Professional Documents
Culture Documents
Abstract
In Standard Yor~b~i, syntax affects the tone of both lexlcal and functional heads m different ways. Before an A c c u s a t i v e - m a r k e d c o m p l e m e n t , the m h e r e n t low tone of a monosyllabic
verb is suppressed. Conversely, m certain empty functional head posit,ons, a ' s p u r i o u s ' high
tone appears Both p h e n o m e n a arguably demonstrate the interaction of labeled phrase-structure with tonal feet. Accordingly, Yor6bfi prosody counts as an example of direct access
by p h o n o l o g y to surface syntax, as proposed by Kalsse (1985) and O d d e n (1990a). 2001
Elsevter Science B.V. All rights reserved
0378-2166/01/$ - see front matter 2001 Elsev,er Science B V All rights reserved
PII. S 0 0 2 4 3 8 4 1 ( 0 0 ) 0 0 0 2 5 - 5
82
pitch-accent. Such phenomena raise the question of whether syntactic objects affect
phonology directly, or whether syntax and phonology are mediated by prosodic representations which are less than fully syntactic, though they may have phrasal characterlstics. Selkirk's (1972) dissertation developed the indH'ect mapping theory of
The Sound Pattern o f Enghsh (Chomsky and Halle, 1968)' only a subset of postcyclic phrase boundaries are relevant to the application of phonological rules. ~ The
contrary view - direct mapping - was arttculated around the same time, in Bresnan's
claim that Chomsky and Halle's (1968) "Nuclear Stress Rule is ordered within the
transformational cycle" (1971' 257, emphasis added).
In its third decade, the debate is far from done (cf. Boolj, 1992). Kalsse (1985)
and Odden (1990a) defend direct mapping, Nespor and Vogel (1986) take the contrary view, and others have proposed hybrid theories (Selklrk, 1984a, 1986; Halle
and Marantz, 1993) Moreover, recent reassignments of the functions of grammahcal
modules in principles and parameters theories have reframed the discussion. In partlcular, economy-based analyses - w h e t h e r derlvatlonal (Chomsky, 1995) or representational (RlZzl, 1997) - force a reappraisal of the architecture of the grammar as
it pertains to overt syntax, w~th consequences In turn for the analysis of syntax-sensitxve phonological processes.
This paper considers audible effects of phrase structure - principally, the suppression of lexlcal low tone m Yorfibfi - and argues that direct mapping permits a 'null'
or nonstlpulative theory consistent with independent generahsatlons m the grammar
(Clnque, 1993). More abstractly, phonological and syntactic representations respect
the same structural condttions on locahty and recoverability. Most abstractly, there
IS a convergence of syntactic and prosodic conditions on the hcensmg of null positions, in the form of the phonosyntacttc ECP which requires that an ungoverned null
functional head be prosodically strong. This result converges with Longobardl
(1994) who argues that null functional heads obey strict recoverability conditions
Understood analogously, the phonosynlacttc EcP is one of the mechanisms available
in Universal G r a m m a r to satisfy recoverabdity.
Section 2 mtroduces the phenomenon Section 3 surveys ItS syntactic environment: the left edge of case-marked DPs, as well as certain CPs. The former are
widely d~scussed m the hterature, the latter less well so. Section 4 derives the effect
from the direct co-representation of phonology and syntax, what Rlzzl and Savola
(1992) call phono~yntax : Specifically, the phenomenon is argued to reflect a recoverabihty condition on null functional heads. Section 5 concludes by considering the
relationship of this condition to other phonological and syntactic constraints.
Selklrk's me&atlng representations are derived according to an algorithm that slmphfles syntactm
bracketing The separate theory ot LexJcal Phonology addresses structule-sensltJve word-internal
phonology, although the distinction between 'lextcal' and 'post-lextcal' phonology remains problematic
(cf Boolj and Rubach 1987) Pulleyblank (1986a 7) dlstmgmshes lextcal fiom phrasal phonology b)
charactensmg the lauer as exceptlonles,, Some exceptlonles,, processe%e g Yor~?.tb~tgerund reduphcatlon, are nonelheles., treated as lexlcal via feedback loops m stratal ordermg (Akmlabf and Oy6bfid6.
1987, Pulteyblank and Akmlabi, 1988)
Also known as wnta.~-phrmolo~,,~ ~ohabltat;on (Ddchalne and Manfredk 1995)
83
(1)
a. m r - m 6
GER-knOW
'knowing'
b. jf-je
GER-eat
'eating'
c. kf-k,6
GER-build
'building'
i
[[rnim55]]
L tone v e r b
[[d3id3~]]
M tone verb
[[k~fi]]
H tone verb
(2)
a. M o m o
i16
e
r~.
ls
k n o w h o u s e GEN 3S
'I know his/her residence'
b. M o j e
ilfi
ls
eat okro
' I ate ( s o m e / t h e ) o k r o '
c. M o k6
i16.
ls
build house
'I built a house'
[[...mS...11
[[...d3~...]]
[[...k5...11
L-drop is o b l i g a t o r y b e f o r e o b j e c t D P s b e a r i n g a n y tones. B u t b e f o r e a n e m b e d d e d
c l a u s e (CP), a n L-tone v e r b surfaces with either L or M, c o r r e l a t e d to a d i f f e r e n c e in
3 Data are presented in the orthography, which marks L and H tones with grave [" ] and acute ['] accents
respectwely (B~i~gb6st, 1970a). M tone is unmarked m the orthography, but where phonetic representations m double brackets are added for clarity, a macron [[-]] m&cates M. Following the convention of the
Yorfibd linguistic community m Nigeria, phonetic contour tones are transcribed as sequences of level
tones. This makes no claim as to vowel length, which is orthogonal to present concerns - although
lmpresslomsucally vowels with contour tones seem longer The superscript exclamation mark represents
downstep - what B~,Tagb6s6 (1966b) calls the "assimilated L tone".
4 O. Aw6b01tlyl (p.c) observes that L-drop falls to occur m eastern dialects such as Ofid6 All data
reported here belong to Standard YorOb& broadly identified with the &alect of Qy6 and its descendant
towns Keeping to the practice of Yorfibfi hngmsts, I m&cate L-drop m orthographic transcription (as
well as m phoneuc double brackets). That this non-lexlcal tonal effect is reflected m the orthography is
notable, smce grammatical tone is usually omitted from wrmng.
84
lnterpretatmn. (3) shows this with an embedded in&catwe clause introduced by the
c o m p l e m e n b e r pd.
(3) a. Mo m6,
[p6 a nflb
iyen]
l s know c 2P have.use that
'I know that we need that"
b. Mo mo
[pC a nil6
~yen].
'What I know ,s that we need that'
[[...m,5..1]
[[.. m,5...]]
The optlonahty of L-drop before clauses Is not hmlted to the indicative pd-complementlser, but extends to the complementlser k[, and to lnterrogatwe clauses introduced by b[. Throughout, the retentmn or non-retentmn of the lexlcal L tone has
interpretive consequences.
(4) a. Mo gb~
[kf 6 lol.
ls
receive c
3s go
'I agree that s/he should go'
b. Mo gba
[kf 6 lo].
'I accept the suggesnon that s/he should go'
'What 1 accept is that s/he should go'
(5) a. Mo m9
[ b f 6 tl pa eklhn].
ls
know way 3s pv kill leopard
'I knew 0t) as soon as s/he killed the leopard'
b. Mo m 9
[ b i 6 tl pa ekfm]
'I know how s/he killed the leopard'
[[
gbh...]]
[ [ . . . g b a . ]]
[[
mS..]]
[[...m5..]]
The m g h t c o n t e x t
[YP
85
For an example like (2a), the derivation proceeds as follows. In the input, only L and
n tones are specified, (7a). L-drop applies to m,b 'know', (7b). Subsequently, toneless
vowels are assigned M by default, (7c).
(7) a. M o [m o Iv [ i l
I
L
b. M o
[m o ]v [ i l
I
c. M o
[m o ]v [ i l
I
I
I
Q~
6 e r
I
~, ]NP
I
6 e r ~, ]NP
I
I
H
L-DROP
6 e r ~ ]N~
I I
I
H M
INPUT
OEFAULTM
A problem for (6) is posed by the fact that embedded CPs can also trigger the
rule. Carstens (1987: 10) attempts to handle this by restating the context in terms of
case features so that a verb's lexical L is deleted before any Accusative XP:
(8) L --'->~ / --Iv [XP+Acc
(8) correctly predicts that since DPs must be case-marked, they obligatorily trigger
L-drop. But (8) is formulated in a framework that adopts the Case Resistance Principle (Stowell, 1981), which bans [-WH] clauses from case-marked positions. This
incorrectly predicts an asymmetry between WH-clauses and non-wrt-clauses: the former should trigger L-drop and the latter should fail to do so. This expectation is confounded by the fact that L-drop is opnonal with embedded CPs of both types.
The present analysis makes five claims. First (following Pulleyblank), L-drop as
syntax-sensitive. Second (following Carstens), it signals case assignment. Third, its
apparent optionality with embedded CPs reflects a structural ambiguity: nominalised
CPs trigger L-drop, bare CPs don't. Fourth, the fact that polysyllabic L-tone verbs are
affected differently from monosyllabic ones reflects independent differences
between Genitive and Accusative case assignment. Fifth (following Manfredi,
1995b), a metrical analysis of tone makes possible a null theory of syntactically conditioned L-drop. Taken together, these claims require a model of grammar in which
syntax and phonology meet directly.
86
for a case-relation to hold. The solution can be cast in a representational or a derivational model. This paper adopts the representational case theory of Bittner and Hale
(1995). Assuming this much, we can ask how a phonological process such as L-drop
can access the syntactic structure relevant for case.
Consider the model in (9), in which the relation between syntactic and phonological representations is mediated by phonosyntax. At such a level, syntactic and
phonological representations may be understood as holding concurrently, or else perhaps as ordered relative to each other.
(9)
Indirect mapping theories exclude the architecture in (9), and account for syntaxsensitive P-rules by appealing to a level of representation in which syntactic information is only partially represented; this corresponds to the adjusted S-representation in (10). The claim is that only syntactic information which is consistent with
independently motivated prosodic constraints is available to the phonology. For
example, it has been argued that phonology is sensitive to edges, but not to phrasal
constituents or to category labels (Nespor and Vogel, 1986; Selkirk, 1986).
(10) Indirect mapping theory
NON-PHONOLOGICALS-REPRESENTATION
I
PROSODICALLY
ADJUSTEDS-REPRESENTATION(S)
NON-SYNTACTIC
P-REPRESENTATION
The schemas in (9) and (10) represent strong versions of direct and indirect mapping. It is also possible to countenance a hybrid theory, as in ( l l ) , where some Prules result from a direct mapping between syntax and phonology, while other Prules result from a mediated S-representation. 5
s Such a theory is adopted by Selklrk (1984a: 34), who recognizes the existence of P-rules that are
directly determined by S-representations(her S-structure), as well as P-rules that apply to adjusted representations (her Sn"and Sn"). In Selklrk's model, Focus rules and the Sense Unit Condition require PF
to feed back into LF. In a theory where phonology is syntax-sensitive,focus intonation would be read
off of overt syntax, and so would automatically have LF consequences.
87
3.1 DP complements
Starting from Carstens' (1987) claim that L-drop is the reflex of Accusative casemarking, a theory of case is introduced and the labeled syntactic bracketing which
triggers L-drop is identified (3.1.1). Confirmation that L-drop applies to labeled syntax comes from its sensitivity to both the complement/non-complement distinction
(3.1.2) and to complement type (3.1.3).
88
3 1 1 Accusative case-markmg
Case can be informally defined as a relation between a head and a dependent,
where head corresponds to a morphosyntactlc syntactic head such as V or Infl, and
dependent to DP The head is the 'source' o f the case feature, e.g. Infl is associated
with Nominative, transitive verbs with Accusative. The dependent DP bears a case
whose value is determined by the position it occupies in the structure, loosely a DP
I'S associated witfi Nominative i f it is in tile vicinity o f f n f f (i.e., in subject position).
and Accusative i f i t is in tile vicinity o f V (i.e., fn object position) Wfien the position that a DP occupies determines the case that it bears, it is often referred to as
'structural case' (Chomsky, 1981). For concreteness, this paper adopts the theory of
Blttner and Hale (1995). Their proposal distinguishes two types of structural case:
unmarked and marked, with the former corresponding to a bare DP, and the latter
corresponding to a DP contained with a Kase Phrase (KP)J ~ In a Nominative/
Accusative configuration, Nominative is the unmarked structural Case (=DP), while
Accusative is the marked structural Case (=DP introduced by KP). Accusative is a
feature associated with a K which heads a KP projection between V and DP. Nominative, by contrast, is a feature associated with some functional head above VP (for
Yorhbfi at least, this seems to be T). K ' s Accusative feature 1s assigned under lexical
head government, while T ' s Nominative feature is assigned under Specifier-head
agreement, cf (12). 7
(12)
TP
DP
T
T
[NOMI
VP
V
KP
K
[ACCl
DP
If the labeled bracketing relevant for Accusative case assignment is (13), and if
L-drop with verbs is the reflex of Accusative case-marking, then this labeled bracketing is the descriptive context for L-drop.
+' Lamontagne- a n d T,-a~ ~ 19~7) dlxo- proW,'~ a" KP" sira~-rmx:-, {Tat h',r fi~m-, all" cw;ewrrdrke~ m-g-am~W,
are KPs F o r Blttner and Hale, only ' m a r k e d siruclural case" ( E r g a i w e . A c c u s a l w e ) a n d mherenl c~se
(Genitive. D a t w e . etc ) is r e a f i z e d t~v K-P H i e presen-Patlorr 111 [fie lIfalll I~JKI- Ix, d SllT[p/:lfit~aI'lolI (Iiz [flu
Blltner and Hale proposal
v Altel'nalveel?; cane" calf be" mwfon'n-ly th'e-ated- as a" i'c,rm-a] ~%-ai'm-e"ow ,c Pan-ct-am-a'l"Ir~ad: so- i'Irai~,all"ca~,e"
relations are S p e c - H e a d c o n f i g u r a t i o n s (e g , C h o m s k y . 1995) A l t h o u g h the KP notation is translatable
into a S p e c - H e a d iheoLy of case, flTe t w o ailalyses m a k e & f f e r e n l Neehcn~ms a N m l possJhle p h o n o s y n tactic interactions e s p e c i a l l y as regards A c c u s a t i v e case This is related to the broader quesiion of how
t;)-treat subjecffobject- a.~?imwetne-~ anal- iread:gx~eTmne-lrc rff m-mmcairq- h=n-rre~u-cks tci- ~k~rn',i~z/m
1995)
89
a. 0 ja
ogun m6ta y]i.
3s fight war three DEM
'S/he waged these three wars' (cf. Abraham, 1958: 337)
b. 0 fi j~
p~ti-p~tL 8
3s IMP fight determination
'S/he is fighting doggedly' (Abraham, 1958: xiil)
(15) a. Mo ta j~ibu-j~bu.
ls sell unimaginable
'I sold unimaginable things' (i.e. everything you can think of)
b. Mo ta j~ibu-jabu.
ls sell unimaginable
'I sold (my wares) incredibly' (i.e. very successfully)
These contrasts show that L-drop Is sensitive to the complement relation. In terms
of bracketed strings, a case-marked complement is separated from V by one righthand bracket (...]v [KP..-), whereas an adjunct is two right-hand brackets away
(...]v]vP [NP...).
A complement/non-complement asymmetry appears with gerunds too (Abraham,
1958: xili, 306,425). Gerunds are formed by prefixal CV reduplication plus a grammatical H tone. Relevant here is the surface tone of the verb root. If L-drop occurs,
the DP following the gerund is construed as the object/theme of the gerund, (16a). If
the lexical L-tone of the verb is pronounced, the DP gets subject/agent construal and
Genitwe case, (16b).9
(16)
sbro.
[[gbfgbS...]]
difficult
[[gblgb~55...]]
s The Afncamst hterature describes items like piti-piti as ldeophones. They are variously translated
into Enghsh as adjectives or adverbs and their syntactic category is a matter of debate (cf Q Stewart,
1998, Aw6yal6, 1998).
9 The vowel of the gerund prefix is z, arguably Yor~b~'s unmarked vowel (Aw6bf~ltlyl, 1978a, Pulleyblank, 1988a, 1998) Gemtlve in Yor~bfi shows up m the form of a M-tone mora, which acqmres segmental content from the preceding vowel In (16b), the relevant vowel is [[5]], which Yor~bfiorthography spells on
90
The complementarity between L-drop and Genitive case confirms the claim that Ldrop requires Accusative case-marking. Moreover, if nomlnahsers freely attach to
different layers of a verb projection (Abney, 1987), then the contrasts in (16) can be
accounted for in terms of where the gerund attaches.
If the gerundive prefix attaches to VP, as in (17a), such a nommalisation has the
external syntax of a nominal (Le., it may occur in argument position) but retains the
internal syntax of VP In particular, VP nominalisation satisfies the structural conditions for L-drop (...]~ [K~. -)- The other possibility IS that the gerundive prefix
attaches to V, as in (17b). This type of nominahzation has the external and internal
syntax of N. Crucially, since KP is not sister to V, the conditions for Accusative Ldrop are not satisfied; instead Genitive is assigned.
(17)
NomP
gerund
f
I
V
I
I
gb[- gbo,n
shaking
b.
NomP
VP
NomP
KP
KP
gerund
DP
I
iy~un ndd
the flour.ace
r
gb[-
i
gbOn
I
on
GEN
I
mdtC) yi[
shaking
tl~e car
These &fferences in case assignment correlate with differences in the thematic role
assigned to the post-nominal KP. In (17a) KP is construed as a theme (the flour is
being shaken), but in (17b) KP is construed as a possessor/agent (the car is doing the
shaking). This leads to the following two generalizations for nominahsed L-tone
verbs (Manfredi, 1992b, Ddchaine, 1993). First, if a nomlnahsed verb loses its lexlcal L-tone, it is construed transitively and the post-nominal KP is interpreted as
object/theme; this is VP-nomlnalisation. Second, if a nomlnalised verb retains its
lexical c-tone, then it is construed intransitively, and the post-nominal KP has a subject/possessor/agent interpretation; this is V-nomlnahsatlon.
These complemenffnon-complement asymmetries support the claim that L-drop is
the reflex of Accusative case-marking. Nominal complements trigger Accusative Ldrop, nominal adjuncts don't. VP-nomlnahsatlon satisfies the structural conditions
for Accusative L-drop, V-nominahsatlon does not.
(18)
a.
b.
c.
d.
e.
Overt DP
Trace of A-bar movement (= WH-trace)
Pronominal clitic
Trace of A-movement (= 'NP-trace')
SubJect PRO
Cased?
YES
YES
YES
NO
NO
P-word?
YES
L-drop?
YES
NO
NO
NO
NO
NO
NO
NO
(n/a)
91
(18) shows that the necessary conditions for L-drop are both syntactic - the trigger must occupy a case-marked position - and prosodic - the trigger must be a
phonological word. Consider how the contexts in (18) satisfy (or fail to satisfy) these
twin conditions. L-drop applies consistently before overt DP's that need case, as
illustrated above. This extends to in sttu WH-words, as in an object echo question:
(19) a. O
ra
kfni?
buy what
'You bought WHAT?' (echo question)
b. *O r?a kfni?
[[...r~...]]
[[...r~...]]
But L-drop fails with A-bar movement (Carstens, 1987: 5; Yfisuf, 1989), e.g. WHmovement, focus-movement and relativisation, (20). The standard view of A-bar
chains is that the head is [-case] and the tail is [+case]. This means that an object
WH-trace is contained in a KP. Although this configuration satisfies the syntactic
condition for L-drop, it fails the prosodic condition: by definition, an empty category
like WH-trace is not a phonological word.
(20) a. Kfi ni
o r~
ti ?
what FOC 2S buy
'What did you buy?'
b. Isui nl
mo r~
ti
yam FO ls buy
'Yam is what I bought'
c. isu, tf
o r~
t~...
yam c
2s buy
'the yam that you bought ...'
Another context where L-drop fails to apply is with object clitics, (21). Chtics are
case-marked argument-type expressions, but they are not phonological words. Since
the trigger must be a cased independent word, clitlcs are correctly predicted not to
trigger L-drop.
(21)
ls
2s
3s
1P
2P
3P
mi
,e/o
O
wa
yfn
won
Kb m6 mr.
Kb m6,6/6.
Kb m6 fin.
Kb m6 w~i.
Kb mb yfn.
Kb m6 w6n.
'S/he doesn't know X'
[[...mSmil]]
[[...mS~/...m555]]
[[...m55]]
[[...mS w~t~i]]
[[...m~ja~]]
[[...mSrjW5 5]]
To verify that the failure of L-drop with these object pronominals is due to their
clitlc status requires independent evidence that that they are indeed clitics. Analyses
of Romance-type clitics put them m an A-position, and posit successive DP and D
movements (Kayne, 1990; Chomsky, 1995; Sportiche, 1998). At first glance,
92
Yorhbfi chtlcs do not exhibit D - m o v e m e n t effects since they occupy the same linear
position as phrasal arguments. Yet many reasons still exist to dlstingmsh them. First,
Yor/abfi chtlcs fail standard syntactic tests of independent words (Bfirhgbds6, 1967:
Pulleyblank, 1986b: 49; cf Kayne, 1975): they can't be modified, conjoined.
focused or topicahsed, while the independent forms permat all these things Second,
chtics differ from the corresponding independent pronouns in terms of their
prosody, most chtlcs lack inherent tone, whale the independent forms consistently
show a LM tonal melody. Further, chtlcs are either V or CV, whale the independent
forms have the canonical V C V stem shape of ordinary nouns'
(22)
ls
2s
3s
lP
2P
3P
ACC CLITICS
mt
,e/o,
O
wa
yfn
won
i
INDEPENDENTPRONOLNS
b-m1
i-wo
6-un
h-wa
~-ym
h-won
I
a. [,~,, 9- [~ c v l ]
b. [D Q [ , c v l ]
INDEPENDENT PRONObN
ACCUSATIVE CLITIC
So far, we have considered A-positions that are case-marked but d o n ' t trigger
L-drop. wH-traces and clitlcs in object position Next, consider how L-drop interacts
with A-positions that are not case-marked If an overt DP must occupy a case-marked
m It may be that the outer functional layer is distract from DP, e g Ddchame and Wlltschko (2000)
treat Yor/~bfipronouns as q~Ps, a proJection located between D and N
t t The syntax of Accusative cllhcs ~s discussed in more detail below The dlstlnchon between phlasal
and non-phrasal pronominal elements recalls Chomsky% (1995 337) distraction between complex and
sample pronominal elements respechvely
93
position, then a DP that occupies an uncased A-position has to move: this is Amovement (or 'NP movement' cf. Chomsky, 1981; Jaeggli, 1986; Baker et al.,
1989). Typical examples occur with verb-forms that fail to assign Accusative case to
their internal argument, e.g. passive, middle, unaccusative and rinsing verbs:
case-marked complement DP
caseless complement DP
[[...ta...]]
b. Bhta~ fi
ta
t,.
shoe NOM sell
'Shoes sold'
(27) a. Mo f,6
,6 ra bath.
ls want H buy shoe
'I want to buy shoes'
b. Mo f,6
batai fi ra t i .
ls
want shoe H buy
'I want to buy some shoes'
c. Mo f,6
bata, rf-ra tj.
ls want shoe GER-buy
'I want to buy shoes' / 'I want shoe-buying to occur'
[[...tfia]]
[[...ra...]]
[[... r~ia]]
[[...r~ia]]
In (26), the DP raises to Spec,TP, where it is assigned Nominative case via Spechead agreement. The DP-movement in (27) has not been extensively treated in the
literature: it minimally involves raising to the specifier of some functional head;
94
Manfredl (1997) treats ~t as movement to Spec,AspectP. In the present analysis, suppressrun of Accusative case correlates with the absence of a KP projection Hence,
these structures involve a caseless complement DP which predictably fails to trigger
Accusative L-drop.
In sum, the claim that L-drop occurs in the context .. ],, [~.. captures the fact that
Accusanve KPs trigger L-drop. It also accounts for the possibility of L-drop with
nomlnalised VPs: wherever the Accusative case configuration is present, L-drop
occurs. The analysis correctly predicts that L-drop will fail to occur if a nominal is
not case-marked: nominal adjuncts don't trigger L-drop, nor do caseless DP-traces
But as noted by Y~suf (1989), the failure of L-drop with a case-marked WH-trace is
not predicted by a purely syntactic analysis, and we have seen that L-drop also fails
with object clltlCS, even though they are case-marked. It Is these two contexts that
establish that L-drop is determined not just in syntax but m phonosyntax. We can
now turn to CP complements
3.2. CP complements
(28)
VP
V
KP
K
DP
Dxom
CP
C
TP
95
syntax in n o m i n a l i s e d clauses. Extraction from a n o m i n a l i s e d clause - with L-drop should violate subjacency, since n o m i n a l l s a t i o n adds a b o u n d m g node, but extraction
should be p o s s i b l e if the matrix verb keeps its lexical L. 12 This is true:
(29)
a. ]w6 ni
m o m6,
p6 a
nfl6.
b o o k FOC l s k n o w c
1P need
' I t ' s a b o o k that I k n o w we n e e d '
b. "]w6 ni
mo mo
p6 a
nfl6.
b o o k FOC lS k n o w c
1P need
[ ' I t ' s a b o o k that I k n o w i t ' s a fact that we n e e d ' ]
a. N -nf
ow6
pfip6, kb {*khn/kan} p6 a se is6 lfle.
GER-have m o n e y plenty NEG reach
c lP do j o b hard
' Y o u d o n ' t n e c e s s a r i l y get rich by e n g a g i n g m hard l a b o u r ' .
(Lit. ' H a v i n g plenty o f m o n e y d o e s n ' t require that we w o r k h a r d ' )
b. I-se
yen {*gbh/gba} 106 kf a jf
lw6 wb. 13
NOM-dO that receive
c c 1P steal b o o k look
' T h a t a s s i g n m e n t requires that w e c o p y from our n o t e s ' .
(Lit., ' D o i n g that requires that we p l a g i a r i s e ' )
I suggest that L-drop is o b l i g a t o r y w h e n e v e r a C P c o m p l e m e n t denotes a state-ofaffairs, which ts distinct from both event n o m i n a l s and p r o p o s i t i o n s (Zucchi,
1993). 14 It seems that if the subject denotes an abstract property, this forces the
e m b e d d e d CP to do the same, w h i c h requtres a state-of-affairs n o m i n a h s a t i o n , w h i c h
in turn triggers L-drop. This is consistent with the fact that the predicates kdn and
gbd n o r m a l l y take an animate subject and select a p r o p o s i t i o n a l c o m p l e m e n t ; in such
cases L-drop is optional, cf. (4) a b o v e for gb~. But when the subject ~s itself a n o m inalisation - as in (30a) n:ni ow6 p @ b ' h a v i n g plenty o f m o n e y ' , and (30b) is(yen
' d o i n g that' - the c o m p l e m e n t denotes a state-of-affairs. A l t h o u g h it is at present
unclear what forces this state-of-affairs constural, its correlation with A c c u s a t i v e
12 To my knowledge, Yor~Jb~idoes not have a contrast between so-called bridge verbs (e.g. Who did
you ~ay left?) and non-bridge verbs e.g (*Who did you whisper left's) See Carstens (1986) for relevant
discussion
i~ The p~ k{ sequence is a series of two distinct Comps' indicative p~, followed by subjunctive ki See
D6chalne (2000) for an analysis based on RlzzFs (1997) exploded Comp, with p~ mstantlatlng a higher
'force' Comp, and k{ a lower 'finiteness' Comp
14 Zucchl's criteria predict the following for Yorl)b~ (1) events should correspond to Gemtlve gerunds,
(u) states-otZafl~urs should be reafised by Accusative gerunds and nomlnahsed Cffs, (in) proposltlonai
entitles should be CPs and thctlve ctauses "Uais classification may provide a way to distinguish the
abstract 'fact' noun posited by Collins (1994) for Gbb from the event nommahser proposed by Manfredi
(1997 96f) for Yor~b The former should act like a proposmonal entity, the latter like a state-ofaffairs.
96
R - ~ 4 D d ~ h a m e / L n q,ua t l l
(2001) 8 1 - 1 3 0
L-drop provMes lndtrect support for the claim that Yonhb;i CPs undergo abstract
n o m m a h z a t m n ~5
3 2 2 N o m m a h s e d C P s with G e n m v e case
By hypothes~s, a n o m m a h s e d CP m obJect positron - triggering L-drop - ts
marked Accusative. This predicts that n o m m a h s e d CPs can occur m other cased
posmons, and th~s ms true Alter a noun that takes a proposmonal argmnent, e g hi
'way/manner" and 0,sdn 'reason', both pd and Lf-clauses can bear overt G e n m v e case
in the form of the ~4-tone mora (Abraham. 1958' 107; OyElfirhn, 1982a 116f )
(31)
As~in an
[p~ Reagan j,d
b'abfi r~,l k6 j d
kf a sun,.
reason GEN C
equal father 3s NEG allow c lp rest
'The fact that Reagan ts his father doesn't allow us to rest"
b Ewfir6 wo
alfihpat'a bf
t
[ki 6 kfi]
goat
look butcher way GEN C
3S die
"The goat looked at the butcher hke it would dte'
While m o n o s y l l a N c verbs assign Accusanve case, polysyllabic verbs assign Gemtive. 132a). With such verbs, Gemtlve Is opnonal before CP. 132b,c)
(32)
a. (5 rfin'ti
]
ml
3s remember GEN lS
"S/he remembered m e '
b. Mo rfin'ti
[pd
w~i].
ls
remember c
2P come
"I remembered that you (pl.) came"
it L-drop is also o b l i g a t o r y belore e m b e d d e d clauses c o n t a i n e d m rhetorical q u e s n o n s (l) Note the
r e m o t e n e s s ol the hteral mterpretatton 0-a), with the e m b e d d e d subject a "we" con,,trued as thslomt t i o m
the matrix wl~-word ta w h o ' More sahent is the bound variable reading (bb)
(1) Ta
m
k~
{*m~jmo}
pd ki a i i
]x~e ~ 6 '~
,aho t o ( Mc,
know.
~ lp steal book look
; e / a ) ' W h o d o e s n ' t know that we cheat >"
= (b) ' W h o doe,,n't know how to c h e a t " '
A q u e s n o n m a y be represenlend m one ol two w a y s ( G r o e n e n d @ and Stokhol. 1983. Engdahl 1986,
C h m r c h m 1993, H o m s t e m 1996) It can ask lor the value ol an m d p d d u a l ,,armble ~, a,, m (n-a) to
w h i c h a fehcmtous a n s w e r provides a hst o[ i n d i v i d u a l s Add, B(>s~ and Mr Friday ( d o n ' t know that we
cheat)' This interpretation is precisely, the one not a w u l a b l e foi O) Or a que',non can ask for the value
ot a funcnon variable / as m (u-b) Iln the hteratme, this is called a tunct2onal rote@re/anon ) Here, the
question being a s k e d is ' I s n ' t it true that ever~ X k n o w s F(X) '~ , 2 e ' l s n t ~t true that ever) ~ k n o w s
cheats '~ This f u n c n o n a l m t e r p r e t a t m n Js the only one a v a i l a b l e for (2) and seem,, to be h n k e d to the
s t a l e - o l - a l t m r s construal associated w n h the n o m m a h s e d ( ' P s that trigger L-drop
(u) ~ ( a ) Q { P P l s t r u e a n d l o r s o m e x
P=xNorknowwecheal}
= ( b ) Q {P P J s t r u e a n d l o r s m n e
f , P = ( N o l some, X k n o , a F I X ) ) }
Note that (u-b) u n p h e s that IP a Is a bound variable This is consp, lent with E k 0 n d a > 6 (1976) w h o
argues that the e x p r e s s i o n gbo~,,h() o wa all ol us c o m b i n e s the umvcr~al quannflet g h o g b o with ~aa
w h i c h he translates as "per,,ons who me here', s u b s u m i n g the d e n o t a n o n of Ip Gbo~,,l~o cannot a p p e a r b 5
itsell
97
c. Mo r~in'tf
i
[p6 e w~i].
ls remember GEN C 2P come
'What I remembered is that you (pl.) c a m e '
The optionahty of Genitive with clausal complements shadows the optlonality of
Accusative L-drop. Such case-marked clauses are often translated into English by
Yorfib~i speakers with a what-cleft, e.g. (3b) and (4b) for Accusative-marked clauses,
and (32c) for Genitive-marked clauses. This is presumably an attempt to render the
state-of-affairs construal associated with CP-nominalisatIon. The claim that nomlnalised complement CPs trigger Accusative L-drop also has implications for the
analysis of embedded wu-clauses, to which I now turn.
3.2.3. On the impossibility of embedded WH-clauses
From examples like (33), Carstens (1988) concludes that L-drop is obligatory with
embedded WE-clauses. The picture is more complicated, however. (34) shows that
such examples can also be pronounced with L tone on the matrix verb m,b ' k n o w ' , so
L-drop is evidently optional, even there.
(33)
(< m,b)
a. Mo mo
[bf 6 ti pa ekfin].
ls know way 3s PV kill leopard
'I know how s/he killed the leopard' (Carstens, 1988)
b. Mo f6
6
mo
[bf o ti rf].
Is want NOM k n o w way 2s pv see
'I want to know how you look' (Abraham, 1958: 104)
c. O mo
[bf a ti fi
dural.
3s know way 1p pv IMP balance
'S/he knows how to keep one's balance' (Abraham, 1958: 104)
d. A m o
[bf a tl se nfl~iti s6].
1p know way 1p PV do must do.3s
' W e know how we must do it' (Abraham, 1958: 104)
a. Mo m6,
[bf 6 tl pa ,ektm].
ls know way 3s pv kill leopard
'I knew (it) as soon as s/he killed the leopard'
b. Mo f6
6
m6
[bf o ti rf].
ls want NOM knOW way 2s PV see
'I want to know how you look' (Abraham, 1958: 104)
c. 6 m6
[bf a ti fi
dura].
3s know way 1P PV IMP balance
'S/he knows how to keep one's balance' (Abraham, 1958: 104)
d. A m6
[bf a ti se nflfitl s6].
1P know way lp pv do must do.3s
' W e know how we must do it' (Abraham, 1958: 104)
i
(34)
A second observation is also pertinent: all the examples in (33) and (34) contain
b/-clauses. Other than b/-clauses, Yon)b~i lacks embedded WH-clauses as a general
98
a. *Add m b / m o [ta
nl Ktinl6 rf].
know
who c
see
[Add knows who Ktinl6 saw]
b. Add mo
em
tf Kfinld rf.
know person ('
see
"Ad6 knows who Ktinld saw'
(Lit 'Add knows the person Ktinl6 saw'.)
,
The above points can be assessed together If L-drop apphes before a KP. it is reasonable to look for Independent evidence that interrogative b/-clauses which trigger
L-drop meet this structural description. The absence o f other types of embedded WHclauses In Yorfibfi suggests that b/-clauses are not [+wn] CPs. But if b/-clauses don't
have the structure of wft-clauses, what ~s their structure '~ There are reasons to tbmk
that b[ is the nominal head of a relative clause:
(36) [,,, b[,
[~,, OP,
[,~.
t, ..]1]
Accordingly, b/-clauses that fail to trigger L-drop will be analysed in the same
way as nominal adjuncts In both instances, the nominal head is introduced as a VPadjunct, so L-drop falls:
(37)
a. Mo m6
[bf 6 tl pa ektm]
Is know way 3s pv kill leopard
'I knew (it) as soon as s/he killed the leopard'
b. Mo t~
jfibu-jhbu.
I s sell
ummagmable
'I sold (my wares) incredibly' (1 e. very successfully)
And b/-clauses that trigger L-drop will be analysed in the same way as nominal complements. Both are in an Accusative configuratlon, so L-drop applies.
(38)
Mo mo
[bf 6 tl pa
ls know way 3s pv kill
'I know how s/he kdled the
b Mo ta
jfibu-j~bu.
ls sell unimaginable
'I sold unimaginable things'
ekan]
leopard
leopard" (Carstens, 1988)
In other words, the contrast between (37a) and (38a) reflects the attachment site of
the b/-clause' as a VP-adjunct m (37a), as a case-marked complement to V m (38a).
99
This is consistent with the interpretive difference between the two examples. Now
consider the evidence that bf heads a relative clause.
The present analysis concurs with the tra&tional view that bf is a noun meaning
'way, manner' (Abraham, 1958; Aw6bhlfiyi, 1978a; Oy~hir~, 1982a). With respect
to its internal syntax, b[ behaves like a noun in its ability to assign Genitive case to
a following noun:
aw6wa].
(39) 13 wf ej6 [bf i
3s say case way GEN grumbler
'S/he complains like a grumbler' (Oyhl~iran, 1982a: 114)
With respect to external syntax, a phrase headed by b[ can stand in the full range of
case positions. In subject position, a b/-clause is marked Nominative by the H tone
spell-out of null T, (40a). A b/-clause in object position can trigger L-drop, which by
hypothesis reqmres Accusative, (40b). A b/-clause is marked Genitive as the complement of the locative nominal in~ 'stomach, inside', (40c). Finally, a b/-clause can
be a prepositional (Oblique) object of s[ 'to', (40d).
(40) a. [BI isu 6
se ta] fi
yh
mf l',6nu.
way yam NOM Pv sell NOM open lS K-mouth
'The way yam sold surprised me'
b. Mo mo,
[bf 6 ti pa ekhn].
ls know way 3s Pv kill leopard
'I know how s/he killed the leopard' (Carstens, 1988)
c. So flin mi ohun tf k6 dfira n'fnti un
tell give ls thing c NEG good P-inside GEN
[bf ~ia tl se lo 6r6,,
yii].
way 1P PV do use word this
'Tell me what is wrong
in the way we have used this word' (Abraham, 1958: 104)
d. 13 pad?~ sf [bf 6 ti wh rf ].
3s return P way 3s PV be.at look
'It changed back to the way it was before' (Abraham, 1958: 105)
This establishes that the external and internal syntax of bf-phrases (including biclauses) is that of a nominal expression. And if b[ is a noun, it can host a relative
clause, so b/-clauses predictably trigger L-drop. This contrasts with Carstens' proposal that b/-clauses are wn-clauses, which leaves unexplained why b/-clauses differ
from all other WH-clauses in YonhNi in that the latter cannot be embedded. What
remains to be shown is that the b/-clauses have the internal structure of relative
clauses.
Yorhb~i relative clauses are head-initial; the head is usually followed by the element tL Obligatory ti is seen in (41a) with object relativisation. However, a restricted
set of head nouns mcluding ~ i 'person', ohun 'thing', ,bnd 'way' and igba 'time',
allow tf to be suppressed as in (41b), cf. Aw6bhlfiyl (1978a: 36).
100
(41)
A stmdar contrast occurs wtth relatlvtsed adjuncts. The adjunct gap ~s licensed by a
preverb, usually tt or se (Carstens, 1986). ~e' Almost all adjunct relatives require t[ as
in (42a) Like other adjunct relatives, clauses headed by the manner noun b[ require
a preverb: but in a b[ relative, 1[ is ungrammancal, (42b).
(42)
[lbl],
[*(ti)B6s~ tt fi
ser6
t,] ..
place REL
PV [MP play
'the place where B6,s6 is playmg ...' (Carstens. 1986" 4)
b. [bfl. [(*tf) B6@ ,6
tl rf
t,]...
way RLL
NOM PV see
'the way Bds6, seemed . ' (Carstens. 1986' 7)
a.
a. NP, [~, t[
b. NP, [~.~ t[
c NP, [~,.(tt')
d. b[, [ , , . 0
[,~,
t, .11
[~,, PV . . t, ...ll
[~
t, . l l
[~e pv . . . t , . . . ] l
In the present analysis, the optlonahty of L-drop before b/-clauses is accounted for
as follows. As a noun. b[ can host a relative clause; this is a h#clause Like other
nominal expressions, hi may be introduced as an adjunct to VP (and so does not trigger L-drop) or as a complement to V (and so triggers L-drop). This closes the discussion of the trtggers of L-drop; now consider the target of the rule.
3 3
L-drop targets an element that bears L tone If the trigger is an Accusative KP,
then only L-tone verbs can be targets Th~s predicts that no other case-asstgnlng head
16 These preverb particles are related to the independent predicates )e 'do' and tt 'from' (Ward, 1952
I091:, 13'-)) /g-'~'~e-' J'~ also iJo-,sd~le-a~ a- pre~erb- wnlr ad]mrct-extracnolr (Ward; 19~2 i09)
17 Ttns accords with AwdbfJlfiy] (1978a 36), who d~stmgmshes three types of relatives according to
whether C is obhgatordy filled(most Ns), opnonaily hiled (clauses headed by ~2" "person , ohun "thing"
tOi pellod'), ol obhgatorfly null (clauses headed by hi "way')
101
in particular L-tone stems which assign Genitive case - will be a target for L-drop
(3.3.1). If the context for L-drop is the head-complement configuration ...]v [KP....
this predicts that no other head-complement structure will be a context for L-drop
(3.3.2). Confirmation of these two predictions provides additional evidence that Ldrop is a phonosyntactic process that refers to labeled syntactic bracketing.
(44)
Accusative KP
Genitive KP
In Yor6b~i, Accusative and Genitive configurations also reflect the prosody and category of the head. Accusative KP is found with all monosyllabic verbs, (45a). Genitive KP is found with all polysyllabic verbs (Oybl~ir~n, 1970; Aw6bfalfiyL 1978a;
Elimelech, 1982; Aw6yal6, 1997a), and all nouns, (45b,c). Is
(45)
a.
monosyllabic verbs
polysyllabic verbs
noutlS
If L-drop is a reflex of Accusative case, then it will occur only in contexts such as
(45a). However, if L-drop marks the presence of KP, then it is expected to occur in
all the contexts of (45). In fact, Genitive KPs trigger something akin to L-drop, but it
does not have the same propemes as Accusative L-drop.
In deliberate speech, before a DP, the overt form of Genitive is an M-tone mora
that borrows the segmental content of the preceding vowel. This holds of nouns, all
of which assign Genitive case:
(46)
a. ilO
u
b~t~i
[[]ltJOb~t~ia]]
drum/beating GEN [type of drum]
'the bdtd type of drum, playing of bg~td drums'
[[kp~tSkiifimi]]
b. p~t~tk)
i
Timi
importance GEN
' T i m i ' s importance'
Js The preose case configuration of Yor/ab~ denomlnal polysyllaNc (Le. Gemtlve-asslgnlng) verbs is
beyond the scope of this paper, Manfredl (1997b 110-112) gives one possible structure
102
a. Mo phthk]
1
T]mL
ls
important GEN
'I m a d e T)m] i m p o r t a n t '
b. Mo fd'rhn an T~mL
ls
love
(;EN
"I love TimV
i
[[...kphthkiifim]]]
[[...f~r55timi]]
(48)
a
b.
c
d
]lu
pht~tkl
M o phthkl
Mof'.ran
bhtfi
Tim]
TimL
Timi.
[[ilfibhtfih]]
[ [kphthkifim] 1]
[[...kpht~tkitimi]]
[[ . f ~ ' r 3 t ~ m i ] ]
These fast-speech forms show that the L tone is j u s t retracted by one syllable, and not
deleted. A c c o r d i n g to the O b l i g a t o r y Contour Principle (Leben, 1973). L-tone stems
have a single L-tone autosegment. If Genitive L-drop deleted the L-tone, this w o u l d
predict the unattested outputs in (49), where all the syllables o f the head surface with
default M-tone.
(49)
a. *itu
bhtfi
b. *patakl
Tim]
c. *Mo pataki Timi
*[[ilabhtfi~t]]
*[[kpfitfikifimi]]
*[[. kpfitfikTfim]]]
The local nature o f G e n i t i v e L-drop - Jt affects only the last v o w e l o f the head suggests that it is a p h o n o l o g i c a l process. This accounts for Its sensitivity to whether
the f o l l o w i n g noun is C-lmtia[ or V'-inltlaL as weir as its sensitivity to tile initial tone
o f V-initial nouns (Bfiffagbds6, 1966a. 100f.). W e have seen Genitive t - d r o p is
optional before a C-initial noun. Before a V-Initial noun, Genitive L-drop is obligatory, as long as the lmtlal vowel does not bear L, (50). If the initial vowel bears L,
then neither the G e n i t i v e m o r a nor Genitive L-drop occur. Instead. the head noun
surfaces with its lexlcal L throughout, (51) 2o
~'~ The-drffeTence be-twee-ct de-aomrrtal- verbs artd V-N compotmds r,; rtor so gweat, strtc-e the ~atter are
semanncally noncomposmonal (OyOlfirhn, 1970, Aw6b~10yi, 1978a, Aw6yal6, 1997 lfi)
-~ The d~rectlon of vowel ass~mflanon m any gwen instance is determined by vowel quahty, among
other condmons (OyOlfirhn, 1970, Bfi~gb6%, 1986, Aw6b~lfiyL 1987. Pulleyblank, 1988a,b, 1998)
(50) a. *phthki
1
emu
importance GEN palmwine
'the importance of palmwme'
b. p h t ~ i emu
(51) a. pht~tki
bgfar6,
importance raphia.wine
'the importance of ~g~trb'
b. *p~t~k~ 1
~)g?ar6
c. *p~t?~kl 5g?ar6
103
*[[kp~t~k~g~m~]]
[[kp?~thk~gm~]]
[[kphtakbbgfar5]]
* [[kp?~t~tkb66gfarS]]
* [[kpht?ak66gt~r5]]
In sum, both Accusative and Genitive L-drop affect stems that end in L-tone, and
both are the byproduct of a case configuration. However, the phonologically conditioned behaviour of Genitive L-drop contrasts with the syntactically conditioned
behaviour of Accusative L-drop. Accusative L-drop is automatic and obligatory
before all overt DP complements, and is insensitive to the prosody of the following
DP complement. In contrast to this, the application of Genitive L-drop is determined
by phonological context: it is optional before consonant-initial stems, obligatorily
before vowel-initial stems that don't begin m L-tone, and ~mpossible before vowelinitial L-initial stems.
3 3.2
104
(52)
a. K6
NEG
'S/he
b. Y66
kUY
"S/he
1o.
go
dtdn't go'
1o.
go
will g o '
i
O 16
lo.
2s able go
' Y o u can g o '
i
L-drop syntax
A summap3'
105
complement must be a full word. Thus L-drop fails to occur before wn-traces and
before Accusative chtics, (54e).
(54)
a . . . . ]v [KF' [K Q~] [DP "'" ] ]
b . . . . ]x [Kp [K~t] [op ... ] ]
C . . . . ]v]vP [NP''"
-.]v [oP
-.]v [vp.-.
..]v [cP...
d . . . . IT [vp...
..]c [Tp''"
'']D [NP''"
..]p [DP''"
e . . . . ]v [Kp [DPt]]
..]v [K [D [N]
NO
NO
NO
n/a
n/a
n/a
NO
NO
The failure of L-drop before WH-trace and object chtics in (54e) can be understood in two ways. One could state L-drop as a phonological rule with syntactic diacrincs. This is essentially the view of Hayes (1990), but with the undesirable consequence of importing syntax into the lexicon. The alternative is to derive the
phonology of L-drop from syntax plus independently needed prosody Such a result,
which would qualify L-drop as a genuine phonosyntactic effect, is the goal of the
next section
} (M=~)
106
a. r i + hp6
b. *r'ap6
c. r~ip6
' s e e + bag"
[[rhkp6]]
[[r~kp66]]
Yonhbfi M can be d t s t i n g u i s h e d from both H and L on the basis o f its inertness wtth
respect to tone spreading. Both within and across words, Yorthbfi H and L spread
a u t o m a t i c a l l y onto each other, (57). By contrast, M ls not affected by spreading (58).
nor does it spread (59)
(57)
(58)
(59)
HL rf-r?a
Ow6 wh.
b LH ]we
Ow6 k6 p6.
a HM jf-je,
A,s6 pupa.
b LM 6fro
bath pupa
a. MH ajfi
aso pdfip6
b ML 9j~
aso o Tim]
'buying'
'There is m o n e y '
'book'
'The money is incomplete'
'eating'
'The cloth is red"
'law'
"red shoes'
"dog'
'short cloth'
'market'
' T i m i ' s cloth'
[[rfr~a]]
[[6w6wfih]]
[[]w66]]
[[6w66kph6]]
[[d3fd3r.]]
[[~ffkpnkpall
1161~11
I[ bhtakp0kpfi]l
l[ada~ill
[[~fakp6rjkp6]]
[[5d3~]]
[[ ~J'5 5 fimill
*[[d3fd3/;g]]
*[[afakp6~kpall
*116fii]1
*[I bhtakp60kpfi]]
*[[adgafill
~[[aJ'akpedrjkpe]]
*[[5d3~h]]
*[[aJ3 5 tfim]]]
2~ In the closely related Kru languages of C6te d'Ivolre and Liberia, Kaye et al (1982) postulate the
existence of a mid tone defined as [+L. +n] - ~e doubly marked. Many Kru languages contrast four lexlcal tone classes (H, "raised' M, M, L), as opposed to three m Yorhbfi 0t, M, i ) and two in most of the rest
of Kwa (H, L)
107
The inertness of Yor~b~i M can be derived by analyzing it as underspecified (Akinlabf, 1985, 1991; Pulleyblank, 1986a). As tonal autosegments, H and L spread, and
are the target of spreading, (60). The absence of an 'M' toneme accounts for its neutral behaviour with respect to tone-spreading, (61).
(60)
a. H
rira
'buying'
[[rfr~i~]]
(61) a. H
I
X
jije
'eating'
[[ dsf d3g ]]
b. L
i w e
'book'
[[iw~6]]
b.
H
I
X
aj
a
'dog'
[[ ad3fi]]
24 Enlarging the sample to the rest of Kwa and Benue-Congo, grammatical L tone IS far less c o m m o n
than grammatical H This IS especially true i f one sets aside tile abstract L tone posited as a diacritic trigger of nonautomatic downstep in ,~k~in (J M Stewart, 1965, Schachter and Fromkln, 1968) and Igbo
(Hyman, 1976, Wllhamson, 1986, Clark, 1989)
108
(62)
CP
C
TP
[Q--~H]
VP
[Q--~H] V
KP
K
DP
[Q--~H] D
..
Ii
[Q---)H]
According to (62) certain empty categories - the functional heads - are active m
phonology. Such PF interpretation of syntax is ruled out m an mdtrect mapping theory, but direct mapping expects it. One context for n-insertion is embedded infinttwes: after a matrix control verb (wtth any tone) a non-lexlcal H-tone appears
( B ~ g b 6 s 6 , 1971, Awdyald, 1974: 338ff., 1983), (63a). Thts infinmval n-tone is
analysable as the spell-out of a non-finite Comp. 25 H-insertion also spells out null T
(D6chame, 1993): in finite clauses whose tense position is not filled by an auxihary,
a non-lexical n-tone appears between the subject DP and the verb, (63b). In the hterature this is often referred to as the "high tone syllable'. Recall that the KP analysis treats Accusative K as an underlylngly empty K: the spell-out of null K ts a plaustble source of the H-tone of Accusative chtics (Manfredi, 1992b, 1995a) (63c).
Spell-out of null D is arguably attested m the H-tone assocmted wtth prefixal redupllcatmn of gerunds (Ddchame, 1993), (63d)
(63) a 6
f,~
~ f;,
a~o,.
11
:s The control 1t could mstantlate a finiteness Comp (hke ltahan de ~.f Rlzzl, 1997). a non-flmte T
(Ddchame, 199~), or some combmauon ot these (like Salentmo ku. ct Calabre~e, 1997) In t~vom ol the
Comp analysp, is the tact that YorOb6 mhmtlva[ 1t is in complementa D distnbuhon with the complementlser ldtl (Aw6yal6, 1974 3381 )
(i) a Olfi fd
6 Io
want II ~20
'Olu wants to go"
b O10 fO
tfih
1o
want for to go
c. r~t ~i
buy 3S.H
'buy it'
d. ri-r~
H.GER-buy
'buying'
109
[[ra~il]
[[rfr~ial]
a. 0 mb
6n.
3s know 3S.H
'S/he knows him/her/it'
b. 0 mo
~w6.
3s know book
'S/he is hterate' (Lit., 'S/he knows book(s)'.)
i
a. * 0
3s
b. * 0
3s
mo
know
m6,
know
p
on.
3s
6n iw6.
H book
I
I now present a theory of tonal prosody that captures the asymmetric behaviour of
tones and explains why, and where, Yonhb~i suppresses L and inserts H.
4.1.3. Tones are organtsed m metrical f e e t
Direct mapping claims that some phonological effects are determined by syntactic structure. A 'null' theory of direct mapping should postulate no mechanisms
beyond those needed independently in the phonology and syntax. Paraphrasing
Cmque's (1993: 239) comments in this regard for Enghsh phrase and compound
stress, the ideal analysis would be one in which no language-specific proviso is
26 Beside Yorfibfi L-drop, L-deletion is widely attested in Niger-Congo, e g ]gbo has many contexts
where a lexlcal L tone is not pronounced, but none where lexlcal H suffers the same fate, cf Clark
(1989).
110
necessary and surface tone can be enttrely d e t e r m i n e d on the basis o f surface syntactic constituent structure, gxven the lexmal tone m e l o d t e s and general principles o f
tone m a p p i n g To this end, the formal treatment o f tone a s s u m e d here ~s c o u c h e d
wtthm a general theory of mtonattonal p r o m i n e n c e that e m b r a c e s both "stress' and
' t o n e ' (Bamba, 1992: L l b e r m a n . 1995)
A metrical analysis o f Yorhbfi tone can capture the m a r k e d n e s s p h e n o m e n a
r e v i e w e d above, specifically the H > L > m rankmg. By means o f underspecificatton,
a u t o s e g m e n t a l theory captures the inertness o f Yor~hbfi M w~th respect to deletion,
msertlon and spreading, but a u t o s e g m e n t a l theory by ttself does no! predtct the other
two basic p r o p e m e s o f Yorflbfi tones: the lnsertabtlity of H (as the spell-out o f null
functtonal heads) and the s u p p r e s s a b t h t y o f L (m an A c c u s a t i v e configuratton)
These a s y m m e t r i e s b e t w e e n H, L and M are, however, ideally suited for metrical
analysts: (i) some item (H) Is u m q u e l y privileged to hcense a s y n t a g m a t l c domain, as
a head: (it) another item (L) Is uniquely d i s p e n s a b l e from the same point o f vtew,
(in) a third ttem (M) is mvistble to s y n t a g m a t l c processes altogether.
A n account of these tone a s y m m e t r t e s ts given by Manfredl (1992a,c, 1995b),
who draws on metrical tone analyses m g o v e r n m e n t p h o n o l o g y (Bamba, 1992), and
on instrumental studies o f Yorfibfi p r o s o d y (Connell and L a d & 1990: Lfinffan,
1992). 2v The m l m m a l prosodtc object ts a b m a r y foot c o m b m m g a strong (,s) and a
w e a k (w) posttton ( L l b e r m a n , 1975; L l b e r m a n and Prince, 1977). In g o v e r n m e n t
p h o n o l o g y ( K a y e et al.. 1990, Charette, 1991; H a m s , 1994), p h o n o l o g m a l d o m a m s
are configuratlonal, e.g. a foot mstanttates the h e a d - c o m p l e m e n t relatton, with ,s the
g o v e r n i n g head, and w the g o v e r n e d c o m p l e m e n t Notattonally, the head o f the foot
is d o m i n a t e d by a v e m c a l hne ( H a m s , 1994 150)'
(66) " ~
3
14'
27 For other analyses positing memcal tone feet see Huang (1980), Odden (1984), KJmenyl (1989) and
Ladd (1990) Essentially the same approach to tone prosody is assumed m register tone theories
(Clements, 1981, lnkelas et al, 1987)
2, Prosodic licensing, which reqmres that every pitch accent of an intonation contour be assigned, is the
tonal counterpart to constraints that associate each segment to a syllable or mola {McCarthy, 1979
SelMrk, 1981, 1984b, It6, 1988, Prince and Smolensky, 1993, Harms, 1994, Myers, 1997)
> Prosodic alignment generahses to 2-tone languages hke ",Ed6,]gbo and Efi k-lbibi6 which have an
H/L opposition In such languages, the weakest tone is predictably L This proposal departs from Mantredi (1995b 175), who proposes two conditions that do the same work ,is ploso&c allgmnent (l) t?;oject*OJl prevents ~i linking to s (Kaye, 1981) *[, M ], in) ,~,Ole;nme;lt torces H to link 1o ~ (Bamba.
1992) [, H ]
c. T o n e
111
In a language where prosodic licensing holds of tones, foot structure is built directly
on tonal elements rather than on rhymal or syllabic constituents. 3 There is typological support for this: tone languages typically lack rhymal weight distinctions, or
lack polysyllabic morphemes, or both, hence there is no rhymal/syllabic work for
metrical structure to perform (Manfredl, 1991: 93f., 1992a). On the same lines, languages with the most lexical tone contrasts (e.g. Mandarin) have less complex words
and syllables. Thus, in tone languages, tones intervene in the prosodic licensing of
feet and rhymes/timing units (cf. Odden, 1997). Now consider how the well-formedness conditions in (67) account for the surface realization of tone in Yor~bd.
Prosodic alignment requires that H -- the strongest tone - aligns with s, and that O
(phonetic M) - the weakest tone - aligns with w. Thus a HM melody (b(tre 'spinach')
and a simple-n melody (pdfild 'stockfish') are both licensed by a single foot, as in
(68a,b). A simple-M melody (saworo 'ritual bell') may invoke no prosodic licensing
at all, on the assumption that, in isolation, a zero element is invisible to licensing
requirements, (68c).31
(68) a.
b.
C.
S
X X
bu
re
p a n la
[[b6re]]
[[kp~fild]]
'spinach (Tahnum Triangulare )' 'stockfish'
sa wo ro
[[fawSr~]]
'ritual bell'
(68b) also illustrates that s implies w even if the latter has no rhymal content
(Bamba, 1992). This follows from the locality of the government relation, (69a).
~0 This departs from Qld (1995) and Ql~i-0n6 and Pulleyblank (1998, 2000), who assume rhymal foot
projection for Yorflbd
~t By hypothesis, non-isolated M - M in a string containing posmvely specified tones - is integrated
into prosodic structure This does not reflect any inherent requirement for prosodic licensing of M, but Is
parasitic on the hcensmg of the skeletal x-slot (moralc timing umt) For example, an MLH melody can be
resolved as LLH, e g Yor~bd is optionally pronounced [[jb(r)hbh~i]], by footing the initial, tonally unspeclfied syllable
b.
["N
O) a.
S
A I
L H ~
I/
X X
Ill
yo (r)u ba
[[J~(r)~bg~ill
P
L H
X X
I I I
yo (r)u ba
[[jb(r)6bS~il]
112
b. R e c o v e r a b d l t v :
This accounts for the prosodic structures associated with tt and g. What of L'~
Prosodic licensing requires all tones to be footed. Smce prosodic a h g n m e n t refers
only to the strongest and the weakest tones, L lS not subject to any alignment condition: ~t c a n ' t be parsed as the head (st because tt ~s not the strongest tone; tt c a n ' t be
parsed as the c o m p l e m e n t (w) because it's not the weakest tone So if L can be neither head nor c o m p l e m e n t , and if it must nevertheless be footed, then the only other
posslbthty ts that L attaches as an adJunct. In prmctple there are four potentml
adjunctlon sites, as In (70). L could left-adjoin, either to s or w, (70a.b) Alternatively, L could rlght-adjom, again rather to s or w, (70c,d).
(70)
'~
S
b
/4'
L H ~
* ('~--.,..
',
1,4
H L ~
c. * ~
5
t! L
d.
14
%\\
,%
HQL
Smce locality requires that a governor be adjacent to its govemee, this means that
adjomed L c a n ' t intervene b e t w e e n s and w, excluding (70b), and (70c). This leaves
two possibilities L adjoins to the left-edge of a foot, (70at: or L adjoins to the rightedge of a foot, (70d) Manfredt (1995b) argues that (70at ts the option selected by
Yor~bfi, and that (70d) is the option selected by [k6m-Y'al'a (cf. Armstrong, 1968)
One argument in favour of (70at ts that It accounts for Yombfi's unique property
a m o n g s t the Kwa languages in permtttmg automatm spread of t, onto H (deriving a
phonetic LH contour) Not only ~s automatic spreading of L onto H typologically rare,
it is perceptually more salient than the reverse This suggests that L-spread and Hspread, lhough they look symmetrical m autosegmental terms, express different
structural relations. M a n f r e d t ' s idea is that automatic spreading of L ~s a s m c t l y local
effect occurring within a smgle, branching metrical position. This is consistent with
the stronger acoustic profile of a d e n v e d LH contour, as compared with derived m .
In (71a,b), the dotted h n e indicates L-spread. (71c) shows Yorfibfi t, alone in a foot:
it locally identifies s, which m turn governs w. Crucmlly, L by itself Is not a prosodic
governor, i.e. it merely adjoms to s ~2
~-' There remains the question ot the lormal status of left-adJoined t Govermnent phonology lecogroses three hcensmg conhgurat~ons constituent hcensmg, rater-constituent licensing, and proJection
licensing (Harris, 1994) The members of a branching con,~tltuentstand m an as3,mmetllc relation, and
one must be the head, such ~on~tgtuc:t: hcetl,~l/tk, is head-mmal There can al',o be as3,mmemc fclahons
between con,,tltuents, such gttte~< oa~ttguent h~ en~taq, is head-final It is ,dso possible lut a head to project, plo/e~ ltoll hc ensmg is subject to parametric variation (head-initial m some language',, head-hnal in
others) As presented in the main text, the [s w] constituent can be defined either by constituent hcensmg or by projection hcensmg It mter-conshtuent licensing wele mw~lved m the [elt-adlomed structure
,a hlch licenses k In Yur~b,i th~s would force I, to head a consltuent (l e to [orm it,, m~'n foot) making I
,1 prosodic governol in its o w n light an unwelcome consequence
(71)
b.
a.
Q~
/q
L
~.~
L
c.
/N
,,4
113
J?~X
I XI
ku
ta
ko ko
[[k6k661]
'cacao'
k e n gbe
[[khl3gbhl]
'winegourd'
Ill
[[Ok~fits]]
' stone'
I I /~
H
I
~
{~
ko
ko
[[k6k66]]
'taro (Colocasta Esculentum)'
I I /q
H
I I I
X
kt ko ro
[[kNOr6]]
'bitterness'
In (72b), the hnkmg of M to w which blocks H-spread is achieved via licensing of the
rhymal timing unit (x). Even though M is zero in YonhbL a rhyme cannot be pronounced unless it is prosodically licensed. This shows that there is a close relation
between prosodic licensing of tone and timing units. Central to this analysis Is the
claim that tones intervene between rhymes and metrical structure, i.e. tonal strength
m tone languages is the counterpart of rhymal/moraic quantity-sensitivity in stress
languages.
Having introduced the organising principles of a metrical theory of tone, I now
show how the effects of L-drop and H-insertion are derived from the interaction of
metrical structure (specifically tonal feet) and recoverability conditions on syntactic
and phonological representahons.
114
Prosodtc Ece:
A null position must be prosodically governed
Not only does phonosyntax predict a parallel between the syntactic and prosodic
Ecp, it predicts that empty syntactic positions must sometimes satisfy the prosodic
ECP. This section argues that null functional heads m YoriJb~i have this property. I
refer to this convergence of syntactic and phonological recoverability as the
phonosyntactlc Ecp, (75) Because the phonosyntactlc ECP lS a byproduct of the syntactic and prosodic EcP, strtctly speaking, it need not be hsted separately. I do so here
for convenience.
(75)
Phonosyntacttc ECP
If a null F-head is not properly head-governed,
then it must be proso&cally governed 33
The phonosyntactlc ECP mentions 'null F-heads', rather than 'null heads" or 'null
syntactic positions'. This is a byproduct of Independent constraints on the licensing
and identification of empty syntactic categories Of the possible null syntactic positions, one may distinguish two classes: empty XPs (phrasal projections) and empty
Xs (heads). Empty XPs arise via chain formation, and their content is recoverable in
the syntax via antecedent-government (Rlzzl, 1990). As for empty Xs, they arise in
one of two ways: via chain formation or they are introduced as such in the structure.
In the former case, like XP-chalns, the content of the tail of an X-chain is recoverable from antecedent-government. What of the empty Xs that are introduced telle
quelle? UG countenances two types of Xs, namely lexlcal and functional. If we
consider which syntactic head positions are likely to be empty, 1[ is clear that null
functional heads are more easily recoverable than null lexlcal heads. All things being
equal, one expects null F-heads to be more pervasive than null L-heads. leading to
the formulation m (75).
The phonosyntactic EC'Phas consequences for the analysis of Accusative L-drop as
a syntax-sensitive P-rule, as the latter revolves a case configuration wtth a null functional head, namely K. More generally, it will be shown that both H-insertion and lodrop are different ways of satisfying the EcP.
115
a. k6
o
build.n 3s
'build it'
b. je
6
eat 3S.H
'eat it'
c. ra
fi
buy 3S.H
' b u y it'
a.
KP
b.
K
[9---)8]
KP
DP
D
K
NP
DP
K
I
N
D [~H]
to
NP
I
tN
Evidence in support o f (77b) comes from the linearization of the H-tone associated
with the null K position. If the Accusative clitic remained in-situ, then the g r a m m a tical n tone would linearise before the clitic, yielding ungrammatical (78a), instead
of the attested (78b). 34
"~4 In Nominative contexts, by contrast, the grammatical H hneanses to the left of the verb, and is pronounced on the final vowel of the subject, 0). The outcome m (I) follows from the absence V-to-T movement (D6chaine, 1992).
(I) imb
6n
le.
knowledge NOMH hard
'Knowledge is difficult'
116
(78)
a. *K6
NEG
b. K6
NEG
'S/he
m6i
6n
wa.
i
know H lP
m6I
w~
know 1P.H
d o e s n ' t know us"
The contextually determined reallsatlon of H tone with object clJtics is the effect of
the phonosyntactlc ECP. In a chtm configuration, null K is ungoverned: V does not
head-govern K, because ~t does not govern all segments of K (K is a two-segment
adjunct). Since null K is not properly head-governed, the phonosyntactlc ECP
requires that it be prosodically governed: this is satisfied by H-insertion) s (79) illustrates the reahsatmn of an object clmc after a H-tone verb. The phonosyntactic ECP is
apparently satisfied by the verb's lexlcal H, so pronunciation of the chtic as M follows. For example, a ls clitm m this context is pronounced [[ .. mi]], while a 3s clitlc
surfaces as an M tone copy of the verb's vowel: -~
(79)
a. ~
S
b.
HI'
- ~
S
14'
i
[, r;]
I
[[[mr ,1 ~o] ~ l
[[ rfmil]
'see m e '
[, r t l
[[O,,1 Q~]
[[ rill]
'see 3s'
N o w consider what happens after an k-tone verb. L is not a prosodic governor, yet
ungoverned null K must be prosodically governed. The only way to satisfy the
prosodic ECP is to insert U. In addJtmn, there is automatic spread of L onto H, Deldm g a surface rising tone, (80a). The 3s chtic copies the vowel of the verb. so ~t ~s the
copy that hosts the inserted H, yielding a surface rising contour, (80b).
~s in this configuration null D is also lexlcally ungoverned, so n-msemon may also be providing a
prosodm governor for D
~' hvthe-followmg-drscu,~slon;l-s ,m-relmeserrts d|ltqre-ch*Jc,~ w(drseg.lrrevgrdlCt)lrim~, as ot~r, ed-t~r3~.
which for Accusative objects contnbutes only vowel length
117
(80) a.
S
[v moj
[[mSmi t]]
'know me'
[[m55]] 37
'know 3s'
[[OD]QKI
After an M-tone verb, i.e. a verb without lexically specified tone, san 'bite', the
phonosyntactic ECP predicts that the clitic will surface w~th H, and it does: 38
a.
U.
S
I
H
[~san]
[[[mi~] ~ ]
[[ sSmll]
'bite me'
I I
~1
We have so far looked at clitics which lack inherent tonal content, but there is one
clitic that has lexical tone: H-tone 2P yfn. Ifyin follows an H-tone verb, the two H'S must
be contained in distinct feet, and a 'buffer' mora with M appears between them, (82a).
On the present analysis, this M-tone is the spell-out of the weak metrical position that
t 18
a n d L - t o n e v e r b s , n o b u f f e r m o r a is n e c e s s a r y , ( 8 2 b - c ) . ~'~
(82)
a. ~
S
~
14'
! I
.0
I I
~-~
b.
/4'
I I
X
[v rt}
It'
Ii
l[ rfisl7]1
'see you(ply
1[m~jah 1]
'know you(pl)'
It'
x
i
I
In addition to accountmg for the presence of H with object clltics, the phonosyntactic ECP generalises to other cases of H-insertion. 4 The grammatical H-tone that
appears with infinitival complements spells out the lower finiteness C. which is not
properly head-governed, (83a). The 'high tone syllable' of null T is reqmred in finite
clauses because T is not properly head-governed, (83b). The appearance of H tone on
reduplicated gerunds also follows: null D is not properly head-governed, so must be
prosodlcally governed, (83c) 41
~9 lgb6m~n'5_differs from Standard Yor/lb4 in treating all object clmcs as inherently toneless Consequently, with H tone verbs, even the 2p verb-chtlc sequence is HM m lgb6mina (Aw6~al6, 1996)
(i) a rf i-yin 'see you(pl)" Standard Yor~bfi
b. rf yin
"see you(pl)' Igb6mlnh
4~.~ Aklniab/ t i 9 9 r j gives subject and; object c[ltlcs inherent H but treats genitive critics as toneless The
present view - that, other than 2P 7h7, all clmcs lack lexlcal tone and the source of H In subject and object
clltiCS is an empty F-head - provides a more general account of the distribution o[ iL as well as a more
consistent treatment of pronominal chtics
41 The phonosyntactic ECP also predacts that the null ncrmmahsmg D a.~socmted wffh cased-marked CP..
(3 2 1) should be spelled out with H-tone since it is lexicall~ ungoverned. (l) However. no detectable
H-insertion occurs in this context The dtlference between the H-tone nommahser ol gerunds and the null
nommahser of case-marked CPs remains to be accounted for
I1
One posslbhty is that the null nommahser is not a governor in the relevant sense, and so is not subJect
to the ECP That something along these lines is going on is confirmed by the margmahty of clausal subjects (cI Koster. 1978) For Yorfibfi. this is illustrated by the contraq between (n) and (ni) A clausal
subject is possible if it is clefted, in which case a resumpUve subject clmc is obhgatory. (n) A clause in
subject position, with ,-tone agreement, is dispreferred. (m)
(11) [P6 kbk6 t5], m
6, wO
mf
(
cocoa sell ~oc 3s please Is
"That cocoa sells well is what pleases me'
(ln) ?[Pal k6k6 th]
J
wiJ
mf
c
cocoa sell AGR please l s
['That cocoa sells pleases me']
The margmahty of (m) may be related to the impossibility of H-insertion with nommallsed clauses
(83)
a ....
[CP [C Q~----gH]
b. [TP [DP ...1 [T Q~--)H]
C. . . .
[DP [D ~---)H]
119
[TP ..- ] ]
[VP ..-]]
[NOMPCV-] [v(P) ""]]
ObJect clittc
Object DP
With object DPs, H-insertion is not possible, instead there is L-drop. This poses two
questions: (i) what prevents H-insertion before object DPs?; (ii) what forces L-drop?
Both answers are syntactically rooted.
Consider the structure of an Accusative KP, (85). V properly head-governs null
K, so prosodic government is not required, and H-insertion is not necessary:
(85) [v-.. [K ~ * H
[Dp ...
The syntactic ECP accounts for the failure of H-insertion before Accusative KPs. But
why is L-drop obligatory in the same context? To answer this, one must consider the
three relevant Accusative case configurations: an Accusative KP introduced by an Htone, an M-tone, and an L-tone verb, as in (86).
(86) a. [vH
[KP[K~]
b.[vM
[~[XO]
C. *[V L [KP [K 0 ]
[OP''"
[D~''"
[DP "'"
As regards null K, the structures in (86) all satisfy the syntactic ECP: null K is properly head-governed by V. L-drop must therefore be an effect of the prosodic ECP,
which requires a null position to be prosodically governed. Thus, while (86a,b) satisfy the prosodic ECP, (86C) must fail to satisfy it in some way. It is instructive to
consider the relevant metrical structures. With H-tone verbs, the prosodic ECP is automatically satisfied, as the H-tone of the verb prosodically governs null K:
120
(87)
a . . ~ ~
b.
14'
["---.
/4'
I /i
/4'
@L
@ @
//'2
X Q~
x@
rl
kl
Ill
rl K (l)3"tl
r[ K be)
'see" "yam'
[[ risa ll
'see" 'shoes"
[[ r ~ f h t h ]]
'see' 'who'
[[rfkfni]]
t~
As for M-tone verbs, prosodic alignment requires that M, as the weakest tone, ahgn
with w. This means that M is always prosodically governed. Consequently, following
an M-tone verb. null K will be prosodically governed. (88) dlustrates this for an Mtone verb before DP objects that start with H, L and M.
(88)
a.
~,,
S
'
H
~
14'
b.
14'
Q~
~J
o
je K bH re
' 3 s ' 'eat' 'spinach'
[[ 6d3e btire ]]
It'
Qx
Jq
14'
14'
17
/I
ga
@ x
1(" K (;)SII
a. *
/?'>
x@
ra
b.
x
K
pan
~
x
la
'buy" 'stockfish'
[[rh~ikp4fil~]]
x
ra
14'
JJ'l
14'
J
@
x x
K ga
rt
'buy" 'tapioca'
* [[ r,aghfif]]
X@
x x x
ra K
sa
wo ro
t
"buy' 'ritual bell*
* [[ r'aJ'~w6r6]]
A minimal w a y to resolve this clash is to fail to parse L into the metrical structure.
This is L-drop
121
(90)
b.
a.
s
I
L
X
ra K
H
~)
C.
w
L
pa n
la
ra K
'buy' 'stockfish'
[[rakpfifil~i]]
~)
I XI
ga ri
ra K
'buy' 'tapioca'
[[rag~fif]]
Q~
s,a w o ro
b.
S
L
ra
c.
sawo
ro
'buy'
'ritual bell'
[[ra JSwor~]]
xOxxx
ra
K saworo
x
[ra]v[Kpt]
'buy'
[[r~]]
To explain the absence of L-drop with WH-trace, one might invoke Nespor and
Vogel's (1986: 48ff.) 'deletion convention' that erases syntactic empty categories
at the input to phonology. However, this won't work for Yo~b~, since both
H-insertion and L-drop crucially depend on the presence of null functional heads.
One could stipulate that the deletion convention only applies to empty phrasal
expressions, but this would import a syntactic distinction - XP vs. X - into the
phonology.
Another way out is to say that WH-traces are not case-marked (Borer, 1983), so
they do not satisfy the context for Accusative L-drop. But language-internal evidence
says the contrary. With object extraction, there is no direct evidence that the gapped
position is [+case], (92a). But subject and possessor extraction respectively require a
Nominative and Gemtive clitic at the extraction site, (92b,c):
122
(92)
a. KI,
nt
o ra t , ?
what Foc 2s buy
' W h a t did you buy?"
b. Ta,
ni
6,
ra
lSU?
who FOC 3S.NOM buy y a m
' W h o bought yam9"
c. Ta, nl
o ra
iw6 e
r~,i?
who FOE 2s buy book GEN 3S.GEN
' W h o s e book did you b u y ? '
(L~t., 'Who, dtd you buy hts, b o o k ? ' )
i
Ki
nl o m o
bf
Oj6 ,se je *(,6) ?
what Foc 2s k n o w way
pv eat 3S.ACC
' W h a t do you know how Oj6 ate it?"
Thus, the failure of L-drop with WH-traces is not accounted for by a deletmn convennon for e m p t y categories, nor is it accounted for by a stipulation about the case
properties of WH-traces. There is a third posslbihty, namely that the failure of L-drop
with WH-trace follows from foot structure. By hypothesis, null metrical posttions are
licensed only if they are recoverable. In particular, L, as the left adjunct to a strong
position, can locally hcense a null posmon. This is precisely what happens ~qth
L-tone verbs followed by a wry-trace. In (91c), the L-tone verb constttutes its own
metrical foot, and L-drop doesn't occur because there is no null K to prosodically
govern. 42
4 3
The possibility of I_-drop, and the absence of H-drop, follow from a metrical
analys~s of tone: all tones are prosodically licensed, tonal prominence ~s determined
by the hierarchy: H > u > O, with the most prominent tone (14) aligning with s, the
weakest (O = M) with w, and L attaching as a left-adjunct to s. A leading idea of the
phonosyntacttc analysis proposed here is that both phonological and syntactic representations respect similar structural conditions regarding the recoverabdtty of null
p o s m o n s ' the syntactic zcp requtres that an empty syntactxc category be properly
head-governed; the prosodic ECP requires that a null position be prosodtcalty governed. W h e n these two conditions converge on the same element, this yields
phonosyntacttc ECP effects' a null F-head which Is not properly head-governed must
42 One might conclude from this that the context for L-drop Is more accurately stated as ( ], [~. [~
O ) rather than ( ], [~p ) However, since the KP analysis treats Accusanve case as an underlymgly
null K, the latter automatically ]mphes the former
123
be prosodically governed. 43 The phonosyntactic ECP accounts for the occurrence and
distribution o f H-insertion. Finally, Accusative L-drop was also analysed as the effect
o f phonosyntax: when a properly governed null K lacks a prosodic governor, the
lexical tone of the governor is not parsed; this is L-drop.
This analysis accords with the general proposal that syntactic null heads obey
recoverablhty, e.g. null D-heads in R o m a n c e and Germanic (Longobardi, 1994). 44
Understood analogously, the phonosyntactic ECP is one of the mechanisms available
in Universal G r a m m a r to satisfy recoverability. These conditions are formulated in a
framework in which the government relation is an organizing principle for both syntactic and phonological representations. An open question is what determines, for a
given language, if the prosodic ECP is active for tone. Speculatively, I would answer
that this is the case when prosodic structure links directly to tones. There is the further question of w h y null syntactic positions are permitted at all. One possibility is
that null positions are never allowed, so languages choose between spelling them out
(as in Yorfab~i) or deleting them, i.e. deletion of null elements is a parametric option.
On another scenario, null positions would be permitted if they are recoverable, with
recoverabtlity defined configurationally: a position m a y be null ff it is governed. The
latter is the view adopted here.
5. Conclusion
The conditions under which a lexical tone is suppressed in Yorfab~i, as well as the
conditions which determine when a tone is inserted in a functional head, disprove the
extreme view that mapping in phrasal phonology is never directly triggered by syntax.
One w a y to reconcile this result with Universal G r a m m a r is to claim that part of
p h o n o l o g y - specifically, rules o f pitch - requires direct mapping between surface
syntactic structure and phrasal phonology, while indirect mapping obtains elsewhere
(cf. Selkirk, 1986: 400f.). Indeed, m a n y o f the clearest examples o f direct syntaxp h o n o l o g y mapping involve tone and/or intonation. Chinese languages are contourtone systems, and tone sandhi in those languages is apparently syntax-sensitive
(Chen, 1990; Lin, 1994). Bantu languages are level tone systems, and m a n y o f their
tone rules have been analysed phonosyntactically (cf. Odden, 1990a,b, 1994, 1995).
In effect, this reclassifies pitch as ' s y n t a x ' and not ' p h o n o l o g y ' - a suggestion which
is not unprecedented. For example, Remhart (1997), updating a tradition that
4"~ If the phonological analysis in this paper were framed in Optimality Theory (OT, cf McCarthy and
Prince, 1993, 1994), both H-insertion and L-drop would violate Faithfulness constraints H-insertion vlofates the constraint banning msertlon ira tone is m the output, then it is also in tile input (DEP-IO(TONE)')"
L-drop violates the the constraint banning d-eli~tion, ltoa tone is in tile input, then it is also in the output
(MAXqO(TONE)) The fact that L-drop applies only when H-insertion fails indicates that the constraint prohlbltlng deletion is more highly ranked than the constraint prohibiting insertion: MAXqO(TONE)
~>> DEPIO(TONE) But even an OT analysis needs to state that the contexts where Faithfulness wolatlons arise are
syntactic, so phonosyntax remains relevant
44 Longobardfs (1994) analysis is extended to |gbo by D6chaine and Manfredl (1998)
124
includes Jackendoff (1972), holds that, because focus intonation has interpretive
effects, semantic interpretation reads both PF (phonology) and LF (syntax). In the
present proposal, this would be captured by saying phonosyntax determines semantic interpretation.
Alternatively, following the sptrIt of Kayne (1994) and Marantz (1995) one could
reclassify the 'syntax' of L-drop, namely the head-complement configuratmn, as part
of morpho-phonology This is possible to the extent that L-drop has no semantic consequences, so the immediate problem would be to find an independent account of the
semantic effect of L-drop before clausal complements.
If one wanted to nevertheless keep L-drop m the phonology, the most opttmlsttc
scenario lS to dispense with labeled syntactic bracketing and claim that the trigger
for the rule is a generahsed head-complement configuratton: L-drop occurs at the
left edge of any complement XP. Any such attempt to save indirect mapping faces
several hurdles. First, there are head-complement structures where a monosyllabic
head bears L-tone, but there ts no L-drop, e.g. 1.-tone auxiliary verbs (cf. 3 3.2).
Sttpulatmg the target as a lextcal head may manage to exclude these, but at a price
which an indirect mapping theory worthy of the name cannot afford. The only open
lexlcal classes in Yor/Jbfi are V and N, so restrtctlng the target of L-drop to "lexlcal
heads' smuggles syntactic category back into the structural description, hence the
indirect alternattve is not so different from what Jt wants to disprove Second,
embedded CPs are head-complement structures which undergo t-drop only optionally Recall that if L-drop does occur the CP is a weak island for extractmn, and conversely there is no island effect if L doesn't drop. Thts is exactly the reverse of what
a generahsed head-complement analysis expects, since fl is obhged to treat the nonL-drop cases as extraposed or adjoined in some way (c) la Carstens). By contrast, the
syntactic solution proposed in 3.2, which posits an abstract nommahser in the
L-drop cases, pushes more, not less, syntactic categorlal infomaation into the phonology. Finally, indirect mapping is less economical than the phonosyntacnc account,
because the prosodic mechanism of tonal feet (4 1.3) is needed lndependentl), to
handle a wide range of facts about the &strlbutlon and reallsatmn of tone in Yor~bfi
and other languages. This last pomt makes any diacritic treatment of L-drop (e.g a
precomptlation analysts along the lines of Hayes (1990)) undesirable smce such an
approach would fail to capture significant generalisations about the relation of Ldrop to H-insertion.
Phenomena discussed throughout this paper point to further simplification of the
syntax-phonology archttecture. For example, stem shape m Yorfabfi correlates with
grammatical category root verbs are canonically monosyllabic CV, while derived
verbs are polysyllabic. This has implications for case: monosyllabic verbs assign
Accusattve; polysyllabic verbs assign Genitive 4s Pronouns also divide into prosodic
4s
The hnk between prosody and category is conhrmed by the behavlour o[ "sphttlng' verbs such as
bgl/ damage', which are synchromcally opaque the interpretation of bd;l is not reducible to its conslituent parts bdr 'come into contact with" and j "be equal to" Such verbs assign Accusatwe or Genitive
according to whether the nommat complement occurs inside or outside the verb complex (Av, db/flfiyL
1969 Awdyal6, 1996 12) If the oblect of hdtl follow~; the first s)llable (glo~,,,ed ~ ,~'~I, it trigger,, I -drop
125
classes, with syntactic consequences: VCV pronouns are phrasal, while (C)V pronouns are clitics. Taken together, these effects suggest that, in addition to a direct
mapping between tonal prosody and syntax, in Yonhb~i there may even be a direct
phonology-syntax connection between syllable-based phonology and syntactic representation. 46 (Dis)confirmation awaits further research.
and bears Accusatwe, (1). Alternatively, for some speakers and some verbs, the object can follow the
second syllable, but then it bears Oblique case, (11) a full DP is case-marked by preposmonal ni, and the
object critic is G e n m v e File object offa splitting verb m a y be fronted, and-L-d~op predictalSl~ tltl[S to
occur, (m)
(1) a. Mo ba 116
j~.
ls cv~ house ACC CV 2
'I damaged the house'
b Mo b'~ ~i
j,6
ls CV~ 3S.ACC CV2
'I damaged it'
(11) a Mo b'hj6
nf
116.
l s damage OBL house
'I damaged the house' (nonstandard)
b Mo bhj6
e,
r~.
l s damage GEN 3S.GEN
'I damaged it' (nonstandard)
(Ul) I16
b~j ,~.
house damage
'The house got damaged'
45 fn addition, vowel-del-etlon, which applies to contiguous vowels at word-6ounditrles, is afso sensitive
to the number of syllables m the first word If the stem is monosyllabic 0 e a CV verb), vowel deletion
applies, (1) In (1), L-drop could hypothencally be said to affect td 'sell' before Og&-,b and ilO,kO, 'beads',
but the lhcts are equivocal- vowel-elision makes Accusative L-drop inaudible betore a vowel--lmtml.
L-mltlaf noun - whether efislon is letiward as m (i-a)` or rightward as in (i-b)
(l) a Kb 'ta'
6gfarb,
[[kb t6g~rS]]
NEG sell r a p h m w m e
'S/he did not sell Og~rb'
b K6 'ta' il~,k~
[[k6 thlkk~]]
NEG sell beads
'S/he did not sell beads'
If the stem IS polysyllabic (nouns and derived verbs), the vowels ass,mdate but there is no ehslon:
(ll) a Mo f,6'rhn 6gOrb
[[.. fkrbbgtar5]]
1s love raphm.wme
'I love ~gftrb'
b Mo p~thki
bg~r6,
[[. kp~thkbbg~r5]]
ls important raphla wine
'I made Ogi~rb important'
Again, we see a correlatmn betwen prosody and case Vowel deletion apphes m contexts where
Accusative is assigned, l e. with monosyllabic verbs And vowel delehon fads to apply in contexts where
Gemtlve is assigned, i e with polysyllabic stems (nouns and derived verbs) The one exception to this
last generahsatlon is b[ ' w a y ' , the only monosyllabic noun m the language. Like all other nouns, It
assigns Genitive case. Ftbwever, its betiawour wltti respect to vowel- e h s m n is compii~x when b f h e a d s
a relative clause it behaves like other monosyllabic stems and undergoes vowel deletion, (rim). However,
when hi is tbllowed by a case-marked Genitive DP, it fads to undergo deietlon, (illb)`
126
References
Abney, S , 1987 The Enghsh noun phrase m its sententlal aspect Dissertation, MIT, Cambrtdge, MA
Abraham, R , [958 Dlchonary of Modem Yorhbfi London Umverslty of London Press
Akfnkugb6, O , 1978 A comparative phonology of Yor/Jbfi dmlects Dissertation, Umverslty of Ib'ad'an
Aklnlabf, A , 1985 Tonal underspeclflcatlon and Yorhbt~ tone Dissertation+ Umverslty of ]b'adan
Akmlabf, A 1991 A two-tone analysis of Yor6bfi Research m Yorflbfi Language and Literature 2,
59-85
Akmlabf, A and F Oyhbfid4, 1987 Lexmal and postlexlcal rule apphcatlon Vowel deletion m Yor~ba
Journal of West African Languages 17, 23-4-2
Armstrong, R 1968 Yhlh (Ikom) A "terraced level' language with three tones Journal of West AfNcan
Languages 5, 49-58
Aw6bhhiyi, O . 1969 Sphttmg verbs m Yor/ab~i Actes du 8e CongrOs de la Socl6t4 Lmgulshque de
l'Alrlque Occidental 1, 151-64 Umversltd d'Abldjan
Aw6bhhiyL Q , t972 Predtcattve a.dject, v e s m Y~thbfi A crtttque In A B',ir~_gbd,~d( e d L The YorhbA
verb phrase, 103-118 [b-aden lbi~d'an Umverslt} Press
Aw6btlhiyi, O , 1978a Essennals ol Yortlbti grammar lb'ad'an Oxlord Umverslty Press
AwOb'-hl~yL. Q . tg-TSb Fov.o:s o:~tt;VJa~ct,:,~,xsa:.;.m3m:t ~kca'5:,; kcrt_$~'J:e;i:ccA_ctdi:yx,:,,4. e*'~-i:i_'4
Aw6b~hlyi, Q , 1987 Towards a typology of coalescence Journal ol West Mrlcan Languages 17 5-22
,'~"~'6"b'i"j~:i'y{. Q. 1:9q'75 [%.~,'ft':.'*. ,:,X 8Xe'_ ~.},:Z.~-; ,;t,*2. '~J'=.~X,~.ks-XO~'k'[,;tO';~2 g.'t,X.. ,:_~a,:t~.i:.,rcX,;5~,;t,;r2;. J:,a,x,~=r;axk ,a~ ~.V+e'~g
African Languages 22, 69-88
Aw6b~hlfiyL Q ~ t992b Aspects of comemporary standard Yodab,:t m &.alecmIogmal perspecttve l.n. A
l~olfi (ed). New findmgs m Y-ori~ba studies (J F Qdtinj<,) Memorial Lecture Serms 3), 8-79 Distributed by Department of Llngmstlcs and African Languages, Umverslty of ]b-adan
Aw6bMfiyL O , 1997 Vowel epenthesls m YorOb~i Manuscript, Umverslty ol ]lorm
Awdyaid, "Y. [974 Stttt}ies tit t[-te syrttay., arut sem.a~tccs of Y-oo'd~:~_m:u:tto:t.t[r..r,~to{tx {)e,,~ect,..~.~ U_rttverslty of IIhnols at Urbana-Champmgn
Aw6yaid, "Y, t983 On the deveMpment of the verb-mfimtlve phrase m ~"o~fibg Studms m Afhcan Lmgmstlcs 14, 71-102
Aw6yal6, ' Y , 1985 Focus as an unbounded movement rule m Yor/abfi Journal ol the Lmgmstlc Association of N~germ 3, 75-84
Aw6yal6, ' Y , 1996 C 'la~sshOleS on Y<~rfib'gphotmlogy aox.t ~ynta.x+ 24 942 Dept of Lmgu!stws and Ph:losophy, MIT Cambridge, MA
Aw6yald, ' Y , 1997a O ~ e c t positrons m Yori]b'g In R -M Ddch~n_e. V M,.mfre& (eds ,L ObNct pos~uons in Benue-Kwa, 7-30 The Hague Holland Academic Graphtcs
Aw6~ral6, ' Y , 1997b The interrelationship between focus and sentence-hnal phenomena m YorObfi
EvMence from w~thm and beyond the syntax Paper presented at the 9th Niger-Congo Syntax dnd
Semanucs Workshop, Umverslty ol Ghana, 30 June-2 July 1997
Aw6yal6, ' Y , 1998 Adtcttonary ~t Y o r ~ , i tde<~ph~me~ M~ Lmgmsttc Data Consort:urn Un:,~ers:ty of
Pennsylvama, Phdadelphla
Baker. M . K Jotmsott aJa_d [ Robert~ 1989 Pass,_ve argumems ra:sed Lmgu>t,~c [nqmry 20, 219-252.
Bamba, M , 1992 De l'lnteractlon entre tons et accent D~ssertahon, Umverslt6 du Qudbec ~ Montrdal
Btiff~gb6s6, A , 1966a A grammar of Yor~)bfi Cambridge Cambridge Umverslt> Press
Btiff~gb6s4. A . t96@ The assimilated low tone m Yorflbfi Lmgua 16. 1-13
(m) a 0 mo
bf
IAd6 se .If
ow6l
[[ bdd6 ]]
3s know way Add pv steal money
'S/he knows how Ade stole the money' (QIfi-(Snh and Pulleyblank, 2000, (9b))
b 6 ri l b'6
tl d~ira
bf
1
6,stlm'ar~.
* [[ b6,~flmar~ 1]
3s see 3s wa3, 3s PV beautiful way <;EN rainbow
He saw th-a-tshe was a', beaatffM ,is gne rainbow' (Abraham, t9.58 [07)
Whether or not these restrmons on vowel deletion are redumble to case conhguratlons remains an open
quesnon See Qlfi-0nh and Pulleyblank (1998, 2000) for prosodic restrictions on vowel deletion
127
128
D~mmendaal, G (1995) Metatony in Benue-Congo Some further evidence for an original augment In
E Em6nanj% Q i'qdlm6le (eds), Issues in African languages and linguistics, 30-38 ~.b~i National
Institute for Nigerian Languages
Ektinday~). A , 1976 The calculus of the Yort)bti umversal quantlfiel Journal of the Yofflbfi Studies
Association ot Nigeria 2, 59-70
Eilmeiech. B-, f982- SyifaNie counting zn Y-orfiba Stucfies in Alrlcan Linguistics 13, 77-88
Engdahl E . 1986. Constituent questnms Dordrecht Reldel
Fukul, N i995 Theory of projectJ~m m ',ynJax QNL1 PuhlJcallons~dJstr!buJ.ed~by Czu~_br!dge Umverslty
Press
Goldsmith. J 1973 Autosegmental phonology New York Garland
Grlmshaw. J , 1991 Extended prolectum Ms Brandeis University, Waltham, MA
Groenendljk, J and M Stokhol. 1983 Interrogative quantlflers and Skolem-functlons In K Ehhch, H
van Rlemsd@ (eds), Sen_tern'e. discourse and mxt T!!burg Tdburg Un:,,erszV Press
tlaegeman. L , 1991 Introduction to government and binding theory (2nd ed ) Oxlord Blackwell
Halle. M and A Marantz 1993 Distributed morphology and the pieces of inflection In K Hale. S J
Keyser (eds), The view from Building 20, 111-176 Cambridge MA MIT Press
Harris, J , 1994 Enghsh sound structure Oxford Blackwell
Hayes. B .. 1990 Pmcamptl_ed pbr_a>,aJ_pbsan).l.o.g~ b.~. S l.o.kN.as..O Z3-..,:(,e:~k~,k. Tbe p.b,~JaN.o.gy-,,yma.x.
connection, 85-108 Chlcagoo, IL University of Chicago Press
Hoekstra, T 1995 The function of functional categories GLOT International 1(2). 3-6
Hornstetn, N, 1995 Logical lorm From GB to mmlmahsm Oxford Blackwell
Huang. C -T J t980 The metrical structure ot terraced-le'~el tones NELS 10, 257-270
Hyman, L, 1976 The great ]gbo tone shift In E Voeltz (ed), Thwd annual conference on African hngulstlCS, 11 t-125 Bloomington, IN Indiana University Press
lnkelas, S, W Leben and M Cobler, 1987 The phonology ol Intonation m Hausa NELS 18, 327342
It6, J , 1988 SyllaNc theory in prosodm phonology New York Garland
Jackendoff, R , 1972 Semantic interpretation in generative grammar Cambridge. MA MIT Press
Jaeggh, O , 1986 Passive Linguistic Inquiry 17 587-622
Kalsse. E , 1985 Connected speech The interaction ol syntax and phonology New York Academic
Pless
Kaye. J . [98 [ Tone sensmve rules in Dlda Studies in African Linguistics, Supplement 8, 82-85
Kaye. J , H Koopman and D Sportlche 1982 Projet sur les langues kru, premlel rappolt Groupe de
recherche en hngulstlque alrlcanlste, Unlverslt6 du Qudbec ~ Montrdal
Kaye, J , J Lowenstamm and J -R Vergnaud. 1990 Constituent structure and government in phonology
Phonology Yearbook 7, 193-23 I
Kayne, R , 1975 French syntax The transformational cycle Cambridge, MA MIT Press
Kayne, R 1990 Romance chtlcs and pro NELS 20, 255-302
Kayne, R, 1994 The antlsymmetry of syntax Cambridge, MA MIT Press
Klmenyl, A , 1989 L'haplogle tonale ou la Iol de Meeussen In F Jouannet (ed). Mod61es en tonologle (Klrundl et Klnyarwanda) Pans Editions du CNRS
Koster, J , 1978 Why su~ect sentences don't exist bt S J Keyse~ (ed)~Rx~_cenJ transformaI!ona! >tudles in european languages, 53-64 Cambridge, MA MIT Press
Ladd D, 1990 Metrical representatmn of pitch register In J Kingston. M Beckman (eds), Papers in
laboratory phonology 1 35-57 Calnbrldge Cambridge University Press
Lamontagne, G and L Travls. 1987 Fhe syntax ol adlacency WCCFL 6, 173-186 Stanford CSLI
Publications
LS.nfran, Y, 1992 Intonation in tone languages, the phonetic implementation o! tones in YorL)bfi Dissertation, Cornell University, Ithaca, N Y
Leben, W , 1973 Suprasegmental phonology New York Garland
Llberman, M , 1975 The intonational system of English Dissertation, MIT, Cambridge, MA
Llberman, M 1995 The sound structure ot Mawu words In L Glettman M Llberman (eds), Invitation to cognitive science, 55--86 Cambridge, MA MIT Press
Llberman M and A Prince, 1977 On stress and llngmstJc lhythm Lmgmstlc Inqulr 3 8, 249-336
129
Lm, J.-W, 1994 Lexical government and tone group formation m Xlamen Chinese Phonology 11,
237-275
Longobardl, G , 1994 Reference and proper names. Ltngmstlc Inqmry 25, 609-665
Lowenstamm, J. 1987 Prosodic government In' J Kaye et al. (eds), Projet de hngulstIque afncamste
l~tudes phonologlques et syntaxiques; Rapport de l'Ann6e 1986-87,248-251 Umverslt6 du Qu6bec
Montr6al
M~anfi'edl, V , f99f ~.gb6 and t~,fiugb6, ]-gbo hngulstJc consciousness, ~ts origins and hm~ts Dissertation,
Harvard Umverslty, Cambridge, MA
Manfredl, V , 1992a Spreading and downstep Prosodic government m tone languages. In: H v d
Hulst, K Smder (eds), The representation of tonal register, 133-184 Berlin Mouton de Gruyter
Manfredl, V , 1992b A typology of Yor~b~i nommahzatlons MIT Working Papers in Lmgmstlcs 17,
205-217
Manfredl, V , 1992c The limits of downstep in A.gb6, sentence prosody In: M LIberman, C
MacLemore (eds.), IRCS Report 92-37, 103-115 Institute for Research m Cognmve Studies, Umversity of Pennsylvania.
Manfredl, V., 1995a Syntactic (de)composition of Yor~b~i 'be' and 'have' In. L Nash, G Tsoulas
(eds.), Langues and grammalre-1, 237-252. D6partement des Sciences du Langage, Umverslt6 de
Paris-8
Manfredi, V , 1995b. Tonally branching s in Yor~bfi Is [LH] Niger-Congo Syntax and Semantics 6,
171-182 African Studies Center, Boston Umverslty
Manfredl, V., 1997. Aspectual hcensing and object shift In' R -M D6chalne, V Manfredl (eds), Object
posmons m Benue-Kwa, 87-122 The Hague Holland Academic Graphics
Marantz, A., 1995 The minimahst program In G. Webelhuth (ed), Government and binding theory
and the mmlmahst program, 349-382. Oxford. Blackwell
McCarthy, J 1979. Formal problems m Semitic morphology and phonology Dissertation, MIT, Cambridge, MA
McCarthy, J and A Prince, 1993. Generalized ahgnment Yearbook of Morphology 1993, 79-153. Dordrecht' Kluwer
McCarthy, J and A Prince, 1994. The emergence of the unmarked NELS 24, 333-379
Myers, S., 1997 OCP effects in optlmahty theory Natural Language and Linguistic Theory 15,
847-892
Nespor, M and I Vogel, 1986. Prosodic phonology Dordrecht Forts
Odden, D , 1984. An accentual approach to tone m Klmatuumbl In' D Goyvaerts (ed), African hngmstlcs, Studies m memory of M.W.K. Semlkenke, 345-419 Amsterdam Benjamms
Odden, D , 1990a C-command or edges m Makonde '~ Phonology 7, 163-169.
Odden, D , 1990b Syntax, lexlcal rules and postlexlcal rules m Kimatuumbi In: S Inkelas, D Zec
(eds), The phonology-syntax connection, 259-277. Chicago, IL University of Chicago Press
Odden, D , 1994 Syntactic and semantic conditions m Klkongo phrasal phonology In J Cole, C.
KIsseberth (eds), Perspectives in phonology, 167-202. Stanford, CA. CLSI Publications
Odden, D , 1995 Phonology at the phrasal level m Bantu In' F Katamba (ed), Bantu phonology and
morphology, 40-68 Mtinchen' Lmcom-Europa.
Odden, D , 1997 Domains and levels of representation m tone In R K Herbert (ed.), African hnguIstics at the crossroads. Papers from Kwalusem, 119-146. Koln. Rudlger Koppe
QI~, O, 1995 Optimality in Benue-Congo prosodic phonology and morphology. Dissertation, Umverslty
of Brmsh Columbm
Ql~i-(3rl~, O. and D. Pulleyblank, 1998. Vowel ehslon is not always onset-driven Ms., Tulane University
and Umverslty of Brmsh Columbia
Qhi-(3n& Q and D Pulleyblank, 2000 Yorfib~i vowel elision. Mlmmahty effects Ms, Tulane Umverslty and University of Brmsh Columbm
Oy~l~irhn, O., 1970. Yor~b~t phonology. Dissertation, Stanford Umverslty
Oy~l~irhn, O., 1982a. On the scope of the serml verb construction m Yor~bfi. Studies in African Llngmstics 13, 109-46
Oy~l~iran, Q, 1982b/1992 The category AUX in Yo~bd phrase structure Paper presented at the 15th West
African Languages Congress, Port Harcourt Research m Yor~bfi Language and Literature 3, 59-86.
130
Oy6l~iran, O, 1989 Morphological and syntacuc constraints on verbal auxlhanes in Yorhb~i. Paper presented at the 4th Niger-Congo Syntax and Semantics Workshop, Tllburg Umverslty, 2 June
Prince, A and P Smolensky 1993 Ophmahty theory Constraint interaction in generative grammar
Ms, Rutgers Unlversaty
Pulleyblank, D, 1986a Tone in Lexacal Phonology Dordrecht Reldel
Pu.lje~bl,'~k~. 1) ,. 1.986h CAJtJc~. ,n. YorJ'lb,'~ 1~. Izl l~o.r.e_r. (~A.).. T.be. ~,y.r~t,tx. o Ws~njlmJnal_ clJJk~-~,,, a.3-6~.
563-605
Selkark, E , 1984a Phonology and syntax The relataon between sound and structure Cambridge. MA,
MlT Press
Selkark, E , 1984b On the major class features and syllable theory In M Aronoff. R T Oerhle (eds).
Language sound structure, 107-136 Cambridge, MA MIT Press
Selklrk, E , 1986 On derived domains In sentence phonology Phonology Yearbook 3, 371-405
S6tal6y6. B. 1997 l~ghngbb synchromc phonology A dmlectal approach Dissertation. University of
]lqrln
Sportache,.D.,. 1998 Par_tatifmsand_at~ms.of_cJanse_~t~lu'Jltr_e_ ~ltt'tje.~s.agA'_e.eJ~e.nt,.case_and_cJJtif.~LJm,
don Routledge
Stewart, J M , 1965 The typology of the Tw~ tone system. Bulletin of the Institute of African Studies
(Legon) 1, 1-27
Stewart, O, 1998 The serial verb constructaon parameter Dassertataon, McGIll Unlversaty, Montrdal
Stowell, T, 1981 OnNns ot phrase structure Dl~sertatian, MJT, Camhndge~ MA
Stahlke, H 1976 The noun prefix in Yorhb~i Studms m African lJngmstm~ SulLo/emenl 6~ Z43 53
Voorhoeve, J , 1961 Le ton et la grammalre dans le Saramaccan Word 17 t46-63
Ward, I C , 1952 An mtroduchon to the YorhNi language Cambridge Heifers
Welmers. W , 1959 Tonemacs, morphotonemlcs and tonal morphemes General Lmgmsncs 4, 1-9
Welmers,.W ,_ 1_9_73 ArJt'.an_LangJutg_e_S}l'_ll[~JlllZe_kBe~r_kele_y..CA_ Llmxer.~aty.of CNJorma_Press
Walhams, E S, 1976 Underlying tone m Marga and ]gbo Llngulshc Inquiry 7. 462-484
WIllIamsfm. K , 1976 No~_m-d~L,,spreD_xes m Proto-Lower Niger Presented at the !2th West African
Languages Congress, I14-If6
r
Wilhamscm:. I~,_ l-9~'r T_he:I~3~: ~k';S~:~:J~m~~unl:,qx'JM~ ~)~'e,-i~:~t.r~,< i~: K B~.~g~.~.~: M: (:e~js):.T.he
phonologacal representation of suprasegmentals, 195-208 Dordrecht Forls
Yhsuf, O, 1989 The denvahon of the focus construction an Yorhbg A problem h)r the trace theory
Journal of the LlngmstlC Association of N~gena 5, 56-71
Zucchl, A , 1993 The language of proposatlons and events Dordrecht Kluwer