LVYB

Variation in the Phase Structure of Applicatives*
MARTHA McGINNIS
University of Calgary
Until recently, cross-linguistic differences in the syntax of applicative

constructions have been attributed to arbitrary variation (e.g. Baker 1988,
Bresnan & Moshi 1990, Marantz 1993). For example, it has been argued that
A-movement, like A-bar movement, respects relativized minimality, and that
cross-linguistic variations in the formation of the double-object construction
arise in part from the presence or absence of an “escape hatch,” which allows
the lower object to leapfrog over the higher one to the subject position (Ura
1996, McGinnis 1998). Like other accounts that assume arbitrary variation,
this account raises a learnability issue: how could a child learn that an escape
hatch is available? The present paper offers the beginnings of an answer, by
arguing that this escape hatch affects a wide range of linguistic phenomena.
The central proposal is that a substantial amount of the cross-linguistic
variation in properties of “applicative” constructions (such as the double-object
construction) is reducible to a lexical parameter. The lexicon contains one or
more applicative (Appl) heads, which may denote either a relation between an
event and an individual, or a relation between two individuals (Pylkkänen
2001). (1a) shows the first type, and (1b) shows the second.
(1) a. ApplHP b. VP
IO ApplH' V ApplLP
ApplH VP IO ApplL'
V DO ApplL DO
*
This is an expanded version of a paper presented at NELS 31. Special thanks go to
Jeffrey Lidz, Youri Zabbal, and the participants of the MIT Appl-Fest workshop for their
feedback. I am also grateful to a number of consultants for their judgement and discussion. My
Italian consultants included Paola Benincà, Pier Marco Bertinetto, Stefano Bertolo, Giancarlo
Buoiano, Nicola Cancedda, Ivano Caponigro, Maria Luisa Ciminelli, Roberta D’Alessandro,
Federica Da Milano, Mariapaola D’Imperio, Michela Ippolito, Gisella Ferraresi, Raffaella
Folli, Nicola Mastidoro, Frida Morelli, Patrizia Paggio, Sandra Paoli, Giovanna Rocca, Andrea
Sansó, Mario Saraceni, and Elina Savino. Other language consultants are cited throughout.
2 Martha McGinnis
The semantic distinction between high and low applicatives has consequences
for phrase structure (note the different structures in (1)), and for locality
domains (‘phases’), which in turn yield a wide range of consequences, relating
to transitivity, A-movement, and phonological phrasing.
1 Asymmetries in Applicatives
There are a number of asymmetries in the syntax of applicatives, both
within and across languages. An example of the kind of variation that arises
can be seen in the differences between Kinyarwanda Benefactive and Locative
applicatives (Kimenyi 1980).
One well-known difference between the two types of applicatives is in
their transitivity properties. An applied Benefactive can be added to a transitive
(2a) or intransitive (2b) predicate. An applied Locative can also be added to a
transitive predicate (3a), but not to an unergative one, even one with an
implicit object (3b).
(2) a. Umugóre a-rá-som-er-a umuhuûngu igitabo.

woman SP-PR-read-APPL-ASP boy book
“The woman is reading a book for the boy.” AK (3,7)1
b. Umugabo a-rá-som-er-a umugóre.

man SP-PR-read-APPL-ASP woman
“The man is reading for the woman.” AK (4,40)
(3) a. Umuhuûngu á-r-íig-ir-á-ho ishuûri imibáre.

boy SP-PR-study-ASP-LOC school mathematics
“The boy is studying mathematics at school.” AK (5.4,12b)
1
Abbreviations for citations are as follows: AK (Kimenyi 1980), AM (Marantz 1984),
AS (Seidl 2000), BJ (Bobaljik & Jonas 1996), BM (Bresnan & Moshi 1990), CF (Falk 1990),
DK (Dalina Kallulli, personal communication), HT (Truckenbrodt 1995), HU (Ura 1996), HV
(Hyman & Valinande 1985), JK (Kanerva 1989), JO (Owens 1985), KA1 (Kisseberth &
Abasheikh 1974), KA2 (Kisseberth & Abasheikh 1977), LR (Rizzi 1986), MB (Baker 1988),
MC (Cuervo 2001), OJ (Ólafur Jónsson, personal communication), SM (Sam Mchombo,
personal communication), TH (Thráinsson 1979), VM (Massey 1992). Where possible,
example numbers are given, preceded by a section number. Numerals in glosses indicate Bantu
noun classes. The following abbreviations from the literature are used. Note that one element
may have different glosses from different sources, for example ASP, T/A and FV in Bantu verbs.
S subject NOM nominative PAS passive
O object ACC accusative NACT nonactive
SP subject pronoun DAT dative APPL applicative
OP object pronoun ABS absolutive LOC locative Appl
CL dative clitic ASP aspect FOC focus
PR present T/A tense/aspect REL relative
PST past FV final vowel FEM feminine
Variation in the Phase Structure of Applicatives 3
b.* Umuhuûngu á-r-íig-ir-á-ho ishuûri.

boy SP-PR-study-ASP-LOC school
“The boy is studying at school.” AK (5.4,8b)
Another difference is in the A-movement properties of the two types of

applicatives. In the passive of a Benefactive applicative, either the Benefactive
(4a) or the Theme (4b) can raise to the subject position.
(4) a. Umukoôbwai a-ra-andik-ir-w-a ti íbárúwa n’ûmuhuûngu.2

girl SP-PR-write-APPL-PAS-ASP letter by boy
“The girl is having the letter written for her by the boy.” AK (6,3c)
b. Íbárúwai i-ra-andik-ir-w-a umukoôbwa ti n’ûmuhuûngu.

letter SP-PR-write-APPL-PAS-ASP girl by boy
“The letter is written for the girl by the boy.” AK (6,3b)
In the passive of a Locative applicative, on the other hand, the Locative can
become the subject (5a), but the Theme cannot (5b).
(5) a. Ishuûrii ry-oohere-j-w-é-ho ti igitabo n’úúmwáalímu.

school SP-send-ASP-PAS-ASP-LOC book by teacher
“The school was sent the book by the teacher.” AK (5.4,19c)
b.* Igitaboi cy-oohere-j-w-é-ho ishuûri ti n’úúmwáalímu.

book SP-send-ASP-PAS-ASP-LOC school by teacher
“The book was sent to school by the teacher.” AK (5.4,24)
The examples in (6) and (7) show a third potential difference between
the two types of applicatives, namely a difference in their pronoun
incorporation properties. When pronominal, either object of a Benefactive
applicative (or both) can be incorporated into the verb (6). In a Locative
applicative, only the Locative argument can be incorporated (7a); the Theme
cannot (7b).
(6) a. Umugóre a-rá-mui-he-er-a ti ímbwa ibíryo.

woman SP-PR-OP-give-APPL-ASP dog food
“The woman is giving food to the dog for him.” AK (4,56c)
b. Umugóre a-rá-bii-he-er-a umugabo ímbwa ti.

woman SP-PR-OP-give-APPL-ASP man dog
“The woman is giving it to the dog for the man.” AK (4,56a)
2
Indexing indicates movement dependencies. No ontological commitment is implied.
4 Martha McGinnis
(7) a. Úmwáalímu y-a-ryi-oohere-jé-ho ti igitabo.

teacher SP-PST-OP-send-ASP-LOC book
“The teacher sent the book to it.” AK (5.4,20)
b.* Úmwáalímu y-a-cyi-oohere-jé-ho ishuûri ti.

teacher SP-PST-OP-send-ASP-LOC school
“The teacher sent it to school.” AK (5.4,25)
Various accounts have been proposed for the set of asymmetries

described above, but all have relied on a formal stipulation—LFG’s functional
requirements (Bresnan & Moshi 1990), GB’s Case properties or government
domains (Baker 1988, Marantz 1993), or Minimalism’s “escape-hatch” EPP
specifier positions (Ura 1996, McGinnis 1998). However, as noted above (1),
Pylkkänen (2001) argues that transitivity properties of applicative
constructions arise from a semantic difference, rather than simply from
arbitrary syntactic variation. She argues that there are two types of
applicatives, which she calls high applicatives and low applicatives. The high
applicative head (ApplH) denotes a relation between an event and an
individual, while the low applicative head (ApplL) denotes a relation between
two individuals.
As a consequence of its semantics, ApplH merges with a VP
complement and a DP specifier, yielding the structure in (1a), while ApplL
merges with a DP complement and a DP specifier, yielding the structure in
(1b). Both types of structures have been proposed elsewhere in the literature as
potentially universal representations of the double-object construction (for
example, by Marantz (1993) for (1a), and by Pesetsky (1995) for (1b)).
The effects of the semantic difference between ApplH and ApplL can
be seen in examples like those below. The Albanian Benefactive, a high
applicative, can be used even when no directional or “prospective possessor”
relation obtains between the two objects. Thus, (8a) is fine, where the
Benefactive Dritës “Drita” is not a prospective Source or Recipient of the
Theme çanten time “my bag”. However, the English Recipient applicative
cannot be used in such a context, since, as a low applicative, it necessarily
involves a relation between the two objects (8b) (see Oehrle 1976, among
others).
(8) a. Agimi i mban Dritës çanten time.

A.NOM CL holds D.DAT bag.ACC my
“Agim holds my bag for Drita.” DK
(e.g., so she can put something in it)
b. *John held Mary the bag.

Pylkkänen’s proposed structures for the two types of applicatives

capture the observation that the semantic difference in (8) corresponds to a
difference in transitivity properties.3 Since it merges with VP, an Albanian-
type applicative (ApplH) can be used with agentive intransitive verb phrases,
including unergatives and VPs with an implicit object. An English-type
applicative (ApplL) cannot be used with unergatives, since it merges with the
DP object. Examples (2)-(3) showed this contrast for Kinyarwanda
Benefactive (ApplH) and Locative (ApplL) applicatives. Similarly,
Benefactive applicatives in Kichaga (9) and Albanian (10) also involve ApplH,
so the Theme argument can be omitted. Recipient applicatives in English (11)
and Icelandic (12) involve ApplL, so the Theme argument is obligatory.4
(9) a. N-a̋-i̋-lyì-í-à m̀-kà k-élyà.

FOC-1S-PR-eat-APPL-FV 1-wife 7-food
“He is eating food for/on his wife.” BM (2)
b. N-a̋-i̋-lyí-í-à m̀-kà.
FOC-1S-PR-eat-APPL-FV 1-wife
“He is eating for/on his wife.” BM (12)
(10) Drita i pjek Agimit (rrepat).

D.NOM CL bake A.DAT turnips.ACC
“Drita bakes (turnips) for Agim.” DK
(11) Mary baked Alicia *(a cake).
(12) Ólafur baka∂i henni *( köku).

O.NOM baked her.DAT cake.ACC
“Olafur baked her *(a cake).” OJ
3
Baker (1996) proposes a distinction between suffixal and prefixal applicatives,
arguing that prefixal applicatives can easily combine with unaccusative VPs, while suffixal
ones cannot. This is not the same as the high/low distinction: both types of applicatives
discussed by Baker can combine with agentive intransitives, a characteristic of high
applicatives. Recent work on Ikalanga suggests that the proposed restriction on suffixal
applicatives does not hold universally (Letsholo 2001).
4
Woolford (1984) points out that there are ditransitive verbs in English for which,
apparently, the Theme argument is optional (i). However, the nominalizations of such verbs
may allow a Recipient argument (ii), unlike the nominalizations of other ditransitives (iii). This
suggests that the Recipient argument in (i) and (ii) is an argument of the lexical root (cf.
Marantz 1997), not an ‘applied’ argument. This account may not cover all of Woolford’s
examples (e.g., write, tell).
(i) I feed cows (hay). / I teach children (French).
(ii) the feeding of hay to cows/of cows; the teaching of French to children/of children
(iii) the baking of cakes for Alicia/*of Alicia; the gift of a book to John/*of John
6 Martha McGinnis
I will adopt Pylkkänen’s proposals concerning the phrase structure and

semantics of the two types of applicatives, and the consequences of these for
their transitivity properties. I will also adopt Chomsky’s recent proposal that
syntactic derivations undergo semantic and phonological interpretation in
incremental chunks or phases (Chomsky 1999, 2000). Phases (i.e., “strong”
phases) can be headed by a number of possible categories, including a
Complementizer (C), Determiner (D), and a light verb (v) which carries active
voice features and projects an external argument in its specifier (see Harley
1995, among many others). This v can also license an object via structural
Case.
Once a phase is complete, movement and agreement operations can
target its head and constituents in its edge—adjuncts and specifiers, like those
circled in (13)—but cannot target constituents in its domain (complement),
marked off by the curved boundary in (13).5
(13) vP
EDGE Obj v'
Subj v'
DOMAIN
v VP
[phase-EPP] V t
In some cases, however, EPP features can be added to a phase before it is

complete, allowing a constituent in its domain to move to the edge. For
example, Chomsky proposes that Germanic object shift involves the movement
of an object from the domain of vP to its edge to check phase-EPP features, as
in (13). There is currently no principled account of when or why EPP features
can be added to a phase; this issue demands further research. We will return to
object shift in the argument for phase-level locality in Section 4.
5
It is assumed here that the domain of a phase is inaccessible to syntactic operations
as soon as the phase is complete, rather than simply at the next phase. This view is supported
by the arguments given below. The contrast between (i) and (ii) provides independent support
for this view. Here, the NP ship originates in the domain of the DP phase the ship. If this
domain is accessible until the next phase (CP), ship should be able to move to subject position
in spec-TP, as in (i), assuming that the EPP feature of T can be checked by NP. However, (i) is
out, suggesting that the domain of the DP phase is inaccessible as soon as the phase is
complete. Of course, the entire DP phase can move to spec-TP, as in (ii).
(i) * [CP [TP [NP Ship] [vP sank [DP the t]]]].
(ii) [CP [TP [DP The ship] [vP sank t]]].
2 Phases and the two types of applicatives

The central proposal of this paper is that the different semantic
properties of applicative heads affect not only phrase structure, but also phase
structure. That is, different XPs may count as phases, depending on whether a
clause contains ApplH, ApplL, or neither. (14) achieves this result. The
boundaries in (15) demarcate the domains of the phases predicted by (14).
ApplH heads a phase, since it is the sister of VP and—perhaps in combination
with VP—assigns a theta-role to the applied argument generated in its specifier
(15a). By contrast, ApplL is not a sister of VP, so it need not head a phase.
Instead, here, as elsewhere, v heads a phase if it has an argument generated in
its specifier (15b). Assuming that no argument is generated in the specifier of
unaccusative and passive v, they also need not head phases.6
(14) The sister of VP heads a phase if an argument is generated in its specifier.
(15) a. High applicative b. Low applicative

vP vP
DP v' DP v'
v ApplHP v VP
IO ApplH' V ApplLP
ApplH VP IO ApplL'
V DO ApplL DO
Ideally, (14) should be derived from some broader generalization.

Chomsky (1999, 2000) proposes that a phase is a proposition; another
possibility is that it is a domain of semantic predication, created by adding
either an ‘external’ argument or an applied argument to the VP predicate. Yet
another possibility is that the constituents here represented as V or N are really
category-neutral lexical roots (Marantz 1997, Harley & Noyer 2000), and that
the head responsible for determining a root’s morphological category can head
a phase. For example, if the lexical root is the sister of D, it is morphologically
nominal; if it is the sister of v or of ApplH, it is morphologically verbal.
However, this sheds no light on why some types of v do not head phases.
6
See Embick (1997) for arguments against theta-assignment by v in unaccusatives
and passives. McGinnis (2001) argues that in some cases ApplL can head a phase, barring
extraction of the DO as well as movement to subject position. See Nakamura (1997) for a more
extensive discussion of DO-extraction in applicatives.
8 Martha McGinnis
The latter proposal is supported by evidence that D does indeed head a

phase. If derivations are strictly cyclic, a branching DP must be constructed
separately before merging with a larger structure—a characteristic of phases
(Chomsky 1999). Moreover, NP cannot be extracted from the complement
position of a DP (see fn. 3). Supposing that DP is a phase, this observation
follows from the claim that the domain of a phase is inaccessible to further
operations once the phase is complete. Plausibly, then, D heads a phase
because it determines the morphological category of the root, and v and ApplH
behave likewise.
For the present, however, any broader understanding of (14) remains
speculative. The focus here will be on the following distinction: in a high
applicative, ApplH heads a phase whose domain contains the Theme argument,
while in a low applicative, v heads a phase whose domain contains both the
Theme argument and the applied argument.
3 Consequences
The proposal in (14) has consequences for A-movement and pronoun
incorporation (or agreement). In a high applicative, only the lower object (the
Theme) is embedded within the domain of the ApplHP phase, so it can check
an EPP feature added to this phase. In a low applicative, both objects are
within the domain of the vP phase, so if only one phase-EPP feature is added, it
can be checked only by the higher, applied object.7
3.1 A-movement
I will assume that A-movement respects relativized locality (Rizzi
1990, Chomsky 1995, McGinnis 1998). Thus, in a passive or raising low
applicative construction, only the higher, applied object can undergo A-
movement to the subject position.
(16) a. vP b. ApplHP
v VP DO ApplH'
V ApplLP IO ApplH'
IO ApplL' ApplH VP
ApplL DO [phase-EPP] V t
Since the higher object is the DP closest to T, it blocks the lower object from
undergoing A-movement to the specifier of T (16a). However, in a passive or
7
If v has two EPP features, other factors come into play; for example, see (44)-(45).
raising high applicative, a lower argument can raise to the subject position.
This is because a phase-EPP feature can (and perhaps must) be added to ApplH
in the passive, allowing the lower argument to leapfrog over the higher one, as
in (16b). Once the DO occupies a higher specifier of ApplH, it is the closest
DP to T, so it can move to spec-T.
As noted above, Benefactives in Kichaga (17) and Albanian (18) are high
applicatives. These applicatives allow the lower DO (Theme) argument to
leapfrog over the higher IO (Benefactive) to the subject position of a passive
(17a)-(18a).8 They also allow the Benefactive argument to raise to the subject
position (17b)-(18b). Kinyarwanda Benefactives are like those in Kichaga, as
was shown in (4). Although word order in Albanian is fairly free, quantifier-
pronoun binding indicates that the Theme is in an A-position c-commanding
the Benefactive in (18a), while reverse is true in (18b). The possibility of
raising the Benefactive to subject position may arise because the Theme
remains in situ, or because it raises to a specifier of ApplH below the IO,
“tucking in” underneath it, rather than stacking up above it as in (16a). We
return briefly to this issue in Section 5.1.
(17) a. K-èlyái k-i̋-lyì-í-ò m̀-kà ti.

7-food 7S-PR-eat-APPL-PAS 1-wife
“The food is being eaten for the wife.” BM (5c)
b. M-kài n-a̋-i̋-lyì-í-ò ti k-èlyâ.

1-wife FOC-1S-PR-eat-APPL-PAS 7-food
“The wife is having the food eaten for her.” BM (5b)
(18) a. Secili libëri iu kthye autorit të tij ti.

each book.NOM CL returned.NACT author.DAT its
“Each book was returned to its author.” VM (3,66)
b. Secilit djalëi iu dha ti paga i tij.

each boy.DAT CL gave.NACT pay.NOM his
“Each boy was given his pay.” DK
By contrast, Recipient applicatives in English and Icelandic are low

applicatives, so only the Recipient can raise to subject position. A phase-EPP
feature added to v can be checked only by the higher argument, since this
argument blocks the lower one from raising past it to spec-T. Since ApplLP is
8
This account does not capture the generalization that in Bantu languages, the lower
object can move to the subject position only if the higher object occupies a focused position
(Juvénal Ndayiragije, personal communication). However, see footnote 13.
10 Martha McGinnis
not a phase, no phase-EPP feature can be added to ApplL to allow the DO to

undergo A-movement to a specifier above IO.
(19) a. Aliciai was baked ti a cake.

b.* A cakei was baked Alicia ti.
(20) a. Honum var gefin bókin.

him.DAT was given.NOM the book.NOM
“He was given the book.” CF
b.* Bókin var gefin honum.9

the book.NOM was given.NOM him.DAT
“The book was given to him.” CF
Suppose, contrary to what has been suggested above, that ApplLP is

indeed a phase, but (I) it has no phase-EPP feature added to it—or (II) such a
feature cannot be checked by its complement, the DO (21). At first glance, this
would seem to allow an alternative account of the ill-formedness of (19b) and
(20b). The Theme would be trapped in the domain of the ApplLP phase,
unable to escape via phase-EPP, while the Recipient would move to spec-T as
described above.
(21) vP
v VP
V ApplLP
DO IO
IO ApplL'
ApplL t
[phase-EPP]
However, there is considerable evidence that (21) is not the correct

account of (19)-(20). Suppose, first, that the DO cannot move past the IO, as in
(21), simply because the complement of a head H cannot move to spec-H
(option II above). Under this view, a lower DP that is not the complement of
9
Although in some cases the nominative Theme can raise to the subject position of
the passive in Icelandic, Falk (1990) argues that it can do so only if it is base-generated above
the dative Recipient.
ApplL should be able to move to spec-ApplL. For example, in a raising

construction with an Experiencer, the lower DP is an embedded spec-T, not a
complement of ApplL. Under the proposal just stated, an embedded subject
should always be able to move to spec-Appl, whether it is ApplH or ApplL.
Instead, we see the familiar contrast between high applicatives and low
applicatives in raising constructions. In Icelandic, the Experiencer (mér) can
raise to the subject position (22a), but the embedded subject (Haraldur) cannot
leapfrog past it (22b).
(22) a. Jón telur [ méri vir∂ast ti [ Haraldur hafa gert óetta vel]].
J.NOM believes me.DAT to.seem H.NOM to.have done this well
“Jon believes Harald to seem to me to have done this well.”
b.* Jón telur [Haralduri vir∂ast mér [ti hafa gert óetta vel]].
“(same as (22a))” TH
In Italian, by contrast, the embedded subject can move to the subject position
(23). This difference follows if the Experiencer construction is a low
applicative in Icelandic, and a high applicative in Italian. Since ApplLP is not a
phase, there is no phase-EPP to allow the embedded subject to move over the
Experiencer in Icelandic. In Italian, the embedded subject escapes the domain
of the ApplHP phase via phase-EPP movement to spec-ApplHP.10
(23) Giannii non gli sembra [ti fare il suo dovere].

G. not him.DAT seems to do his duty
“Gianni does not seem to him to do his duty.” LR (22b)
Further evidence against (21) (under either option I or II above) comes

from other types of movement. For instance, the Theme can undergo wh-
movement and quantifier raising (QR), which would be impossible if the
Theme were trapped within the domain of ApplLP. Wh-movement of the
Theme is shown in (24).
(24) Which medali did Reuben award Ben Johnson ti ?
Bruening (1999) notes that in examples like (25a), the Theme must also
be able to undergo QR to resolve the antecedent-contained deletion (ACD) of
the VP. He points out that, as a result, this Theme must take wide scope over
the intensional verb refused, as shown in the LF representation (25b).
10
It should be possible to find independent evidence from semantics and/or argument
structure for the high/low split in Experiencer constructions. I leave this matter for further
research.
12 Martha McGinnis
(25) a. Reuben [VP refused to award Ben Johnson every medal that Eva did
[VP refused to award Ben Johnson]].
b. Reuben [every medal that Eva did [VP refused to award Ben Johnson]]i
[VP refused to award Ben Johnson ti ].
To verify this claim, consider (26), which allows either a narrow-scope de

dicto or a wide-scope de re reading for the quantifier. Under the de dicto
reading, Reuben made a categorical refusal to award Ben Johnson any medals
that Eva had told him to buy, without necessarily knowing which medals these
were. Under the de re reading, Reuben refused to award Ben Johnson a set of
medals, each of which Eva had told him to buy. In (25), however, every medal
must take wide scope over refused to resolve ACD. Thus the ‘multiple
refusals’ de re reading is available for (25), but the ‘categorical refusal’ de
dicto reading is not.
(26) Reuben refused to award Ben Johnson every medal Eva told him to buy.
These examples show that the Theme of an English Recipient

applicative can undergo syntactic movement, even though it cannot undergo A-
movement to the subject position. Thus, we can conclude that the Theme is not
trapped within the domain of an ApplL phase.
3.2 Object agreement and pronoun incorporation

The phase account of applicatives makes predictions for object
agreement and pronoun incorporation as well. Suppose that in some
applicatives, a phase-EPP feature is optionally added to ApplH in the active
voice as well as in the passive. Alexiadou & Anagnostopoulou (1998, fn. 7)
have suggested that in some cases an EPP feature can be checked by feature-
movement. If phi-features of the DO move to check a phase-EPP feature on
ApplH, this will appear as object pronoun incorporation or agreement (27). As
noted previously, I will not attempt to give a principled account of when EPP
features can and cannot be added to a phase.
(27) vP High applicative (DO agreement/incorporation)

DP v'
v ApplHP
IO ApplH'
ApplH VP
[phase-EPP] V DO
For example, in a Kichaga Benefactive, the Benefactive (28a), the Theme

(28b) or both (28c) can undergo pronoun incorporation. This follows if
Benefactive incorporation involves a checking relation with v, while Theme
incorporation involves a checking relation with ApplH.
(28) a. N-a̋-i̋-m-lyì-í-à k-èlyâ.

FOC-1S-PR-1O-eat-APPL-FV 7-food
“He/she is eating food for/on him/her.” BM (7a)
b. N-a̋-i̋-kì-lyì-í-à m̀-kà.
FOC-1S-PR-7O-eat-APPL-FV 1-wife
“He/she is eating it for/on the wife.” BM (7b)
c. N-a̋-i̋-kì-m̀-lyì-i̋-à.
FOC-1S-PR-7O-1O-eat-APPL-FV
“He/she is eating it for/on him/her.” BM (7c)
In a low applicative, however, if v has only one phase-EPP feature,

only the higher object can check it by feature-movement. Because of locality,
the higher IO blocks v from attracting features from the lower DO (29).
(29) vP Low applicative (*DO agreement/incorporation)

DP v'
v VP
V ApplLP
IO ApplL'
ApplL DO
*
14 Martha McGinnis
Examples are given in (30) from the Chi-Mwi:ni: Recipient applicative. Here
the Recipient (wa:na) can trigger agreement (30a), but the Theme (cha:kuja)
cannot (30b).
(30) a. HamadÙi Ø-wa-pik-il-ile wa:na cha:kuja.

Hamadi SP-OP-cook-APPL-T/A children food
“Hamadi cooked food for the children.” KA2, AM (7.12b)
b.* HamadÙi Ø-sh-pik-il-ile wa:na cha:kuja.

Hamadi SP-OP-cook-APPL-T/A children food
“Hamadi cooked food for the children.” KA2, AM (7.20)
4 The MLC: Independent evidence for the ApplHP phase

As we have seen, the proposal that ApplH heads a phase makes it
possible to account for a number of morphosyntactic asymmetries between
high and low applicatives. Further evidence that ApplHP is a phase can be
drawn from Chomsky’s (1999) proposal that the Minimal Link Condition is
evaluated at the phase level.
4.1 Phase-level locality in vP

Recall that Chomsky proposes that Germanic Object Shift (OS)
involves the movement of an object from the domain of vP to its edge to check
phase-EPP features, as in (13), shown again below.
(13) vP
EDGE Obj v'
Subj v'
DOMAIN
v VP
[phase-EPP] V t
OS languages like Icelandic contrast with non-OS languages like

English. In Icelandic, an object can shift out of its base position, into a position
to the left of negation (31). Under the traditional assumption that negation
diagnoses the upper left boundary of the VP, this means that the object can
shift out of VP, as shown in (13). By contrast, OS is generally not possible in
English (32).
(31) Jólasveinarnir bor∂u∂u bú∂inginni [VP ekki tV ti]

the.Christmas.trolls ate the.pudding not
“The Christmas Trolls did not eat the pudding.” BJ (1)
(32)*The Christmas Trolls did [the pudding]i [VP not eat ti].
However, the object is not trapped within transitive VP even in English. For
example, an object can move out of VP via wh-movement (33a) or
topicalization (33b).
(33) a. Whati did they [VP eat ti]?

b. [The pudding]i, they [VP ate ti].
Chomsky (1999) suggests that (32) is ungrammatical because in

English, a shifted direct object blocks the subject from moving to the specifier
of T. Assuming that an outer specifier of v can check a phase-EPP feature of v,
object shift obeys locality, as schematized in (34). Subject movement,
however, violates locality: the shifted object is closer to T than the logical
subject, so this derivation is ruled out.
(34) TP
Subj T'
vP T vP
Obj v' 2 Obj v'
Subj v' t v'
1 v VP 1 v VP
V t V t
The remaining details of Chomsky’s account of (33) are cast in a

feature-checking theory, in which uninterpretable features must be checked in
an appropriately local relation in order for the derivation to satisfy Full
Interpretation and converge at LF. Uninterpretable features include Case
features on DPs, as well as EPP-features and nominal (phi-) features on certain
heads. An EPP feature on a head H is checked by moving a DP into spec-H. As
noted above, EPP features can be added at a phase boundary; T also has an
obligatory EPP feature, although it does not head a phase. The uninterpretable
phi-features of a head are checked by a local relation with the interpretable phi-
16 Martha McGinnis
features of a DP, and the uninterpretable Case feature of a DP is checked by a

local relation with a Case-checker, such as finite T, or active transitive v. It will
be assumed below that the high applicative head ApplH is a Case-checker as
well.11
A constituent is “active” for (certain types of) movement by virtue of
having (certain types of) uninterpretable features. For example, the object is
rendered inactive for EPP-movement in (34) when it checks its uninterpretable
Case feature on v. As a result, the object cannot raise to spec-T. We have seen
above that the shifted object also blocks the subject from raising to spec-T in
(34). However, if nothing raises to spec-T, the uninterpretable EPP and phi-
features of finite T go unchecked, causing the derivation to violate Full
Interpretation and crash at LF. Even if an expletive can be inserted to check
these features, the shifted object blocks the logical subject from checking its
Case feature on T or on any other head, so the derivation crashes anyway.
Given this account, the well-formedness of (33) provides evidence that
locality is evaluated at the phase level, rather than at each step of movement.
The derivation of (33) passes through a stage identical to the ill-formed
derivation in (34). The three relevant steps of movement in (33a) are shown
schematically in (35).
(35) CP
wh-Obj C'
C TP
3 Subj T'
T vP
2
t v'
t v'
1 v VP
V t
In the first stage, the object moves to spec-v, escaping the domain of the vP
phase by checking an EPP feature on v. In the second stage, the subject moves
11
The Case-checking heads presumably also include P and complementizer for. I
assume that the verbal ApplL assigns inherent Case to the DO, like a preposition.
past this shifted object. If locality is a condition on Move, this movement

should be blocked by the intervening object, as in (34). However, in the third
stage, the shifted object wh-phrase moves to spec-C, checking a [+wh] feature
on C. If the Minimal Link Condition is not evaluated until this move takes
place, it will be satisfied by the resulting configuration.
Consider the first chain link, created by movement of the object from
the complement of V to spec-v. Suppose this link satisfies the MLC because
there is no closer candidate for movement to this position. The third chain link,
between this spec-v position and spec-C, also satisfies the MLC. The object
wh-phrase is the closest candidate for movement to spec-C, provided that the
subject in spec-T is not itself a wh-phrase. Now consider the second link,
created by movement of the subject from the lower spec-v to spec-T. Once the
object has moved to spec-C, this link also satisfies the MLC: the trace of the
object is not a closer candidate for movement to spec-T, since traces cannot
move. In brief, if locality is computed at each step of the derivation, (35)
should be as bad as (34), but if it is computed only after each phase is
completed, (35) is correctly predicted to satisfy the MLC.
We can now formulate the MLC as follows. Given a constituent α in
the edge of an HP phase, the chain <α, tα > formed by movement is
interpretable only if it respects (36). Let us assume that a phrase α is more
local than another phrase β to a c-commanding head if α asymmetrically c-
commands β.
(36) Minimal Link Condition

No element β, more local to H than tα , is a candidate for the movement
triggered by the uninterpretable features of H.
Note that this account of the contrast between ungrammatical OS in

(32), and grammatical OS in (33), relies crucially on the claim that the shifted
object ‘stacks up’ in the higher specifier of v, above the base position of the
subject. Because of this, the second step of movement in (34) violates the
MLC. The shifted object bears phi-features and c-commands the base position
of the subject, so it is a closer candidate for movement triggered by the
uninterpretable EPP and phi-features of T. If, instead, the shifted object were to
‘tuck in’ underneath the base position of the subject, movement of the subject
to spec-T would not violate the MLC, and the account of the contrast between
(32) and (33) would be lost.
Nothing has yet been said here about why OS is grammatical in
Icelandic. According to Chomsky (1999), in OS languages a shifted object can
shift to a syntactically invisible position after moving to spec-v, so it will not
block the subject from moving to spec-T (see also Holmberg 1997). On the
other hand, Hiraiwa (2001) argues that the object itself moves to spec-T in
these languages. I leave this issue for further research.
18 Martha McGinnis
4.2 Phase-level locality in ApplHP

We have reviewed Chomsky’s proposal that locality need only be
satisfied at the phase level. This proposal accounts for a contrast between
grammatical and ungrammatical movement of the object to the edge of the vP
phase in English transitive clauses. A parallel contrast in French and Italian
Experiencer+Raising derivations provides evidence that ApplHP is a phase. In
this case, however, the raised DP (here, the embedded subject) tucks in
underneath the base-generated DP (the matrix Experiencer), as we will see
below.
Consider the French examples in (37). If there is no matrix
Experiencer, the embedded subject can raise to spec-T of the matrix clause
(37a). If there is a matrix Experiencer, and this Experiencer remains in its base
position, like au garçon “to the boy” in (37b), movement of the embedded
subject to spec-T is impossible (see Chomsky 1995:305).
(37) a. Ellei semble [ti avoir du talent]].

she seems to.have of.the talent
“She seems to have talent.”
b.* Ellei semble au garçon [ti avoir du talent]].

She seems to.the boy to.have of.the talent
“She seems to the boy to have talent.”
The contrast in (37) is accounted for if the Experiencer is a closer candidate for
movement to spec-T, so that movement of the embedded subject in (37b)
violates the MLC. Recall that a shifted object blocks movement of the subject
to spec-T in English, even though the object cannot itself move there. The
same appears to be true in (37b): the Experiencer blocks movement of the
embedded subject, even though the Experiencer itself cannot move to spec-T.
We may suppose that the Experiencer cannot move to spec-T for Case reasons:
it checks Case below T (presumably on v), or has inherent Case assigned along
with its theta-role. By the time T is merged with vP, the Experiencer bears no
uninterpretable Case feature, and thus is not “active” for movement to spec-T.
On the other hand, no locality violation arises if the Experiencer moves to
spec-C via wh-movement (38). These sentences are better than (37b).12
12
Thanks to Marie Claude Boivin, Marie-Hélène Côté, Estelle Dansereau, Gilles
Mossière, Claude Romney, and Philippe Schlenker for judgements and discussion. Judgements
given are contrastive. One speaker commented that (i) is fine and better than (ii), as long as à
Jean in (i) receives contrastive focus. All speakers found (iii) better than the others. Here the
Experiencer clitic adjoins to a head before the subject raises, so it would create no violation
even if locality were evaluated at each step of the derivation (McGinnis 1998).
(i) A Jean, elle semble avoir du talent. “To John, she seems to have talent.”
(ii) Elle semble à Jean avoir du talent. “She seems to John to have talent.”
(38) a. À qui semble-t-elle avoir du talent?

to who seems-she to.have of.the talent
“To whom does she seem to have talent?”
b. Voici le garçon auquel elle semble avoir du talent.

here.is theboy to.whom she seems to.have of.the talent
“Here is the boy to whom she seems to have talent.”
The contrast between (37b) and (38) is parallel to the contrast between
grammatical and ungrammatical object shift in English. We can account for
this contrast in essentially the same way, as follows. Suppose that ApplHP is a
phase, and the matrix Experiencer is generated in its edge. If so, the embedded
subject must move to spec-ApplH if it is ever to escape the domain of the
ApplHP phase. However, such movement does not allow the derivation in
(37b), since the intervening Experiencer blocks a well-formed dependency
between the subject elle in spec-T and its trace in spec-ApplH (39).
(39) ApplHP
Exp ApplH'
Subj ApplH'
ApplH VP
V TP
t …
In (38), however, only a trace of the Experiencer will occupy spec-

ApplHP at LF. Let us consider the step-by-step derivation of the relative clause
in (38b). The Experiencer is generated in spec-ApplH (40a). The embedded
subject moves to a lower specifier of ApplH to check a phase-EPP feature
(40b). It then moves to spec-T to check the uninterpretable EPP and phi-
features of T (40c). If locality were evaluated at each step of the derivation,
this step would violate locality, since the Experiencer is a closer candidate for
movement to spec-T, as in (37b). However, locality is not evaluated until the
next phase level, upon completion of the next CP. By this stage, the
Experiencer has raised to spec-C (40d). Its trace in spec-ApplH is not a closer
candidate for movement to spec-T, so the MLC is respected.
(iii) Elle lui semble avoir du talent. “She seems to him to have talent.”
20 Martha McGinnis
(40) a. [ApplP auquel [TP elle avoir du talent]]

b. [ApplP auquel ellei [TP ti avoir du talent]]
c. [TP ellei semble [ApplP auquel ti [TP ti avoir du talent]]]
d. [CP auquelj [TP ellei semble [ApplP tj ti [TP ti avoir du talent]]]]
Rizzi (1986) notes that an Italian Raising+Experiencer derivation is

improved if the Experiencer topicalizes. The account of the contrast between
(37b) and (38) can be applied to this contrast as well. In (41a), the Experiencer
intervenes between the raised subject and its trace, while in (41b), only the
trace of the topicalized Experiencer does so.
(41) a. *Giannii sembra [ApplP a Piero ti [ti non fare il suo dovere]].
Gianni seems to Piero not to.do the his duty
“Gianni seems to Piero not to do his duty.” LR (fn. 9)
b. A Pieroj, Giannii non sembra [ApplP tj ti [ti fare il suo dovere]].

to Piero Gianni not seems to.do the his duty
“To Piero, Gianni does not seem to do his duty.” LR (22b)
This account assumes that the subject in spec-T is not a closer candidate than
the topicalized experiencer for movement to spec-C. We might conjecture that
the topicalized phrase is generated with a feature [Topic], which other phrases
lack and which satisfies an uninterpretable feature on C. In some languages,
the [Topic] feature is realized morphologically on the topicalized phrase (as
with the Japanese topic marker wa, or the non-manual topic marker in ASL). In
Italian, as in English, [Topic] is not realized morphologically, but its presence
allows the movement of the topicalized phrase over other phrases, as in (41b).
We can ask whether there is any independent evidence that an
Experiencer is base-generated in a position between the embedded subject and
the matrix subject in French and Italian. Such evidence arises when the
embedded clause is finite, as in (42). Here, the embedded subject need not
raise into the matrix clause, since it receives Case in the embedded clause. In
French the subject is an expletive (EXPL) which takes the form of a third-
person masculine singular pronoun.13 Italian is a pro-drop language, so I
assume that the subject of (42b) is a null expletive pro. In both cases the
Experiencer can immediately precede the embedded clause, since no ill-formed
dependency is created. In (41a), however, an Experiencer in this position will
block raising of the embedded subject.14
13
Even speakers who found (42a) awkward preferred it to the raising example.
14
Thanks to Ur Shlonsky for raising this issue. I checked the full set of contrasts with
eighteen native speakers who answered a request for judgements posted on the Linguist List.
(42) a. Il semble à Jean [qu’elle a du talent].

EXPL seems to Jean that-she has of.the talent
“It seems to John that she has talent.”
b. (pro) Sembra a Piero [che Gianni non abbia fatto il suo dovere].
EXPL seems to Piero that Gianni not has done the his duty
“It seems to Piero that Gianni has not done his duty.”
Like the account of object movement in English given above, the

account of the Raising+Experiencer derivation in French and Italian depends
crucially on the relative position of the two highest arguments after the lower
argument has undergone phase-EPP movement. We have seen that the shifted
object must stack up above the base position of the subject if the account is to
go through. Here, however, the lower argument (the raised subject) must tuck
in underneath the base position of the higher one (the Experiencer). If the
raised subject could stack up above the base position of the Experiencer,
further movement of the raised subject to spec-T would not violate the MLC,
and the relevant contrasts would not be captured.
We can adopt the following proposals:
(43) a. A constituent can check a feature of a head H in a specifier either

above or below a constituent base-generated in spec-H.
b. When a head has two features of the same type, the two constituents
that check these features preserve their existing hierarchical order.
Independent support for (43a) comes from the variable (S>O, O>S) order of
the thematic subject and a shifted object in Icelandic transitive expletive
constructions (Jonas 1996). However, it remains to be seen what determines
which of the options in (43a) is available for a given phase head. The
restriction in (43b) yields rigid ordering and scope effects discussed by
Richards (1997), Bruening (1999), and Rezac (2000), among others. (43)
dispenses with the notion that specifiers of the same head are equidistant for
the purposes of further syntactic movement. Rather, the highest specifier is the
most local to a c-commanding head.15
All respondants found (41a) worse than (41b). Some respondants commented that (41b) is by
no means perfect, however. All eighteen also found (41a) worse than (42b). Again, some
speakers still found the better example somewhat marked. Several of these speakers were also
asked about (i). Most found (i) better than (41a), though again, (i) was judged far from perfect.
(i) A chi Gianni sembra non fare il suo dovere?
“To whom does Gianni seem not to do his duty?”
15
The tucking-in option may make it possible to capture the generalization alluded to
in footnote 6. Suppose the DO always tucks in under the IO in Bantu languages. Then the DO
could move on to the subject position only if the IO underwent focus-movement, avoiding an
22 Martha McGinnis
It is worth pointing out that wh-moving the Experiencer in Icelandic

does not allow the embedded subject to raise to spec-T, as it does in French
(Holmberg 1997, Hiraiwa 2001). This is illustrated in (44a). Instead, the
embedded subject remains below spec-T, while the Experiencer checks the
EPP feature of T and moves on to spec-C (44b).
(44) a. *Hverjumj hefur Ólafuri alltaf virst tj [ti. vera gáfa∂ur]?

who.DAT has Olafur.NOM always seemed to.be intelligent
“To whom has Olafur always seemed to be intelligent?”
b. Hverjumi hefur ti alltaf virst ti [Ólafur vera gáfa∂ur]?

“(same as (44a))”
The difference between Icelandic and French can be attributed to their different
Case and EPP properties. In Icelandic, a dative Experiencer can check EPP on
T, while in French this is impossible. We have assumed that the Experiencer
cannot raise to the subject position in French because it bears no
uninterpretable Case feature. Suppose that the ‘quirky’ dative experiencer in
Icelandic does bear an uninterpretable Case feature, so it can move to spec-T,
and here must do so to check the EPP feature of T. Moreover, the embedded
subject checks Case below T, on matrix ApplH, so it cannot raise to the subject
position in (44a). Once it has checked Case, the embedded subject is inactive
for EPP-movement.16
Under this analysis, the Icelandic contrast in (44) is parallel to a
contrast in simple transitives with a wh-object. For example, in (45) the direct
object cannot become the subject of a passive even if wh-movement of the
logical subject would allow the derivation to respect the MLC. (45a) shows the
ungrammatical case in which the object moves to the subject position,
checking the EPP and phi-features of T, while the logical subject undergoes
wh-movement to spec-C. This CP phase would satisfy the MLC, since the trace
of the logical subject is not a candidate for movement to spec-T, and the non-
wh object moved to spec-T is not a candidate for movement to spec-C. Given
the current analysis, however, the object is not active for movement to spec-T,
as in (45a), because its Case feature is checked within vP. The logical subject,
however, does have an uninterpretable Case feature, so it is active for
MLC violation when the DO moves to spec-T. This would assimilate the Bantu facts to object
shift in Icelandic, where Chomsky (1999) suggests the shifted object must undergo focus-
movement in order for the subject to move to spec-T without violating the MLC (see Section
4.1). This possibility remains to be explored.
16
Hiraiwa assumes (following Chomsky 1999) that the nominative object checks
Case on T; however, if it checks Case on ApplH, (44a) is correctly predicted to be
ungrammatical. Under this proposal, however, the object’s structural Case (checked on ApplH)
is dissociated from its morphological case (determined by a local relation to finite T).
movement to spec-T. Thus the logical subject can and must move to spec-T,
and then on to spec-C (45b).
(45) a. *[CP Whoj did [TP Johni [vP ti tj [VP see ti ]]]]?
b. [CP Whoi [TP ti [vP ti [VP saw John]]]]?
To recap, the claim that ApplH heads a phase is supported by the parallels
between movement from spec-v in English transitive clauses, and movement
from spec-ApplH in French or Italian Raising+Experiencer clauses. We have
seen that a shifted object in spec-v blocks the logical subject from moving to
spec-T, and an Experiencer base-generated in spec-ApplH blocks an embedded
subject from moving to spec-T. On the other hand, wh-movement or
topicalization of the “blocking” argument allows the otherwise “blocked”
argument to move, since the resulting derivation will not violate the MLC
when it is evaluated at the CP phase level. The assumption that ApplHP heads
a phase makes it possible to test another prediction of the phase-level theory of
locality, namely that the MLC is evaluated at the vP phase level as well as the
CP phase level.
4.3 Evaluating locality at the vP phase level

If locality is evaluated at the phase level, it should also be evaluated
when a vP phase is completed. This prediction is somewhat difficult to test, but
confirming evidence can be drawn from the Kichaga Benefactive applicative.
In the active version of this applicative, the word order is fixed: the
Benefactive precedes the Theme (46a), but the Theme cannot move and stack
up above the Benefactive (46b).
(46) a. N-a̋-i̋-lyì-í-à m̀-kà k-élyà.

FOC-1S-PR-eat-APPL-FV 1-wife 7-food
“He is eating food for/on his wife.” BM (2)
b.* N-a-i-lyi-i-a k-elyai m-ka ti.

“(same as (46a))” BM (37a)
Recall that the Kichaga Benefactive is a high applicative (47a).

Suppose the Theme checks Case on ApplH, while the Benefactive checks Case
on v. In (47a), all feature-checking relations obey the MLC: no closer DP
intervenes to block feature-checking between ApplH and the Theme, or
between v and the Benefactive. On the other hand, if the Theme were to stack
up on top of the Benefactive, as in (47b), Case and phi-feature checking
relations between v and the Benefactive would violate the MLC. Since the
Theme checks Case on ApplH, it is itself inactive for a feature-checking
24 Martha McGinnis
relation that could check uninterpretable phi-features of v. As a result, both

these features of v and the uninterpretable Case feature of the Benefactive
argument go unchecked, and the derivation crashes. This account is parallel to
the account of ungrammatical object shift in English.
(47) a. b. vP
vP v ApplHP
v ApplHP Theme ApplH'

Benefactive ApplH' Benefactive ApplH'
ApplH VP ApplH VP
V Theme V t
On the other hand, the Theme can undergo the movement in (47b) if it
undergoes another step of movement before the next phase is complete. In an
active clause like those in (46), the next phase above ApplHP is the active vP;
in a passive, however, it is the next CP. Consider first the passive in (48). Here
the Theme moves to spec-ApplH, as in (47b), but then it moves on to spec-T,
leaving a trace in spec-ApplH. This trace is not a candidate for checking the
phi-features of v, since a trace cannot enter into checking relations. Thus, by
the time the CP phase is complete, the checking relation between v and the
Benefactive satisfies the MLC.
(48) [CP [TP K-èlyái [vP k-i̋-lyì-í-ò [ApplHP ti m̀-kà [VP ti]]]]].
7-food 7S-PR-eat-APPL-PAS 1-wife
“The food is being eaten for the wife.” BM (5c)
The Theme can move through spec-ApplH in an active benefactive

clause as well, provided that it is a wh-operator (49). If so, it will undergo
another step of movement to spec-vP before the vP phase is complete.17
(49) K-èlyá [á-i̋-lyì-í-à m̀-kà] ki̋-pùsù.

7-food 1S.REL-PR-eat-APPL-FV 1-wife 7-rotten
“The food which he is eating for the wife is rotten.” BM (38b)
17
Presumably, high applicatives that allow the DO-IO order also have the DO move
to spec-v before the vP phase is complete. This DO-IO option is not permitted in Kichaga
(Bresnan & Moshi 1990).
The derivation of the relative clause in (49) is schematized in (50). In (50a),

the Benefactive is generated above the Theme, which is a relative operator. In
order to escape the ApplHP phase, the Theme operator must undergo phase-
EPP movement to spec-ApplH, where it checks Case and the uninterpretable
phi-features of ApplH (50b). Active v is then merged, and a (here, null) subject
is base-generated in spec-v (50c). If no further movement of the Theme takes
place before the vP phase is complete, this Theme will block the necessary
feature-checking relation between v and the Benefactive, and the derivation
will violate the MLC and crash, as it does in (46b). However, if the Theme
operator is to achieve its scope position in spec-C, it must escape the active vP
phase as well. Phase-EPP movement of the Theme to spec-v is shown in (50d).
Once the Theme occupies this position, the feature-checking relation between v
and the Benefactive respects the MLC, since the intervening trace of the
Theme is not a closer candidate for checking features of v. Thus the vP phase
in (50d) satisfies the MLC. The derivation continues as in (50e), with the
subject moving to spec-T and the Theme operator moving to spec-C of the
relative clause. This CP phase also satisfies the MLC, and the relative clause in
(49) converges, as predicted.
(50) a. [ApplHP Benefactive [VP Op-Theme]]

b. [ApplHP Op-Themei Benefactive [VP ti]]
c. [vP pro-Subject v [ApplHP Op-Themei Benefactive [VP ti]]
d. [vP Op-Themei pro-Subject v [ApplHP ti Benefactive [VP ti]]
e. [CP Op-Themei [TP pro-Subjectj [vP ti tj v [ApplHP ti Benefactive [VP ti]]
The account above assumes that, although the Theme operator checks
its uninterpretable Case feature in spec-ApplH, it does not become inactive for
the phase-EPP movement to spec-v shown in (50d). We can attribute the active
status of this operator to an uninterpretable [WH] or [scope] feature, which is
eventually checked by the head of the CP in which the operator takes
scope—in this case the relative CP in (50e). The same assumption is necessary
to allow successive-cyclic wh-movement out of a CP phase. For example, the
wh-phrase in (51) checks Case in the lower spec-v, but nevertheless it can
escape the embedded CP (and the higher vP) via phase-EPP movement.
(51) [CP Whoi did [TP you [vP ti say [CP ti that [TP you [vP ti [VP met ti ]]]]]]]?
Again, we can assume that this wh-phrase bears a feature that is not checked
until it reaches its scope position in matrix spec-C.
26 Martha McGinnis
5 Phonological phrasing
In Sections 1-4 it was shown that asymmetries in transitivity, A-
movement, object agreement, and pronoun incorporation can be derived from
the semantic difference between ApplL and ApplH, in combination with the
proposal that ApplHP heads a phase. I will now extend this proposal to capture
asymmetries in phonological phrasing in applicatives. Seidl (2000) makes the
striking observation that phonological phrasing in Bantu languages is closely
related to variation in the derivation of the passive. If an applicative allows
symmetric passives, both objects are generally bracketed in a phonological
phrase with the verb. If only the higher object can become the subject of the
passive, the two objects are generally in separate phonological phrases.
Example (52a) is from Kinande, a language with symmetrical passives.
In Kinande, a process of Penultimate Vowel Lengthening (PVL) applies only
at the right edge of a phonological phrase (Hyman & Valinande 1985). PVL
applies to the Theme in (52a), but not to the Recipient. This indicates that the
two objects are in the same phonological phrase ([V IO DO]). By contrast, the
Chi-Mwi:ni: Recipient applicative has an asymmetric passive: only the higher
object can raise to the subject position. In Chi-Mwi:ni:, a process of Vowel
Length Shift (VLS) applies only at the right edge of a phonological phrase
(Kisseberth & Abasheikh 1974). Since VLS applies to the Recipient in (52b),
we can conclude that it is phrased separately from the Theme ([V IO] [DO]).
The chart below (52), from Seidl (2001), suggests that the relation between the
passive and phrasing facts holds up cross-linguistically.
(52) a. [Tu-ká-βi-túm-ir-a omúkali valinánde]. Æ

Tu-ká-βi-túm-ir-a omúkali valiná:nde.
we-PST-T-send-APPL-FV woman Valinande
“We have just sent Valinande to the woman.” HV, AS (5.7)
b. [Ni-mw-andik-il-ile nuru:] [xatí]. Æ

Ni-mw-andik-il-ile nu:ru xatí.
SP-OP-write-APPL-FV Nuru letter
“I wrote Nuru a letter.” KA1, AS (5.9)
LANGUAGE PASSIVE PHON. PHRASING

Kikuyu Symmetric [V NP NP]
Kinyarwanda Symmetric [V NP NP]
Kinande Symmetric [V NP NP]
Haya Symmetric [V NP NP]
Xhosa Symmetric [V NP NP]
Chi-Mwi:ni: Asymmetric [V NP] [NP]
Kiswahili Asymmetric [V NP] [NP]
To account for this generalization, Seidl argues that in a symmetric

applicative (here, a high applicative), the DO raises to spec-Appl.18 In an
asymmetric applicative (here, a low applicative), the IO raises to spec-v, but
the DO remains in situ. Under the account given here, both types involve
movement to the edge of a phase to check phase-EPP. The Theme/DO of a
high applicative raises to the edge of the ApplHP phase (53a), while in a low
applicative, only the more local IO can raise to the edge of the vP phase. In this
case the Theme/DO is left in the domain of the vP phase, while the IO is at its
edge, as in (53b).
(53) a. vP High applicative (Symmetric)

DP v'
v ApplHP
IO ApplH'
DO ApplH'
ApplH VP
[phase-EPP] V t
b. vP Low applicative (Asymmetric)

DP v'
IO v'
v VP
V ApplLP
[phase-EPP] t ApplL'
ApplL DO
The phonological phrasing facts follow if we assume that the domain of

a phase is phrased separately from its edge. The correlation between phases
18
Seidl also assumes that the IO moves to spec-v in symmetric applicatives, but this
assumption is unnecessary under the approach adopted here.
28 Martha McGinnis
and phonological phrases supports the view that phases are units of the syntax
interpreted separately by the phonological component.
This line of reasoning brings us back to an issue raised in Section 3.1.
We have seen (as in (4), repeated below) that a high applicative may allow a
‘short’ passive, in which the higher object moves to subject position (4a), as
well as the ‘long’ passive permitted by a phase-EPP feature on ApplH (4b).
(4) a. Umukoôbwai a-ra-andik-ir-w-a ti íbárúwa n’ûmuhuûngu.

girl SP-PR-write-APPL-PAS-ASP letter by boy
“The girl is having the letter written for her by the boy.” AK (6,3c)
b. Íbárúwai i-ra-andik-ir-w-a umukoôbwa ti n’ûmuhuûngu.

letter SP-PR-write-APPL-PAS-ASP girl by boy
“The letter is written for the girl by the boy.” AK (6,3b)
There are a couple of possible accounts of the optionality in (4). One

possibility is that the phase-EPP feature on ApplH is optional, so the DO
remains in situ. Another possibility is that the phase-EPP feature on ApplH is
obligatory, but that the DO can raise to a specifier either above or below the
base position of the IO. When the DO raises to the higher specifier of ApplH, it
can become the subject of the passive. When the DO raises to the lower
specifier, the IO can become the subject of the passive.
A number of observations combine to support the latter option: phase-
EPP on ApplH is obligatory, but the DO and IO in spec-ApplHP may be in
either c-command order (recall that a higher specifier c-commands a lower
one). First, many languages with symmetrical passives have both IO-DO and
DO-IO order in the active voice, which suggests that phase-EPP on ApplHP is
available in the active.19 Now, recall that in an active high applicative,
phonological phrasing of the IO and DO does not simply vary optionally:
generally, the two are phrased together. Under the account given above, this
means that phase-EPP on ApplH is obligatory in the active, with DO either
stacking up above IO or tucking in underneath IO.20 If so, it is plausible to
assume that the same options arise in the passive, with obligatory phase-EPP
on ApplH here as well.21 Moreover, according to Seidl (2000), the DO of a
symmetric (high) applicative remains in the same phonological phrase as the
19
See footnote 15.
20
However, there are cases in which ApplH lacks phase-EPP in active ditransitives
with no movement of either object. For example, although the syntax and semantics of the
Kichaga Benefactive indicate that it is a high applicative, in the active its phonological
phrasing is [V IO][DO] (Seidl 2000). As expected, Kichaga also disallows DO-IO word order
(see footnote 15).
21
Thanks to Christina Tortora for raising this issue.
verb, even when the IO raises to the subject position. This observation also
supports the view that the DO raises to spec-ApplH in the passive.22
As we have seen, however, stacking and tucking are not always freely
available options. For example, a shifted direct object cannot tuck in
underneath the base position of the subject in English. If it did, movement of
the subject to spec-T would not violate the MLC, and object shift would be
permissible in English. Similarly, a raised embedded subject cannot stack up
on top of the Experiencer in a French or Italian Raising+Experiencer
construction. If it did, movement of the embedded subject to the matrix spec-T
would not violate the MLC, and raising past an Experiencer in spec-ApplH
would be permissible in these languages. Another case of obligatory ordering
in phase-EPP movement can be seen in Chichew◊a.
The Chichew◊a benefactive has the phrasal phonology of a high
applicative, as shown by penultimate vowel lengthening in (54). Since the
vowel lengthens on the DO, but not on the IO, we can conclude that the two
objects are phrased together: [V IO DO].
(54) [a-dzá-ónetsa mfúmú yá álenje gálu] Æ

[a-dzá-ónetsa mfúmú yá álenje gaálu]
he-will-show chief of hunter dog
“He will show the chief of the hunters the dog.” JK, HT (3,37b)
However, the Chichew◊a benefactive has an asymmetrical passive, like a low

applicative (55). The Benefactive can become the subject of the passive (55a),
but the Theme cannot (55b) (based on Marantz 1993:127).
(55) a. AtsíkaÌnai a-na-gúl-ír-idw-á ti mphâtso (ndí chítsîru).

2.girls 2S-PST-buy-APPL-PAS-FV 9.gift by 7.fool
“The girls were bought a gift (by the fool).”
b.* Mphâtsoi i-na-gúl-ír-idw-á átsíkaÌna ti (ndí chítsîru).

9.gift 9S-PST-buy-APPL-PAS-FV 2.girls by 7.fool
“A gift was bought for the girls (by the fool).”
Prima facie, this combination of facts is problematic for the analysis presented
here. If the Chichew◊a benefactive has an asymmetrical passive, we expect it to
be a low applicative. If so, the IO and the DO should be in separate
phonological phrases, but in fact they are phrased together.
Interestingly, however, the Chichew◊a benefactive has the semantics of
a high applicative (Sam Mchombo & Liina Pylkkänen, personal
22
The same phrasing arises in a passive low applicative, where vP is not a phase,
since no argument is generated in spec-v (recall (14)).
30 Martha McGinnis
communication). The applicative can be used even when there is no direct

relation (such as intended possession) between the IO and the DO, as shown in
(56).
(56) a. A-na-dy-er-a mkazi wake chakudya.

SP-PST-eat-APPL-FV woman his food.
“He ate food for his wife.”
b. A-na-gwir-ir-a mkazi wake thumba.

SP-PST-hold-APPL-FV woman his bag
“He held the bag for his wife.” SM
This combination of phonological and semantic properties suggests that

the Chichew◊a benefactive is actually a high applicative. If so, the fact that
Chichew◊a benefactives allow only the Benefactive to become subject of the
passive suggests that the DO must tuck in under the IO (57).
(57) ApplHP
IO ApplH'
DO ApplH'
ApplH VP
V t
As a result, the DO-IO order does not arise in the active counterpart, even
though the DO moves out of the domain of the ApplHP phase.
(58) a. Anyani a-ku-pang-ir-a atsikana mauta.

baboons SP-PR-make-APPL-FV girls bows
“The baboons are making bows for the girls.”
b.* Anyani a-ku-pang-ir-a mauta atsikana.

“(same as (58a))” MB (28)
Under this analysis, the phonological phrasing of a high applicative is correctly

predicted. The DO escapes the domain of the ApplHP phase, so it is phrased
together with the IO.
6 Non-asymmetries
As we have seen, there are a number of asymmetries between high and
low applicatives. I will now present several apparent non-asymmetries between
high and low applicatives. These include non-asymmetries in quantifier scope
possibilities and morphophonological properties.
6.1 Quantifier Scope

It has long been observed that quantifier scope is ‘frozen’ in the
English double-object construction (Aoun & Li 1989). For example, the
double-object construction in (59a) allows the direct scope reading, in which
the same child receives all the dolls, but not the inverse scope reading, in
which each doll goes to a different child. This construction contrasts with the
prepositional dative in (59b), which does allow an inverse scope reading, in
which each child may receive a different doll.
(59) a. I gave a child each doll. ∃ » ∀, *∀» ∃

b. I gave a doll to each child. ∃ » ∀, ∀» ∃
Bruening (to appear) argues that quantifier scope is frozen because QR

respects the MLC. A lower quantifier cannot undergo QR over a higher one to
take wide scope. Assuming that QR is a type of phase-EPP movement, the
restriction follows from the low applicative structure given for English
Recipient applicatives (60b). The DO cannot undergo QR over the IO to the
edge of the vP phase.
(60) a. High applicative b. Low applicative
ApplHP VP
Q ApplH' V ApplLP
IO-Q ApplH' IO-Q ApplL'
ApplH VP ApplL DO-Q
[phase-EPP] V DO *
If this account is correct, then, just as high applicatives allow the lower
object to raise to the subject position in a passive, they should also allow the
lower object to take scope over the higher one. This prediction appears to be
borne out in Albanian high applicatives. (61a) does allow the inverse scope
reading, in which each book goes to a different student. However, note that the
32 Martha McGinnis
inverse scope reading is available only when the direct object is clitic-doubled;
(61b) allows only the direct scope reading.23
(61) a. Profesori i-a dha një studenti çdo libër. ∃ » ∀, ∀» ∃

professor.NOM CL gave a student.DAT every book.ACC
“The professor gave a student every book.” DK
b. Profesori i dha një studenti çdo libër. ∃ » ∀, *∀» ∃
Further complications arise in Spanish. Here the DO can raise to the

subject position of the passive, but cannot undergo QR over an IO quantifier in
the active. Let us begin with the claim that Spanish has a dative alternation (see
Cuervo 2000, 2001, among others). In the prepositional dative (62a), the Goal
argument is not clitic-doubled, while in the double-object derivation (62b) it is.
In the clitic-doubled double-object variant in (62b), the doubled dative is
restricted to potential beneficiaries or recipients.
(62) a. Pablo mandó una carta a Andreína / a Barcelona.

Pablo sent a letter to Andreína / to Barcelona
“Pablo sent a letter to Andreína / to Barcelona.” MC
b. Pablo lei mandó una carta a Andreínai / *a Barcelonai.

Pablo CL sent a letter to Andreína / to Barcelona
“Pablo sent Andreína / *Barcelona a letter.” MC
The two structures also show a contrast in scope possibilities. In the

prepositional variant, scope is free, as in English. (63a) shows that the
universally quantified DO (cada cuadro “each painting”) can take scope over
an indefinite PP, and (63b) shows that the universally quantified PP (a cada
revista “to each museum”) can take scope over an indefinite DO.24
(63) a. Andrés mandó cada cuadro a un museo (distinto). ∃ » ∀, ∀» ∃

Andrés sent each painting to a museum different.
“Andrés sent each painting to a (different) museum.” MC
b. Carolina llevó un artículo (distinto) a cada revista. ∃ » ∀, ∀» ∃

Carolina took an article (different) to each magazine
“Carolina took a (different) article to each magazine.” MC
23
Kallulli (1999) discusses Albanian clitic-doubling and quantification in more detail.
24
Similar options are available for French datives (Youri Zabbal, personal
communication).
In the double-object variant, however, scope is frozen, again as in

English, although in Spanish the DO linearly precedes the IO. A universally
quantified DO cannot take scope over an indefinite IO (64a), by contrast with
(63a): the IO must take higher scope. Even if distinto “different” is included in
(64a), it does not allow the distributive reading; instead, it is taken to indicate
that the single museum that received all the paintings is different (say, from
some other museum in the context). In (64b), the universally quantified IO
takes scope over an indefinite DO.
(64) a. Andrés lei mandó cada cuadro a un museo (#distinto)i. ∃ » ∀, *∀» ∃

Andrés CL sent each painting to a museum different.
“Andrés sent a (#different) museum each painting.” MC
b. Carolina lei llevó un artículo (distinto) a cada revistai. ∀» ∃

Carolina CL took an article (different) to each magazine
“Carolina took each magazine a (different) article.” MC
Like Albanian, Spanish allows the DO to become the subject of a

passive double object construction (65).
(65) El premio lei fue entregado a Valeriai.

the prize CL was given to Valeria
“The prize was given (to) Valeria.” MC
Thus we find that in Spanish, the DO can undergo movement past the IO in the
passive, but cannot take scope over the IO in the active. This contrast is not
predicted if A- and A-bar movement (such as QR) are subject to the same
restrictions. One possible account of frozen scope in the Spanish double-object
construction is that the dative clitic contains the quantificational features of the
dative argument. Since the DO does not c-command this clitic, it cannot take
scope over it. On the other hand, if a clitic can contain quantificational features
of the doubled DP, then the possibility of wide scope for a clitic-doubled DO
in Albanian can be captured without recourse to the high/low asymmetry. The
relation between A- and A-bar movement in applicatives remains to be fully
explored.
6.2 Morphophonological properties

High and low applicatives apparently cannot be reliably distinguished
according to morphological case or applicative marking. For example,
applicatives whose applied object has dative morphological case may be high,
as in Albanian benefactives (66a), or low, as in Icelandic benefactives (66b).
34 Martha McGinnis
(66) a. Drita i pjek Agimit.

D.NOM CL bake A.DAT
“Drita bakes for Agim.” DK
b. Ólafur baka∂i henni *( köku).

O.NOM baked her.DAT cake.ACC
“Olafur baked her *(a cake).” OJ
Likewise, applicatives in which the applied object lacks dative morphological

case may be high, as in Kinyarwanda benefactives (67a), or low, as in English
benefactives (67b).
(67) a. Umugabo a-rá-som-er-a umugóre.

man SP-PR-read-APPL-ASP woman
“The man is reading for the woman.” AK (4,40)
b. Mary baked Alicia *(a cake).
Moreover, applicatives marked by an overt applicative morpheme may

be high, as in the Kinyarwanda benefactive (67a), or low, as in the
Kinyarwanda locative (68).
(68) Umuhuûngu á-r-íig-ir-á-ho ishuûri * (imibáre).

boy SP-PR-study-ASP-LOC school mathematics
“The boy is studying mathematics at school.” AK
Applicatives not marked by an overt applicative morpheme may also be high,

as in the Italian Experiencer construction (69), or low, as in the English
benefactive (67b).
(69) Giannii non gli sembra [ti fare il suo dovere].

G. not him.DAT seems to do his duty
“Gianni does not seem to him to do his duty.” LR (22b)
In the data discussed here, high applicatives are always associated

either with dative morphological case or with an overt applicative affix, while
a low applicative may lack both. However, this distinction is apparently not
universal. Applicatives without applicative marking or dative case on the
applied argument show a cross-linguistic distinction between symmetric and
asymmetric passives (Ura 1996). Assuming that symmetric passives involve a
high applicative, (70) shows a high applicative in Oromo which lacks both
applicative marking and dative case: both objects in (70a) are in the absolutive
case. Nevertheless, the subject of the passive counterpart of (70a) can be either
the IO (70b) or the DO (70c).
(70) a. Innii na gaafii gaafat-e.

he.NOM me.ABS question.ABS ask-PST
“He asked me a question.”
b. An gaafii gaafat-am-e.
I.NOM question.ABS ask-PAS-PST
“I was asked a question.”
c. Gaafii-n na gaafat-am-t-e.
question.NOM me.ABS ask-PAS-FEM-PST
“A question was asked (of) me.” JO, HU (5A9)
While there may be a positional distinction between high and low

applicative morphemes within a given language, such distinctions do not seem
to hold up cross-linguistically (Marantz 2001). The benefactive ApplH head in
Kinyarwanda is an affix -er, which comes before the aspectual suffix (see
(67a)), while the locative ApplL head is a clitic ho, which follows the aspectual
suffix (see (68)). However, recall the argument from Section 5 that the Chi-
Mwi:ni: applicative is also low (71); here the ApplL head is an affix, -il, not a
clitic.
(71) Ni-mw-andik-il-ile nu:ru xatí.

SP-OP-write-APPL-FV Nuru letter
“I wrote Nuru a letter.” KA1, AS (5.9)
In short, the distinction between high and low applicatives is primarily a

syntactic and semantic one. The distinction appears to have no universal
morphological correlate.
7 Concluding remarks
The foregoing is only a preliminary sketch of the syntactic properties
that appear to be connected to the semantic difference between high and low
applicatives. It would be misleading to suggest that all applicatives fall neatly
into this characterization. Still, it is clear that the phase account of applicatives
makes it possible to express important new generalizations, spanning the
continuum from phonology to semantics. This account also makes it possible
to formulate clear hypotheses about the kinds of evidence learners use to set
the lexical parameter that underlies much of the cross-linguistic variation in the
syntax of applicatives. As such, it constitutes a crucial step towards an
explanatory account of such variation.
36 Martha McGinnis
References
Alexiadou, Artemis, & Elena Anagnostopoulou. 1998. Raising without

infinitives and the nature of agreement. In S. Bird, A. Carnie, J.
Haugen, & P. Norquest (eds.), WCCFL 18 (West Coast Conference on
Formal Linguistics). Somerville, MA: Cascadilla Press, 14–26.
Aoun, Joseph, & Yeh-hui Audrey Li. 1989. Scope and constituency. Linguistic
Inquiry 20, 141–172.
Baker, Mark C. 1996. The Polysynthesis Parameter. New York: Oxford
University Press.
Baker, Mark. 1988. Theta Theory and the syntax of applicatives in Chichew◊a.
Natural Language and Linguistic Theory 6, 353–389.
Belletti, Adriana, & Ur Shlonsky. 1995. The order of verbal complements: A
comparative study. Natural Language and Linguistic Theory 13, 489–
526.
Bobaljik, Jonathan, & Diane Jonas. 1996. Subject positions and the roles of
TP. Linguistic Inquiry 27, 195–236.
Bresnan, Joan, & Lioba Moshi. 1990. Object asymmetries in comparative
Bantu syntax. Linguistic Inquiry 21, 147–185.
Bruening, Benjamin. 1999. QR obeys Superiority: Frozen scope and ACD.
Ms., MIT. To appear in Linguistic Inquiry.
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger
Martin, David Michaels, & Juan Uriagereka (eds.), Step by step: Essays
on minimalist syntax in honor of Howard Lasnik. Cambridge, MA:
MIT Press.
Chomsky, Noam. 1999. Derivation by phase. MIT Occasional Papers in
Linguistics 18. Department of Linguistics, MIT.
Chomsky, Noam. 1995. Categories and transformations. In The minimalist
program. Cambridge, MA: MIT Press, 219–394.
Cuervo, Maria Cristina. 2001. The dative alternation in Spanish. Paper
presented at the Appl-Fest Workshop, MIT, January 27–28.
Cuervo, Maria Cristina. 2000. First things first in dative experiencers. Ms,
MIT.
Embick, David. 1997. Voice and the interfaces of syntax. Doctoral dissertation,
University of Pennsylvania.
Falk, Cecilia. 1990. On double object constructions. Working Papers in
Scandinavian Syntax 46, 53–100. Department of Scandinavian
Linguistics, Lund University.
Harley, Heidi. 1995. Subjects, events, and licensing. Doctoral dissertation,
MIT.
Harley, Heidi, & Rolf Noyer. 2000. Licensing in the non-lexicalist lexicon. In
Bert Peeters (ed.), The Lexicon/Encyclopaedia Interface. Amsterdam:
Elsevier Press, 349–374.
Hiraiwa, Ken. 2001. EPP and Object Shift in the Scandinavian Languages:
Localization of Locality. Paper presented at WCCFL 20 (West Coast
Conference on Formal Linguistics), University of Southern California,
February 23–25.
Holmberg, Anders. 1997. The true nature of Holmberg’s Generalization. In
Kiyomi Kusumoto (ed.), Proceedings of NELS 27. Amherst, MA: GLSA.
Hyman, Larry, & Nzama Valinande. 1985. Globality in the Kinande tone
system. In D. Goyvaerts (ed.), African Linguistics: Essays in memory of
M. W. K. Semikenke. Amsterdam: Benjamins.
Jonas, Dianne. 1996. Clause structure and verb syntax in Scandinavian and
English. Doctoral dissertation, Harvard University.
Kallulli, Dalina. 1999. The comparative syntax of Albanian: On the
contribution of semantic types to propositional interpretation. Doctoral
dissertation, University of Durham.
Kanerva, Jonni M. 1989. Focus and phrasing in Chichew◊a phonology.
Doctoral dissertation, Stanford University.
Kimenyi, Alexandre. 1980. A Relational Grammar of Kinyarwanda. Berkeley:
University of California Press.
Kisseberth, Charles, & Abasheikh, Mohammad Imam. 1977. The object
relationship in Chi-Mwi:ni:, a Bantu language. In Peter Cole & Jerrold
Sadock (eds.), Syntax and Semantics, Vol. 8: Grammatical relations.
New York: Academic Press, 179–218.
Kisseberth, Charles, & Abasheikh, Mohammad Imam. 1974. Vowel length in
Chi-Mwi:ni:—a case study of the role of grammar in phonology. In M.
W. la Galy, R. A. Fox, and A. Bruck (eds.), Papers from the
parasession on Natural Phonology. Chicago Linguistic Society.
Letsholo, Rose. 2001. On the (universal, non-?) existence of unaccusative
applicatives: the case of Ikalanga. Paper presented at the Appl-Fest
Workshop, MIT, January 27–28.
Marantz, Alec. 2001. Prefix/Suffix Asymmetries in a Late Insertion/No
Lexicon Framework. Paper presented at the 2001 Asymmetry
Conference, Université du Québec à Montréal, May 7–10.
Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis
in the privacy of your own lexicon. In Alexis Dimitriadis, Laura Siegel,
Clarissa Surek-Clark, & Alexander Williams (eds.), Proceedings of the
21st Annual Penn Linguistics Colloquium. Penn Working Papers in
Linguistics, 4.2. Department of Linguistics, University of
Pennsylvania, 201–225.
Marantz, Alec. 1993. Implications of asymmetries in double object
constructions. In Sam Mchombo (ed.), Theoretical aspects of Bantu
grammar. Stanford: CSLI Publications, 113–150.
Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge,
MA: MIT Press.
38 Martha McGinnis
Massey, Victoria Walker. 1992. Compositionality and constituency in

Albanian. Doctoral dissertation, University of North Carolina, Chapel
Hill. MIT Occasional Papers in Linguistics 3. Department of
Linguistics and Philosophy, MIT.
McGinnis, Martha. 2001. Phases and the syntax of applicatives. In Minjoo
Kim & Uri Strauss (eds.), Proceedings of NELS 31. Amherst, MA:
GLSA.
McGinnis, Martha. 1998. Locality in A-movement. Doctoral dissertation, MIT.
Nakamura, Masanori. 1997. Object extraction in Bantu applicatives: Some
implications for Minimalism. Linguistic Inquiry 28, 252–280.
Oehrle, Richard. 1976. The grammatical status of the English dative
alternation. Doctoral dissertation, MIT.
Owens, Jonathan. 1985. A grammar of Harar Omoro. Hamburg: Buske.
Pesetsky, David. 1995. Zero syntax: Experiencers and cascades. Cambridge,
MA: MIT Press.
Pylkkänen, Liina. 2001. What applicative heads apply to. In Michelle Minnick
Fox, Alexander Williams & Elsi Kaiser (eds.), Proceedings of the 24th
Annual Penn Linguistics Colloquium. Penn Working Papers in
Linguistics, 7.1. Department of Linguistics, University of
Pennsylvania.
Rezac, Milan. 2000. Case checking in ditransitive constructions. In
Proceedings of the CLA Annual Conference (Canadian Linguistics
Association). Distributed by Cahiers Linguistiques d’Ottawa.
Department of Linguistics, University of Ottawa.
Richards, Norvin. 1997. What moves where when in which language? Doctoral
dissertation, MIT. [Revised version to appear as Movement in
Language: Interactions and Architectures. Oxford University Press.]
Rizzi, Luigi. 1990. Relativized minimality. Cambridge, MA: MIT Press.
Rizzi, Luigi. 1986. On chain formation. In Hagit Borer (ed.), The syntax of
pronominal clitics: Syntax and Semantics 19. New York: Academic
Press.
Seidl, Amanda. 2000. Minimal Indirect Reference: A theory of the syntax-
phonology interface. Doctoral dissertation, University of Pennsylvania.
Thráinsson, Höskuldur. 1979. On complementation in Icelandic. New York:
Garland.
Truckenbrodt, Hubert. Phonological phrases: Their relation to syntax, focus,
and prominence. Doctoral dissertation, MIT.
Woolford, Ellen. 1984. Dative verbs with unspecified objects: Optional
subcategorization versus lexical detransitivization. The Linguistic
Review 3, 389–409.
Ura, Hiroyuki. 1996. Multiple feature-checking: A theory of grammatical
function-splitting. Doctoral dissertation, MIT.

LVYB

Uploaded by

Copyright:

Available Formats

You might also like

LVYB

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

LVYB

Uploaded by

Copyright:

Available Formats

Variation in the Phase Structure of Applicatives*

Until recently, cross-linguistic differences in the syntax of applicative

(2) a. Umugóre a-rá-som-er-a umuhuûngu igitabo.

b. Umugabo a-rá-som-er-a umugóre.

(3) a. Umuhuûngu á-r-íig-ir-á-ho ishuûri imibáre.

b.* Umuhuûngu á-r-íig-ir-á-ho ishuûri.

Another difference is in the A-movement properties of the two types of

(4) a. Umukoôbwai a-ra-andik-ir-w-a ti íbárúwa n’ûmuhuûngu.2

b. Íbárúwai i-ra-andik-ir-w-a umukoôbwa ti n’ûmuhuûngu.

(5) a. Ishuûrii ry-oohere-j-w-é-ho ti igitabo n’úúmwáalímu.

b.* Igitaboi cy-oohere-j-w-é-ho ishuûri ti n’úúmwáalímu.

(6) a. Umugóre a-rá-mui-he-er-a ti ímbwa ibíryo.

b. Umugóre a-rá-bii-he-er-a umugabo ímbwa ti.

(7) a. Úmwáalímu y-a-ryi-oohere-jé-ho ti igitabo.

b.* Úmwáalímu y-a-cyi-oohere-jé-ho ishuûri ti.

Various accounts have been proposed for the set of asymmetries

(8) a. Agimi i mban Dritës çanten time.

b. *John held Mary the bag.

Pylkkänen’s proposed structures for the two types of applicatives

(9) a. N-a̋-i̋-lyì-í-à m̀-kà k-élyà.

(10) Drita i pjek Agimit (rrepat).

(11) Mary baked Alicia *(a cake).

(12) Ólafur baka∂i henni *( köku).

I will adopt Pylkkänen’s proposals concerning the phrase structure and

In some cases, however, EPP features can be added to a phase before it is

2 Phases and the two types of applicatives

(14) The sister of VP heads a phase if an argument is generated in its specifier.

(15) a. High applicative b. Low applicative

Ideally, (14) should be derived from some broader generalization.

The latter proposal is supported by evidence that D does indeed head a

(17) a. K-èlyái k-i̋-lyì-í-ò m̀-kà ti.

b. M-kài n-a̋-i̋-lyì-í-ò ti k-èlyâ.

(18) a. Secili libëri iu kthye autorit të tij ti.

b. Secilit djalëi iu dha ti paga i tij.

By contrast, Recipient applicatives in English and Icelandic are low

not a phase, no phase-EPP feature can be added to ApplL to allow the DO to

(19) a. Aliciai was baked ti a cake.

(20) a. Honum var gefin bókin.

b.* Bókin var gefin honum.9

Suppose, contrary to what has been suggested above, that ApplLP is

However, there is considerable evidence that (21) is not the correct

ApplL should be able to move to spec-ApplL. For example, in a raising

(23) Giannii non gli sembra [ti fare il suo dovere].

Further evidence against (21) (under either option I or II above) comes

(24) Which medali did Reuben award Ben Johnson ti ?

To verify this claim, consider (26), which allows either a narrow-scope de

These examples show that the Theme of an English Recipient

3.2 Object agreement and pronoun incorporation

(27) vP High applicative (DO agreement/incorporation)

For example, in a Kichaga Benefactive, the Benefactive (28a), the Theme

(28) a. N-a̋-i̋-m-lyì-í-à k-èlyâ.

In a low applicative, however, if v has only one phase-EPP feature,

(29) vP Low applicative (*DO agreement/incorporation)

(30) a. HamadÙi Ø-wa-pik-il-ile wa:na cha:kuja.

b.* HamadÙi Ø-sh-pik-il-ile wa:na cha:kuja.

4 The MLC: Independent evidence for the ApplHP phase

4.1 Phase-level locality in vP

OS languages like Icelandic contrast with non-OS languages like

(31) Jólasveinarnir bor∂u∂u bú∂inginni [VP ekki tV ti]