Instrument subjects are agents or causers
A: In this talk we argue against the view that the thematic variation of subjects of many causative
verbs (cf. 1) can be captured by the underspecified notion of an abstract causer or initiator of an event
(cf. van Valin & Wilkins 1996, Ramchand 2003 for this position). Arguments come from two
domains. First, this view runs into problems when faced with languages that restrict the interpretation
of the external argument to that of either an agent or a causer. This has been reported for e.g. Hebrew,
which selects different verbal templates for the two roles (Doron 2003) and for the Greek passive,
which licenses only agentive PPs (Alexiadou, Anagnostopoulou & Schäfer (AAS) 2005, Zombolou
2004). Moreover, while it is well known that anticausatives across languages do not license agents,
they do allow causers (AAS 2005). (The behavior of the Greek passive (cf. 2) and of anticausatives in
English Greek and German (cf. 3) will be used as diagnostics). Second, the underspecification view
cannot explain the crosslinguistic restrictions on instrument subjects on which we focus here. Taking
up the observation by Cruse (1973), De Lancey (1984, 1991) and Schlesinger (1989) among others,
we note that in English instrument subjects often sound unnatural but become acceptable under one of
two licensing conditions. We show that the same two conditions hold for instrument subjects in Dutch,
German and Greek. We re-interpret these conditions and argue that they trigger a causer-like and an
agent-like interpretation of the instrument respectively. This provides a further argument in favor of a
formal distinction between agents and causers.
B: The first condition states that if an instrument can be understood as eventive then it can serve as
subject. This condition is fulfilled by machines (called intermediary or instrument-causers by Levin &
Rappaport (1992) and Kamp & Rossdeutscher (1994)) which act on their own, but not by facilitating
or pure instruments which are under permanent control by a human agent (cf. 4a vs.b). Non-
mechanical instruments show the same partition. Axe is typically understood as a pure instrument but
in an eventive construal axe is an acceptable subject (cf. 5a vs. 5b) (De Lancey 1984). Borrowing a
term from van Valin & Wilkins (1996) we call the instruments in (4a) and (5b) self-energetic, thereby
being comparable to causers/natural forces as wind or earthquake (cf. Talmy 2000).
The second condition says that if one element in the sentence bears contrastive focus then again the
instrument subject can become acceptable (Schlesinger 1989, De Lancey 1991). This condition does
not apply to all kinds of instruments (cf. 6 vs. 7) which indicates that it is not focus alone that brings
about the effect in (6b-c). Further, the fact that focus on the instrument, the verb or the object supports
this effect suggests that it is not a property of the instrument itself that is responsible for the
amelioration, but rather a property relating the instrument and the VP.
C. Following AAS (2005), we assume the structure in (8) for change of state verbs. Voice introduces
external arguments (agent/causer) in causatives and passives. CAUS introduces a causal relation
between a causing event (its implicit argument) and the resultant state denoted by the Root + Theme
complex. The general presence of CAUS is diagnosed by the anticausatives in (3) which lack voice.
On the basis of (8), we formalize the two licensing conditions on instrument subjects as follows. We
assume two Voice heads differing in their semantics: following Kratzer (1993) and Pylkkänen (2002)
the Voice head in (9a) introduces agents via event identification and the Voice head in (9b) introduces
causers/natural forces by equating the eventive causer/force with the CAUS event in the VP of (8).
Instrument subjects licensed by the first condition are (re-)interpreted/coerced as causers/forces with
an event variable and are introduced by (9b). This predicts that such types of instruments should be
able to appear as PPs in anticausatives where causers are licensed but agents and pure instruments not.
This is borne out crosslinguistically and shown here with English in (10). Turning to the second
condition (11) shows that focus cannot turn instruments into causers. We think that focus stresses a
non-trivial relation between (a property of) the instrument and the change of state of the theme, which
exists in (6) but not in (7). The fact that contrastive focus can also fall on the verb or the theme shows
that the instrument is evaluated against the whole VP, something argued to be characteristic of agents
(Marantz 1984, Kratzer 1993). We argue therefore that instrument subjects licensed by the second
condition are agents under a stripped down notion of `agent'. Evidence for this view gives the Greek
passive which licenses agents via apo and pure instruments via me but not causers. But if contrastive
focus is involved, the instrument is introduced via apo just like human agents (cf. 12).
The stripped down notion of agentivity does not entail volition. This feature can only be
implicated if the subject is a human agent. Some roots (murder) are [+volition] by their encyclopedic
meaning thereby forcing the subject to be a human agent.
(1) Max / The storm / The key opened the door (agent / causer / instrument)
(2) I porta anihtike apo ton filaka / me, *apo to klidi / *apo, *me ton aera
The door opened-Pass by the guardian / with, by the key / by, with the wind
(3) a. The vase broke from the earthquake / *from John / *from, with the axe
b. Die Vase zerbrach durch das Erdbeben / *durch Hans / *mit, durch die Axt (German)
The vase broke through the earthquake /through John / with, through the axe
c. To vazo espase apo ton aera / *apo to Jani / *apo, me to tsekuri (Greek)
The vase broke by the wind / *by the John / by, with the axe
(4) a. The dishwasher cleaned the dishes
b. *The rag cleaned the dishes
(5) a. *The axe broke the window
b. The falling axe broke the window / The axe fell off the shelf and broke the window
(6) a. ?(?)The key opens / opened the door
b. THIS key opened the door / The KEY opened the door
c. The key OPENED the door / The key opened THIS door
(7) a. *The rag cleans / cleaned the table
b. ??THIS rag cleaned the dishes / The RAG cleaned the dishes
c. ??The rag CLEANED the dishes / The rag cleaned THIS table
(8) [(Voice) [ CAUS(e) [Root + Theme]]] (cf. AAS 2005)
Voice-agent: x e [agent (x) (e)]
(9) a. (cf. Kratzer 1993)
Voice-causer: x e [(x) = (e)]
b. (cf. Pylkkänen 2002)
(10) a. The vase broke from the earthquake / *from John / *from the axe / *with the axe
b. The window broke from the falling axe
(11) *The / THIS door opened from THIS key / from the KEY
(12) I porta anihtike apo ton filaka / *apo, *me ton aera / me, *apo to klidi / apo AFTO to klidi
The door opened-Pass by the guardian / by, with the wind / with, by the key / by this the key
References:
Alexiadou, A., Anagnostopoulou, E. & F. Schäfer (2005). The properties of anticausatives cross-
linguistically. To appear in M. Frascarelli (ed.). Phases of Interpretation, Berlin: Mouton.
Cruse, D. A. (1973). Some Thoughts on Agentivity, Journal of Linguistics 9, 11-23.
De Lancey, S. (1984). Notes on Agentivity and Causation. Studies in Language 8, 181-213.
De Lancey, S. (1991). Event construal and Case Role Assignment. BLS 17, Berkeley Linguistics
Society, Berkeley, CA, 338-353.
Doron, E. (2003). Agency and Voice: The Semantics of the Semitic Templates. Natural Language
Semantics 11 (1), 1-67.
Kamp, H. & A. Rossdeutscher (1994). Remarks on lexical structure and DRS construction.
Theoretical Linguistics 20 (1), 97-164.
Kratzer, A. (1993). The Event Argument and the Semantics of Voice. Ms., University of
Massachusetts, Amherst.
Levin, B. & M. Rappaport (1992). -er Nominals: Implications for a Theory of Argument Structure. In
T. Stowell and E. Wehrli (eds.), Syntax and Semantics 26: Syntax and the Lexicon, 127-153,
Academic Press, New York, NY.
Marantz, A. (1984) On the Nature of Grammatical Relations. Cambridge/Mass: MIT Press.
Pylkkänen, L. (2002). Introducing Arguments. Ph.D. diss., Department of Linguistics, MIT.
Ramchand, G. (2003). First phase syntax. Ms., University of Tromsoe.
Schlesinger, I. M. (1989). Instruments as Agents: On the Nature of Semantic Relations. Journal of
Linguistics 25, 189-210.
Talmy, L. (2000). Toward a Cognitive Semantics I: Concept Structuring Systems. MIT Press,
Cambridge, MA.
Van Valin, R. D., Jr. & D. P. Wilkins (1996). The Case of Effector: Case Roles, Agents, and Agency
Revisited. In M. Shibatani & S. A. Thompson (eds.), Grammatical Constructions. 289-322.
Oxford UP.
Zombolou, K. (2004) Verbal alternations in Greek: a semantic approach. Ph. D. diss., Department of
Linguistics, University of Reading.