There are a number of theories concerning how (or whether) people establish common ground in a conversation. We briefly review these, and then present several experiments relevant to Clark's Collaborative Communication theory. According to Clark, participants in a conversation constantly monitor one another at multiple levels to ensure common ground; the conversation does not go forward until each participant is satisfied that both are in agreement regarding what ought to be shared understandings. A conversation is thus a series of joint actions in which the participants continually negotiate and assess comprehension.
One of the strongest supports for this theory is the Overhearer Deficit, reported by Schober and Clark (1989). In their version of the referential communication task, two people attempt to select and arrange tangram figures to match that of a third person known as the Director. One of these, the Matcher, can speak to the third person , but the other (the Overhearer) can only listen in on the conversation, rather than contributing to it. In accordance with Shober and Clark's claim that they have not negotiated common ground with the Director, Overhearers place fewer figures correctly.
The current experiments explore several themes related to this finding, and to Clark's model. The first is the extent to which language is public versus private. Overhearer deficits imply a private language not readily comprehended by non-participants, as opposed to (less interactive) public language such as speeches, lectures, and monologues. We port a version of the Schober and Clark task to a computer-mediated environment. Given that understanding in spoken language is signaled by verbal and non-verbal cues in spoken discourse (backchannels; eye contact; body posture; etc.), the lack of these in computer-mediated discourse ought to result in greater difficulties in comprehension. If participants are aware of these difficulties and compensate for them, then we would expect computer-mediated language to be more public. A natural prediction is thus a reduced Overhearer deficit compared to a control condition in which Directors and Matchers speak with one another.
The second theme involves a possible alternative explanation for the Overhearer deficit. We can term this explanation the Linguistic Precision Hypothesis (LPH), and trace its roots back to a weak version of Whorfian Linguistic Determinism. According to LPH, the process of encoding semantic content into linguistic form requires more effort, and results in a more precisely formulated meaning. Speakers will need to clarify their thoughts, and subject them to possible challenge by others, resulting in yet further modification and clarification. Schober and Clark's Overhearers did not speak, and so we explore whether part of the Overhearer deficit might be due to LPH, rather than a failure to establish common ground. To test this possibility, we compare the performance of strict Overhearers to Overhearers we will refer to as Chatters. Chatters can sidebar with one another to help one another understand the conversation of the Director and Matcher. LPH predicts that Chatters ought to perform better than Overhearers.
A final theme concerns the factor of co-presence. In some of our experiments, Overhearers and Chatters work in real time with a Director-Matcher pair; in others, they work from a transcript of an earlier conversation. Are there motivational factors having to do with co-presence that influence performance on the referential communication task?