differential branching to other segments of instruction or to drill-and-practice routines . Although this basic behavioral model continues to dominate mainstream educational applications of computers such as integrated learning systems ( Bailey , 1992 ), interactivity in some of today ’ s most innovative applications , such as constructivist learning environments ( Wilson , 1996 ), is based upon advances in cognitive psychology and constructivist pedagogy ( Coley et al ., 1997 ) ( see Section Three of this report ).
Research Results
The good news is that even with a primarily behavioral pedagogy , computers as tutors have positive effects on learning as measured by standardized achievement tests , are more motivating for students , are accepted by more teachers than other technologies , and are widely supported by administrators , parents , politicians , and the public in general ( Coley et al ., 1997 ; President ' s Committee of Advisors on Science and Technology , 1997 ). These conclusions about the effectiveness of computers in classrooms in the USA are in agreement with the conclusions of similar reports in Australia ( Directorate of School Education , 1994 ), Canada ( Bracewell & Laferriére , 1996 ), and the United Kingdom ( Department for Education and Employment , 1996 , 1997 ). Regrettably , the impacts of CBI in countries such as Brazil ( Chaves , 1993 ), Chile ( Oteiza , 1993 ), China ( Makrakis & Yuan-tu , 1993 ), and Malaysia ( Shahdan , 1993 ) are less clear .
Meta-analyses of research examining the effectiveness of computers in schools have illuminated their advantages and limitations ( Kulik & Kulik , 1987 , 1991 ). Meta-analysis is a statistical procedure that allows researchers to synthesize the results of numerous research studies comparing different treatments ( e . g ., CBI versus teacher-centered instruction ) by reducing the results to a common “ effect size ” ( Hunt , 1997 ). Although some proponents of computer-based instruction have promised a 2.0 effect size ( representing an improvement of two standard deviations between CBI and traditional instruction ), a more reasonable expectation is in the range of 0.5 to 1.0 . The results of meta-analyses for computer-based instruction show an interesting pattern , as illustrated in Table 1 .
Table 1 . Effect sizes for CBI computed from meta-analysis studies . Academic Level Effect Size Source Elementary School 0.47 ( Kulik , Kulik , & Bangert-Drowns , 1985 ) Secondary School 0.36 ( Bangert-Drowns , Kulik , & Kulik , 1985 ) College and University 0.26 ( Kulik & Kulik , 1986 ) Adult Basic Education 0.42 ( Kulik , Kulik , & Schwab , 1986 )
As indicated in Table 1 , the effectiveness of CBI over traditional teacher-centered instructional methods appears to decline as the level of education increases with
11