ABSTRACT
In Pfister & Mühlpfordt (2002) a study was presented showing that chat discussions with a strict turn order combined with the requirement to assign a type and an explicit reference to each message lead to a higher learning score than discussions in a normal chat or in a chat with strict turn order only. Due to the experimental design it was not possible to judge the role of explicit referencing. Now we present the "missing" data: The higher learning score can be explained just by the explicit referencing. We argue that this is an important design issue for chat applications, because it seems that explicit referencing leads to a more homogeneous discourse behavior (more homogeneous participation, more participation in parallel discussion threads) and a better grounding. A case study explored the use of the referencing function in a less restricted everyday collaborative situation.
