record the active referents in the discourse. A new referent is introduced by adding a new index that is not marked
coreferential with any of the previous indices.
We now impose some structure on the list. Consider the event of a woman buying a car. In order for this event to be
conceptualized, there must also be a conceptualized woman who does the buying and a conceptualized car that she
buys. Of course, this woman and this car might be on the list of available referents even if the event of her buying it
never took place. Thus there is a sense in which the event isreferentially dependenton its characters, while the characters
are not referentially dependent on the event.
Let us notate this asymmetric referential dependence in the referential tier like this:
Thearrows pointingfro m3 to1 and 2 can beunderstoodineither oftwoways: (a) Theclai med existenceoftheevent
depends ontheclaimed existenceof thecharacters, (b)The claimed existenceof theevententailstheclaimed existenceof
the characters. For the moment we leave open which interpretation is best. (This notion of referential dependence
comes from Csuri 1996; a similar notion appears in Erteschik-Shir 1998.)
Now consider what happensif (35) is turned intoa question:Did a woman buy a car?Suddenlythere is no claimthat the
event took place; rather the speaker wishes the hearer to assert whether or not it did. In the absence of this claim,
suddenlya womananda carlosetheirexistential force as well:weneed notbeabletoidentify either ofthem. Thus there
is a difference in the referential structure of assertions and questions. We will express this in the referential tier by
adding a furtherarrow that denotesassertiveforce,terminatingin theindex for theeventvariable. Thus thereferential
tier of (35) will become (36), while the corresponding question lacks the assertion arrow.