Adjectival participles revisited
Artemis Alexiadou1, Berit Gehrke2, Florian Schäfer1
(1University of Stuttgart, 2Universitat Pompeu Fabra)
Kratzer (2001) shows that participles denoting states resulting from prior events do not form a
homogeneous class from a semantic point of view. They are divided into two subclasses: target
and resultant state participles (Parsons 1990: 234-235). The former describe states that are in
principle reversible; the latter introduce states that hold forever after the event that brings them
about. The test to differentiate between the two types of participles is their (in)compatibility
with immer noch ‘still’ ((1a) vs. (1b)).
(1) a Die Geißlein sind (immer noch) versteckt.
Target state
the goats
are still
hidden
b. Das
Theorem ist (*immer noch) bewiesen.
Resultant state
the theorem is still
proven
Building on Kratzer, several authors proposed that there are structural differences between two
types of participles (e.g., Anagnostopoulou 2003, Alexiadou & Anagnostopoulou 2008,
Lundquist 2008; cf. Embick 2004). In particular, target state participles are sometimes claimed
to lack any eventive layer, i.e. they are directly derived from roots (2a, or stative projections ResultP). Resultative participles involve a verbal/eventive layer on top of the root ((2b)). In
some languages, the latter can also have a Voice layer (Greek), and thus inherit the complete
argument structure of the underlying verb (Anagnostopoulous 2003).
(2) a. [Asp [Root]]
b.
[Asp [VoiceP/vP [Root]] ]
The absence of a v layer is further argued for by Embick (2004) with examples like those in (3).
Similar to the adjective in (3a), the participle in (3b) expresses a situation in which the door
never participated in a change-of-state event, and thus it should involve a root-derivation.
However, as we will show, such root-derived participles cannot be equated with target state
participles per se.
(3) a. This door was built open/*opened.
b. This door was built closed.
In the recent literature, however, the above picture has been challenged. For instance,
Anagnostopoulou & Samioti (2009) have already provided evidence that in Greek (at least
some) target state participles must involve a verbal layer. McIntyre (2011) shows that English
adjectival participles can license by-phrases. In this paper, we provide further arguments that
adjectival passives can actually involve more verbal functional structure than assumed so far, by
focusing on German, which we compare to English and Greek. Importantly, we a) identify
target state participles in German and English, which involve verbal projections, and b) we show
that German adjectival participles like their Greek and English counterparts can license byphrases. We then address some questions of variation with respect to the nature of by-phrases,
the licensing of adverbial modification and control into purpose clauses in adjectival participles
in the languages under discussion. Note that our claim is not that ALL adjectival participles
involve a verbal eventive layer; sometimes adjectival participles lack any eventive verbal layers,
as suggested by (3), for which we assume the structure in (2a).
Our arguments for the presence of verbal functional structure in adjectival participles
come from the presence of (i) verbalizers as well as (ii) transitivizing morphology. First, we
show that target state participles can involve overt verbalizing heads. In English, many verbs are
derived from some non-verbal source (category-neutral Roots in our terminology) by the
addition of verbalizing affixes. From the perspective of Distributed Morphology (DM),
verbalizing affixes are the spell-out of a v-head as their presence is clearly related to the
verbal/eventive nature of the verbs. Harley (2009b) discusses in detail that affixes like -ify, -ate
and -ize are specific verbalizing morphology. As is shown in (4), adjectival participial
morphology attaches to these affixes that have verbalized the bare root; this suggests that the
verbalizing head is still present.
(4) Root:
!COLON !MOBIL !DICT
!HTML !SATIS
Root + v: colon-ize mobil-ize dict-ate
html-ify satis-fy
participle: coloniz-ed mobiliz-ed dictat-ed htmlifi-ed satisfi-ed
Nevertheless, a number of participles involving v layers pattern with target state participles in
being compatible with still ((5)); this makes a root-derivation of these participles implausible.
(5)
a. The boss is still satisfied.
b. The patient is still hospitalized.
c. The city is still electrified.
d. The country is still colonized.
The logic of this argumentation predicts that adjectival participles involving verbalizing suffixes
should not be compatible with Embick’s context in (3). This is borne out, see (6).
(6)
*The city is built electrified.
Second, German (also English rise vs. raise, lie vs. lay) has a few verbs undergoing the
causative alternation which show a stem alternation; in the anticausative variant we find the
stem vowel i, while in the causative and passive variant the stem vowel shifts to e ((7)).
(7)
a. Das Schiff versinkt/*versenkt.
b. Hans versenkt/*versinkt das Schiff.
the ship sinks.intrans/sinks.trans
John sinks.trans/sinks.intrans the ship
c. Das Schiff wurde (von der Marine) versenkt/*versinkt.
the ship was (by the marine) sinked.trans/sinked.intrans
From a DM perspective, this shift can be related to the presence vs. absence of higher verbal
structure, i.e. verbal structure on top of the first verbalizer/eventivizer. While both causatives
and anticausatives are eventive, only the former are transitive, i.e., involve an external
argument. Following Kratzer (1996), we assume that Voice introduces external arguments, and
we argue that the presence vs. absence of this layer triggers the stem alternation. Specifically,
with ver-sinken/senken, we assume that ver introduces the result state, sink modifies v and senk
reflects the Voice + v structure. Crucially, not only the anticausative but also the causative can
form an adjectival passive. While das Schiff versenken does not really form a target state
participle ((8a)), as still-modification is only possible if the state is in principle reversible, the
example in (8b) is fine.
(8)
a. *Das Schiff ist immer noch versenkt.
b. Die Münze ist immer noch im Aquarium versenkt.
the coin is still in the aquarium sunk.trans.participle
The above suggests that target state participles can even contain Voice, the layer introducing the
external argument, in addition to the v layer. The availability of by-phrases in German (e.g.,
Rapp 1996, Maienborn 2007) and English (9) adjectival participles support this.
(9)
The dictator remained propped up by the warlords. (McIntyre 2011)
However, while both Greek and English/German adjectival passives are compatible with
manner modifiers, instruments, and by-phrases, the availability of these modifiers is much more
restricted in English/German than in Greek. In particular, English/German only allow for byphrases that introduce agents, whose nature is detectable from the nature of the state. Similar
facts have been observed for Hebrew (Meltzer 2011). This is not the case in Greek.
(10) a.
Ta
lastixa ine
fuskomena apo
tin
Maria.
The tires are
inflated
by
the
Mary
‘The tires are still inflated by Mary’
b.
*Der Fisch
war von Maria
gebraten.
The fish
was by
Mary
fried
‘The fish was fried by Mary’
Moreover, while Greek participles allow control into purpose clauses, German/English/ Hebrew
ones do not. We argue that this provides further evidence for assuming two types of by-phrases
associated with two different heights of attachment: an argumental one licensed by Voice in e.g.
Greek, and a modifier one licensed by Aspect in languages like German/English/Hebrew.
Building on Meltzer (2011), we argue that in Greek by-phrases attach to Voice prior to
stativization, while in English/German/Hebrew by-phrases combine with Aspect. Thus the
former are unrestricted, while the latter are severely restricted, cf. McIntyre (2011) who
proposes that in English, Asp is a special kind of adjectival Voice head, which merges with
transitive verbs and projects no overt external argument in the syntax.