Further, non-argument wh-words
suggests ‘What
that there
DPdid
coordination
cannot be similarly coordinated:
b. Mit és hol javítottál meg?
what and where repaired.indef.2sg pv
‘What and where did you repair?
(34) *Kofi didii brebɛn na dɛn?
(34) *Kofi dii dɛn na brebɛn?
Kofi ate when and what
Kofi ate what and when
This suggests that both wh-words are actually undergoing movement
‘When and what did Kofi eat?’
‘When and what did Kofi eat?’
(34)
Whati and whenj Igor repairs ti tj ?
In this section, we saw that Twi does not seem to employ sluicing and as such a sluicing
approach
to coordinated-wh
is implausible.
The analysis in which coordinatedBut in Germanic,
we seemquestions
to have this
sort of pattern
wh questions differ from sluicing makes the correct split.
Verb-Type: eat
Verb-Type: fix
Wh-order: Arg first
Good
Bad
adjseen
firstevidence from
Goodverbal morphology, coordination
Good type, and the
--Wh-order:
We have
availability of sluicing. All of which point to the analysis in (35)
This is best accounted for with an analysis like the following:
(35)
(35)
CP
CP
whati
&P
C
&
CP
when
C’
C
TP
Ivan
T’
T
VP
eat-xi
t
9 "!
!
Given this structure, the shared material ‘the book’ is c-commanded by elements
!
!!!!!!/5!!"#$%&'(&%)$*+(),"-.)!'%/)0.&1"23))
!
MD analyses (Gracanin-Yuksek 2007, Citko & Gracanin-Yuksek 2010, Ratiu 2009)
!!!!!!/5!!"#$%&'(&%)$*+(),"-.)!'%/)0.&1"23))
!
! !
!
!!
(13)%$!!
!
!
&'(! !
!
+!
#$!
!
!%)!!%$!!! !
!
!
!
!
!
!
!
!
!#+!!!! !
!
%$! !#)!
+!
!#! ! ! !!%)!%$!
% ! &'! ! ! !! !%)!!! !! !!!&'
4 Another
possible analysis
(
*
+!
! ! ! % ! !! ! ! !! ! ! !!%+!!!!&'
! *! ! !!,$!!!%)!
The! data
above
could
also
be
explained
+ by a multidominance approach to coordinated-wh
! 4'*! !!,$!
!
! ! ! ! !! ! ! !! ! ! ! ! ! !!!%
!34'(! !R!3
questions following
Gracanin-Yuksek 2007 and Citko & Gracanin-Yuksek 2013
!
!
!
!
!!
!
!
!
! ! !3!4'(!R!34'*!
!
!
!!!!!
!! !! ! #$! !#)!
!
!
65!!"#$%&'(&%)$*+(),"-.)452#!'%/)0.&1"23!
!!!!!
(14)
(36)
!
!
!
#$!
65!!"#$%&'(&%)$*+(),"-.)452#!'%/)0.&1"23!
!
!
!
&'!(!
!
!
!
!
!
!
!
!!
!
!
!
!
!
!
!
!
%$!! ! ! ! ! %$!#$!
!
!
!
!
%)! %$!
! &'!*! ! ! %)! %$!
!
%+!! &'!(!! ! ,$!%)! ! ! &'*!! ! ,$ %)!
!
+!
! ! -./0%1!!! ! ! ,)! ,$! ! ! ! ! ,)! ,$
!
!
!
! 2$!,)!
! ! ! ! ,+ -./01!!! ! 2$! ,)!! ! !
!
!
!
!
!!
!! !
!
!
!
!
!
! 2+! ,+ !!!!!!!!!!3
!
2$! ! ! !
34'*2$!
!
4'(!
!
!
!!
!! !
!
!
!
+
!!!!!!!!!!34'(!
!
!!
2!
!
34'*!
786'!9:!3';!3'<;;!-3<.63.<;-!'8-!/;;=!1=>;?;=>;=3@A!?<9?9-;>!:9<!%&B-5!&;!>;?8<3!:<9C!
!
"!
786'!9:!3';!3'<;;!-3<.63.<;-!'8-!/;;=!1=>;?;=>;=3@A!?<9?9-;>!:9<!%&B-5!&;!>;?8<3!:<9C!
?<;D19.-!8669.=3-!1=!3'83!4;!8<E.;!3'83!1=!9<>;<!39!86'1;D;!>;-6<1?31D;!8>;F.86AG!
Grossly simplified, this approach holds that so long as the verb in question is optionally
?<;D19.-!8669.=3-!1=!3'83!4;!8<E.;!3'83!1=!9<>;<!39!86'1;D;!>;-6<1?31D;!8>;F.86AG!
transitive, then all else being equal, the sentence should be acceptable. The verb ‘shares’
H=1D;<-8@!I<8CC8<!'8-!39!8@@94!8@@!3'<;;!3A?;-5!,';!8D81@8/1@13A!9:!8!E1D;=!-3<.63.<;!1=!8=!
the to-be-moved argument wh-word and the to-be-moved adjunct wh-word.
H=1D;<-8@!I<8CC8<!'8-!39!8@@94!8@@!3'<;;!3A?;-5!,';!8D81@8/1@13A!9:!8!E1D;=!-3<.63.<;!1=!8=!
1=>1D1>.8@!@8=E.8E;!>;?;=>-!9=!1=>;?;=>;=3!:8639<-G!-.6'!8-!3';!8D81@8/1@13A!9:!C.@31?@;!
1=>1D1>.8@!@8=E.8E;!>;?;=>-!9=!1=>;?;=>;=3!:8639<-G!-.6'!8-!3';!8D81@8/1@13A!9:!C.@31?@;!
4'J:<9=31=E5!&;!@1=K!3'1-!?<9?;<3A!39!3';!8/1@13A!9:!%!8=>!6!';8>-!L3';!349!?'8-;!';8>-!
Optionally transitive verbs like eat work:
4'J:<9=31=E5!&;!@1=K!3'1-!?<9?;<3A!39!3';!8/1@13A!9:!%!8=>!6!';8>-!L3';!349!?'8-;!';8>-!
1=D9@D;>!1=!4'J;M3<86319=N!39!'9-3!C.@31?@;!4'J-?;61:1;<-5!&;!:.<3';<!8<E.;!3'83!3'1-!1-!
1=D9@D;>!1=!4'J;M3<86319=N!39!'9-3!C.@31?@;!4'J-?;61:1;<-5!&;!:.<3';<!8<E.;!3'83!3'1-!1-!
(37)
4'83!@1C13-!3';!-3<.63.<;-!1=!LO8N!8=>!LO/N!39!C.@31?@;!4'J:<9=31=E!@8=E.8E;-5!!!
4'83!@1C13-!3';!-3<.63.<;-!1=!LO8N!8=>!LO/N!39!C.@31?@;!4'J:<9=31=E!@8=E.8E;-5!!!
eat
what
when
,';!-3<.63.<;!1=!LO8N!1-!3';[email protected]@!-3<.63.<;!8<E.;>!:9<!C9-3!<;6;=3@A!/A!
,';!-3<.63.<;!1=!LO8N!1-!3';[email protected]@!-3<.63.<;!8<E.;>!:9<!C9-3!<;6;=3@A!/A!
I<1/8=9D8!L*++PN5!,';!<;C81=1=E!349G!=8C;@A!LO/N!8=>!LO6N!8<;!/[email protected]@!/.3G!6<.618@@AG!
Obligatorily transitive verbs like fix do not:
I<1/8=9D8!L*++PN5!,';!<;C81=1=E!349G!=8C;@A!LO/N!8=>!LO6N!8<;!/[email protected]@!/.3G!6<.618@@AG!
>9!=93!1=D9@D;!/86K48<>-!;@@1?-1-!9:!3';!K1=>!8<E.;>!:9<!/A!I18==8K1>9.!8=>!Q;<6'8=3!
(38)
>9!=93!1=D9@D;!/86K48<>-!;@@1?-1-!9:!3';!K1=>!8<E.;>!:9<!/A!I18==8K1>9.!8=>!Q;<6'8=3!
*fix
!
what
when
"
!
"
This makes the correct predictions for both English (noted by Whitman 2002, GracaninYusksek 2007, and Larson 2012) and Twi:
10 (39)
(40)
a.
What and when did Dana eat?
b.
*What and when did Dana fix?
a.
dɛn ne brebɛn na Kofi dii?
b.
*dɛn ne brebɛn na Kofi bɔɔ?
what and when C Kofi hit?
‘What and when did Kofi hit?’
The multidominance account predicts that only verb-type matters. But the account
presented here predicts the wh-word order matters as well.
Assuming that adjunct wh-words can always find a null variable to bind, they should be
acceptable as the first wh-word no matter the verb-type:
(41)
Wheni and whatj did Dana fix tj time-xi
And this is the case:
(42)
(43)
a.
When and what did Dana eat?
b.
When and what did Dana fix?
a.
brebɛn ne dɛn na Kofi dii?
(from Larson 2012)
when and what C Kofi eat
‘When and what did Kofi eat?’
b.
brebɛn ne dɛn
na Kofi bɔɔ?
when and what C Kofi hit?
‘When and what did Kofi hit?’
11 5 Conclusion
-- I have argued that coordinated-wh questions in Twi do are not derived via ellipsis or
multidominance
-- Instead, I posit that an approach to the construction wherein only one wh-word moves
is correct. Evidence from this view comes from four types of evidence:
1)
The transitive version of the verb arises in such constructions.
2)
The particular coordinator employed is the one that joins sub-clausal elements
3)
Sluicing does not seem to be available in Twi
4)
The order of the wh-words crucially determines the acceptability of the sentences.
6 References
Bánréti, Zoltán. 1992. A mellérendelés. [Coordination.] In Strukturális Magyar nyelvtan I. Mondattan. [Strucural
Hungarian Grammar I. Syntax.], ed. by F. Kiefer, 715–797. Budapest: Akadémiai Kiadó.
Browne, Wayles. 1972. Conjoined questions and the limitation on English surface structure. Linguistic Inquiry 3
Citko, Barbara & Martina Gracanin-Yuksek. 2013. Towards a new typology of coordinated whquestions. Journal of
Linguistics.
Chomsky, N. 1995. The Minimalist Program. MIT Press.
Giannakidou, Anastasia and Jason Merchant. 1998. Reverse sluicing in English and Greek. The Linguistic Review 15.
Gracanin-Yuksek, Martina. 2007. About sharing. Doctoral dissertation, MIT.
Kobele, Greg and Harold Torrence. 2004. The Syntax of Complement Clauses in Asante Twi. paper presented at the
35th Annual Conference on African Linguistics.
Kush, D., S. Lewis, and B. Larson. 2013. The what and when of processing coordinated-wh questions. WCCFL 31.
Larson, B. 2012. Wh-dependencies without movement in Germanic. Comparative Germanic Syntax Workshop
(CGSW) 27. Yale University, New Haven, CT.
Larson, B., D. Kush, and S. Lewis. 2012. Psycholinguistic evidence for non-syntactic dependencies in coordinated whquestions. Workshop on the Syntax & the Semantics of Sharing. Université de Nantes.Nantes, France.
Ofori, Seth 2006. Topics in Akan grammar. PhD dissertation, Indiana University.
Whitman, Neal. 2002. What and How We Can Learn from Mixed-Wh Interrogatives.Paper presented at Chicago
Linguistic Society, University of Chicago.
12

Download Report