This study examined the impacts of educational interface agents with different attributes on achievement, attitude and retention of elementary school students in their science and technology courses. The study was implemented in four different eighth- grade classes (aged 13-14) of an elementary school. Four different types of educational software, covering living things and life unit of 8th graders, were developed to analyze the impacts of educational software agents with different attributes on the results of learning. The study was conducted according to experimental model with pretest and posttest control group, which is one of the quasi-experimental designs. For five consecutive weeks experimental and control group students used the software designed for this research. The students were asked to fill an attitude scale on science and technology both before and after the implementation. And following completion, an achievement test was applied. The results of the study revealed that students who used a human-like educational interface agent were more successful than both the rest of the experimental groups and the control group in terms of achievement, attitude and retention of learning in science and technology classes. It is also revealed that the software to be developed for elementary school 8th graders (aged 13-14) should be supported with human-like educational interface agents. (Contains 5 tables and 4 figures.)