Following up on the idea that information learned under particular conditions is better recalled in those same conditions, Horton (2007) tested whether language recall is sensitive to audience. Specifically, is it easier to find a word that you recently used with a particular addressee if that addressee is present? If so, that would imply that our referential choices can make use of addressee-specific encodings.
For Horton's study there was a pre-task to build up partner-specific name associations. Participants in the study played a game where they had to identify a partially obscured word after being prompted with a cue:
fill in B _ _ J O after partner1 says "a musical instrument"
fill in R _ _ I N after partner2 says "a kind of bird"
Then, for the main task, the participants had to do some picture naming. What varied was the presence/absence of the associated partner. The results showed faster picture naming if the associated partner was present in the room (*present*, not even part of the communication task). So if the participants saw the following object highlighted in the array of objects, they were faster to say "banjo" if partner1 was in the room and if banjo was the word that partner1 gave them the cue for in the fill-in-the-blank pre-task. The effect was apparent for associations between specific partner~label pairs (e.g., partner1~banjo, partner2~robin). Interestingly, the effect was also apparent for even broader associations between partners and object categories (e.g., partner1~instruments, partner2~birds)
Horton's results suggest that domain-general memory processes can increase the accessibility of information associated with specific partners.
If domain-general memory functions do indeed play a role in audience design, then individuals with memory impairment should fail to benefit from common ground. Who has memory impairments? Often the elderly. This is what Horton and Spieler (2007) investigated with the following results. For young adult speakers, there was evidence for audience design such that descriptions of objects for familiar addresses were shorter, initiated more quickly, and more similar to previously established descriptions compared to descriptions for unfamiliar addressees. Older speakers, however, showed less sensitivity to the status of the addressees. They referred to objects as if everything were "new" for both familiar/unfamiliar partners. They produced relatively long, idiosyncratic descriptions for both familiar and unfamiliar addressees.
As a last note, what I particularly like about the Horton and Spieler study is that it's an example of the use of a unique population to test predictions of competing models. Studies with specific populations sometimes risk asking the mere question "how does language ability X function in group Y? (e.g., "can children use implicature and at what age?"), but it furthers our understanding of language more to be able to test predictions -- in this case, the predictions made by a model of audience design which posits a crucial role for memory capacity.
To go on to section 7.5 on "Uniform Information Density", click here.