The morphology of transitivity*

Naoko Tomioka
McGill University

In this paper, I present a puzzle that a class of V-V compound

(aspectual V-V compound) in Japanese poses. I first show that
with causative constructions, the morphological shape of the
base verb and the adverbial interpretation provide useful
evidence in determining the size of the embedded structure. In
the context of aspectual V-V compounds, however, the two
tests show conflicting results.

1. Introduction

Traditionally, all embedded clauses were assumed to have a full clause structure
such as CP or IP. Recent developments in syntax have challenged this idea and instead,
postulate that some embedded clauses are bare verbal projections (VPs), lacking
functional projections (e.g. Baker 1996, Wurmbrand 2002). One of the positive
consequences of this proposal is that the similarities between lexical causative
constructions and syntactic causative constructions can be restored. In both constructions,
a causative predicate embeds a structure denoting the caused event, but they differ from
each other in whether this structure contains a functional category or not. While lexical
causative predicates embed bare VPs, syntactic causative predicates embed a full clause.
This idea, however, requires us to take an extra precaution in determining what
exact embedded structure we might have whenever we have an embedded structure. It is
during such a process that the puzzle I report here was discovered.
The puzzling piece of evidence presented in this paper is exemplified in (1).

(1) Kotaro-ga yukkuri-to gohan-o tabe-hazime-ta.

K-NOM slowly-ADV meal-ACC eat-begin-PAST
Kotaro began to eat the meal slowly.

In this paper, I show that this construction has the morphological appearance that makes
it pattern with the syntactic causative. In other words, the morphological shape of this
construction suggests that the embedded structure is a full clause. However, the
interpretation of the adverb yukkurito slowly, as indicated in the translation makes this

*
I would like to thank the participants of the Workshop on Japanese Phonology and Morphology at
University of Toronto. The research was funded by J. W. McConnell McGill Major Scholarship to the
author and SHRRC grant # 41020040966 to Professor Lisa Travis.
T W Toronto Working Papers in Linguistics 26: 115122
P L Copyright 2006 Naoko Tomioka
 NAOKO TOMIOKA

construction pattern with the lexical causative, in which the embedded structure is a bare
VP.

1.1 Types of causatives and transitivity

Causative predicates can be classified based on the type of causation they express:
lexical causative predicates express direct causation (as defined in Lewis 1973,
McCawley 1976, Bittner 1999) and syntactic causative predicates express non-direct
causation.1 This lexical-syntactic distinction is discussed extensively in Shibatani (1976).
He examines the behavior of adverbs, VP anaphora (so su) and event-referring pronoun
(sore) and argues that lexical causatives, unlike syntactic causatives, do not involve full-
clausal embedding.
Moreover, the lexical/syntactic distinction correlates with the morphological
shape of the base the causative predicate selects for: the lexical causative suffix selects
for a root (Shibatani 1976) or an unaccusative verb in certain contexts (Miyagawa 1984),
and the non-lexical causative suffix selects for verbal stems of any transitivity
(unaccusative, unergative, transitive), but not a root.
Syntactically, lexical causative verbs take a VP or rootP complement, and
syntactic causatives take a vP or a larger constituent as their complement (e.g. Pylkknen
2002).

1.2 Adverbs

The lexical/non-lexical classification of causative predicates correlates with

whether a manner adverbial element can modify one of the events (see Shibatani 1976).
In a lexical causative construction, a manner adverb cannot modify only one of the events
(the causing event or the caused event). Instead, a manner adverb must modify the entire
event. In a syntactic causative construction, in contrast, a manner adverb can modify just
the caused event on its own, or the causing event on its own.
This behavior of a manner adverb is generally attributed to a connection between
adverbial modification and event-ness. Originally, Davidson (1967) argues that adverbs
are predicates of events, and that an action verb introduces such an event. Davidsons
idea is implemented in the current syntactic theory by associating voice with event-ness
(see also Harley 1995, Kratzer 1996, and Travis 1994).
Based on the behavior of the adverbs in the context of causative predicates, one
may generalize that manner adverbs attach to vP, and not to VP.

1.3 Aspectual V-V compounds

In Japanese, aspectual verbs (such as hajime begin) form compounds with an

embedded verb. I will show that morphologically, aspectual verbs pattern with syntactic
causatives. They select for verbal stems of any transitivity, but not a root. When the

1
This is a language-particular feature of Japanese (see Shibatani 1976). In English, a syntactic causative
may express direct or non-direct causation.

116
 THE MORPHOLOGY OF TRANSITIVITY

adverbs are applied, however, the aspectual compounds pattern with lexical causatives in
not allowing the modification of part of the complex word.
The aspectual V-V compound thus presents a puzzle for an analysis of the
embedded clause structure. The morphological form and the adverbial test provide
conflicting evidence about the size of the embedded structure.

2. Lexical causatives

As I mentioned in the introduction, the lexical causative construction always

expresses direct causation. Moreover, adverbs treat the lexical causative construction as
the unit of modification, providing no ambiguity as shown in (2).

(2) a. Kotaro opened the door slowly.

b. Kotaro made the girl sad gradually.

In English, direct causation is expressed with a phonologically null element (), or make
+ adjective. The adverbs in these examples modify the progression of the whole event,
rather than just the caused event or the causing event.
Next, we see that the same condition holds in the Japanese lexical causative
construction. In Japanese, direct causation is expressed with a lexical causative
morpheme (irregular forms, or sasi).

(3) a. Kotaro-ga isu-o yukkuri-to tao-si-ta.

K-NOM chair-ACC slow-ADV fall-L.CAUSE-PAST
Kotaro toppled the chair slowly.
b. Kotaro-ga bo:ru-o sotto koroga-si-ta.
K-NOM ball-ACC gently roll-L.CAUSE-PAST
Kotaro rolled the ball gently.

As in the English examples, the adverbs yukkuirito slowly and sotto gently
unambiguously modify the whole event rather than just the causing event or the caused
event.
We should also note the morphological shape of the base of the lexical causative
construction. The lexical causative morpheme attaches to a root (3a) or a
monomorphemic unaccusative verb (3b), but not to a transitive (4a,b), or bimorphemic
unaccusative verb (4c).

(4) a. * Kotaro-ga Naoko-ni isu-o kowasi-si-ta.

K-NOM N-DAT chair-ACC break.TRANS-L.CAUSE-PAST
b. * Kotaro-ga Naoko-ni isu-o kowas-asi-ta.
K-NOM N-DAT chair-ACC break.TRANS-L.CAUSE-PAST
c. * Kotaro-ga Naoko-o tao-re-si-ta.2
K-NOM N-ACC fall-UNACC-L.CAUSE-PAST

2
The verb root tao fall appears with -re when it is unaccusative and -s when it is transitive.

117
 NAOKO TOMIOKA

In addition, the caused event in a lexical causative construction is always non-agentive.

The sentences in (3) imply that the accusative casemarked element did not participate
willingly.
Syntactically, lexical causative predicates are analyzed as VP-taking, and are
distinguished from syntactic causative predicates, which subcategorize for vP or IP (see
Pylkknen 2002). This idea is consistent with the behavior of the adverbs shown in (3)
and the morphological restrictions on the verbal base (4). I have argued elsewhere that
manner adverbs target vP, and not VP (see Tomioka, forthcoming). The behavior of the
adverbs in (3) thus confirms the general idea that a lexical causative construction embeds
a structure smaller than the structure embedded in a syntactic causative construction.
Moreover, the lexical causative suffix may not attach to a verbal base which contains a
voice feature (transitivity), because the structure that a lexical causative predicate embeds
should not contain such information.

3. Syntactic causatives

Japanese has, in addition to the lexical causative suffix, a suffix that realizes a
syntactic causative predicate.3 Syntactic causatives in Japanese always express non-direct
causation (Shibatani 1976). In other words, the syntactic causative construction expresses
that, in addition to the causer argument of the causing event, there is a volitional agent
that carries out the caused event. Moreover, in both English and Japanese, the same
adverb that cannot modify just the caused event or the causing event with a lexical
causative construction can do exactly that with a syntactic causative construction, as
shown in (5).

(5) a. Kotaro caused the ship to sink slowly (by chewing a small hole in its hull).
b. Kotaro quickly caused the ship to sink (slowly).

In (5a), the sentence asserts that the caused eventthe sinking of the shipwas slow,
without entailing that what caused the ship to sink was slow. In (5b), the sentence asserts
that what Kotaro did, which eventually led to the sinking of the ship, was quick, but the
sentence is true even if the caused eventthe sinking of the shipitself was slow. The
same pattern is observed in Japanese, as shown in (6).

(6) a. Kotaro-ga Naoko-o yukkuri-to taore-sase-ta.

K-NOM N-ACC slow-ADV fall-S.CAUSE-PAST
Kotaro made Naoko fall slowly (only Naokos action is slow).
b. Kotaro-ga Naoko-ni inu-o sotto okos-ase-ta.
K-NOM N-DAT dog-ACC gently awaken.TR-S.CAUSE-PAST
Kotaro made Naoko awaken the dog gently (Naokos action is gentle).

Moreover, in Japanese, the difference between the lexical causative construction and the
syntactic causative construction surfaces in their morphological shape as well. Unlike the

3
However, see Miyagawa (1984) for an argument that in some limited contexts, this syntactic causative
morpheme may realize a lexical causative predicate)

118
 THE MORPHOLOGY OF TRANSITIVITY

lexical causative morpheme, the syntactic causative morpheme attaches to a bimorphemic

unaccusative (6a), or transitive verb (6b), but not to a root (7).

(7) * Kotaro-ga Naoko-o tao-sase-ta.

K-NOM N-ACC fall-NL.CAUSE-PAST (cf. 6a, 3a)

Syntactically, non-lexical causative predicates are treated as vP- or IP-taking verbs (see
Pylkknen 2002). This assumption about the non-lexical causative predicate is consistent
with the behavior of adverbs with this construction. As the embedded clause of a non-
lexical causative predicate is at least a vP, manner adverbs can attach to and modify just
the embedded event (Tomioka, forthcoming).
Based on the examples presented so far, one might come up with the following
generalizations:
A predicate that selects for a VP (or a smaller constituent) appears with an
unaccusative verb or a root, and adverbs cannot modify just the embedded event
(the pattern of lexical causatives).
A predicate that selects for a vP (or a larger constituent) appears with a verb of
any transitivity, and adverbs can modify just the embedded event (the pattern of
non-lexical causatives).

4. Aspectual Verbs

In this section, I show that aspectual V-V compounds are problematic to the
classification presented above. In a way, the aspectual V-V compound appears to embed
a full clause, like the syntactic causative. Some tests, however, show that the embedded
clause of the aspectual V-V compounds cannot be a full clause.

4.1 The morphological shape of the aspectual V-V compound

Aspectual verbs select for an event-denoting complement and express the

temporal aspect of that event its complement denotes. In this section, I show that the
aspectual V-V compound has a morphological shape closer to the syntactic causative than
to the lexical causative.
The following examples show that aspectual verbs form a compound with a
transitive (8a), unergative (8b), or unaccusative verb (8c) (see Kageyama 1989).

(8) a. Kotaro-ga gohan-o tabe-hajime-ta.

K-NOM meal-ACC eat-begin-PAST
Kotaro began to eat the meal.
b. Kotaro-ga warai-tuzuke-ta.
K-NOM laugh-continue-PAST.
Kotaro continued to laugh.
c. Kotaro-ga taore-kake-ta.
K-NOM fall-almost.do-PAST
Kotaro almost fell.

119
 NAOKO TOMIOKA

We should recall that this is the pattern we saw with the syntactic causative suffix. In
addition, like syntactic causative suffixes, aspectual verbs do not combine with a root (9).

(9) * Kotaro-ga tao-kake-ta.

K-NOM fall-almost.do-PAST (cf. 7)

These examples thus show that morphologically, aspectual verbs pattern with the
syntactic causative morpheme -(s)ase, but not with the lexical causative suffix.

4.2. Adverbs

However, aspectual verbs pattern with lexical causatives when it comes to

adverbial modification. The following examples illustrate that aspectual verbs and their
V1 are inseparable for adverbial modification.

(10) a. Kotaro-ga gohan-o yukkuri-to tabe-hajime-ta.4

K-NOM meal-ACC slow-ADV eat-begin-PAST
Kotaro slowly began to eat the meal.
Not Kotaro began to eat the meal slowly (i.e. he slowed down).
b. # Kotaro-ga gohan-o satto tabe-kake-ta.
K-NOM meal-ACC swiftly eat-almost.do-PAST
Kotaro swiftly almost ate the meal
Not Kotaro almost ate the meal swiftly.
(Kotaro is trying to correct his fast-eating habit)

We should recall that in the syntactic causative construction, the same adverbs can
modify the caused eventi.e. the embedded structure, but not in the lexical causative
construction. The behavior of the adverb in (10) thus suggests that the complement of
aspectual verbs is a constituent too small for adverbial modification (VP).
The properties of the aspectual V-V compound presented in this section pose a
question for the relation between morphology and syntax. If the aspectual verbs
subcategorize for VPs, why do they combine with transitive verbs? If aspectual verbs can
subcategorize for VPs and appear with a transitive verb base, why cant a lexical
causative morpheme attach to a transitive verb base?

References:

Baker, Mark. 1996. The Polysynthesis Parameter. Oxford, UK: Oxford University Press.
Bittner, Maria. 1999. Concealed causatives. Natural Language Semantics 7: 178.
Davidson, Donald. 1967. The logical form of action sentences. In The Logic of Decision and Action,
Nicholas Roscher (ed.) pp. 8195. Pittsburg: University of Pittsburg Press.
Harley, Heidi. 1995. Subjects, Events and Licensing. Doctoral dissertation. MIT.

4
There is an additional complication here. If we change the form of the adverb yukkurito slowly to
yukkuri -, some speakers are able to get the intended reading of the adverb. However, this variation in
form does not affect the interpretation of the adverb in the lexical causative context, and hence, this issue is
left aside in this paper. I would like to thank the audience at the workshop for raising this issue.

120
 THE MORPHOLOGY OF TRANSITIVITY

Kageyama, Taro. 1989. The place of morphology in the grammar: verb-verb compounds in Japanese.
Yearbook of Morphology 2, 7394.
Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase Structure and the
Lexicon, Johan Rooryck & Laurie Zaring (eds.) pp. 109137. Dordrecht, the Netherlands: Kluwer.
Lewis, David. 1973. Causation. Journal of Philosophy 70: 556567.
McCawley, James. 1976. Remarks on what can cause what. In Syntax and Semantics VI: The grammar of
causative constructions, Masayoshi Shibatani (ed.) pp. 117129. New York: Academic Press.
Miyagawa, Shigeru. 1984. Blocking and the Japanese causatives. Lingua 64: 177207.
Pylkknen, Liina. 2002. Introducing arguments. Doctoral dissertation. MIT, Cambridge, Massachusetts.
Shibatani, Masayoshi. 1976. The grammar of causative constructions: a conspectus. In Syntax and
Semantics VI: the grammar of causative constructions, Masayoshi Shibatani (ed.) pp. 142. New
York: Academic Press.
Tomioka, Naoko. forthcoming. Resultative constructions: Cross-linguistic variation and the syntax-
semantics interface. Doctoral dissertation. McGill University, Montreal, Canada.
Travis, Lisa. 1994. Event phrase and theory of functional categories. In Proceedings of Canadian
Linguistics Association (CLA 1994), pp. 559570.

121
NAOKO TOMIOKA

122