VIRTUAL REALITY AUTHORING FOR SCHOOLS Raul Sidnei Wazlawick i , Tereza Kirner ii , Lea da Cruz Fagundes iii , Bernd Heinrich Storb i i Federal University of Santa Catarina, Florianópolis, Brazil ii São Carlos Research Center, São Carlos, Brazil iii Federal University of Rio Grande do Sul, Porto Alegre, Brazil Introduction Many authors have identified Virtual Reality (VR) authoring as a promise for educational purposes (Youngblut, 1998). Nevertheless, due to the complexities involved in authoring virtual worlds, most applications consist merely in predefined worlds where students can walk (or fly) and, in some cases, do some kind of interaction (Allison, et al., 1997; Dean, et al., 2000; Salzman, et al., 1999). Virtual Reality authoring in schools is still a challenge. In this paper, we explore some of the research themes on that field that are addressed by the museuVirtual project developed in Brazil. The museuVirtual project aims to study and develop virtual reality authoring tools to be used in schools. This project started in 1999 and involved two Brazilian universities, one research centre and six schools. Wazlawick, Kirner and Fagundes (2001) reported a number of results when the project finished in 2001. Many prototypes were implemented. One of them, built in the squeak- Smalltalk environment, allows scenario authoring and behaviour definition. This prototype generates VRML code to define scenarios, and Smalltalk code to define behaviour. The combined VRML/Smalltalk code can be, then, loaded to a server that runs the environment. In this paper we summarize some results of that project, especially the use of design patterns to allow non-expert users to create behaviour for VR dynamic objects. VR Authoring When we think about VR authoring, the following challenges come to mind: a) Creating 3D objects, b) Placing 3D objects on a scene, c) Programming reactive behaviour of 3D objects, d) Programming proactive behaviour of 3D objects, e) If available, programming chatterbots to answer questions about the environment. Then, the question is: how to provide authoring tools with enough power to deal with those tasks, and still being easy to use and understand, so that they can be used by non-specialists, such as, for instance, school students. The literature addressed exhaustively questions related to authoring 3D objects and scenes. However, behaviour authoring is a quite more complicated task. It is very hard for a non-expert user to use the VRML nodes to create complex (an even simple) behaviour for 3D objects. Then, we have to explore more