Saturday, March 29, 2008

Three types of causation

Now we come to the more complex events. These verbs encode the broad idea of causation, and have three case slots; not only the THEME and LOC of Simple events, but a AGENT (or CAUSER) slot.

Delancey has the general AGENT slot, witha single CAUSE meaning. As discussed by Delancey, AGENT in this meaning is anything that can be construed as causing an action. In [1], Delancey presents examples of "His attitude infuriates me" (AGENT = "His attitude"), "This mess really bothers him." (AGE = "This mess"), "The beauty of this vista has inspired many artists." (AGENT = "The beauty ..."), and "The look on her face would curdle milk." (AGENT = "The look ...").

Janckendoff [2], cited in a previous post, recognizes two kingd of grammatical causation, CAUSE and LET, with the AGENT for CAUSE "bringing the event about", and for LET, "ceasing to prevent the event".

As noted in a previous post, Steven Pinker, in The Stuff of Thought, recognizes what seems to be this basic distinction between CAUSE and LET.
Jackendoff's "ceasing to prevent the event" for let appears to be not quite right; it is included in, but not completely covering, the range of LET. THe basic difference between CAUSE and LET seems ot be that with CAUSE, the CAUSE'd event is construed as not happening on its own, and would not have happend but for the (unspecified) action of the AGENT. LET, in the other hand, deals with an event that is construed as happening, on their own, but with the (unspecified) action of the AGENT somehow implicated in its happening. This can be either by removing a condition that prevents them from happening, or by doing something that establishes a condition that allows it to happen.

Jackendoff's brief definition of LET shows there is another type of causation - PREVENT, which is a sort of opposite of LET. Prevent, like LET, deals with events that are construed as (potentially) happening on their own, but through the action of the AGENT is construed as not happening.

PREVENT may be able to be treated something along the lines of AGENT CAUSE NOT simple-event, and it may at some points be useful to view it in this manner. But it seems that a certain clarity is obtained by keeping it separately.

So we not have three types of complex events
  • AGENT CAUSE simple-event
  • AGENT LET simple-event
  • AGENT PREVENT simple-event
where simple-event is any of the four types of event discussed in the previous post.

[1] Delancey, Scott, Event Construal and Case Role Assignment, in Proc. of the 17th Annual Meeting of the Berkeley Linguistics Society (1991), pp. 338-353

[2] Jackendoff, Ray, A System of Semantic Primitives (pdf), in Theoretical Issues in Natural Language Processing, 10-13 June 1975

Labels:

Causation

Now we come to the more complex events. These verbs encode the broad idea of causation, and have three case slots; not only the THEME and LOC of Simple events, but a AGENT (or CAUSER) slot.
Delancey has the general AGENT slot, witha single CAUSE meaning. As discussed by Delancey, AGENT in this meaning is anything that can be construed as causing an action. In [1], Delancey presents examples of "His attitude infuriates me" (AGENT = "His attitude"), "This mess really bothers him." (AGE = "This mess"), "The beauty of this vista has inspired many artists." (AGENT = "The beauty ..."), and "The look on her face would curdle milk." (AGENT = "The look ...").
Janckendoff [2], cited in a previous post, recognizes two kingd of grammatical causation, CAUSE and LET, with the AGENT for CAUSE "bringing the event about", and for LET, "ceasing to prevent the event".
As noted in a previous post, Steven Pinker, in The Stuff of Thought, recognizes what seems to be this basic distinction between CAUSE and LET.
Jackendoff's "ceasing to prevent the event" for let appears to be not quite right; it is included in, but not completely covering, the range of LET. THe basic difference between CAUSE and LET seems ot be that with CAUSE, the CAUSE'd event is construed as not happening on its own, and would not have happend but for the (unspecified) action of the AGENT. LET, in the other hand, deals with an event that is construed as happening, on their own, but with the (unspecified) action of the AGENT somehow implicated in its happening. This can be either by removing a condition that prevents them from happening, or by doing something that establishes a condition that allows it to happen.
Jackendoff's brief definition of LET shows there is another type of causation - PREVENT, whic is a sort of opposite of LET. Prevent, like LET, deals with events that are construed as (potentially) happening on their own, but through the action of the AGENT is construed as not happening.
PREVENT may be able to be treated something along the lines of AGENT CAUSE NOT simple-event, and it may at some points be useful to view it in this manner. But it seems that a certain clarity is obtained by keeping it separately.
So we not have three tyoes of complex events
  • AGENT CAUSE simple-event
  • AGENT LET simple-event
  • AGENT PREVENT simple-event
where simple-event is any of the four types of event discussed in the previous post.
[1] Delancey, Scott, Event Construal and Case Role Assignment, in Proc. of the 17th Annual Meeting of the Berkeley Linguistics Society (1991), pp. 338-353

[2] Jackendoff, Ray, A System fo Semantic Primitives (pdf), in Theoretical Issues in Natural Language Processing, 10-13 June 1975

Labels:

Monday, March 03, 2008

The four types of simple events

Simple event structures, those without an AGENT, are represented by Jackendoff as GO and STAY, and by Delancey as GOTO. Following Delancey, each as a THEME and an expressed LOC; we can also see that they include an implict LOC. The LOC of GO expresses the final, or ending state or position, and encodes the meaning that the THEME ended up at that LOC. There is, however, the impicit state or location that the THEME started from, and with GO this implicit state is different than the ending one.

Jackendoff, in [1] cited in the previous post, notes almost in passing while discussing the sentence "Lyn caused Bob to be happy", that with his GO "Source and Goal must be disinct". Reading the senetence as Lyun causes Bob to become happy", then following more the lines of Delancey than the presentation of Jackendoff here, I would parse that sentence Lyn CAUSE Bob GOTO Happy, in which we may conclude (as does Jackendoff), that Bob was not happy prior to whatever Lyn may have done to cause him to be happy.

With Jackendoff's STAY, there are also the two states, but here beginning and ending states are the same.

This leads to the classification of these simple events by examining the beginning and ending states, and the explictness of the LOC in the grammar. IN GO (or GOTO), THEME GOTO LOC, with the beginning and ending states being different and the ending state is expressed; with STAY, THEME STAY LOC, the beginning and ending states are the same. THERe are two other possibilites, THEME LEAVE LOC, in which, like GOTO, the two states are the different but it is the beginning, rather than the ending, state that is made explicit. Then there is THEME AVOID LOC, in which the THEME does not GOTO, STAY, or LEAVE the LOC. Here, since neither the beginning or ending point are explictly expressed, it is also somewhat unexpresses whether they are the same or different - one may avoid going to New York by staying in New Jersey, or by going form new Jersey to Pennsylvania.

The there seem to be four possibilites:

  • THEME GOTO LOC

  • THEME STAY LOC

  • THEME LEAVE LOC

  • THEME AVOID LOC



In use, the word choice is made in part by the salience of the manner in which the event is being construed, it one uses an AVOID verb, then it is somehow salient that a reasonably expected THEME GOTO LOC did not occur.

It seems that Jackendoff would handle the LEAVE and AVOID verbs by using negatives : THEME LEAVE LOC would be considered THEME GOTO NOT LOC, and AVOID as something along the lines of THEME STAY NOT LOC. At this point, however, my feeling is that having all four will be useful, although I do not deny these types of conversions between them may be also useful in some circumstances.

Labels: