5.1.1.2.3. The Causal Approach

Croft (1991) is the main proponent of the causal approach to event structure. He builds from Talmy’s (1988) work on the dynamics of force and takes as his approach’s philosophical underpinnings Donald Davidson’s 1969 "The Individuation of Events", where it is argued that causal structure defines events. Croft proposes to apply Davidson’s ideas to lexical items. The first example he presents to support this move is precisely one of the verbs that Talmy analyzed as incorporating both "manner" and motion. Croft’s example is the following (1991: 160):

He claims that the conflation of both meanings, manner and motion, is only possible because there is a causal relationship between the activity of sailing and the motion. The activity denoted by the verb has to be the cause of the motion in order for the sentence to be felicitous. If it is not, the conflation cannot take place. Croft illustrates this with the following unacceptable example:

According to Croft this sentence cannot mean that the boat "was burning while entering the cave." Only when the burning is the cause of the motion, the combination becomes acceptable:

Further evidence for Croft’s position comes from examples like the following, which illustrate a contrast between manner verbs and path verbs:

Since in (b) the manner of motion must be the cause of the entering, a further means of entering by breaking the door cannot be specified.

For Croft, then, the lexical semantics of verbs reflect causal structure. The main assumptions of his causal approach are the following:

The event type that fits this model best is the prototypical transitive verb in which unmediated volitional causation brings about a change in the entity acted upon. A sentence like the following exemplifies this prototype:

It involves a three part causal chain, with Sandy exerting force on the window, the window changing state, and ending up in a resulting state (i.e. broken). There is an asymmetric transmission of force from Sandy to the window, with Sandy as the initiator and the window as the endpoint. This is why they appear as subject and object, respectively.

In order to accommodate the many verbs that do not denote this asymmetric transfer of force, Croft stipulates a ‘coercion’ mechanism by which they are conceptualized as if they did. This process compares with the process by which the non-local values are derived from the local ones in the localist approaches.

Many of the semantic role list approaches carry with them a causal perspective on event structure. Names like Agent and Patient make this manifest. Nevertheless, as Croft remarks, his approach delineates an explicit model of event structure and organizes the relationships between the participants in an event in a way that semantic role lists cannot. He achieves this by giving definitions of the semantic roles in terms of configurations in causal chains. His argument realization rules also make reference to the causal chain. Thus, for example, subjects precede objects causally. Even obliques are accounted for by appealing to the causal chain denoted by the verb. He subclassifies oblique roles into antecedent and subsequent roles according to whether they precede or follow the endpoint of the causal chain. Instrument, Manner, Means, Comitative, and Cause are antecedent roles, whereas Benefactive, Recipient and Result are subsequent roles. As support for this distinction, Croft points out that across languages there are case syncretisms involving roles within the antecedent and the subsequent classes, but no case syncretisms across the divide.

Comparing both the aspectual approach and the causal approach, it can be argued that as far as the subject function and the oblique functions are concerned, the causal approach makes a more substantive contribution. With regard to the realization of arguments as objects, both identify the patient argument of causative change of state verbs as the canonical objects, but whereas the causal account is better suited for affected objects, the aspectual alternative seems to handle effected objects more adequately.