During the past fifty years many educational researchers have investigated the effectiveness of alternative methods of instruction in science. These investigations have resulted in the publication of numerous studies which frequently report contradictory findings. The purpose of this study was to synthesize the studies conducted on individualized instruction in secondary school and collegiate science courses through a meta-analysis, the statistical integration and analysis of research studies. The major goals of the meta-analysis were to determine the experimental effect of individualized instruction in science and to explain the variation in the effect sizes obtained.

This dissertation included an analysis of 115 studies comparing individualized instruction with traditional lecture instruction. The methods of individualized instruction incorporated were audio-tutorial instruction, computer-assisted instruction, Keller's personalized system of instruction and programmed instruction. Instruction which was individualized but not clearly identified as one of the previously listed methods was included under a combination category. The effectiveness of individualized instruction was measured in terms of effect size, a standardized difference in means between the individualized and the lecture classes. Effect sizes were calculated using the pooled standard deviation.

A number of instructional, design and miscellaneous variables were identified in the research studies. Instructional variables identified were mastery orientation, self-initiated testing, self-pacing of instruction, and choice of instructional delivery systems. Design variables identified were self-selection, historical effect, subjective rating of study, instrument development, equivalence of subjects, and instrument reliability. Miscellaneous variables identified were level of instruction, nature of instruction, source of publication, year of publication, setting of the study, and subject of instruction.

Data analysis involved: (a) descriptive statistics for the entire data set, (b) descriptive statistics for each of the individualized instructional methods studied, (c) analysis of variance and t-tests to test the effect of selected variables on instructional effectiveness, and (d) regression analysis in which effect sizes were regressed onto selected independent variables.

The findings of this meta-analysis suggest that individualized instruction in science is somewhat more effective than traditional instruction. All five methods of individualized instruction analyzed were not differentially effective when compared to traditional instruction. The instructional, design, and miscellaneous variables identified in this meta-analysis contributed little to the variance in effect size.