As the use of Web-based instruction increases in the educational and training domains, many people have recognized the importance of evaluating its effects on student outcomes such as learning, performance, and satisfaction. Often, these results are compared to those of conventional classroom instruction in order to determine which method is "better." However, major differences in technology and presentation rather than instructional content can obscure the true relationship between Web-based instruction and these outcomes. Computer-based instruction (CBI), with more features similar to Web-based instruction, may be a more appropriate benchmark than conventional classroom instruction. Furthermore, there is little consensus as to what variables should be examined or what measures of learning are the most appropriate, making comparisons between studies difficult and inconclusive. In this article, we review the historical findings of CBI as an appropriate benchmark to Web-based instruction. In addition, we review 47 reports of evaluations of Web-based courses in higher education published between 1996 and 2002. A tabulation of the documented findings into eight characteristics is offered, along with our assessments of the experimental designs, effect sizes, and the degree to which the evaluations incorporated features unique to Web-based instruction.
Keywords: Web-based courses; benchmarks; CBI
Authors' NoteWe would like to caution readers against drawing inappropriate cause-and-effect conclusions from the results presented in this paper. A goal of this paper is to present data based on existing empirical literature. As demonstrated by our finding of only one study in which random assignment of individuals to conditions occurred, there is a tremendous need for experimental studies of Web-based instruction if we are to draw any definitive conclusions about the effectiveness of Web-based instruction in comparison to other methods. The conclusions that one can draw from the results of a meta-analysis are highly dependent on the research designs of the individual studies examined. For example, if all the studies included are corelational, as were many of the studies reviewed in this paper, associational rather than causal conclusions are more appropriate. In this paper, the terms "effect," "effect sizes," and "effectiveness" are used in accordance withOlson & Wisher -The Effectiveness of Web-Based Instruction: An initial inquiry 2 the accepted technical language of meta-analysis and are not meant to necessarily imply the existence of causal relationships.
The Effectiveness of Web-Based Instruction: An initial inquiryThe World Wide Web can be used to provide instruction and instructional support. Webbased instruction offers learners unparalleled access to instructional resources, far surpassing the reach of the traditional classroom. It also makes possible learning experiences that are open, flexible, and distributed, providing opportunities for engaging, interactive, and efficient instruction (Kahn, 2001). Phra...