Posted by on Sat 08 Dec 2007 at 9:23 AM:
It seems like the effects on the shared conditions are quite solid, and the Empirical Logit Model seems to be a more effective interpretation tool.
One question is that why the privileged object draws additional attention of the speaker? (How the privileged conditions is created? What effects would have had if the objects were to be increased to six or the privileged object was to be increased to the same number as the non-privileged objects?)
Posted by dalebarr on Tue 11 Dec 2007 at 10:48 AM:
Thanks for your question.
There are two possibilities why the interference in the privileged condition would be slightly more than in the shared conditions. The first possibility is that the additional interference is an "ironic" effect, of the sort discussed by Wegner and colleagues (e.g. Wegner, Schneider, Carter & White, 1987): the attempt to actively suppress the information actually makes it more rather than less available.
The second possibility (and the one that I favor) is that the additional interference is a by-product of the memory task. Listeners believed that they were playing a computerized card game with the speaker, and that cards that were marked as 'hidden' were not visible to the speaker. After some trials containing a hidden card, listeners were asked to report the identity of the hidden card. This was to make sure they were distinguishing between shared and privileged knowledge. However, it is very likely that it had the by product of making them focus more on privileged objects, making these objects (competitors and noncompetitors alike) more available to the reference resolution process.
The manuscript I am working on that reports these results has an additional experiment that supports this second interpretation. Specifically, when one removes the memory task, identical interference is observed in the shared and privileged conditions.
I think that increasing the number of objects or the ratio of shared to privileged objects within a trial should dilute the overall main effect of competition, but I see no reason to believe that it would differentially affect the shared or privileged conditions. Maybe an extreme situation in which there are many privileged objects and say, only one or two objects might do so. But this should be taken with the caveat that I've learned not to trust my intuitions regarding how even seemingly minor task variables might affect the results!
Posted by on Sun 23 Dec 2007 at 8:43 AM:
The additional interference might be due to the "ironic" effect or due to the by-product of the memory task. But it might also be possible that it is due to the effect of an "artificial" shared condition, depending on how the privileged condition is created.