MSU LIBRARIES RETURNING MATERIALS: PIace in book drop to remove this checkout from your record. FINES wiII be charged if book is returned after the date stamped beIow. THE IMPACT OF STUDENT LOCUS OF CONTROL ON ACADEMIC ACHIEVEMENT AS A FUNCTION OF LECTURE VERSUS COMPUTER-ASSISTED INSTRUCTION By Gregory Chase Hamilton A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of College and University Administration 1984 ABSTRACT THE IMPACT OF STUDENT LOCUS OF CONTROL ON ACADEMIC ACHIEVEMENT AS A FUNCTION OF LECTURE VERSUS COMPUTER-ASSISTED INSTRUCTION By Gregory Chase Hamilton The infusion of computers into the educational system raises questions regarding the appropriate use of such technology. This study, formulated in the aptitude— treatment interaction model, investigated the functional dependence of academic achievement on the personality vari— able of locus of control. The study contrasted computer- assisted instruction with traditional lecture. The literary review providing background support was drawn from three areas: 1) locus of control; 2) computer— based instruction; and 3) aptitude-treatment interaction studies. The intent was to document the contribution these three areas have made in academic achievement. The study utilized two independent variables: 1) locus Of control (internal/external); and 2) instructional method- ology (CAI/lecture). The Intellectual Achievement Responsi- bility (IAR) Questionnaire was used as the measure of locus 0f control serving to identify the internally and externally oriented students. The dependent measure was academic Gregory Chase Hamilton performance on a teacher-made test covering topics in College Algebra. Subjects were 51 students enrolled in two intact classes (32-Winter 1983, 19-Spring 1983) of College Algebra. These subjects were predominantly Black, freshmen students enrolled in the Michigan State University College of Engin- eering. The IAR Questionnaire was administered to a larger group of predominantly white engineering students as a means of assessing representativeness of the experimental samples. The 2-by-2 design was analyzed using variance, covari- ance, and linear regression techniques. Analysis of covari— ance was performed using the covariates of MSU math place- ment score, ACT math score, formal instruction time, exter- nal study time, home work grade, quiz score, course test scores, and the previous math course grade. Analyses were performed for each trial separately and for a combined sample. The study found no significant differences in achieve- ment for locus of control orientations or for instructional methods. No significant interactions were found. Analysis Of covariance revealed one significant main effect for instructional method when the MSU math placement score was controlled. Linear regression analysis indicated that treatment regression lines were statistically parallel with Slopes equivalent to zero. Although the lecture method tended to produce higher achievement scores, the CAI method reQuired 422 less instructional time. ACKNOWLEDGMENTS This research effort would not have been possible with- out the assistance and cooperation of several very important individuals. First and foremost is my wife, Joanne, whose love, understanding, and support were constant sources of encouragement and motivation. Her patience throughout the entire degree program and her editorial skill in the pre— paration of this dissertation were invaluable and greatly appreciated. Drs. Richard Featherstone, Frederick Ignatovitch, and Stephen Yelon, as members of the guidance committee, pro- vided valuable insights into the design and analysis aspects of this study. Their critical review and comments were both beneficial and educational. I would like to especially thank Dr. Howard Hickey who, as committee chair, provided not only professional expertise but acted as both mentor and friend--I am thankful for both. I would also like to thank two colleagues, Thad Roppel and Pam Reisner. Thad assisted in the development of several sections of the CAI software system and served as a sounding board for many of the computer applications used in this study. Pam contributed to the design of the courseware materials and acted as the classroom lecturer. Finally, I ii wish to recognize members of the College of Engineering and the Office of Minority Student Education who supported this project through their contribution of financial and physical resources . iii TABLE OF CONTENTS Page LIST OF TABLES. . . . . . . . . . . . . . . . . . . . . vii LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . viii CHAPTER I: INTRODUCTION. . . . . . . . . . . . . . . . 1 Identification of the Problem . Definition of Terms . . . . . . Purpose of the Study. . . . . . Statement of Research Questions Importance. . . . . . . . . . . Generalizability. . . . . . . . Limitations . . . . . . . . . . o O o O o 0 o 0 o o o o o o o o o O O O 0 0 O O o O o O o 0 o o O C O O O O O O O O O O o O o O o o o o o o o o H # Overview of Subsequent Chapters 19 CHAPTER II: REVIEW OF THE LITERATURE . . . . . . . . . 21 Introduction. . . . . . . . . . . . . . . . . . . 21 Student Locus of Control. . . . . . . . . . . . . 22 Theoretical Background . . . . . . . . . . . 22 Assessment Measures. . . . . . . . . . . . . 29 Ethnicity. . . . . . . . . . . . . . . . . . 30 Sex. . . . . . . . . . . . . . . . . . . . . 37 Anxiety. . . . . . . . . . . . . . . . . . . 41 Summary. . . . . . . . . . . . . . . . . . . 45 Computer-Based Instruction. . . . . . . . . . . . 47 Historical Overview. . . . . . . . . . . . . 47 Hardware Considerations. . . . . . . . . . . 54 Software Capabilities. . . . . . . . . . . . 57 Comparative Studies. . . . . . . . . . . . . 65 Strategies for Optimizing Instruction. . . . 72 Feedback Options . . . . . . . . . . . . . . 82 Anxiety and CAI. . . . . . . . . . . . . . . 85 Summary. . . . . . . . . . . . . . . . . . . 88 Aptitude-Treatment Interactions . . . . . . . . . 90 Research Methodology . . . . . . . . . . . . 9O Ability by Treatment Interactions. . . . . . 96 Personality by Treatment Interactions. . . . 103 summary. 0 O O O O O O O O O O O O O O I O O 109 (3HAPTER III: DESIGN AND PROCEDURES . . . . . . . . . . 114 Introduction. . . . . . . . . . . . . . . . . . . 114 Research Questions. . . . . . . . . . . . . . . . 114 Research Hypotheses . . . . . . . . . . . . . . . 115 iv Design. . . . . . . . . . . . . . . . . . . . . . 118 CAI System Design. . . . . . . . . . . . . . 118 Courseware Design. . . . . . . . . . . . . . 120 Research Design Over Time. . . . . . . . . . 123 Variable Selection and Analysis Techniques . 124 Significance Level . . . . . . . . . . . . . 128 Reliability and Validity Conc cerns. . . . . . 129 Predictions . . . . . . . . . . . . . . . . . . . 133 Procedures. . . . . . . . . . . . . . . . . . . . 136 Population and Sample. . . . . . . . . . . . 136 Treatment Description. . . . . . . . . . . 139 Instrumentation and Data Collec ti ion. . . . . 141 Achievement Measure. . . . . . . . . . . . . 143 Additional Concerns. . . . . . . . . . . . . 144 Summary . . . . . . . . . . . . . . . . . . . . . 145 CHAPTER IV: ANALYSIS OF THE DATA . . . . . . . . . . . 147 Introduction. . . . . . . . . . . . . . . . . . . 147 Sample Representativeness . . . . . . . . . . . . 147 Treatment Assignments . . . . . . . . . . . . . . 150 Combining Samples . . . . . . . . . . . . . . . . 152 Analysis of Variance. . . . . . . . . . . . . . . 154 Analysis of Covariance. . . . . . . . . . . . . . 161 Regression Analysis . . . . . . . . . . . . . . . 166 Johnson-Neyman Procedure. . . . . . . . . . . . . 173 Stability Across Samples. . . . . . . . . . . . . 174 Summary . . . . . . . . . . . . . . . . . . . . . 177 CHAPTER V: SUMMARY AND CONCLUSIONS . . . . . . . . . . 181 Introduction. . . . . . . . . . . . . . . . . . 181 Overview of the Study . . . . . . . . . . . . . . 181 Discussion of Research Questions. . . . . . . . . 184 Reflections and Observations. . . . . . . . . . . 195 Implications for Continued Research . . . . . . . 200 APPENDIX A: COMPUTER-ASSISTED INSTRUCTION SYSTEM DESCRIPTION 0 O O O C O O C O O O O O O 207 Overview. . . . . . . . . . . . . . . . . . . . . 207 Software Programs . . . . . . . . . . . . . . . . 209 Data File Contents. . . . . . . . . . . . . . . . 210 Organizational Chart. . . . . . . . . . . . . . . 211 Flowchart and Data Linkages . . . . . . . . . . . 212 Text Processing Commands. . . . . . . . . . . . . 213 Examples. . . . . . . . . . . . . . . . . . . . . 215 APPENDIX B: UNIVERSITY CORRESPONDENCE. . . . . . . Memoranda to University Committee on Research Involving Human Subjects (UCRIHS). . . . UCRHIS Response . . . . . . . . . . . . . . . Memorandum to Committee on Release of Confidential Information . . . . . . . . . Committee on Release of Confidential Information Response . . . . . . . . . . . APPENDIX C: RESEARCH FORMS AND QUESTIONNAIRES. . . Research Project Consent Form . . . . . . . . Intellectual Achievement Responsibility (IAR) Questionnaire. . . . . . . . . . . . . . . Math 108 Research Study Time Survey . . . . . APPENDIX D: ACHIEVEMENT INSTRUMENTS. . . . . . . . Instructional Objectives. . . . . . . . . . . Winter Trial Exam and Statistics. . . . . . . Spring Trial Exam and Statistics. . . . . . . APPENDIX E: JOHNSON—NEYMAN PROCEDURE . . . . . . . Introduction and Discussion . . . . . . . . . Calculations for 1+ Composite Data. . . . . . Calculations for 1- Composite Data. . . . . . REFERENCE NOTES 0 O O O O O O I O O O O O O O O O O BIBLIOGRAPHY. O O O O O O O O O O O O O O O O O O 0 vi 217 217 220 221 222 223 223 224 229 230 230 231 236 241 241 248 252 256 257 p ¢ L.“ 12. 13. 14. 15. 16. 17. 18. LIST OF TABLES Analysis of Variance--IAR Scores Winter, Spring, and MMM 160 Groups. . . . . . . . Treatment Assignments, Final Distribution . . . . Sample Combination, t-test Summary Statistics IAR and Achievement Scores. . . . . . . . . . . . Design Matrix, Achievement, Winter 1983 Trial . . ANOVA Results, Achievement, Winter 1983 Trial . . Design Matrix, Achievement, Spring 1983 Trial . . ANOVA Results, Achievement, Spring 1983 Trial . . Design Matrix, Achievement, Composite Sample. . . ANOVA Results, Achievement, Composite Sample. . . Achievement ANCOVA Results, Primary Covariates Winter, Spring, and Composite Groups. . . . . . . Achievement ANCOVA Results, Secondary Covariates Winter, Spring, and Composite Groups. . . . . . . Linear Regression Coefficients Achievement Versus Locus of Control Scales Winter, Spring, and Composite Groups. . . . . . . Test of Parallelism, 1+ and I- Regression Equations, Composite Sample . . . . . . . . . . . Mean Achievement Scores, Students Selected for CAI Electing Lecture Method . . . . . . . . . . . Illustration of the Johnson-Neyman Procedure: Equations and Solutions for Regions of Significance, 1+ and I- Regressions . . . . . . . Locus of Control Conditions Predicting Lecture Agaignment. O O O O O O O O O O O O O O O O O O O I+ Composite Data Work Sheet. . . . . . . . . . . I- Composite Data Work Sheet. . . . . . . . . . . vii Page 149 151 153 155 155 156 156 157 157 162 163 167 173 176 243 246 248 252 Figure LIST OF FIGURES Research Design Matrix. . . . . . . . . . Achievement Means by Design Matrix Cell Winter, Spring, and Composite Groups. . . Regression Lines for 1+ Subscale Composite Sample. . . . . . . . . . . . . Regression Lines for I- Subscale Composite Sample. . . . . . . . . . . . . Region of Significance for 1+ Subscale Composite Sample Illustration of Johnson-Neyman Procedure. Region of Significance for I- Subscale Composite Sample Illustration of Johnson-Neyman Procedure. viii Page 125 160 169 170 244 245 ‘ A‘- E . IE '1' b .13 ‘F .4 a: 9x, _ CHAPTER I INTRODUCTION The educational system in the United States is in the midst of a revolution. Although this revolution is quiet and benevolent it will nonetheless leave few educators and administrators untouched. Ashby (1967) has identified this revolution as the fourth to affect education. The organiza- tion of schools wherein verbal instruction, exemplified by Socratic dialogue, was the delivery mode represented the first revolution. The second and third revolutions are characterized by their use of hand written and printed materials respectively. The fourth and current revolution is distinguished by its use of electronic technology. Ashby (1967) states that: ...new technologies are being adopted in teaching which will certainly transform the whole process of education, though what the transformation will be is still a matter for speculation. (p. 361) Indeed, the use of electronic media has been slowly incor- porated into the educational system over the past four decades. The advent of computer technology and specifically microcomputer technology, however, has brought the revolu- tion into a new focus and perspective. The Carnegie Commission (1972) addressed the impact of electronics on the educational system and made recommen- dations that would promote and enhance the acceptance, applicability, and use of the tools of the fourth educa- tional revolution. The authors of that report did not forsee the development of microcomputers in 1975. Since that time, inexpensive and surprisingly powerful computer systems have become obtainable for a few hundred dollars. Educational theorists have also been developing method- ologies for effective learning. The mastery learning tech- nique advocated by Bloom (1971) and the Personalized System of Instruction approach of Keller (1968) emphasize the development of learning packets designed for individual use. The intent of both methods is to allow students to learn and master instructional materials at their own pace. These two approaches- are among several that promote learning in a flexible and individualistic manner. These models address the issue of individual differences between students by restructuring the educational materials and environment to suit the learners' needs. In addition to advancements in technology and learning theory, a third factor impacting on the educational system has been the continued research into how affective charac- teristics contribute to academic achievement. Most notable among these is the aptitude-treatment interaction research. This research attempts to determine the relationship between affective or behavioral variables and instructional methods, educational environment, and academic achievement. The philosophy is that there is no one best method for teaching. Indeed, the best method may be a combination of several available techniques. Again, the issue of individual dif— ferences is addressed but from a perspective that tries to match students to the appropriate instructional method on a situational basis. The development of the microcomputer has, in some quar- ters, been heartily embraced as the tool for truly indivi- dualizing instruction. Computer-assisted instruction (CAI) has made significant inroads into the pre-college educa— tional system. It is also heavily used in business and industry and the military for job training and advancement. Additionally, there is small but growing use of computers for the preparation and delivery of instruction at the col- legiate level. Other educators view computers as a real threat to their jobs, authority, and freedom. The majority of teachers, however, are taking a wait and see approach. Oettinger and Marks (1969) do not consider computers as either a boon or a boondoggle. They will take their place in the educational system along with the carousel slides, movie projectors, and video tape players. Computer systems, however, are perceived by some as a threatening and controlling influence. The "big brother" concept is readily associated with computers and the fact that computer storage systems can record seemingly every- thing reinforces the perception of power and control. The issue in question is how an individual's sense of personal control determines subsequent outcomes in different situa- tions. The concept of locus of control, formally defined by Rotter (1954), provides a means for investigating student performance in the novel environment of CAI. Do students believing their successes and/or failures are primarily due to external controlling factors have different levels of achievement than individuals who assume personal responsi- bility for their successes and/or failures? In what way does the locus of control factor interact with instructional methodology, particularly CAI? Locus of control does appear to be one personality variable which merits investigation relevant to its interaction with instructional methodology and the combined effect on achievement. Identification of the Problem The challenge facing educators is to define, shape, and guide the impending transformation of the educational sys— tem. This involves understanding the capabilities of the new computer technology and its potential use in education. It transcends simply designing computer-based instructional materials for it includes determining when, under what cir- cumstances, and for whom the use of computers is warranted. The classroom in the year 2000 will be vastly different from what it is today. By that time the fourth educational revo— lution will be over. What happens now, during the revolu- tion, will have a profound impact upon the quality of the future educational system. Research into the means, methods, and necessity of computerizing education needs careful investigation and YES C" AI p b resolution. If computer usage in education is viewed on a continuum with total absence and total implemetation at the extremes, the global issue is to determine the optimal usage of the technology. To adopt a Luddite philosophy and ignore the computer and its potential use for the development, delivery, and management of instruction is impractical. This is a real possibility if the design of hardware systems is inadequate and, more importantly, if the quality of the instructional software is inferior to the more conventional printed materials. The progression is clear. Verbal instruction became enhanced with the advent of hand written materials that could be shared between individuals. The development of the Gutenberg press allowed for the mass pro- duction and widespread distribution of information. Compu- ter technology therefore, can and should be utilized in the management of educational materials, resources and facili- ties, and in the actual delivery of instruction. Embracing the opposite extreme by viewing computer technology as a panacea to current educational problems is also impractical. The educational system, described by Kozma, Belle, and Williams (1978), is an integration of instructors, students, the subject matter, and instructional delivery methods--all within the boundaries of a learning environment. Computer technology alone is not the answer to current problems in any one of these areas. If viewed as a multidimensional space, the problem becomes at once more complex (adding more dimensions) but also more manageable by investigating the optimal effectiveness (cost, resource uti- lization, learning, time reduction, etc.) of computer tech- nology along any particular dimension. These dimensions are not independent, however, and any two can be thought of as defining a plane in the complex educational space. As one moves off the axes that define a plane, the issue becomes one of the interaction between the two dimensions. This study will consider the dimensions of the student, the instructional delivery method, and their interaction. The plane formed by these two variables pro- vides research opportunities for identifying the student characteristic-instructional methodology matches that enhance learning. The specific problem considered herein is to determine the extent that student locus of control can be used to predict academic achievement based on the avalabil- ity of lecture and computer-assisted instructional method- ologies. Definition of Terms Many of the terms in this study encompass a broad range of activities as reported in the literature. Salisbury (1971) listed 21 terms synonymous to computer-assisted instruction in his attempt to standardize the terminology associated with the usage of computers in education. More recently, Burke (1982) provides an excellent glossary of the terminology used in this rapidly expanding area. The defi- nitions listed here, however, are solely for the purpose of this study. Artificial Intelligence (AI): A subfield of computer science concerned with the concepts and methods of symbolic inference by a computer and the symbolic representation of the knowledge to be used in making inferences. A computer can be made to behave in ways that humans recognize as "intelligent" behavior in each other (Fiegenbaum & McCor- duck, 1983). Aptitude-Treatment Interaction (ATI): An ATI occurs whenever the regression line of the outcome from one treat- ment, based upon some kind of information about the stu— dent's pretreatment characteristics, differs in slope from the regression line of the outcome from any other treatment (Cronbach & Snow, 1977). In this study the two treatments are lecture and computer-assisted instruction and the pretreatment characteristic is student locus of control. The intent of ATI research is to generate predictive models (regression equations) that reflect the interdependance of student characteristics and instructional methods. Branching, Program: Any program which uses built-in branching. Branching is usually designed to allow some students to bypass some of the material of the program based on their performance in the program to that point (Burke, 1982). Also known as intrinsic, Crowderian, or scrambled programs. in: 50f 501 St: is g. ”‘3. Computer—Assisted Instruction (CAI): Any method of learning in which a computer is the primary delivery system (Burke, 1982). The term has evolved to imply direct, on- line, interactive use of computer facilities. Also known as Computer-Assisted Learning (CAL) or Computer-Assisted Train- ing (CAT). Computer-Based Instruction (CBI): Nearly synonymous with CAI. However, some people may reserve the term CBI for cases in which there is less direct interaction with the computer (Burke, 1982). The term often represents the gen- esis of all computer related activities in education. Also referred to as Computer—Based Education (CBE). Computer-Managed Instruction (CMI): CMI has come to mean the systematic control of instruction by the computer. It is characterized by testing, diagnosis, learning pre- scriptions, and thorough record keeping (Burke, 1982). Courseware: It has become popular to refer to CAI les— sons as courseware rather than software (Burke, 1982). The instructional materials specifically prepared for delivery by a computer system will be considered courseware. The particular design of the instructional system used in this study necessitates this distinction between courseware and software. Disordinal Interaction: One of the possible outcomes of aptitude-treatment interaction research. An interaction is disordinal if the regression lines for the different treatments cross within the range of the aptitude variable Iq- I ‘14 9 ma u A, KL- 9:“) 1831 (Cronbach & Snow, 1977). This type of interaction indicates that a differential application of treatments based on apti- tude measure may be of benefit. Hardware: The physical parts or components of a compu- ter system. Hardware includes the electronic elements, chips, keyboard, video monitor or VDT, and perpherial devices such as printers, disk drives, and telephone modems (Toong & Gupta, 1982). Interactive: A term which describes a learning process in which the student and the system alternate in addressing each other. Typically, each is capable of selecting alter- native actions based on the actions of the other (Burke, 1982). Linear Prggram: A program which contains little or no branching. In other words, every student who goes through the lesson sees exactly the same information and questions. The logical branching which occurs in providing feedback to either a correct response or an incorrect response does not disqualify a lesson as linear (Burke, 1982). Programs of this type are also known as extrinsic programs. Microcomputer: A small machine that receives, stores, manipulates, and communicates information under the direc- tion or control of a microprocessor. A microprocessor is a single integrated circuit on a square chip of silicon that is typically a quarter of an inch on a side (Toong & Gupta, 1982). 10 Ordinal Interaction: One of the possible outcomes of ATI research. In an ordinal interaction one regression line remains above but not parallel to the other regression line (Cronbach & Snow, 1977). If the two regression lines are parallel to each other then there is no interaction between treatments and characteristics. This type of interaction may indicate a differential application of treatments but not based solely on the aptitude variable. Programmed Instruction (PI): Instructional materials that use the principles of programmed instruction: 1) small steps, 2) active responding, and 3) prompt feedback (Burke, 1982). Software: The sequence of instructions that direct the hardware components to perform a certain task is known as the software. The core of the software is an operating sys- tem that controls computer operations and manages the flow of information. Software typically refers to programs or application programs which enable the computer to perform specific tasks (Toong & Gupta, 1982). The model used in this study distinguishes software as the data processing program from courseware which is the data being processed. Student Locus of Control (LOC): An affective variable indicating the degree to which an individual perceives that reinforcements are contingent upon his/her own behavior. Internal individuals assume responsibility for subsequent events while externally oriented subjects feel powerless and Dr 11 often manipulated by others. A more formal definition and discussion of this concept is provided in the review of the literature. Purpose of the Study Research into the three basic components of this study-~computer-based education, role of affective charac— teristics in educational achievement, and aptitude-treatment interactions--has proceeded largely in isolation. Much of the research into affective variables has focused on the identification of such variables, development of reliable assessment procedures, and correlation with other affective measures. Aptitude-treatment interaction studies have been overshadowed by correlational studies that attempt to show significant treatment effects. Of the aptitude-treatment interactional studies that have been performed very few involve computer-assisted instruction as one of the treat- ments. The major reason for this is that the availability of inexpensive microcomputer systems is relatively new and, therefore, still subject to the correlational or "one best method” type of research. The literature is replete with studies that compare CAI to traditional or other non tradi- tional (PSI, PI, etc.) methods. The particular research method used in most of these studies attempts to determine which of the treatments provides for more effective learning for all students. The availabiltiy of microcomputer systems 5:1; 5‘' 5- :10 tio: iflde 50:5 1385 12 has simply changed the equipment used in the research and not the research approach. Empirical evidence on the relationship between the con- tent and structure of computer-based instructional mate- rials, the procedures by which computers can or will be per— mitted to modify those materials, and the short and long term effects on the students is required for the proper design and utilization of the new technology. Regardless of the advances made in the technological area, computer-based education is not necessarily the best method for teaching all students. Until educators can understand the relation— ships between computers, instructors, students, and the structure of the subject matter, and can define the condi- tions for effective and efficient use of the emerging compu- ter technology, they run the risk of having the commercial entrepreneurs determine the quality of educational software and hardware. Indeed, there is a rapidly increasing market for ready to run microcomputer based instructional materials and this demand is only surpassed by the abundance of soft- ware vendors and products. It is the emergence of microcomputer technology that is enabling state-of-the-art educators to investigate the uses of these machines for the preparation and delivery of educa— tional materials on an individual basis. Microcomputers can indeed be directed to reconfigure units of instruction to be more attuned to the specific needs of the learner. However, questions arise as to the intended use and design of [in (I) p.1- in 13 instructional software as well as the quality and acceptance of the methodology itself. The foremost purpose of this study is to investigate the existence of an interaction between student locus of control and the method of delivering instruction. Addition- ally, analysis procedures will indicate if there is a dif- ference between delivery modes themselves. Moreover, the study consisted of two experimental trials as a means of replication and examination of the stability and generaliza- bility of any aptitude-treatment interaction. The two trials are identical in all respects except that different subjects were used. Statement of Research Questions The major emphasis of this study was to determine the extent that student locus of control predicts academic achievement based upon differences in the methods of instruction. Two distinct instructional methodologies are compared: lecture and computer-assisted instruction. The issue was not to determine which method is best but to determine whether the interaction of locus of control and instructional technique has a predictable relationship with achievement. Results of this and similar aptitude-treatment interaction studies will help decide both to what extent and in which manner computers should be used in instruction. The independent variables for the study consist of the two instructional models, lecture versus computer-assisted > —4 E4? fox de1 Sin 6 Us 14 instruction, and a student locus of control measure divided into internal and external subgroups. The dependent vari- able is academic achievement on the material covered during the experiment. Four main research questions are addressed. First, to what extent is achievement effected by the method of instruction? Second, is student locus of control 3 via- ble affective characteristic for predicting achievement? Third, are there significant interactions between locus of control and instructional method? Lastly, since there are two experimental trials using different subjects, are aptitude—treatment interactions generalizable across sam— ples? Importance The study is important for four reasons. First, computer-based education is neither a panacea nor a placebo for the problems confronting educators. It is important to determine how students interact with computers and what role student attributes play in the interaction. The aptitude- treatment interaction studies focus on generating regression equations for prediction purposes. The development of con- sistent and stable regression equations could play a key role in providing students with effective instruction geared to situation variables and their learning style, of which locus of control is one possible factor. The new technology has the capability to truly address individual differences for instructional purposes. 15 Secondly, the viability of locus of control as a suit- able student attribute for use in predictive regression equations needs to be determined. There is an abundance of affective variables that undoubtedly impact upon academic achievement and indeed the eventual form of effective regression equations may involve several of them. The cir- cumstances in which student locus of control is one of the components needs to be determined. Thirdly, the design of the computer-based materials is intimately related to the overall level of achievement. The quality of the instruction and the procedures by which the computer system manipulates, restructures, and presents the material warrant attention. Computer technology itself is advancing at an ever increasing rate and as newer and better computers become available, researchers will be able to develop new educational applications. The design of computer—based instruction should not automatically embrace the latest in electronic gadgetry. It must be developed from philosophical and theoretical bases and resist the almost seductive nature of electronic capabilities. Finally, the electronic revolution is proceeding with proponents claiming less expense, greater flexibility, opportunity for students to learn at their own pace and cog— nitive level, and less preparation time. It is important that potential users of the electronic hardware and software systems understand the capabilities and limitations of these sys OI] 16 systems. Educators and administrators will need information on the design and effectiveness of these systems in order to make prudent decisions regarding their adoption and utiliza- tion. Bell (1979) states that: ...personal computers with their low costs, easy accessibility, total dedication to the user, and person-on-the-street popularity may provide the long awaited catalyst that is needed to make some dramatic change in how computers are used in schools. (p. 70) Computer-based educational systems will not be adopted solely for the reasons listed above. They will be accepted on the basis of their capacity to provide high quality instruction, and personal computers will indeed provide the catalyst for the prerequisite research. This study repre- sents one step toward determining some of the ingredients necessary for the development of materials that are not only effective and efficient but also acceptable to the instruc- tors and students. Generalizability This study examines important issues for both the users and developers of computer-based instructional materials. The end users of these materials are the students. There- fore, recognition that affective characteristics influence academic achievement and design of materials that incorpor- ate these variables are issues that should be of concern to all educators. The locus of control attribute has been observed and measured on the basis of age, sex, race, and PI- III“ ali p 5 ca and has 511': tn t1? di '1: dc fie l7 nationality. The presence of locus of control as a person- ality trait suggests that the findings of this study can be applied across populations. Computer technology has at least a foothold in all edu- cational arenas and its use is not expected to decline. The results of this study, therefore, contribute to the growing pool of knowledge necessary for the integration of computers and education. This study also presents a philosophical basis from which instructional materials for a variety of subject areas and cognitive levels can be developed. Future trends regarding computers in education demand that this type of study be conducted. Other studies investigating different affective variables, computer applications, and implementation procedures also need to be performed. Limitations The study is limited to the investigation of only two main treatments: lecture and computer-assisted instruction. It is also limited to the single affective variable of stu- dent locus of control. Whether there exists a significiant interaction between these variables does not preclude the existence of significant interactions between these two treatments and other affective variables such as anxiety, impulsivity-reflectivity, or introversion-extroversion. Nor does it preclude the interaction of locus of control and methodologies other than lecture and computer-assisted instruction. 18 The study does not focus on the design and structure of the computer-assisted instructional materials or their delivery system. Computer-based materials can be prepared in any number of ways. They can be presented to the user exactly as they were prepared and entered into the machine or they can be written to require restructuring or configur- ation by the computer prior to presentation. Although research into the design of computer-based educational sys- tems is needed, this study uses the linear programming instructional format. As previously mentioned, the study was conducted twice. The first trial consisted of college students enrolled in a College Algebra course during the Winter 1983 term. A second trial was conducted the following term with a dif- ferent student sample but using the same instructional mate- rials and methods. The purpose of the dual trials was to investigate the consistency of the experimental findings. Additionally, the study is limited to the subject of mathe- matics which is highly adaptive to computerization. Gener- alizing the results of this study to other academic subjects is cautioned. The student sample used consisted of minority students enrolled at Michigan State University. The combined sample consisted of 51 predominantely Black freshmen students enrolled in the College of Engineering. The fact that the sample was composed of a small, narrowly defined group of 19 students could possibly limit the generalizability of the findings. Unfortunately, only one microcomputer station was available for the computer-assisted instructional (CAI) component of the experiment. This limited the number of students who could access the CAI materials and in some cases affected the length of time available for student usage of the system. Logistical and scheduling problems were minimized, but were not entirely eliminated. One final point of concern regards the content and con- duct of the study. The author designed and wrote the compu- terized delivery system as well as all of the instructional materials presented in the CAI trials. The lecturer for the mathematics control groups is one of the author's colleagues and has been teaching the subject material for the past eight years. Both have been working closely together in mathematics education and collaborated on the design and content of the CAI materials. They were not unbiased observers of the experimental events and the possibility exists that their enthusiasm for both the course and the computer system may have influenced the results. Overview of Subsequent Chapters The second chapter contains a comprehensive review of the literature pertinent to this study. This review is intended to provide the background information dealing with the theory, methods, trends, and experimental results of 20 research into student locus of control, aptitude-treatment interactions, and computer-based education. Chapter III presents a formal statement of the research hypotheses, a discussion of the CA1 system and courseware, and the exper- imental design employed. The third chapter will also pre- sent the design matrix, statistical analysis procedures, and reliability and validity concerns of the study. The fourth chapter presents the results of the study. Each of the research hypotheses is restated and translated into statis- tical form providing the basis on which to discuss the sta- tistical analyses. Finally, Chapter V presents a summary of the study with a discussion of the conclusions drawn from the research. Implications for future research into the area of computer usage in education and instruction are also forwarded. An insightful reflection on what transpired during this research project is given, for it assists in understanding and deciding the direction to take in this latest educational revolution. CHAPTER II REVIEW OF THE LITERATURE Introduction A review of the literature pertinent to this study comes from three distinct areas: 1) student locus of con- trol; 2) computer-based instruction; and 3) aptitude- treatment interaction research. The first section focuses on the definition and research into the affective variable of student locus of control. The discussion of computer- based instruction presents a review of the theoretical foun- dations and empirical findings from over two decades of research. For both of these areas the review begins by pre— senting a theoretical or historical background of the sub— ject. The major emphasis, however, is the relationship each area has with academic achievement. The effects of ethni- city, sex, anxiety, and other mediating variables are included in the discussion as a means of defining and clari- fying the impact of this study. The last area is concerned with the techniques and principles of aptitude-treatment interaction research. This selective review of the litera- ture is intended to provide the theoretical and philosophi- cal basis underlying the research methodology of this study. 21 22 Student Locus of Control Theoretical Background There is little argument that academic achievement is directly related to a student's desire or motivation to learn. With an increasing emphasis on mastery learning, independent study, self-paced instruction, and alternate teaching technologies, how students perceive their ability to control these school related activities has a strong bearing on their academic performance. Hence, the notion of personal control over one's behavior and environment, or lack of it, is applicable to education. The degree of personal control is addresed in Heider's (1958) analysis of action in which the "results of an action is felt to depend on two sets of conditions, namely factors within the person and factors within the environment" (p. 82). The concepts of "trying", "ability", "can", "wants", "difficulty", "opportunity", and "luck" are interwoven to describe the outcome of some action. Whether one attributes the outcomes to personal or environmental factors is indica- tive of one's perception of the locus of control for those outcomes. This formulation further posits that "behaviors can be accounted for by relatively stable traits of person- ality or by factors within the environment" (p. 56). Weiner (1979) extends Heider's theory by including the dimensions of stability and control in addition to locus of causality. In his formulation locus of causality is viewed as either internal (personal) or external (environmental). 23 Control, on the other hand, is a separate concept which is controllable (e.g., effort; mood) or uncontrollable (e.g., ability; task difficulty). According to this taxonomy, ability is classified as internal and stable but not under ones control while luck is considered as external, unstable, and also uncontrollable. Even though the basic theory is generalizable across most situations, the degree or relative importance of causal relationships attributed to these three dimensions are situation specific. This three factor model is an expanded version of a two component paradigm advocated earlier by Weiner, Nierenberg, and Goldstein (1976). In this earlier formulation the locus of causality and control dimensions were considered as a single locus of control attribute. Indeed, the three factor model has been criticized for non-orthogonality of the fac- tors. Stipek and Weisz (1981) comment that the control dimension is highly correlated to both the locus of causal- ity and the stability dimensions. Events that are typically considered external are usually not under the control of the individual. Similarly, unstable causes such as effort or mood are likely to be under the direct control of the person while stable causes (e.g., ability) are largely uncontrol- lable. There appears to be an unequal weighting in this causal matrix design with most unstable events being con— trollable and most stable events being uncontrollable. DeCharms (1968) in his discussion of personal causation states: "Man strives to be a causal agent, to be the primary 24 locus of causation for, or the origin of, his behavior; he strives for personal causality" (p. 269). He describes the personal-environmental basis of causality using an Origin- Pawn metaphor: An Origin has a strong feeling of personal causation, a feeling that the locus for causation of effects in his environment lies within himself. The feedback that reinforces this feeling comes from changes in his environment that are attributed to personal behavior. This is the crux of the concept of per- sonal causation and it is a powerful motivational force directing future behavior. A Pawn has a feel- ing that causal forces beyond his control, or per- sonal forces residing in others, or in the physical environment, determine his behavior. This consti- tutes a strong feeling of powerlessness or ineffec— tiveness. (p. 274) Minton (1967) in his discussion of latent power pro- vides another link between power and locus of causality by stating that: ...an attitude of powerfulness or high power is consistent with an action outcome of success ascribed to the person; an attitude of power- lessness or low power is consistent with suc— cess ascribed to the environment. (p. 233) Thibaut and Kelley (1959) also separate power into two cate- gories which they label fate control and behavior control. Although these two types of control are presented in refer- ence to human dyadic relationships the basic idea can be generalized to person-environment interactions. It is the social learning theory of Rotter (1954) that provides the framework for much of the research dealing with the perception of personal control. Indeed, the concept of locus of control was generated by advances made in the development of social learning theory. Its widespread 25 acceptance as a measureable quantity, in addition to the contributions of the aforementioned theorists, has refined the original theory (Rotter, Chance, & Phares, 1972). The concept of locus of control, however, has actually existed informally for thousands of years. The phrases "fate of the gods", "lady luck", or more recently "the devil made me do it" all refer to how people perceive the relationship between their actions and the subsequent events. Rotter (1966) provided a more formal definition of locus of control when he wrote: When a reinforcement is perceived by the subject as following some action of his own but not being entirely contingent upon his action, then, in our culture, it is typically preceived as a result of luck, chance, fate, as under the control of power- ful others, or as unpredictable because of the great forces surrounding him. When the event is interpreted in this way by an individual, we have labeled this a belief in external control. If the person perceives that the event is contingent upon his own behavior or his own relatively permanent characteristics, we have termed this a belief in internal control. (p. 1) Although there are subtle differences between social learning, attribution, and intrinsic motivation theories these authors ascribe to, the theme of locus of control is omnipresent. Whether it is labelled as locus of control or locus of causality, or whether it is viewed as fate versus behavior control, person versus environment control, or as Origins and Pawns, it can be interpreted as the perceived causal source for the interaction between action and rein- forcement. 26 Early formulation of social learning theory (Rotter, 1954, 1966) promoted the concept of internal-external (I-E) locus of control (LOC) as a generalized expectancy. As such the concept spawned research efforts in numerous areas. A bibliography by ThrOOp and MacDonald (1971) lists 339 LOC studies performed in the period of 1954 to 1969. The extent of the research is also represented by the fact that 11 dif- ferent locus of control assessment instruments have been devised and implemented in a variety of settings. Joe (1971) reviews locus of control as a personality variable and subdivides the research into 12 major application areas. Lefcourt (1972) also reviews LOC research making eight gen- eral categories. These reviews and others (Lefcourt, 1966, 1976; Phares, 1973, 1976; Rotter, Chance & Phares, 1972) present research findings in the areas of antecedent varia- bles, risk—taking behavior, societal influences, cognition, anxiety, and achievement to name a few. In general, all of the reviews lend support to the social learning theory of motivation. The research has suffered, however, in its attempt to consistently identify 'locus of control as a significant determinant of behavior. Rotter (1975) critiques I-E research by stating: Expectancies in each situation are determined not only by specific experiences in that situation but also, to some varying extent, by experiences in other situations that the individual perceives as similar. (p. 57) He further elaborates by stating that: 27 ...the relative importance of generalized expect— ancy goes up as the situation is more novel or ambiguous and goes down as the individual's experi— ence in that situation increases. (p. 57) One should, therefore, not assume a measure of generalized expectancy to be highly predictive of academic performance, especially as the amount of formal education increases. Possibly more important in predicting behaviors than a measure of generalized expectancy is the "value" of the reinforcement in a specific situation. The concept of rein- forcement value has been explicately stressed by Perlmuter and Monty (1977) and Rotter (1975), and implicitly mentioned by Hamsher, Geller and Rotter (1968), McGhee and Crandall (1968), Minton (1967), and Mischel (1977) in their discus— sions of possible mediating variables influencing locus of control research. The concept of "congruence" also plays an important role in understanding behavior. Much of the early LOC research focused on performance of internal/external indi- viduals under skill or chance conditions. Phares (1957) conducted a study in which subjects were required to match cards of slightly varying colors. Half of the subjects were told that performance on the task was due to luck while the other half of the subjects were told that performance was a matter of skill and ability. James and Rotter (1958) studied skill and chance conditions in the extinction of reinforcements in a card guessing task. Minton (1967) in a study of schizophrenics, and Watson and Baumal (1967) in an 28 investigation of learning nonsense syllable pairs emphasize that the performance level is dependent upon the congruence between the individuals' general expectancies and the con- ditions under which the task is performed. All of these studies support the hypothesis that internally oriented sub- jects perform better under skill conditions while external subjects do better under chance conditions. Petzel and Gynther (1970) studied performance in solv- ing anagrams under skill and chance conditions. Contrary to the social learning theory predictions, their results indi- cated that externals performed better under the skill condi— tions while internals solved more anagrams under the chance conditions. The seeming contradiction is partially explained by inconsistencies in defining internal and exter- nal groups. Procedural differences between researchers in defining two or three groups of subjects on the I-E Scale tend to confound even significant experimental finding. The preceding discussion has defined locus of control and commented on some of the related concepts. The discus- sion has not focused on academic achievement specifically but on those factors present in any task. The concepts of reinforcement value and congruence are important and indeed have an impact in the specific area of academic performance. Based on this preliminary discussion, academic achievement is a function of specific situational variables (course, physical location, teacher, etc.), the value students have 29 for the learning task, and whether the instructional method- ology is congruent with personal characteristics. Assessment Measures Since locus of control research covers a multitude of student variables and achievement tasks, the instruments for measuring LOC warrants a short discussion. The review by Throop and MacDonald (1971) lists 11 different LOC measures and a more current review by Stipek and Weisz (1981) con- tains references to 13 instruments that have been used in academic settings. The Internal-External Locus of Control Scale (Rotter, 1966) is the most widely used assessment tool. The scale consists of 29 agree/disagree items of which eight reflect an internal orientation, seven indicate external beliefs, and six are filler items used to mask the purpose of the questionnaire. This instrument was designed to assess the generalized expectancies toward internal or external control over subsequent reinforcements. High scores on the I-E scale indicate high levels of externality. Two closely related scales for assessing generalized expectancies are the Children's Nowicki-Strickland Internal- External Control Scale (CNSIE) and an alternate form called the Adult Nowicki-Strickland Internal-External Control Scale (ANSIE). The CNSIE (Nowicki & Strickland, 1973) and the ANSIE (Duke & Nowicki, 1974) are both 40 yes/no item ques- tionnaires with 17 internal, 12 external, and 11 neutral 30 items. The difference between the forms is that the ANSIE has an upgraded vocabulary for the adult audience. Scoring is in the external direction. A third popular measure is the Intellectual Achievement Responsibility (IAR) Questionnaire (Crandall, Katkovsky & Crandall, 1965) This instrument consists of 34 choices of attribute items. Each question has two alternatives indica- tive of internal or external orientations. Consequently, the IAR has two subscales; one representing acceptance of responsibility for success (1+) and one for failure (I—). The composite I score (1+ plus I-) indicates a person's acceptance for both success and failure. This measure was specifically designed for academic settings and is scored in the internal direction. Ethnicity According to Coleman, Campbell, Hobson, McPartland, Mood, Weinfeld, and York (1966) in their comprehensive study of racial issues in education: ...a pupil attitude factor, which appears to have a stronger relationship to achievement than do all the "school" factors together, is the extent to which an individual feels that he has some control over his own destiny. (p. 23) They conclude that a large percentage of the variability in school grades is attributable to this factor. Their find- ings indicate specifically that "minority pupils, except Orientals, have far less conviction than whites that they can affect their own environments and futures" (p. 23). The 31 three items used in this study constitute a general percep- tion, agree/disagree questionnaire which is now known as the Fate Control measure. The results clearly indicate that minorities, and Blacks in particular, have an external ori- entation toward school. They perceive their performance, as measured by course grades and standardized tests, as under the control of others and they do not believe that they have any real power or ability to improve their performance. This control factor was the predominate variable in predict- ing the performance of Black students in the Coleman report. It was not a significant factor, however, in determining the performance of white students. Alker and Wohl (1972) used urban (70% minority) and suburban (100% white) school settings to determine the extent to which LOC accounted for variability in grade point averages. Students were administered Rotter's I-E scale and a cross school comparison showed no significant differences in LOC mean scores. Locus of control was a significant fac- tor in predicting achievement in the urban school. Although LOC was not significantly related to achievement in the sub- urban school, a clear relationship between internality and higher GPA's was found. Moreover, an interaction between locus of control and school setting was present in which GPA's were more dependent on externality for the urban than the suburban school. Even though race was not factored out in this study, the authors noted that equivalence of LOC 32 means between school settings implied that their results need not be interpreted on racial grounds. In a similar study, Lessing (1969) administered the Personal Control Scale (PCS) and a Delay of Gratification Scale to 558 eighth and eleventh grade students. The PCS was specifically used as a measure of internal/external locus of control. While the Black students had lower GPA's and IQ scores, and showed significantly less control over their lives than their white counterparts, the relationship between academic performance and locus of control was not definitive. The PCS and Delay of Gratification measures did correlate with achievement but the effect was greatly atten- uated when IQ was controlled. Locus of control was not found to be a major contributor in predicting achievement. Moreover, race was accountable for only a small amount of the variance in PCS scores. The study did, however, lend support to the high performance-internality connection. Shaw and Uhl (1971) report on the relationship between LOC and reading scores for second grade students. Locus of control was assessed by the Bialer-Cromwell Children's Locus of Control Scale which yields a measure of generalized expectancy. Black subjects had higher external scores than white subjects but only in the upper-middle socioeconomic levels. There was no racial pattern for the lower SES levels. Locus of control scores were significantly related to reading scores only in the white upper-middle SES group. 33 Again, no racial patterns were evident in reading scores for the other SES groups. Karmos, Bryson, and Tracz (1982) found locus of con- trol, as measured by the I-E Scale, to be unrelated to the college GPA's of Black and white graduates. University graduates are extremely familiar with the academic environ- ment and therefore, a measure of generalized expectancy looses its power as a predictor. This result is not entirely unexpected in view of Rotter's comments regarding the novelty of any given situation. Despite the absence of any relationship, however, locus of control scores were more external for the Black subjects in their sample. Using the IAR Questionnaire, Soloman, Houlihan, and Parelius (1969) analyzed school grades for 262 fourth and sixth grade students. They found no significant differences on the basis of race. Katz (1967), also using the IAR scale, concurs with the findings that race does not appear to be a factor in the IAR's predictive power of school grades. DuCette, Walk, and Soucar (1972) studied the interac- tion of nonadaptive classroom behavior and locus of control. Black and white students were administered the IAR in addi— tion to being classified as adaptable or non—adaptable according to their ability to adjust to classroom situa- tions. Significant main effects for race were found, with Blacks exhibiting more externality on both the 1+ and I- subscales. Moreover, the existence of a three factor 34 interaction between race, behavior classification, and locus of control indicates that race had a moderating effect on the relationship between locus of control and conformity in school settings. The study also indicates that the connec- tion between locus of control and behavior is very complex. In straight comparative studies between Black and white samples, Rotter (1966) reports inconsistent racial trends in I-E scores. Race by socioeconomic status interactions indi- cate that lower class Blacks were distinctly more external than either middle class Blacks or middle and upper class whites (Rotter, 1966; Shaw & Uhl, 1971). Studies of Black and white inmates indicate mixed results. Black inmate scores on the I-E scale were found to range from significantly more external than their white counterparts (Lefcourt & Ladwig, 1965) to no appreciable difference (Kiehlbauch, 1968). Interpretation of these two studies, and indeed many of the studies involving race, need to be viewed in the context of the larger societal events of the time. The first study was conducted during the early stages of the civil rights movement when racial repression was just becoming a major social issue. The latter study, however, took place during the height of the militant phase of that movement. Lefcourt and Ladwig (1965) prognosticated that it was: ...possible that in the current Negro mass movement for civil rights, there will be greater opportunity for Negroes to witness concrete changes deriving from their social actions. (p. 380) 35 One problem associated with ethnicity and locus of control research lies with the interpretation of LOC scores. Rotter (1966) presents evidence for a unique, single factor interpretation of the I—E scale. Critics of the measure (Abrahamson, Schludermann, & Schludermann, 1973; Katz, 1969: Mirels, 1970) forward evidence for two distinct subscales within the measure. One factor is identified with personal control and the second, which contains all of the items worded in the third person, constitutes a political/social factor. Gurin, Gurin, Lao, and Beattie (1969) provide evidence which taints much of the race-locus of control research in which LOC is measured by the I-E Scale. A 39 item question- naire (23 from Rotter's I-E Scale, 3 from the Personal Effi- cacy Scale, and 13 items that were specifically written to elicit racial beliefs) was administered to 1,695 subjects. Factor analysis of the responses identified the four sub- scales of: 1) Control Ideology, 2) Personal Control, 3) Sys— tem Modifiability, and 4) Race Ideology. Further factor analysis of the System Modifiability and Race Ideology com- ponents produced four more factors labeled: 1) Individual- Collective Action, 2) Discrimination Modifiability, 3) Indi- vidual-System Blame, and 4) Racial Militancy. The I-E items were concentrated in the two subsections of Control Ideology and Personal Control factors lending support for a two fac- tor interpretation of the measure. 36 Lao (1970) further fueled the controversy of interpret- ing Black responses to the I-E measure. She surveyed 1,493 Black male college students and reproduced the Control Ide- ology and Personal Control subscale separation on the I-E scale. Indeed, the correlation between these two subscales indicated clear support for their independence. Rotter (1975) attributes the emergence of these factors as either temporal or population artifacts which he views as possibly helpful features "if it can be demonstrated that reliable and logical predictions can be made from the subscales to specific behaviors..." (p. 63). In summarizing the findings between locus of control and ethnic status two trends emerge. In simply comparing Black versus white locus of control scores, Black subjects tend to be more external in their beliefs. This trend appears to be largely independent of assessment measure or situational factors. However, on occasion, racial differ- ences as measured by instruments of generalized expectancies are susceptible to both temporal and social issues. Racial differences seem to be eliminated, however, if the IAR Ques- tionnaire is used in academic settings. The second trend is for locus of control to be moderately related to achievement measures with higher performance levels corresponding to higher levels of internality. This trend appears to be race independent. The concerns raised by Abrahamson et a1. (1973), Gurin et al. (1969), Katz (1969), Lao (1970), and Mirels (1970) 37 highlight the difficulty in using a generalized expectancy measure to predict outcomes in specific situations. This type of incompatability may, in part, be responsible for the diverse and often confusing results from interpreting LOC scores and their impact on other measures with reference to ethnicity. 'Sgg Interpreting locus of control scores on the basis of sex is no more clear than ethnicity. Studies that employ the I-E Scale indicate that female subjects are slightly more externally oriented than male subjects (Cellini & Kan- torowski, 1982; Rotter, 1966; Strassberg, 1973). Data gen- erated by the IAR Questionnaire indicates, however, that females are more internal than males for all grade levels on the I total and the I+ scales and for the I- scale in grades six through 12. The I- scores for males in grades three to five tend to be more internal than female scores although the difference did not reach significance (Crandall et al., 1965). Results obtained using the Bialer Locus of Control Scale also indicate that females were significantly more internal than males in grades six, seven, and eight (Prawat, 1976). Fendrick-Salowey, Buchanan, and Drew (1982), how- ever, report no sex differences in the fifth or sixth grades. This sample was very small and results that are non-significant are not surprising. Prawat, Grissom, and Parish (1979), using the CNSIE measure, reported that 38 females are more internally oriented than males in all grades (3-12) with the lone exception of grade nine. Studies investigating locus of control and school achievement are dependent on both the achievement tests and the locus of control measures. Barnett and Kaiser (1978), using the IAR Questionnaire, found significant relationships between the I total score and IQ, school GPA, and a battery of achievement test scores for males only. Buck and Austrin (1971) found significant correlations for both males and females when the Iowa Test of Basic Skills was the Achieve- ment measure. Brown and Strickland (1972) found internal scores on the I-E Scale to be related to college grades and frequency of extracurricular campus activities for males only. Male I total scores were significantly and positively correlated to the reading and arithmetic sections of the California Achievement Test in grades one through three (Crandall, Katovsky, & Preston, 1962). Negative, though non-significant, correlations for the female subjects were measured in this study. Crandall et a1. (1965) and McGhee and Crandall (1968) report positive and significant correla- tions for the I total score in relation to scores on the Iowa Test of Basic Skills and school GPA's in grades three to five. The I+ scale was a better predictor for third and fourth grade female students while the I- was a stronger indicant of achievement for fifth grade males. For grades 39 six through 12 the California Achievement Test scores showed no consistant relationship to the IAR scores. Messer (1972) reported that fourth grade GPA's and Stanford Achievement Test scores were best predicted by the I+ for male subjects and by the I- for females. I total scores predicted school grades better than they predicted standardized test scores for both sexes. Clifford and Cleary (1972), promoting their Academic Achievement Accountability (AAA) measure of LOC, found sig- nificant correlations with vocabulary achievement tests for both males and females and spelling achievement for males only. Math achievement scores were not correlated to AAA scores for either sex. Overall, achievement scores were better predicted by the AAA for male subjects. Nowicki and Strickland (1973), using the CNSIE and an unspecified achievement measure report better predictability for male subjects than for females in grades three through 12. DeCharms and Carpenter (1968), however, found non- significant relationships between spelling and math test scores with the Bialer-Cromwell LOC measure for male sub— jects. They did find significant effects for the female subjects in their sample. Duke and Nowicki (1974) support studies indicating that the I-E Scale does not correlate with college achievement measures (Hjelle, 1970: Rotter, 1966). Their investigation did find that ANSIE scores were correlated with college GPA's. In support of many studies, male internality was 40 associated with higher college GPA's. However, in contrast to other studies (McGhee & Crandall, 1968; Nowicki & Segal, 1974; Prawat, 1976; Prawat et al., 1979) a significant cor— relation between female externality and higher GPA was measured. In recapping the literature on sex differences in locus of control and the impact on academic achievement, three trends emerge. Contrary to the conclusions of some researchers, notably Lefcourt and Nowicki, there appears to be no consistent pattern establishing internality as a stronger predictor of academic achievement for males than for females. This hypothesis receives support only if the CNSIE scale or its variants are used. Critics point to a social desirability factor within the scale which may elicit the sexual differentiation. As a mediating variable, social desirability may influence female subjects to respond to the LOC items according to their perceptions of the appropriate response rather than with their true feelings. There is almost uniform agreement, however, that inter- nality is related to higher levels of achievement for both sexes. Studies that fail to find any relationship, espe— cially those using the I-E Scale, propose that "defensive externals" sufficiently externalize LOC scores to obfuscate any correlations between internality and academic achieve- ment. Defensive externals are described as "people who have arrived at an external view as a defense against failure but who were originally highly competitive" (Rotter, 1966, p. 41 21). The effects of defensive externals enter as a viable explanation for non-significant results in studies at the high school and most definitely the college levels. College students are usually highly competitive, goal oriented indi- viduals who may indicate external orientations as a means of shifting responsibility for their failures from themselves to other external forces. Therefore, studies that fail to find significant relations between LOC and achievement, par- ticularly at the college level, may have a sufficient number of defensive externals in the sample to reduce the cor- relations between LOC and achievement to a point of non- significance. Lastly, local school grades and GPA's are more closely related to LOC scores than standardized examinations such as the Iowa Test of Basic Skills, the California Achievement Test, or the Stanford Achievement Test. The strength of the relationship, of course, depends on the LOC measure used, with the IAR being the best (Kennelly & Kinley, 1975; Stipek & Weisz, 1981). This is due, in part, to the fact that the IAR Questionnaire elicits responses within a specific con- text while other LOC measures tend to be more generalized. Anxiety Rotter (1975) stressed the differences between general- ized and situational expectancies. As the situation becomes more familiar, the strength of a situation dependent expect- ancy shows mediation. The distinction between generalized 42 and situational expectancies parallels, in many respects, trait and state anxiety (Spielberger, 1966, 1972). Spiel- berger (1966) contrasted these anxieties with: ...anxiety states (A-states) are characterized by sub- jective, consciously perceived feelings of apprehen- sion and tension, accompanied by or associated with activation or arousal of the autonomic nervous system. Anxiety as a personality trait (A-trait) would seem to imply a motive or acquired behavioral disposition that predisposes an individual to perceive a wide range of objectively nondangerous circumstances as threatening, and to respond to these with A-state reactions dis- proportionate in intensity to the magnitude of the objective danger (pp. 16-17). One might, therefore, be tempted to assume relation- ships between generalized expectancy and trait anxiety and between situational expectancy and state anxiety. Indeed, in a review by Archer (1979), 18 of 21 studies reported sig- nificant interactions between trait anxiety and locus of control. Empirical findings overwhemingly indicate that greater externality is related to higher levels of trait anxiety. The studies, typically using the I-E Scale, strongly correlate LOC scores with almost all of the popular trait anxiety measures. Research into the relationship between LOC and specific kinds of state/trait anxiety becomes cloudy due to the diversity of the research settings. Archer (1979) lists 13 studies involving college students where measures of test anxiety (considered a measure of trait anxiety) were cor- related with locus of control scores. Significant relation- ships between higher test anxiety and externality were reported in seven of the studies. State anxiety measures It ‘I 43 are even more suspect for interpretation than LOC measures. Of eight studies listed that involved college students, only three reached significance. The lack of any conclusive relationship between LOC and anxiety measures, particularly state anxiety, may be attributable to the incongruity of generalized LOC expectancy, as measured by the I-E Scale, and the situation specific nature of anxiety. The strong situational expectancies tend to mask or dilute any effects that a generalized expectancy might have. Crandall et a1. (1962) report significant correlations between IAR scores and academic achievement measures (time and intensity in academic free play, IQ, and reading and arithmetic tests). Their research also concluded that the more typical achievement related correlates such as the need for achievement (McClelland, Atkinson, Clark & Lowell, 1953) and manifest anxiety (Sarason, Davidson, Lighthall, Waite & Ruebush, 1960) bore no relation to their achievement vari- ables. Katz (1967) studied school achievement of Black youth and the influence of test anxiety. Scores on the Test Anx— iety Scale for Children showed significant differences for high versus low achievers with highly anxious subjects being poorer performers. This relationship is strongest for male subjects, leading to the conclusion that anxiety was an extremely important factor in understanding behavior, espe- cially for low-achieving males. Although correlational 44 analysis was not presented, higher anxiety levels were associated with externality. Strassberg (1973) also measured significant relation- ships between locus of control and anxiety with extreme externality associated with higher levels of anxiety. Morris and Carden (1981) corroborate with findings that relate externality to trait anxiety (neuroticism). Their study also indicated that LOC was the best predictor of academic achievement and that anxiety measures were uncor- related with grades. Watson (1967) and Feather (1967) sup- port the connection between LOC and anxiety by reporting positive correlations of externality on the I-E Scale for both manifest and debilitating anxiety. Both of these authors also found that facilitating anxiety was not related to I-E scores. They did report, however, negative correla- tion coefficients thereby indicating a tentative mapping of internality to facilitative anxiety and externality to debilitative anxiety. Using a modified form of the I-E Scale, Powell and Vega (1972) also correlate locus of con- trol and anxiety in a sample of teachers and teacher aides. The connection between anxiety and locus of control, while consistent, appears to be an enigma. Certain forms of anxiety stongly correlate with locus of control and there is a general trend for highly anxious people to be more exter- nal. However, whereas locus appears to be a fairly good predictor of academic achievement, anxiety seems to be unre- lated (Crandall et al., 1962; Morris & Carden, 1981). The .9 .1 L- 45 causal relationship between locus of control and anxiety has been questioned by several authors (Feather, 1967; Joe, 1971; Watson, 1967). Joe (1971) ponders: ...whether the belief in external control is a reaction (defense) against anxiety learned on the basis of past experiences in stressful sit- uations or whether anxiety is a reaction to the perception that the world is unpredictable, pre- determined, or controlled by powerful others. (p. 626) The relationship between these two characteristics appears to be very complex indeed. Summary Locus of control has been recognized as an important motivating factor in school achievement. It has been researched in a wide variety of settings using an assortment of measuring devices for the last thirty years. As a result of the diversity of the research studies, clear consistent findings are scarce. There are a few important trends in the literature, however, which are pertinent to this study. In academic settings the IAR Questionnaire tends to be consistently superior at predicting measures of academic performance than other LOC instruments. Predictions are better for school grades than for standardized test scores. Locus of control scores tend to become more internal as a function of age and, therefore, lose their predictive power. The IAR, however, remains a strong predictor of achievement because of its situation specific design. 46 Whereas other measures of LOC, notably the I-E Scale, are susceptable to differences in race, SES, IQ, need for achievement, sex, and even history, the situation specific nature of the IAR coupled with its dual subscale format tends to reduce the impact of these confounding variables. The research on anxiety and locus of control indicates a strong relationship between trait anxiety and an external generalized expectancy. State anxiety research is not con- sistently correlated with LOC measures, possibly due to the specific versus general incongruency of the measuring devices. Indeed, several studies have indicated that while LOC may predict achievement, anxiety measures do not (Cran— dall et al., 1962; Morris & Carden, 1981). There is abundant evidence, though not entirely in agreement, to indicate that the design of the IAR Question- naire does differentiate subjects on the basis of Rotter's (1966) definition of locus of control. The situation speci- fic nature of the IAR also mediates possible confounding variables such as race, sex, and anxiety. Therefore, in disagreement with the Stipek and Weisz (1981) statement that situation specific measures of LOC are no better nor worse than generalized measures at predicting academic achieve- ment, this author concludes that the IAR Questionnaire is a superior instrument for use in academic settings. This author is in agreement, however, with their conclusion that there is little support for LOC as a stronger predictor of male academic performance than for females. 47 Computer-Based Instruction Historical Overview Computers, as electromechanical devices, are a rela- tively recent invention having been developed during the second world war. Despite their recency, most experts in the field of computer technology would agree that by 1990 so called fifth generation computers will be commonplace (Avoli, 1981; Fiegenbaum & McCorduck, 1983). The character- istics of these generations and their approximate time frames are: 1. Vacuum Tube Machines 1945-1960 2. Transistor Based Machines 1960's 3. Integrated Circuitry and Large 1970's Scale Integrated Chip Structure 4. Very Large Scale Integrated 1980's Chip Structures 5. Artificial Intelligence and 1990's Pseudo—human Machines It is interesting to note that the first four generations are differentiated primarily by hardware capabilities while the fifth generation is distinguished by advances in both hardware and software. Hardware changes consist of the pro- posed use of cryogenics, fiber optics, bubble memory, and microcode. Software innovations include specifically designed programming languages, knowledge systems, heuristic logic, and programs that learn from their mistakes. As state-of-the-art as these terms may sound, current computer 48 technology has been compared to the automotive industry in that "...the computers that most of us are familiar with right now, they aren't horseless carriages. They're no more than bicycles" (Fiegenbaum & McCorduck, 1983, p. 11). As advanced computer systems are used to design more efficient computers, the speed at which computer technology and com- puter applications evolve will far outpace the evolution of the automobile. Computers, however, are extremely versatile machines and are now practically ingrained one way or another into our daily lives. The use of machinery for educational purposes has four distinct phases (Gable & Page, 1980). Linear Systems Intrinsic Branching Systems Adaptive or Extrinsic Systems Generative Systems #UJNH coco The first phase is characterized by the strictly mech- anical teaching machines of Pressey (1926) and Skinner (1958). Initially developed for use as testing or evalua- tion systems, these machines were also useful in presenting course material. The instruction was presented in a linear sequence of "frames." Each student moved through the same sequence of frames regardless of ability or performance, hence the linear system denotation. It should be pointed out that in some respects these early machines actually imposed restrictions upon the learner; one could at least quickly scan through sections of a textbook. 49 The second phase is characterized by the "scrambled textbook" approach of Crowder (1959, 1962). Early devices were electromechnical machines with 35-mm filmstrips pres senting instructional frames requiring the student to press one of several buttons. Each button engaged a different sequence of instructional material. The machine automati— cally branched to whatever sequence was engaged. Moreover, each sequence had to be carefully prepared and organized by the curriculum developer. Some of the more advanced systems included machine con— trol features. Internal film control codes provided for such features as engaging all user buttons, returning to the preceeding frame only, entering a correctional sequence only, and permitting backward film motion only. These internal control codes, in effect, overrode the user response and their use was predetermined by the curriculum developer. Much of the computer—assisted instruction cur- rently available is responsive to student input but the level of branching ranges from non-existent to highly sophisticated. A majority of the systems, even though computer-based, are still considered to be of the linear model. Adaptive or extrinsic systems are similar to Crowder's intrinsic model except that the branching techniques are based on student response histories. Computer-managed instructional systems which employ databases to record and 50 analyze individual student responses and learner character- istics can, through Bayesian statistics for example, adapt instructional sequences to the individual learner. These systems in essence formulate a model of each student which is electronically stored between instructional sessions. The last phase is currently under development and represents the state—of—the-art in computer applications in education. Generative systems utilize complex artificial intelligence principles to actually construct problems and answers for presentation to the student. These systems usually incorporate relational databases or semantic net- works of the subject material. Semantic networks are soft- ware structures where relationships between various data items are explicitly stored as part of the instructional system. Using heuristic search and sorting techniques these programs can generate new questions or even respond to questions posed by the student (Gallagher, 1981; Mitchell, 1981). The SCHOLAR program (Carbonell, 1970) was an early pro- totype of the knowledge based systems. This program, teach- ing South American geography, permitted both the computer and the student to ask and answer questions. This two-way interaction constitutes a mixed-initiative system. The very fact that computers are becoming capable of formulating, asking, and answering questions is a marked departure from the previous phases where subject content, sequence, and 51 branching aspects were predetermined and explicitly program- med into the system. In a sense, these generative systems "understand" the subject material. Carbonell's programming strategy also marks a departure from the traditional CAI instructional format. Typical CAI can be described as ad hoc-frame-oriented (AFO) in which instructional material consists of specific frames of text material, questions, answers, and diagrams. These frames have to be prepared in advance by the curriculum developer. The generative systems are considered as information- structure-oriented (ISO) CAI. These ISO-CAI systems utilize an information network of facts, concepts, and decision- making capabilities. Since these systems are not designed according to the AFO-CAI format, they are able to generate text, pose questions, and respond with the appropriate answers. Coupled with a structural parser for analyzing user input, these systems can even respond to questions input in natural language. Xoffman and Blount (1975) and Koffman and Perry (1976) incorporate models of the student in their generative sys- tems. These models enable the systems to tailor the diffi- culty level of the instruction to the level of the learner. As the student becomes more proficient with the material, the computer automatically adjusts the model parameters to reflect the increased ability. The preceding discussion indicates that through the use of artificial intelligence techniques, relational databases, 52 and student modeling methods, computers may eventually serve as personal tutors. The progression from the very early linear models to the generative systems currently under development represents a natural incorporation of technology into the instructional process. Education may be coming full circle with the rejuvination of the Socratic dialogue-- a computerized Socrates to be sure. Historically, there have also been two predominant philosophies regarding CAI implementation-~learner versus machine control (Roblyer, 1981; Splittgerber, 1979). The early mechanical systems and the more recent computer-based Stanford/CCC model are representative of the machine control paradigm (Lysiak, Wallace, & Evans, 1976). The Stanford/CCC system consists of almost total machine controlled drill and practice exercises. The use of feedback, graphics, and ani- mation is limited in these materials. Consequently, the system and the courseware are relatively simple from both developmental and operational aspects. Learner controlled systems are exemplified by the PLATO (Progammed Logic for Automatic Teaching Operations) system and the TICCIT (Time-shared, Interactive, Computer- Controlled Information Television) project. Both of these systems utilize programmed instructional techniques for courseware development. Bunderson (1974) defines learner- controlled courseware as having 1) a heirarchial level of student-machine discourse: 2) a modularized structure replete with relationships to instructional taxonomies: and 53 3) an interface between instructional system components and the learner. The TICCIT project (Faust, 1974) provided for user control over the selection and sequencing of instruc- tional courseware. The computer-controlled and television aspects simply represent the particular hardware delivery modes. PLATO system (Alpert & Bitzer, 1970) courseware like- wise permits the student to choose the instructional mate— rials via menu selections. Prompt feedback, help tables, forward and backward motion, graphics and animation are standard features of PLATO courseware. Although the student can select the topics to be studied, the actual instruction has been prepared in advance and tends to be rigid. Learner control aspects relative to PLATO and TICCIT simply refer to the sequencing of materials and to the amount of drill and practice desired. The diametric philosophies represented by these CAI systems are not related to their effectiveness as instruc- tional tools. Their applications are as distinct as their underlying approaches. The PLATO and TICCIT systems have been used at all levels of education and training while the Stanford/CCC model works well in those areas where mastery of basic facts and principles is required. These philosoph- ical approaches have spawned research into the optimization of man-machine control and its effects on student perform- ance which will be discussed in a later section. The gen- erative CAI models and the systems forthcoming in the next 54 decade will possess a blend of man and machine control. Intelligent systems, through self-regulation, will be able to either impose or to relax machine control on the basis of student performance. Hardware Considerations As with any new invention or application of technology, research into the effects of the physical devices and their operation on the intended users is of great importance. The computer systems now being employed in education are no exception and this section is intended to illuminate some of the hardware aspects involved. The ergonometric considera- tions have received much attention regarding the use of microcomputers and particularly video display terminals (VDT's) in the work place. These man-machine interactions are also pertinent to the student learner. There are other concerns regarding equipment usage relative to the educational process. Bevan (1981), for example, investigated the relationship between performance in a learning task and the speed of presenting textual information on a video display unit (VDU). Lesson comple- tion times, error rates, recall, and attitudes toward com- puterized instruction were analyzed. The empirical results were that a presentation speed of between 10 and 15 charac- ters per second jointly optimized the four factors studied. A 480 character per second presentation speed, i.e. full screen display, detracted from the overall efficiency by 55 increasing the reading error rates. The process of filling the VDU screen appeared to introduce extraneous factors adversely affecting the results. Bork (1981) and Jenkin (1981) comment that newer hard- ware technology is not only creating problems but also expanding capabilities in the preparation and presentation of instructional materials. The use of color screens, ani- mation and graphics, white space, speech synthesis devices, and character intensity control must now be considered by curriculum developers. The inclusion of these emerging hardware capabilities in instructional materials has gener— ated research into the issues of screen design, screen man- agement, and visual aesthetics. Moore and Nawrocki (1978) investigated the use of high resolution graphics in Army training procedures. Proponents of graphics tout perceived efficiency, realism, increased student performance, accommodation of student preferences, and the provision of a system with a little panache. None of the popular reasons for including high resolution graphics were supported empirically. Research into the effectiveness of varying grades of graphics; boxed alpha- numeric labeling and schematics, simple line drawings, and line drawings with animation, resulted in no statistical difference in academic performance or unit completion times (Moore, Nawrocki, & Simutis, 1979). Despite the fact that high resolution color graphics are immensely popular, these 56 researchers found that simplicity was best. Color and real- istic simulations failed to improve performance and even though subjects expressed personal preferences, there was little relationship to their performance or motivation. The overall conclusion was that the use of graphics does not, by itself, guarantee inprovement in either completion time or achievement. Many of the newer sophisticated computer systems sup- port foreground/background color features. Ohlsson, Nils- son, and Ronnberg (1981) investigated the interaction of text/background color combinations as they affected the speed and accuracy in scanning a matrix of letters for a specified character. Although no simple color combination was the best, several schemes were better than others in optimizing speed. The results supperted the contention that the greater the difference in color wavelengths between text and background, the higher the ratings on contrast, spacing, and overall readability of the displayed material. Two optimal color combinations were forwarded for reducing the error rates in reading; green lettering on a white back- ground or magenta on green. Oddly enough, the reverse pairing of white letters on a green background produced one of the highest error rates. The study did not investigate the joint optimization of reading speed and accuracy nor did it report on the effects of prolonged use of the VDU on eye fatigue and/or strain. 57 Thomas (1979) studied whether multiple-choice items were better keyed with a letter or a numerical response relative to computer keyboard input. Problems were pre- sented to the students along with the answer that was to be input. In one study an interaction effect occurred in which typists were quicker with the letter input while non-typists were more adept at using the numerical input. Fewer errors, however, occurred when letter input was required. The errors made by non-typist subjects included pressing the space bar, the return key, or a key adjacent to the correct answer key. Although these results may have more relevance to courseware development, the design, layout, and usage of computer keyboards may be a factor. This particular study also highlights the existence of interaction effects between the hardware and the user. This rather brief section only scratches the surface of the research into man-machine interactions. Computer hard- ware systems and their capabilities, especially the use of color, graphics and animation, and the variable intensity features of video monitors, have become a boondoggle for curriculum developers. These diverse capabilities, however, represent a boon for research into determining which fea— tures are effective in an instructional setting. Software Capabilities The power behind computer systems is the design and structure of the software programming. It is the computer 58 programs that inform the hardware system to display infor- mation on the video monitor, to wait for user input, to retrieve or transfer information to any number of peripher— als. Much of the CAI research, therefore, has been on the efficacy of different software features. Magidson (1978) comments on the diversity of computer usage in education and the various techniques used in the preparation of CAI materials. Instructional materials are most often written using drill and practice, tutorial, or simulation paradigms. Although roughly 80% of CAI course- ware is in the subject areas of mathematics, science, and the study of computers, CAI is making inroads into virtually every discipline at every educational level. Rushby (1979) provides a slightly different but paral- lel classification of CAI courseware. The British equiva- lent of CAI is computer—assisted learning (CAL) which is subdivided into four general paradigms: instructional, revelatory, conjectural, and emancipatory. The emphasis of the instructional CAL model is on the subject material and the student's mastery of it. This model is similar to the drill and practice and the tutorial formats. The revelatory design supports a guided learning, dis- covery type of process. The subject material and the basic theory are slowly revealed to the student in such a manner that the student can formulate or deduce the essence of the instruction. This model is analogous to the simulation approach mentioned earlier. 59 The conjectural model allows the student to formulate and test hypotheses he or she might be researching. The focus of this model is to promote student experimentation and verification or rejection of ideas or theories related to secific topics. The level of this type of CAL software is generally postsecondary because it requires a substantial knowledge base as well as abstract reasoning abilities (Dwyer, 1974; Dwyer & Critchfield, 1981). Emancipatory CAI simply refers to using the computer as a means of reducing the student's workload. Software pro- grams that perform data analysis and numerical calculations, word processing systems, and electronic scratch pads or spreadsheets are examples of the emancipatory usage of com- puters. Burke (1982) provides a more comprehensive classifi- cation by distinguishing between the functional, physical, and logical designs of CA1 materials. Functional designs consist of the drill and practice, tutorial, gamelike (simu- lation), and problem solving methods. The problem solving design employs the computer as an intelligent calculator and monitors the student's actions often providing guidance and redirection. This type of CAI model is similar to the con- jectural paradigm previously discussed. Examples of the probelm solving CAI design are the BLOCKS (Gallagher, 1981), GIANT (Wexler, 1970), MALT (Xoffman & Blount, 1975), SCHOLAR (Carbonell, 1970), and SOPHIE (Brown & Burton, 1974) pro- grams. 60 The second design variable pertains to how the instruc- tional material is to be presented (and consequently pre- pared) with particular emphasis on computer usage. The six basic physical designs are: 1) linear, 2) spiral, 3) branch- ing, 4) multitrack, 5) regenerative, and 6) adaptive. The linear and branching models are the most common and have already been described. The spiral design is akin to Bruner's (1966, 1977) spiral curriculum. Each time through the material a different property of the subject is brought into focus and highlighted. The multitrack design is aligned with Gagne's (1975, 1977) learning hierarchy whereby the lowest level material may concentrate on basic facts while the highest level may be written more abstractly requiring analysis and/or synthesis of the subject matter. The regenerative design utilizes the computer's ability to generate different numerical values, key textual phrases, or even entirely different problem sets for each student or for each time the unit is presented to the same student. Lessons written according to this design appear to be dif- ferent each time they are viewed. This type of model does not "know" who is using the system; it simply generates new values or phrases each time the program is requested. CAI lessons prepared in the adaptive model, however, utilize information about the student and can tailor the instruction to that particular individual's needs or history. These systems are still very rare and are com— parable to the adaptive or extrinsic systems espoused by 61 Gable and Page (1980). None of the physical models discus- sed by Burke are as sophisticated as the generative, artifi- cial intelligent systems discussed earlier. Logical designs address the manner in which the instruction is presented from a cognitive psychological viewpoint. The five prominent logical designs are: 1) didactic, 2) discovery, 3) EGRUL, 4) RULEG, and 5) fading. The didactic model is typically used because it provides a convenient method for assessing student understanding. It also fulfills the perceived need for interactive dialogue between man and machine. Even though the level of dialogue is restrictive, the didactic method is one that actively engages the learner. The discovery design involves artificially creating conditions in which the student can discover relationships and develop an intuitive understanding of the intended subject matter. Papert (1980) is a strong advocate of the LOGO language and the use of video and mechanical "Turtles" to promote discovery learning. The EGRUL model presents a series of examples designed to facilitate in the development of a rule that connects all of the examples. Both the dis- covery and the EGRUL paradigms rely on an inductive thought process. The RULEG method, in contrast, is a deductive process. Students are first taught a specific rule and are typically required to apply the rule in distinguishing examples from 62 nonexamples. In essence, it is the reverse of the EGRUL model. The fading design is one which initially contains strong, forceful prompts to direct the learner. As the les- son proceeds, however, the prompts become weaker, eventually disappearing by the end of the unit. This particular design is useful for materials requiring memorization. Vinsonhaler and Bass (1972) reviewed ten major drill and practice CAI studies. The studies were selected because evaluation involved standardized tests and the studies com- pared CAI supplementing traditional lecture to traditional lecture alone. All of three language arts studies indicated positive gains in grade-year equivalents for the CAI supple- mentation model of instruction. Similar results were noted for a majority of the mathematics studies. Most of the studies showed statistically significant differences favor- ing CAI augmentation of traditional instruction over tradi— tional instruction alone. A comparison of various logical designs has been inves- tigated by Lahey (1979, 1981). He reports on the effective— ness of different instructional sequences on academic per- formance in electronics laboratories. The three instruc- tional activities of examples, rule statement, and practice were combined to form four experimental sequences; rule- example-practice (RULEG): example-rule-practice (EGRUL): practice-example-rule: and a randomized sequence. Four experimental groups completed 23 CAI lessons on the PLATO 63 system. As might be expected, the group receiving the ran- domized sequence required more time to complete the assign- ments and answered more questions. However, there wasn't any statistically significant difference in the overall per- formance of the four groups. There was an interesting trend that the randomized method may be advantageous in the joint optimization of time and performance. Park and Tennyson (1980) contrasted two designs within the RULEG paradigm in the presentation of six concepts drawn from the field of psychology. The order in which examples of the various concepts were presented was either response- sensitive (dependent on student input) or response- insensitive (randomized). The six psychology concepts were defined and discussed prior to the students task of matching examples to the appropriate concept. If a student's classification of the example was correct another example of any of the concepts was presented. If the classification was incorrect, an example of the incorrectly chosen concept was immediately displayed in the response-sensitive mode, the hypothesis being one of providing a contrast of the concepts to facilitate in the removal of any over or undergeneralizations. In the response-insensitive mode the computer simply ignored the student input and presented another example at random. The results indicated that the response-sensitive mode of operation was superior with respect to achievement, reduction of instructional time, and the number of examples needed. The fine tuning capabilities 64 afforded by computer technology are evident in this type of learning situation. It is not uncommon for combinations of the various functional, physical, or logical designs to appear at dif- ferent points within any CAI lesson. The lesson content, its proposed use, and the level of the students for whom the material is intended are three crucial factors in deciding the appropriate functional, physical, and logical designs of the CAI courseware. There are, of course, hardware capa- bilities that have to be considered when selecting or pre- paring courseware. Not all computer systems are equipped to process some of the more complex or sophisticated materials. The sequencing of materials, use of perpherial devices, and the incorporation of adjunctive media are also important factors in CAI courseware design and implementation. Diversities of implementation strategies are almost as numerous as the individual researchers. Today computers influence practically every aspect of education. They are being used to teach a wide variety of subjects and are even assisting in the development and preparation of instruc- tional materials. Westrom (1983) discusses the advantages and the disadvantages associated with CAI (and CMI) develop- ment and implementation as they relate specifically to the functional designs mentioned earlier. Gleason (1981) also gives an overview of microcomputer uses in education from the viewpoints of hardware, software, computer literacy programs, and research efforts. The infusion of computers 65 into the educational system has brought a number of arti- cles, position papers, and books speculating on their poten- tial for both good and bad education (Chambers & Sprecher, 1980; Ellis, 1974; Hofmeister, 1982; Holmes, 1982; Leiblum, 1982). The Carnegie Commission on Higher Education (1972), Hunter, Xastner, Rubin, and Seidel (1975), Levien (1972), and Rushby (1981) provide excellent background materials on computer based instruction. Comparative Studies Considerable research has been conducted that compares CAI to other instructional methodologies. Avner, Moore, and ' Smith (1980) contrasted active and passive CAI techniques in chemistry laboratory preparation classes. The passive CAI units were essentially electronic page turners and were con- sidered analogous to the programmed or self-paced modes of instruction. The difference between these models was that the active model required students to respond to questions and thereby demonstrate their understanding of the material while the passive units had no such questioning. Two instructional units of each type were administered to 700 college undergraduate students. One unit was of the follow- the-instructions mode while the other required the students to make decisions. Experimental results indicated that fewer errors were made in actual laboratory exercises by those students receiving the active CAI materials. This finding was restricted to the decision-making task only. No 66 significant differences were associated with the follow-the- instructions exercise. Although the students having the interactive training took less time in the laboratory, they took more time working through the CA1 materials. In a study of nursing education Boettcher, Alderson, and Saccucci (1981) compared the CAI and programmed instruc- tional techniques. The dependent variables of interest were knowledge and skill acquisition. Analyses of posttests showed significant increases in learning for both groups followed by significant drops in achievement on a delayed retention test. There were no statistically significant differences between the two methods on any of the cognitive variables investigated. The authors concluded that CAI can be as effective as more traditional approaches for teaching factual knowledge and applications of learned material. They also stressed that controlling the content of CAI mate- rials may inhibit effective utilization of CAI: it is how CAI is used as opposed to its usage that may determine its efficacy. Brebner, Hallworth, Woetowich, Mah, and Huang (1981) report on three experimental studies into the effectiveness of remedial CAI mathematics programs. The first study used a standard pre—post—retention testing assessment to measure performance and attitudes of students receiving adaptive CAI and traditional math instruction. The two groups were sta- tistically equivalent on all performance measures. The CAI group did take less time and had more positive attitudes 67 toward math. A second study compared adaptive CAI to indi- vidualized booklet instruction. A second independent varia- ble of student versus computer controlled routing (teacher control for the booklet group) through the material was con- sidered. Again, no significant differences between method of control conditions were found. Trends indicated, how- ever, that CAI instruction resulted in higher performance, that student controlled routing was inferior, and that the student controlled routing booklet group expressed negative attitudes toward that method. The third study, comparing adaptive CAI to linear CAI, resulted in no significant dif- ference. The overall conclusions indicated that CAI is no more effective than other techniques in terms of perform- ance, but does reduce instructional time. The authors also suggest that too much branching (adaptive strategy) may be counterproductive and confusing to the student. Deignan and Duncan (1978) compared CAI with programmed instructional text (PIT) and the traditional lecture method for medical training. Evaluations were made regarding per- formance, time savings, and attitudinal acceptance. CAI was superior over lecture in a medical laboratory training course and outperformed PIT in a radiology course. CAI time savings were 142 in the medical laboratory lecture and 122 over the programmed text group. Kamm (1983) developed 50 CAI tutorial units in physics which he used in comparison with a mastery model version of 68 the traditional lecture, i.e. lectures with retesting capa- bilities to ensure mastery of the unit material. The CAI model produced a decrease in course attrition and the number of unit assessment test retakes. These studies do not conclusively indicate that CAI is a superior instructional methodology. Many studies conclude that CAI is just as good as any other instructional tech- nique in terms of achievement. Indeed, when CAI is employed as a substitute to other strategies it is just an also ran. Edwards, Norton, Taylor, Weiss, and Dusseldorp (1975) found that 45% of the research studies they considered showed achievement gains when CAI was used to supplement tradi- tional instruction, 40% showed no difference, and 15% reported mixed results. A comprehensive survey of alternate instructional media conducted by Jamison, Suppes, and Wells (1974) concurs. They concluded that CAI was effective in reducing instructional time while maintaining achievement. Moreover, when CAI replaces the traditional lecture, higher achievement does not necessarily result. CAI augmentation of the traditional lecture appeared to be consistently pre- ferred over the extremes of all or no CAI. These authors also conclude that there is an overabundance of no signifi- cant difference studies and that CAI had not fulfilled its envisioned role. Despite an intervening decade, their conclusion is still appropriate. In a review of 92 studies involving comparisons of various educational techniques, Kulik and Jaska (1977) found 69 on the basis of final exam performances that 54 indicated superiority over the lecture format, 34 were equivalent, and only three were inferior. 0f the five studies specifically involving CAI, two were superior and three showed no differ- ence in final exam performance when compared to conventional methods. On average there was only a 4% increase in achievement over the lecture mode. In a meta-analysis of 59 CAI research studies, Kulik, Kulik, and Cohen (1980) found that a majority (37 of 54) favored CAI as a means of improv- ing student achievement. In 14 of 54 studies there were statistically significant differences favorng CAI. Thirteen of the 59 studies provided data regarding course completion rates. The results were inconclusive in that no systematic increase or decrease in attrition was found for CAI as com- pared to conventional instructon. The most dramatic find- ings came in the area of reduced instructional time for CAI. On average, when CAI is substituted for traditional lecture, instruction time is reduced by one—third. In one of the more comprehensive reviews of the liter- ature Rapaport and Savard (1980) compiled data on the issue of CAI as a supplement to or replacement of traditional instruction, its effects on student retention, and its efficiency as an instructional delivery system. As a sup- plement to more traditional methods, 114 studies reported increased academic performance, three showed declines, and six found no significant difference. Nine studies indicated that CAI as a replacement was superior, 11 had inconclusive 70 results, and there wasn't even one study that showed a decline in performance. Long term retention was negatively affected in 11 studies, positively affected in only two, and 22 investigations showed no appreciable difference. With respect to instructional time, all 21 studies that provided data found that students in the CAI environment worked approximately twice as fast as their lecture-based counter- parts. The overall conclusion was that CAI as a supplement was superior to a total replacemet of conventional methods and that it proved to save instructional time without com- promising academic performance. Orlansky and String (1981) investigated computer—based instruction for military training. The intent behind computer-based training (CBT) is to educate to the level of conventional methods but in less time. Less instructional time implies less cost (instructor and student wages) and more time to engage in field training and applications. This review aggregates 48 Army, Navy, and Air Force studies in which CAI or CMI methods were used. Of the 40 CAI studies 39 showed equivalent or superior levels of achieve- ment when compared to "lock step" lecture method. All eight of the CMI studies were equivalent to the traditional tech- niques. Since saving time is crucial to the military, all but four of the 48 investigations reported reductions in formal instructional time. CAI instructional time ranged from 31% longer to an incredible 89% shorter than the time necessary in the lecture—based sections. The median was a 71 30% reduction in time when the computers were used. Although CAI saved time (and consequently a projected $13 million per year) with no detrimental effects on achieve— ment, the drop out rates were substantially higher than for the lecture groups. While retention may be a higher pri- ority in academic institutions it is not of primary concern in military settings. The authors point out that the attrition/retention studies are suspect due to a generally higher attrition rate in the general military student popu— lation. These studies, however, lend support to the argument that computer—based education leads to higher attrition. In summarizing the experimental results regarding com— parisons of CAI to other instructional procedures there does appear to be a consensus on some issues. With few excep- tions, CAI is as good as if not superior to conventional methods on the basis of academic achievement. When CAI is implemented in a supportive, supplementary role, achievement gains over non-CAI paradigms are convincing evidence of CAI's effectiveness. When used as a replacement the results indicate that CAI is equivalent to any other technique. Although few studies have investigated long term knowledge retention, there is modest support for the claim that instruction via computers has a negative impact (Edwards et al., 1975; Rapaport & Savard, 1980). One of the major advantages for using CAI is the con- sistent finding of a time compression effect without any 72 immediate decrease in performance. Although savings of instructional time have their extremes, an average savings of 30—35% is frequently reported. One could argue that because there is a substantial reduction in formal instruc- tional time requirements, student retention in CAI courses should increase. The research literature does not support such a hypothesis. Indeed, the literature reveals that CAI has no systemic impact on student retention. Even though the results are mixed, drop out rates are probably influ- enced by factors other than instructional methodology. Strategies for Optimizing Instruction As mentioned in the historical perspective, there have been two predominant schools of thought on the control of instructional content and sequencing. Since the computer has the capability to make decisions, then why not utilize this feature to determine not only when but what instruc- tional event to present to the student? These decisions can be made at that point in the learning activity when and where it is optimal. Other researchers do not wish to second guess individual learners and therefore leave such decisions as what to study, in which order, and when to take the assessment tests entirely to the student. The debate becomes one of program versus learner control or to what extent should each one have its influence. Atkinson (1972) contrasted three CAI sequencing stra- tegies in the learning of German-English vocabulary pairs; 73 random order; learner-controlled; and response-sensitive. The experiment consisted of four learning trials followed by a delayed retention test. The random order strategy was superior in terms of initial performance while the response— sensitive method was the worst. Just the reverse was true on the delayed retention test. The results were not unexpected. During instruction the learner-controlled strategy permitted the student to select which items to study-—usually those that were missed on a previous trial. The same applied to the response- sensitive strategy. The randomized procedure not only pre— sented cases which were initially answered incorrectly but also pairs which the student correctly matched. The delayed retention test presented all of the vocabulary pairs, thus measuring how much of the list was actually mastered. The response-sensitive or adaptive model had proved its effec- tiveness. The fact that the learner-controlled strategy was weaker was attributed to the students being poor judges and consequently poor decision-makers regarding their level of understanding or actual progress. Brebner et al. (1981) reported similar findings in a comparison of computer versus student controlled routing through mathematics material. Judd, Bunderson, and Bessent (1970) developed three units of mathematics instruction in which four control stra- tegies were compared. The two extremes of program control and learner control were studied, as well as two strategies mixing the two. Total learner control consisted of being 74 able to select the instructional unit and which segments within that unit to study. The student could also skip around within the segment, i.e., skip over questions, jump to new problem sets, go back to the beginning of the seg- ment, etc. The order of the instructional units and the sequence of events within the units were predetermined for the other three control versions. On the basis of a pre- test, both of the mixed control strategies were routed through the materials under program control. Both of these groups, however, were advised as to whether the instruc- tional segment should be studied. If the learner elected to enter the segment the instruction was subject to the prede- termined order. One of the groups could skip around within the segment while the other was forced to proceed through the segment under program control. The order in which the segments were presented, however, was according to the pre— programmed sequence. Under complete program control the order of units, unit segments, and problem sets were pre- sented in the preset order. Although there were minor differences in achievement between the various control models for the three mathematics units, the authors concluded that all of the learner-control conditions were equivalent. The equivalence of the control strategies indicated that the students were capable of making sound decisions regarding the amount of practice they required on specific topics. From a programmatic vantage point the learner-controlled systems were of little academic 75 benefit over predetermined sequencing. The recommendation forwarded was that a certain degree of program cotrol should be used to direct students through material that may be new or of proven difficulty for the individual learner. Following this recommendation, decision-making capa- bilities were eventually designed (Tennyson, 1975) and incorporated into the management of instruction through the use of Bayesian statistics (Rothen & Tennyson, 1978). Unlike the program control strategies, adaptive control sys~ tems constantly adjust the number of instructional events the student will receive on the basis of the individual's on-task performance. As students work through the instruc- tional units, those students doing poorly will automatically receive more exercises, problems, or explanations; students doing well will receive less. Tennyson and Rothen (1977) investigated the effective- ness of adaptive systems by contrasting two adaptive designs and a non-adaptive model. A full Bayesian adaptive model, using data from a pretest and on-task information, and a partial adaptive design using pretest information only were compared with a program control paradigm in which students receive the exact same sequence of instruction. The results of the study convincingly demonstrate the effectiveness of the full adaptive control condition. Higher achievement levels in less time were statistically significant over the other two techniques. 76 Park and Tennyson (1980) extend this line of research by comparing different methods for initializing the adaptive control strategies. The number of examples selected in a concept learning activity was determined on the basis of pretest data only, on-task student input only, and a combi— nation of the two. For the pretest only group, the number of instructional events once calculated remained fixed through- out the experiment. The other two groups had the number of events modified during the learning activity. The empirical data indicated that while pretest information reduced the on-task time and the number of examples presented to the students, the on-task only data source proved more efficient in terms of total instructional time and total number of questions needed (the pretest made the difference). There was no difference in the overall performance of any of the groups. Although the three information sources used to drive the adaptive control models were equivalent in terms of achievement, the concept of adaptive control proved efficient with respect to the time and the number of learn- ing events. Tennyson, Tennyson, and Rothen (1980) investigated the effectiveness of two dichotomous control strategies. The type, amount, and sequencing of instruction according to a totally adaptive procedure were compared to a system under complete learner control. Students receiving materials administered under the adaptive CAI model outperformed stu- dents in the learner-control group. The learner-control 77 group took less time to complete the materials due in part to their ability to leave the instructional units and pro- ceed to the posttest. The reduction in time resulted in poorer achievement, however. In this study, students given absolute control over the instructional process tended to make inappropriate decisions regarding their level of under— standing and did not effectively use the instructional time or the available CAI capabilities. Even though the extreme case of total adaptive control appears to optimize performance, there are some drawbacks. Instructional time is not necessarily optimized and, more importantly, the student is relieved of all decision-making responsibilities for his or her own learning. If adaptive and learner control strategies represent the extremes of a continuum, how effective would a hybrid of these two models be? The problem with the learner-controlled system seems to be the students' inability to make appropriate decisions, usually by overestimating their level of understanding. If an adaptive control system can make decisions, then a learner-adaptive control strategy in which the learner makes decisions based on diagnostic and prescriptive information may optimize time and performance (Tennyson & Rothen, 1979). Tennyson (1980) experimented with the three management strategies of adaptive, learner, and learner—adaptive methods. The learner-adaptive and learner-controlled con— ditions permitted students to make decisions regarding the sequencing and the amount of instruction they received. The 78 distinction between these two strategies is that students in the learner-adaptive group were advised of their learning needs in relationship to an established criterion level. Essentially the computer operated in the adaptive mode "thinking out loud", making recommendations, and giving the learner the option of taking or ignoring the advice. The learner-adaptive model not only gave students control over the amount and sequence of instructional material but also provided advisement, diagnosis, and prescriptions relative to a preset achievement level on which to base their deci- sions. The learner-controlled strategy indeed proved to be inferior to the adaptive models with respect to achievement. The learner-adaptive model required less instructional time and fewer examples than the adaptive strategy, but more than the learner control method. In a subsequent study Tennyson (1981) replicated and expanded these findings. An experiment was conducted in which the adaptive model was replaced by a learner-partial control method whereby the introductory sections of the material were under strict program control but the practice sections were under student control. Using three units covering rule learning in English, the learner-adaptive strategy maintained its superiority in terms of performance. On-task time for the learner-adaptive group was longer than for the learner-control group, but less than the learner- partial control subsample. Moreover, the performance scores 79 of the learner-partial control group were not consistantly better than those of the learner-control group. In a review of learner control in CAI environments, Steinberg (1977) identified certain trends even though nothing was of statistical significance. In a learner con- trolled setting students took longer to complete a course (not necessarily CAI lessons) and did not perform as well as students in computer controlled situations (program or adap- tive control). Students did not seem to accurately assess their knowledge of, or progress toward, course objectives. If students were given control over the level of instruc- tional difficulty, they tended to make improper choices regarding their ability by working on units that were either too hard or too easy. Obviously these observations are more acute for the poorer learner or when the subject material is new to the students. Decision-making procedures in the adaptive paradigm focus on previous and/or current performance within the learning task. Information processing or mathematical learning models, in contradistinction, rely on the modeling of cognitive structures and methods of thinking (Atkinson, 1972; Suppes, Fletcher, & Zanotti, 1976). The emphasis of these models is to develop predictive mechanisms for indivi- dual student progress within a subject area. Predictive capabilities would then permit control over time/resource allocation so as to optimize the grade-placement gain of an 80 individual student. Suppes et al. (1976) developed a sto- castic differential equation indicative of the progress students make through a CAI curriculum. The quantitative model produced an equation characteristic of the CAI mate- rial but constrained by student parameters. The model was tested using 297 deaf students involved in a 14 strand CAI elementary mathematics package. The theory does surpris- ingly well in predicting the number of CAI lessons completed to grade placement. Predictive equations based on individual parameters were significantly better than the predictions generated from population parameters. Population parameters, however, produced an equation better suited for prediction purposes. The trajectory model, based on population parameters, pro- vided a means of investigating student progress through the instructional materials and may impact on the quantitative details of course organization. In a subsequent study Larson, Markosian, and Suppes (1978) used the trajectory model to fit performance data of college undergraduates in a logic course. The model was able to fit time data rather well, lending support to the use of such predictive measures as a control mechanism in the allocation of CAI access time given resource limita- tions. The trajectory model was found to be relatively stable after the first third of the course and may prove useful to students in assessing their progress within a course 0 81 A theoretical extension of the trajectory approach is forwarded by Malone, Macken, and Suppes (1979) in their discussion of six CAI time allocation strategies. Following Atkinson's (1972) lead, each strategy was geared to the optimization of some aspect of class performance. The six models investigated were 1) maximize mean grade placement, 2) minimize variance in grade placement, 3) maximize mean grade placement without an increased variance, 4) maximize the number of students at or above grade level, 5) maximize the number of students making a specific gain, and 6) an equal time paradigm for all students. These researchers used a grade placement equation based on CAI access times and individual student parameters to make the necessary predictions. The theoretical results indicated that little was to be gained by using anything other than the equal time option. A simple increase of CAI access is probably more effective than utilization of specialized allocation strate- gies that would benefit only a few students. Clearly, different control strategies focus on distinct aspects of instruction. As internal management strategies of CAI lessons go, the learner—adaptive technique appears to optimize both on—task time and performance while providing students the freedom and responsibility to make decisions regarding their education. The learner control aspect of instructional management can be of benefit provided that the students are continually informed of their progress and that the system provides meaningful advice on which to base sound 82 decisions. Decisions made in the absence of information (learner control) are clearly less desirable and inefficient at improving performance than the data driven adaptive models. The other extreme of program control, albeit adap- tive control, also seems ineffective at minimizing on-task instructional time. Since the adaptive strategy yields no apparent gain in performance, any management model that minimizes time would be preferable. The learner-adaptive technique provides such joint optimization of achievement and time reduction. If college and university administrators are confronted with insufficient resources, then trajectory models would be of benefit in the planning, allocation, and management of those resources in a way that is justifiable and equitable to all parties. The control strategies of Tennyson and his colleagues focus on optimizing student learning and knowl- edge acquisition within CAI lessons. A coupling of these two techniques providing learners information on their pro- gress within CAI lessons as well as within the larger con- text of a course or curriculum would be of obvious value. Feedback Options One of the highly touted features of CA1 is the ability to provide immediate feedback. Linear systems can inform the student of the correctness of an input answer. Branch- ing systems are capable of diagnosing errors and moving to the appropriate remedial section of the material. Some of 83 the most sophisticated systems of the knowledge based or intelligent generation will actually question the learner about his or her input and/or direct the student through the subject matter. The unique features of computer technology, however, make possible the parameterization of feedback. Gaynor (1981) contrasted four feedback conditions with performance on long and short term retention of mathematics material. The four treatment groups received 1) no feedback at all, or feedback 2) immediately after each question, 3) thirty seconds after each item, and 4) at the end of the CAI unit. On the seventh day of the study a short term reten- tion test was administered and a long term retention test followed two weeks later. When student entry levels were equated there were no significant differences in either long or short term retention of the material. Two trends did emerge from the analysis. Immediate feedback produced slightly better retention scores for lower ability students while and of unit feedback benefitted the more capable learners. The second trend was a slight decline in perform- ance scores for the 30 second delay group. Although not significant in a statistical sense, it did raise the possi- bility that such a delay may be counterproductive. When large computers become overloaded due to a high number of users, the delay induced may be detrimental to effective learning. Of interest, however, is the unexplained result that the group which received no feedback whatever performed just as well as students in the feedback conditions. 84 Rankin and Trepper (1978) set up immediate, 15 second delayed, and end of session feedback conditions in their study of knowledge acquisition of sexual facts. Their dependent measure was a 24 hour delayed retention test. The experimental results indicated that all three groups per- formed the same during the instructional trials but that the delayed feedback groups did significantly better on the retention test: there was no significant difference between the two feedback groups. The longer delay group, however, did have higher retention test scores and a smaller standard deviation than the 15 second delay group. Sturges (1978) used two second, end of session (20 min- utes), 24 hour, and no feedback conditions. The dependent measures were an initial test score, the score on a crite- rion test given 24 hours after the initial session, and the score of a retention test given one to three weeks after the conclusion of the instructional period. Long term retention was not enhanced by immediate feedback. Delayed feedback, however, promoted long term retention without retarding learning. The results supported what is known as the delay- 2° retention effect (DRE) observed in other studies (Kulhavy Anderson, 1972: Markowitz & Renner, 1966; Sassenrath & Yonge, 1968; Sturges, 1969, 1972: Sturges, Sarafino & Donaldson, 1968; Surber & Anderson, 1975). These studies have far reaching ramifications in learn- ing theory, curriculum construction and presentation, and the design of learning environments. Of particular interest 85 is the implementation of delayed informational feedback within CAI environments. All of the above referenced studies adjusted the time of the delay and in all cases feedback was automatically administered. None of these studies allowed learner control over the timing or amount of feedback. The unique capabilities of computer systems for controlling the delay and extent of feedback as well as the use of remedial sequences will certainly facilitate further research into the delay-retention effect. Anxiety and CAI Since 1956 the United States has been an information based society in that white-collar workers employed in the creation, processing, and transmission of information out- numbered blue-collar workers (Naisbitt, 1982). The change from an industrial society to an information society has progressed slowly, being barely noticed. The advent of microprocessors has changed an evolutionary process into a modest revolution. Microprocessors are used for control systems in practically every electromechanical device. Despite this influx of technology, major portions of the American society remain apprehensive about computer systems. There is a widening gap between the technically literate and technically illiterate members of our society, hence, the term "computerphobia" has been added to the vernacular. In recognition of this fear of computers, educational systems 86 and software companies have created computer literacy courses and user friendly systems. One major area of research has been based on the hypothesis that performance in a CAI environment could be improved by designing instruction responsive to learner characteristics. Anxiety has been one of the most fre- quently researched personality variables. Spielberger, O'Neil, and Hansen (1972) report on four studies involving CAI and anxiety levels. One study compared anxiety levels in a CAI task versus subsequent performance in a laboratory situation. The lab setting was more anxiety provoking as measured by A-trait and A-state scales. Other experiments investigated performance within CAI lessons of high and low A-trait students. Performance on CAI tasks was a function of A-state level (high A-trait students typically have ele- vated A-state levels) and task difficulty with high A-state students doing poorer on the harder tasks; no relationship was observed for easy tasks. None of the studies found any systematic relationship between A—trait and achievement in CAI units. Results of two experimental studies are presented by Leherissey, O'Neil, Heinrich, and Hansen (1973) which focus on the interaction of anxiety and several CAI design para— meters. One of these variables was user response mode to questions presented. Four possible options were used: no response; covert (blank spaces for students to "think" the answer); modified multiple-choice, and constructed response 87 (student types in the answer). Instructional materials con- sisted of familiar subject matter or technical information. High A-trait subjects had higher levels of A-state anxiety and the technical materials invoked a higher A-state differ- ential (between high and low A-trait subjects) than the familiar subject matter. Students in the no response and constructed response groups performed better than subjects in the covert or multiple-choice treatments. Students in the constructed response group had consistently higher A- state levels. A second design parameter studied was the length of the CAI units. The no response and constructed response groups were used in this study in which long and short versions of the instructional materials were compared. Students in the no response groups performed better than the students in the constructed response groups and completed the assignments in about half the time. Shortening the length of CAI units was not effective in reducing state anxiety. The no response mode was less anxiety producing and led to better perform- ance. In these studies on-task instructional time was not considered a critical variable for reducing state anxiety or improving achievement. Steinberg (1977) reviews several studies investigating anxiety levels under CAI learning conditions. One finding, which is not endemic to CAI, is the strong relationship between A-trait and A-state levels. On-task errors of high A-state students can be reduced by the use of various 88 feedback techniques (Leherissey, O'Neil & Hansen, 1971) and A-state levels can be reduced by providing feedback on the more difficult tasks (Hansen, 1974). The overall conclu- sion, however, is that anxiety has little usefulness in the design and presentation of CA1 materials. Tobias (1973a) states: "...that anxiety, while useful in other areas, has limited utility in the area of individualized instruction" (p. 237). Summary The research opportunities afforded by computer-based systems have opened diverse areas of design and application. This review has not focused on any specific academic disci- pline, grade level, or implementation strategy. The intent has been to introduce the reader to the diverse research areas, factors that can and are being studied, and some of the empirical data available to date. There are several reviews and meta-analyses of the literature that address the efficacy and effectiveness of CA1 as a practical instruc- tional methodology. Interested readers are referred to the reviews of Burns and Bozeman (1981), Chambers and Sprecher (1980), Edwards et al. (1975), Jamison et al. (1974), Kulik and Jaska (1977), Kulik et al. (1980), Orlansky and String (1981), Rapaport and Savard (1980), Steinberg (1977), and Vinsonhaler and Bass (1972). There is overwhelming evidence supporting the position that computer-assisted instruction is extremely efficient at 89 delivering instruction or training in a shorter time without compromising achievement. There is also an increasing num- ber of studies which support the premise that CAI is an effective educational tool, but little evidence that CAI is superior to other methods. Research into instructional con- trol strategies indicates that a learner-adaptive approach maximizes achievement while minimizing on-task instructional time. Control techniques for the timing and amount of informational feedback and its impact on the delay-retention effect can be easily studied using computer technology. Research into how student characteristics (anxiety, general ability, introversion-extroversion, field-dependence, etc.) interact with CAI, although limited, has failed to identify any specific affective variable that could be used in a pre- dictive sense. Regardless, the time compression and delay- retention effects as well as instructional control tech- niques and hardware capabilities have been the focal point for many CAI activities and the development of more sophis- ticated computer-managed instructional systems (Baker, 1981). Kearsley, Hunter, and Seidel (1983) summarize two decades of computer based instruction research. A few of their conclusions are pertinent to this review. 1. Computers can provide for more efficient and effec- tive instruction (time, cost, achievement, resource usage, etc.). 2. Despite claims for individualization we really know very little about how to do it. 9O 3. We have little knowledge regarding the impact of graphics, speech, motion, and other instructional variables. 4. Computer-based instruction has involved all areas of instruction and training for both research and applications. 5. The use of computers for education is still in its infancy and as such offers tremendous potential, which even after twenty years we don't fully recog- nize or completely understand. Only through continued research efforts can we gain the insights and knoweldge necessary for the full development and utilization of computer-based education. Aptitude-Treatment Interactions Research Methodology The previous two sections of this literary review have presented few conclusive results regarding achievement, despite the abundance of research data. A possible explana- tion for the inconclusive findings is that much of the research tries to determine the "best" method of instruc- tion. Indeed, much of the CAI literature seeks to determine whether CAI is better than alternate methods or to decide which design features work best for "most of the students." Irrespective of the potential offered by computers for indi- vidualizing instruction, efforts to actually do so have been largely unsuccessful. The problem may not be with the method or the design features but in the application. 91 Recognizing and planning for individual differences and administering the "package" to all students seems contra- dictory. If instructional systems are designed to differ- entially impact on the student population then it would be appropriate to apply the methods in a differential fashion. Given the presence of a diversity of instructional tech- niques and an equally diverse student audience, it is highly improbable that one method works best for all students in all situations. An effort directed at determining the optimal match of learner and instructional characteristics is generally known as aptitude-treatment interaction (ATI) research. ATI is not a separate, clearly defined research area but a method- ological approach to doing research. Cronbach and Snow (1977) provide a description of the method and evaluation techniques used in aptitude-treatment interaction research. They also provide a synopsis of ATI research findings and their implications for education. The goal of ATI research is to generate regression equations that can be used for the prediction of effective and efficient teaching methods on the basis of individual differences. The intermixing of student learning styles, teaching technique, curriculum design features, and situational factors influences achieve- ment. Hence, the ATI approach attempts to empirically define the optimal mixture. ATI research data are often statistically cast into linear (occasionally curvilinear) regression equations. 92 Regression equations are generated for each of the experi- mental treatments. There are three basic possibilities available when these regression lines are plotted on an aptitude versus outcome graph. If the two regression lines are parallel, then no interaction exists and the treatment that yields the desired results should be used for all students. If interactions occur they can be either ordinal or disordinal. In an ordinal interaction the regression lines are neither parallel nor do they cross each other within the range of values present for the aptitude measure. In this case a differential utilization of the treatments may be in order but the decision cannot be based on the aptitude measure alone. Other factors, such as treatment costs or duration times, resource availability, etc., must be considered in deciding whether multiple treatments are going to be supported, and if so, what student assignment procedures are to be used. A disordinal interaction occurs when regression lines cross within the range of the aptitude measure. Differential effects of the treatments, however, are not only dependent upon those factors mentioned for the ordinal interaction, but also on the statistical and practi- cal significance of the interaction itself. ATI research covers such a wide range of activities that the terminology is not uniform across studies. Ber- liner and Cahen (1973) in their review use the term trait- treatment interaction (TTI) although they prefer trait, treatment, and task interaction (TTTI). Tobias (1976) 93 replaces aptitude with achievement in the ATI acronym and has proposed the term attribute as one connoting a more uni- versal meaning. Even though these terms appear in the literature and have subtly different meanings, they all con- vey the idea that individual characteristics and educational treatments may interact and that these interactions may have potential use in how we, as educators, can better teach our students. Salomon (1972) argues that ATI research serves the two functions of improving instruction and developing principles concerning the nature of instruction and learning. He elu- cidates on three ATI paradigms: remedial, compensatory, and preferential. The remedial model assumes some missing ele- ment is responsible for a student's lack of progress through hierarchical or structured materials. The thrust of this model is to identify and fill the knowledge gap while mini- mizing time (and cost). This approach, however, is applica- ble if the variance in learning outcomes is directly related to task-specific capabilities, if the subject matter is hierarchical in nature, and if the subordinate skills are indeed learnable through instruction. Notice that this model does not necessarily rely on cognitive or personality characteristics, but focuses on expedient techniques for overcoming educational deficiencies. The compensatory model employs various treatments which have been designed to circumvent ineffective and unproduc- tive aspects of other treatments or learner abilities. The 94 treatment compensates for student deficiencies by presenting the material in an organized way; one that the student may not be able to provide for himself. For example, a well structured presentation with detailed feedback and a high degree of redundancy may benefit students who are unable to synthesize or organize the material on their own. Such a treatment, while helpful to one type of student, may be det- rimental to the learner who can effectively provide the necessary structure on his or her own. In this model, apti- tudes play a greater role than in the remedial model. This model is not concerned with changing student capabilities, just compensating for the deficiencies. The choice between the remedial and compensatory models rests on whether the learning capability (if identified) can be taught or simply averted. The preferential model is used to capitalize on exist- ing student capabilities. It does not make up for deficien- cies or circumvent them but attempts to match treatment characteristics with individual learning styles. The apti- tude measures for this paradigm typically consist of general ability, mode of information processing, and motivational factors. The intent is for each student to maximize his or her learning through the appropriate assignment of treatment conditions. Tobias (1976) labels this approach as the alternative abilities model, in which instructional treat- ments are designed to engage different student abilities and 95 interact with student aptitudes to produce higher perform- ance scores. From a theoretical point of view, ATI research would appear to hold the key to determining the method-situation- student mix that is optimal for learning. Tobias (1976) laments, however, at the lack of interest in ATI by writing, "Despite this persistent interest in individualized instruc- tion, there are few systematic attempts to adapt the method of instruction to student characteristics" (p. 61). McCombs and MacDaniel (1981) state: While much has been written concerning the desirability of an aptitude—treatment-interaction (ATI) approach to the individualization of instruction...there has been a notable lack of effort toward addressing the way educa- tors could implement such an approach. (p. 11) They proposed a computer-based training program which was an adaptive system couched in the preferential ATI model. Stepwise regression procedures on selected cognitive and affective pre-course characteristics were used to predict performance scores and unit completion times. Alternate instructional strategies differed in media form, format or style, difficulty, and/or use of special learning aids. Design features of the system allowed for low reading/ processing ability, low memory ability, and high test anxiety. The adaptive nature of this system did accommodate individual differences and led to higher achievement in less time. This type of approach parallels the adaptive system techniques of Tennyson and his colleagues discussed in the previous section. 96 The preferential ATI model appears to be receiving much of the research attention, especially for CAI applications. Research findings, however, are far from corroborative, much less conclusive. The next two sections will present some of the research findings, particularly those involving locus of control and/or computer-based instruction. For a more detailed treatment of the theoretical background and analy— sis techniques as well as a comprehensive review of the literature, interested readers are referred to the Cronbach and Snow (1977) text and the Berliner and Cahen (1973) article. Ability by Treatment Interactions A wide range of studies which have some measure of ability as the predicting variable can be grouped together. Ability may be formally measured via standardized tests (SAT, ACT, IQ, etc.), teacher made pretests, or GPA's. One of the larger areas of ATI research involved programmed and computer-based instruction. In most of the studies these instructional methods have been compared to other tech- niques, thus constituting treatment contrasts. However, there is a significant number of reports that use one speci- fic delivery mode and investigate design variations. Tobias (1969) and Tobias and Abramson (1971) used programmed instruction to study the interactions of response mode (reading the answer versus constructing the response) and the level of familiarity with the subject material. The 97 programs consisted of approximately 55 frames of familiar material and roughly 90 frames of technical material of which the students knew very little. Both studies indicated an attribute-treatment interaction on the unfamiliar techni— cal material favoring the constructed response mode. There was no difference in posttest scores based on response mode in the more familiar materials. This form of ATI would pos- sibly be classified as an ordinal interaction in which the constructed response mode is clearly favorable for material that is foreign to the students. Tobias (1973b) investigated the use of scrambled versus structured frame presentations on familiar and unfamiliar materials in a PI format. As one might expect, significant attribute-treatment interactions were found for the sequencing of PI frames but on the unfamiliar materials only. The study also considered whether generalized ability, as measured by the SAT, would serve as a stable predictor of achievement. No ATI involving SAT scores was observed. Tobias and Duchastel (1973), working in a CAI environ- ment, questioned whether the use of behavioral objectives interacted with the sequence of familiar and technical materials. The expected ATI between the use of behavorial objectives and sequencing was not found. Sequence, however, did have a significant main effect for students receiving the technical materials. Despite the absence of any ATI's, 98 the authors suggest that the CA1 and PI techniques demon- strate strong sequence effects on materials of a technical nature. In an early study involving CAI, PI, and standard textbook instruction, Schurdak (1967) used scores on the college level form of the Henmon-Nelson Mental Ability test as the main determinant of achievement. Although CAI was clearly the best treatment on average, no significant ATI's were found. The analysis revealed, however, that all three methods were equivalent for students with Henmon-Nelson scores above 80, while CAI was superior for students with scores below 80. More recently, Deignan and Duncan (1978) conducted a study comparing the effectiveness of CAI, programmed instructional text, and lecture methods. A series of pre- treatment aptitude measures was used to artificially define a tertiary aptitude scale. Although the design and analysis of this study was not in the preferred ATI context, results indicated that low level aptitude CAI students had an 182 higher achievement level over their lecture counterparts and 72 higher achievement with 17% greater time savings than the programmed text group. High aptitude CAI students had a 332 time savings over the lecture group. These inferred dis- ordinal ATI's support the recommendations to assign CAI for low ability students and programmed text to middle and high ability students if achievement is the criterion. CAI would 99 be used regardless of ability level if time was the cri- terion. Masuo and Furuta (1981) designed and conducted an ATI study involving CAI and PI as treatment variables and a pretest of the subject matter as the ability or aptitude measure. Regression analysis indicated a significant disor- dinal ATI for posttest scores on the basis of pretest scores and treatment method. CAI was a superior method for the low pretest scorers while high ability students benefitted from either instructional mode. The pretest measured the stu- dents' levels of previous knowledge and can be interpreted as a level of familiarity with the subject matter. Viewed in this context, this study supports the work of Tobias in that differential treatment effects occur in unfamiliar subject areas while no one particular instructional system is preferred by students of higher ability. The possibility exists that the differential effects are indeed present but the more capable students are able to compensate for the variations in technique. In a study of mathematics instruction Battista (1981) presented materials requiring spatial visualization (rota- tions of three dimensional objects). The ability measure was a specially constructed exam assessing student capabil- ity for spatial visualizations. One group of students received verbal instruction without the use of any visual aids. The second treatment group received verbal lectures supplemented by as many visual-spatial aids as possible. No lOO significant ATI's were observed although the treatment regression lines did cross, indicating a possible disordinal interaction. The ATI suggested that students of high pre- treatment spatial visualization ability performed better in the verbal only method of instruction, a reversal of the predicted result. Explanations for the non—significant, reversed ATI trend included a subject material dependence, differing mental process requirements of the treatments, and task difficulty differences. Janicki and Peterson (1981) investigated large versus small group instruction for the presentation of math mate- rials (fractions). A three level blocking on ability was employed based on scores from Raven's Progressive Matrices and the Sequential Tests of Educational Progress. The large group treatment involved lectures and individual seatwork. The small group treatment used lectures but seatwork was done in groups of four students: one high ability, two med- ium ability, and one low ability. With respect to achieve- ment, an ability by treatment interaction was non-existant. In a similar study involving geometry instruction, Peterson, Janicki, and Swing (1981) measured a curvilinear ATI between achievement and ability. The small group approach was pre- ferred by students in the low and high ability groups while the middle ability students performed equally well in either treatment. In a review by Cronbach and Snow (1977) 32 studies com- paring programmed instruction and conventional techniques 101 are documented. The studies ranged from one hour to one year in duration. Thirteen studies reported PI regression lines with slopes less than those for conventional instruc- tion. Four studies reported significant ATI's, eight showed weak or non-significant ATI's, and one was noncommittal. Students of lower ability performed better in all 13 of the non-conventional modes. High ability students were not dif- ferentiated by treatments on 12 of the studies. Five of the 32 studies reported PI regression lines with slopes greater than conventional instructional methods. Two studies quoted significant ATI's while the remaining three showed weak or non-significant interactions. In four cases low ability students performed better in Iconventional classes. High ability students had higher achievement levels in all five of the non-conventional methods. Fourteen of the 32 reports showed no interactions. In 25 of the 32 studies the ability measure was useful in predicting academic achievement in the non-conventional instructional treatments. Cronbach and Snow (1977) also report that ATI research into variations within the PI treatment has generally ended in failure. Effects of branching versus linear programming, small detailed steps versus large steps, scrambled versus logical sequencing, reading versus constructed response, etc., have generally proved inconclusive. The generalized claim that PI would allow low ability students to overcome any difficulty and achieve as much as high ability students receives little support. There have been isolated studies 102 that support this contention but a comprehensive review of the literature most often produces studies with no inter- actions, or, when interactions exist they are usually weak and rarely consistent from study to study. Computer-based instruction is the natural extension of programmed instruction and as such, research findings in which CAI is one of the treatments are no more definitive than those of PI. Since a majority of educational research efforts are correlational or of the "one best method" type, many of the interactions reported are not the product of an overt ATI design. Moreover, regression equations are gen- erally not presented in the event of significant interac- tions nor is the necessary data for generating them. Most studies, even in those where non-significant yet non- parallel treatment regression lines occur, do not apply the Johnson-Neyman (1936) procedure to determine those extreme ranges of the ability measure which do produce significant differential treatment effects. Berliner and Cahen (1973) report on one study in which 842 of the students had ability scores in a non-significant region but where the Johnson- Neyman technique identified the extreme regions where signi- ficant differences were observed (8% of the sample at each end of the ability scale). Nevertheless, Burns and Bozeman (1981) found in a meta-analysis of 40 studies that when achievement is the dependent variable, the CA1 treatment favored high achieving and/or disadvantaged students. Stu- dents of average ability performed well regardless of 103 instructional mode. Edwards et al. (1975) and Jamison et al. (1974) agree that CAI appears to promote higher levels of achievement for low ability or disadvantaged students. Whether the ATI is linear or curvilinear remains open for further research. Personality berreatment Interactions A second major variable category for predicting achievement is personality traits. Personality variables are difficult to isolate because of their interactions with other personality measures. Anxiety is one of the most researched areas with respect to its effect on P1 or CAI methodologies. The studies of Tobias (1973b) in a PI format and Tobias and Duchastel (1973) with CAI, used A-trait and A-state as the aptitude measure. Both of these studies failed to find ATI's for performance. Anxiety was influen- tial in increasing on-task error rates in the CAI materials but not on posttest scores. In their study of subject familiarity and response modes in a PI environment, Tobias and Abramson (1971) used Alpert and Haber's Achievement Anxiety Test (AAT), composed of facilitating and debilitating anxiety subscales. Facili- tating anxiety did interact significantly with response mode (reading versus reinforced constructed). Debilitating anx- iety was neither a significant main effect nor was it involved with any interactions. The authors conclude that their study along with others (Ripple, Millman & Glock, 104 1969: Steinberg, 1977) provides little support for aptitude- treatment interactions between anxiety and programmed instruction. In a PI environment Tobias (1969) investigated the applicability of creativity, measured by the Remote Associ- ates Test, for predicting achievement in different instruc- tional response mode treatments. The hypothesis that highly creative individuals would do poorly in a constructed response format (as opposed to simply reading the answers) while less creative students would do better was not sup- ported. Creative students obtained higher achievement scores regardless of response mode. Ripple et al. (1969) contrasted conventional instruc- tional and PI strategies in an attempt to identify disor- dinal interactions based on exhibitionism, compulsivity, and convergent minus divergent thinking. Programmed instruction was hypothesized to produce higher achievement scores for students with low exhibitionism, high compulsivity, and high levels of convergent minus divergent thinking. None of the 36 separate analyses dealing with these three characteris- tics were significant: conventional instruction proved to be superior to P1 in all cases. Domino (1971) sought interaction effects between achievement orientation and teaching style. Extremely high and low scoring students on the achievement-via-conformity and the achievement—via-independence scales of the Califor- nia Psychological Inventory were assigned to lecture 105 sections taught in either a conforming or independent man- ner. A significant disordinal interaction resulted, as well as a consonance between student orientation and instruc- tional methodology which led to higher academic performance. The "congruence" concept discussed in the review of the locus of control literature seems to parallel these find- ings. Internal students performed better under skill condi- tions while external students worked better under chance conditions. Hoffman and Waters (1982), in one of the few studies of the relationship between student affective characteristics and CAI, used the dichotomous scales of extroversion- introversion, sensing—intuition, thinking-feeling, and judging-perception as independent predictors. These four scores were derived from the Myers—Briggs Type Indicator instrument. Results of a seven week course showed that sensing type students had higher retention rates and quicker completion times. Extrovert/perceptive type students had the highest attrition. The thinking-feeling scale didn't appear to be an important factor. Even though the study is not of an ATI design, specific changes to the CAI materials for the extroverts, intuitive, and perceptive type of stu- dents were indicated in order to keep them motivated and interested. Goldberg (1973) reports on an extensive search for ATI's between student personality measures and learning 106 conditions. Over 800 students were assigned to two instruc- tional conditions (lecture versus self-study) crossed with two methods of assessment (multiple-choice quizzes versus integrative papers). The personality measures, gathered from an extensive battery of questionnaires, inventories, and tests, yielded over 350 scores (students answered roughly 3500 items). Data from the three general dependent variable categories of course content knowledge, amount of extra-curricular reading, and degree of student satisfaction were collected. Despite what can only be considered a "shot-gun" research approach, the number of significant ATI's identified were less than the number expected by chance alone. Specially constructed personality assessment scales designed to elicit ATI's failed to do so in a cross- validation study. Unfortunately, the analysis technique used in the report was largely correlational and not regres- sional, even though the author addresses the different methods. The paucity of ATI's generated by such an approach is not only disappointing but confusing and difficult to interpret within any meaningful context. In somewhat of a variable role reversal, Corno, Mitman, and Hedges (1981) questioned whether different instructional procedures could change levels of anxiety, self-esteem, locus of control, and attitude, and whether a measure of general mental ability was a viable predictor of such changes. A three level teacher training program and a two level learning skills program (administered by parents) were 107 used as treatments. Significant ATI's were found for atti- tude toward school and anxiety as a function of general men- tal ability and the learning skills program. The relation- ship involving anxiety was a three-way interaction between general ability, anxiety, and the learning skills program. The results also showed no significant changes in the stu- dent locus of control variable. The study, despite the com- plexity of the results, lends support for the existence of ATI's between general mental ability, motivational varia- bles, and instructional treatments. This and similar studies are important in that most ATI research uses measures such as anxiety and attitude as predictors rather than as dependent variables. This particular study, how- ever, did not differentiate between state and trait anxiety, thus raising a question of the permanence and generaliza- bility of the results. The fact that locus of control, con- sidered a relatively stable personality measure, was unaf- fected suggests that a distinction between "state" and "trait" measures is in order. Smith (1973) also investigated possible changes in student personality measures due to experiences in a CAI setting. The subjects were junior high school students of whom approximately 752 were from Mexican-American back- grounds. The sample was given the Sears Self-Concept Inven- tory, Coopersmith Self-Esteem Inventory, and a modified version of the Crandall Locus of Control Instrument prior to 108 the beginning of a ten week mathematics course and immedi- ately afterwards. Pretest-posttest scores for the non-CAI group were relatively stable. Posttest scores for students in the CA1 group were less predictable for the self-concept scales and the locus of control measure. The author con— cludes that for the locus of control data, the slopes of the CAI and non-CAI regression lines were significantly dif- ferent. The locus of control instrument consisted of three sections, modified versions of five items from each of the IAR+ and IAR- subscales, and the three item Fate Control Scale (Coleman et al., 1966). The locus of control measure is, therefore, a shorter, highly modified version of the original IAR instrument. Moreover, pretest-posttest cor- relations for the IAR+ and IAR- subscores were not signifi- cantly different for the non-CAI and CAI groups. The signi- ficance appears to be due to the three Fate Control items. Regrettably, no further discussion of what appears to be a significant ATI was presented. Janicki and Peterson (1981) and Peterson et al. (1981) used the Academic Achievement Accountability locus of con- trol questionnaire as a predictor of achievement for small versus large group instructional models. Both of these studies investigated several other personality and attitu- dinal measures. To reduce the multicollinearity of the variables used, a factor analysis was performed and a single factor of attitude toward math and locus of control was constructed. The factor loadings were .94 and .52 (Janicki 109 & Peterson, 1981) and .47 and .60 (Peterson et al., 1981) for the attitude and locus of control components respec- tively. The first study reported a significant disordinal ATI on achievement for the small versus large group instruc- tion predictable on the basis of the attitude/locus of con- trol factor. The small group method proved successful for 35.5% of the students in the study. The large and small group approaches were statistically equivalent for the rest of the students. The attitude/locus of control measure revealed a signi- ficant disordinal interaction for scores on an attitude toward teaching approach scale. Students with positive attitudes toward math/internal locus of control preferred the large group method. Students with poorer attitudes/ external orientations preferred the small group approach. No ATI between achievement, the attitude/locus of control factor, and instructional approach was found; the achieve- ment ATI reported by Janicki and Peterson (1981) was not replicated in the Peterson et al. (1981) study. Summary The intent of this brief review of ATI research was to provide a modicum of background information on the basic concept and goals of this research approach. If one assumes a normal bell-shaped distribution of performance scores, then one would not be in error in stating that 17% of the students do quite well, 66% do average work, and 17% never 110 really get on track. Some would point to such distributions as an indictment of the educational system's inability to adapt instructional practices that account for individual differences. It is indeed foolish to attempt to eliminate or even mollify individual differences in learning. More- over, educators can no longer ignore such differences. Bell-shaped distributions will always exist; it is the posi- tioning of the distribution on the achievement scale that is important. It is also important to identify those ATI's which will truly benefit the students and the educational system in general. Bracht (1970) reviews 108 ATI studies. Each study was classified according to three dichotomous scales: treatments (controlled or uncontrolled); personological variables (fac— torally simple or complex); and dependent measures (general or specific). Of the 108 research studies documented, 103 of them reported ordinal or no interactions and only five showed disordinal interactions. On the basis of these data disordinal ATI's were more probable for controlled treat- ments, i.e., subject to little external influence, and more probable for factorally simple personological variables, i.e., variables having low correlations with other persono— logical variables. Since most of the studies had a specific dependent measure, little is known regarding the effects of the dependent measure. The ordinal interactions should not have been grouped with the no interaction studies because the possibility exists that significant treatment effects 111 may be present for certain extreme sections of the aptitude variable (determined by the Johnson-Neyman technique) or that other indirect factors may have a substantial impact upon any decisions regarding support for different treatments. Berliner and Cahen (1973) conclude their review of the literature by stating: In general, significant interactions are not a rare occurrence, and interactions have important impli- cations for the design of instructional treatments. ...most studies of interaction have not been repli- cated; when replicated, interactions have not been confirmed. (pp. 84-85) The research reviewed herein is clearly in support. The diversity of aptitude and personological measures, and the variations in instructional treatments creates an extremely complex network held together solely by the ATI philosophy. There appears to be very few rules; personality measures are used as predictors as well as dependent variables. Highly unstable and/or temporal variables are used as predictors, not easily generalizable even in the event of significant interactions. When significant interactions do occur they are rarely the product of an overt ATI design and they are seldom fully developed into useful regression equations. For the factors pertinent to this study, CAI as a treatment is beneficial for students of lower ability or in need of remedial assistance. Most studies using CAI, or its predecessor PI, as one of the treatments typically fail to have any impact for high ability students. The reasons for 112 these differential treatment effects is not clearly under- stood. Research into variations within the CA1 methodology has been unsuccessful in identifying the causal agents. It may simply be the entire approach that assists less capable students to show increased performance. Lower ability stu- dents are probably utilizing CAI materials in the remedial or compensatory formats which appear to be unnecessary for students of higher ability. Despite the desirability of the preferential model, little is actually known regarding its effective implementa- tion; we simply don't have the knowledge or the ATI's on which to intentionally make student-treatment assignments. Reece and Gable (1982) developed an attitude survey which was used to elicit student feelings about computers. Factor analysis reduced a 30 item survey to a ten item general attitude toward computers questionnaire. This instrument was promoted as a means of identifying students for CAI assignments or general computer usage. Posner and Osgood (1980) comment that computer availability alone was not suf- ficient to attract student use and that many students were reluctant to use computer facilities. To overcome this pro— blem they had to prepare a special course to provide a "threshold of familiarity with the computer" (p. 92). The task consists of formulating a theory which can be used to prescribe the effective and efficient use of current educa- tional technology in a manner that is non-threatening and even inviting. 113 The locus of control variable has been used as both an independent measure and a dependent variable. Studies which used locus of control as an independent measure confounded the experimental results by pairing it with other persono- logical measures. These studies, however, do imply that locus of control may interact with treatment methods to pro- duce differential achievement effects. Studies that treated locus of control as a dependent variable find it to be relatively stable over the duration of the experiment. These findings are important in that the locus of control concept appears to be a stable "trait" type of measure potentially capable of predicting treatment- dependent achievement while remaining resistant to change over the duration of the treatment. The focus of this study is to investigate whether the personality measure of student locus of control interacts with the instructional treatments in a differential fashion. The ATI research model provides the requisite underpinnings that connect the variables, methods, and evaluation pro- cedures. Tobias (1976) provides an appropriate quote to close this review of the literature by stating that: ...the bulk of the work remains to be done, and the viability of the ATI construct for the illumination of our understanding of instructional events, as well as for advancing practice to the point where instructional prescriptions can be made, is still to be demonstrated. (p. 63) CHAPTER III DESIGN AND PROCEDURES Introduction This chapter presents a discussion of the research questions and hypotheses, the experimental design used to resolve the questions, and the procedures followed in the preparation and execution of the study. Design aspects for this investigation involves the CAI delivery system, the courseware used, and the experiment itself. After a brief discussion of the research questions and the related hypotheses, these three design issues are presented. The experimental design over time and variables is developed. The statistical analysis procedures employed are then dis- cussed as are the reliability and validity data of the assessment instruments. Following specification of the independent, dependent, and covariate measures, predictions of the research results are presented. The chapter concludes with a discussion of the specific procedures involved in the preparation and implementation of the study and is provided to assist in the replication and/or continuance of this line of research. Research Questions The primary focus of this study was to investigate the interaction between student characteristics and the mode of 114 115 instruction. The study specifically addresses the following questions: 1. Will the distinctly different strategies of tradi- tional lecture and computer-assisted instruction differentially affect academic achievement? 2. Can levels of academic achievement be predicted for the different instructional techniques based on student locus of control measures? 3. Does there exist an interaction between instruc- tional methodology and student locus of control? 4. Are the results generalizable across student sam- ples? Research Hypotheses The review of the literature identified several factors that are results of, or directly related to, previous research efforts in the area of CAI. Prior knowledge, gen- eral ability, and on-task time have been linked to perform- ance levels in CAI settings. Study or non-formal instruc- tional time is one variable that is rarely considered or controlled in research studies. These variables as well as class performance levels, measured by scores on homework, quizzes, and unit tests, were measured and considered as possible covariates in the data analysis procedures. Addi— tional measures of prior knowledge and ability considered were ACT and SAT math scores, the MSU math placement score, and the previous math course grade. The independent measures were scores on the Intellect- ual Achievement Responsibility Questionnaire (Crandall, et al., 1965). This particular scale contains two subscales. 116 The I+ subscale is a measure of a student's acceptance of success. The I- subscale assesses the level of responsibil- ity a student assumes for failures. A third score, I total, is the sum of the two subscale scores and indicates the degree of acceptance for successes and failures in academic settings. High scores on these scales represent an internal orientation, that is, success/failure is a direct result of the student's actions. Low scores indicate a student's belief that success/failure is not a direct result of his or her actions, but primarily due to external agents. These three scales were used in the analysis. The treatments were the distinctly different educa- tional delivery methods of traditional lecture and computer- assisted instruction. The dependent variable was student achievement on a specially constructed unit examination for a College Algebra course, Michigan State University course Math 108. The following hypotheses are grouped according to the order of the research questions and the statistical design used in the analysis of the data: analysis of variance: analysis of covariance; and linear regression analysis. For the analysis of variance and covariance, the locus of con- trol measure was used as a blocking variable in which stu- dents scoring above the sample mean on the IAR total score were classified as internal subjects. Those below the sample mean were classified as external subjects. This bi- level blocking is preferred for ATI research (Cronbach & 117 Snow, 1977) because it improves the power of the statistical analysis This is students external research research 1. procedures by balancing treatment group cell sizes. particularly important when the total number of involved in the study is small. The internal- delineation is also typical of locus of control (Rotter, 1975). Subject to these conditions, the hypotheses for this study are: There will be no difference in the mean achievement scores for students in the traditional lecture and computer-assisted instructional treatment groups. There will be no difference in the mean achievement scores for students classified as internally or externally oriented on the basis of the total IAR score. There will be no achievement interactions between instructional treatments and locus of control classifications; all students will perform equally well regardless of instructional method or locus of control orientation. There will be no difference in the mean achievement scores for students in the traditional lecture and CAI treatments after covariate adjustments. There will be no difference in the mean achievement score for internally or externally oriented stu- dents after covariate adjustments. There will be no achievement interactions between instructional treatment and locus of control classifications after covariate adjustments. There will be no difference in the slopes of the regression lines (academic achievement as a func- tion of locus of control measures) for the lecture and CAI treatments. There will be no extreme locus of control condi- tions where a differential application of instruc- tional methodology is warranted based on signifi- cant differences in achievement scores. 118 222132 CAI System Design The CAI system used in this study was developed and written by the author (Hamilton, 1981). The system is designed to be curriculum independent and to optimize com- puter capabilities via graphics, calculator functions, and instructional tailoring based on the academic history of the user. The system uses a 48K Exidy SorcererTM microcomputer and the Exidy Disk Display Unit. This unit contains a 12 inch P31 video monitor and a dual disk drive system. The system uses single sided, soft sectored, 5.25 inch disketts and operates under CP/M version 1.42/3. All software and courseware are stored on floppy disks as are all ancillary data management, student tracking, and analysis programs. All programs, with two exceptions, are written in Micro— soft's MBASIC version 5.03. The exceptions are the main instructional delivery program and the instructional author- ing program, both of which are written in Z-80 assembly language. The goals of the computer driven delivery system are: 1. To provide a versatile instructional system for stand alone use or as a supplement to traditional information delivery methods. 2. To provide instruction on an individual or "tail- ored" basis. 3. To incorporate an instructional management system where the sequencing of materials is determined by diagnostic analysis of student input and perform- ance history. 119 4. To Optimize existing computer capabilities for edu- cational purposes. 5. To create a system easy for student and instructor use. The CAI delivery system does not contain the courseware used in the study. The delivery system simply processes the courseware units which have been created and stored separ- ately. This separation of functions highlights the distinc- tion between software (data processing programs) and course— ware (instructional units). This separation also allows for a wide variety of subject matter and instructional designs to be prepared and investigated. The CAI management system can present materials at three levels of difficulty depending upon the student's current capabilities. The system can raise or lower the difficulty parameter on the basis of student responses. The system uses this difficulty parameter to pre-configure an instructional lesson prior to its actual presentation, so there are intra-lesson adjustments. The program does not permit intra-lesson branching although inter-lesson branch- ing is allowed. Inter-lesson branching can be forward, backward, or through the same lesson at the same or a dif- ferent difficulty level. Sequencing decisions are based on algorithms provided by the curriculum developer and the decision-making procedure occurs when a lesson is completed. Decision algorithms can be prepared for sequencing, diffi- culty level adjustments, and in extreme cases, preparing a 120 printed copy of the lesson for the student to take as study material. The CAI delivery system also contains calculator capa- bilities, allowing students requiring assistance to use the computer to perform numerical computations. Mathematical expressions can be input by the user and evaluated by the system. The calculator mode supports six mathematics opera- tions and ten functions. The curriculum developer or course instructor can spe- cify whether students have access to the calculator mode, control of forward and/or backward frame advances, and if the unit is to be presented according to the user's assigned difficulty level. These control options are an integral part of each instructional lesson. This organizational structure provides for a CAI delivery system which is subject indepen- dent, responsive to instructor control, yet sensitive to student ability. Courseware Desigg Courseware preparation often uses special processing codes (Hamilton, 1981) or "dot" commands (Jelden, 1981) for the identification and specification of text processing pro- cedures. The CAI system developed for this study supports 18 such text processing functions. Appendix A provides a brief overview of the CAI system configuration and a description of the special text processing capabilities. 121 The courseware selected for this study covers subject material from the College Algebra course (MTH 108) offered under the auspecies of the Michigan State University Mathe- matics Department. The material covered during the last two weeks of the course was selected for the study. Three one- hour CAI lessons were prepared by the author which utilized the processing features of the software system. The subject matter was based on the lecturer's notes and the textbook (Hestenes & Hill, 1981). The six major topics and the CAI lessons were: Complex Numbers and Complex Roots of Equations (Lesson 1); Polynomials, the Remainder and Factor Theorems, and Synthetic Division (Lesson 2); and Zeroes of Polynomials and the Rational Root Theorem (Lesson 3). The courseware was prepared using a learning system design methodology (Davis, Alexander & Yelon, 1974). Each instructional lesson was composed of a linear series of frames or pages (a page could consist of several complete video screens of informa- tion) and the three lessons constituted a hierarchical sequence of learning materials. The first few pages of each lesson presented a list of objectives for the lesson and a list of references where supplemental information could be obtained. The body of the materials centered on the stated objectives as did the examples and problems within the les- sons and the unit exam questions. Required student response to questions and problems was analyzed for correctness. Most of the questions used a multiple-choice format. Occasionally a problem requested a 122 numeric answer which had to be specifically typed in. Feed- back was tailored to the specific response given. Correct answers were indicated by a short word of encouragement. Incorrect answers were countered with a brief discussion of the error and often helpful hints regarding the intent of the question. Most of the questions were capable of this diagnostic feedback approach. Feedback, however, was pro- vided immediately after each item. Since the instructional materials were of a mathematics nature, a random number generator was used to prepare dif- ferent numerical values or to select textual phrases for insertion into problem or question statements. Students could work through the same lesson several times and, although the same problem might be presented, it would con- tain different numerical values, forcing the student to rework the problem. Courseware design often utilized a fea- ture for randomizing the multiple-choice foils, thereby pre- venting memorization of the correct answer key or its posi- tion. One unique capability of the computer-based instruction system is the ability to control the display of the steps presented in a problem-solving sequence. The computer would present one line of a multi-step sequence and wait for the student to respond before the next line was displayed. A textbook simply displays all the lines in the solution. This control aspect not only requires active student parti- cipation but it gives the learner the opportunity to work 123 out the next step of the solution prior to having it dis- played. The courseware was also designed to allow students access to the calculator mode and to skip around within the lesson. The last page of each lesson contained a summary of the topics discussed and references to the position in the lesson of the major topics. This was included so that stu— dents wanting to review the materials and rework the prob- lems could do so. Research DesiggyOver Time The design over time describes the temporal sequence of experimental events (Campbell & Stanley, 1963). The nota- tion used provides the relationship between treatments, measurements, and time. The dependent variable, achievement on a unit exam, was studied for each of the two treatment groups. Each of the groups were created by a stratified random assignment based on a dichotomized locus of control (LOC) scale. The IAR Questionnaire (Crandall et al., 1965) was used to measure student locus of control. Permission to administer this instrument had to be acquired from the Uni- versity Committee on Research Involving Human Subjects (UCRIHS). Permission was granted and the appropriate docu- mentation can be found in Appendix B. The following graphi- cal representation describes the research design over time. Lecture SR X1 0 CAI SR X2 0 124 The "SR" indicates that students previously classified as internal or external on the basis of their locus of con- trol scores were assigned to treatments by a stratified random technique. "X1" represents the lecture method of instruction while "X2" is the CA1 system. Lastly, the "0" indicates the achievement observation, i.e., performance on a unit exam. There are no subscripts for the observation because the same exam was taken by all students regardless of treatment assignment. The time line proceeds from left to right. Variable Selection and Analysis Techniques The research hypotheses presented earlier require three distinct methods of analysis, those being analysis of vari— ance, analysis of covaraiance, and linear regression analy- sis. This experiment used the two distinct treatments of lecture and CAI and the artificially constructed two-level blocking variable of internal and external locus of control. The experimental matrix, therefore, takes on the form of a 2-by—2 completely crossed design. The design matrix is illustrated in Figure 1. A two—way analysis of variance (ANOVA) performed on the achievement measure provides the means of addressing the first three stated hypotheses. The results of an ANOVA generate information on the instructional mode main effect (independent of locus of control), a locus of control main effect (independent of instructional method), and the Nahum: mwfimmn :uummmmm .H ouswwm 125 when mum: cog when com Hangman: uoa mansmm HmeumueH Hmuoh Hmeumuxm Hench uummmm samz mama muemvsum mucousum HmH mo._uqo mes um sauusmsamfim * oo.m Ho.sH oo.mo oH mascam *Noo.o as om.m owo.o 0H.H on.~ on.mH mm.om mm House: musmmmz unmam>mwnu< «5.0 s~.m mo.s~ as mascam sso.o as om.o smo.o so.H mm.o om.m mm.s~ mm Hoses: macaw sauce m .noum m=Hm> uoupm .>mn new: mmmmu mo HmHuH HfimHIN H HHmBIN m .cum .uum nonssz umm mucmfium> umaoom mououm uemso>mfinu< was mmwcu< 92H me 'I L :5 III musuumq I l L .1 OH ON on ow om 00 on ow ow OOH 91038 nuamaAarqgv 92H N whawflh mmafi wcwuqm Hmanu< "show me» mo macaumcvm .mucz om.¢h wa.l om.oo wc.l os.ow h~.I a ucmsm>manu< mucmauammmco scammmuwmm unmeaa Na maan 168 the slope of the line and where it crosses the vertical axis. These are listed as "Slope" and "Const." respec- tively. The ANOVA and ANCOVA analyses indicated that ATI's are non-existant if the IAR total score is used as the indepen- dent, predictor measure. Indeed, the slopes of the regres- sion lines under the "IAR Total Score" heading are all nega- tive and roughly equal; the lines are almost parallel. Fig- ure 2 gave preliminary indications of the parallel nature of the IAR total score regression lines. Linear regression analysis using the 1+ and I— scores as predictors produced interesting results. The regression lines for each treat- ment are quite similar for each trial separately. Comparing the lines between terms shows a reversal from Winter to Spring. The Winter sample is described by descending lines for the I+ scale and ascending lines for the I- scale. Just the opposite is true for the Spring term sample. The com— posite sample has the most interesting combination. For both of the subscales the regression slopes are of opposite sign. One line is rising while the other is falling. Fig— ures 3 and 4 graphically show the regression lines. If 1+ is used as the predictor, the lines indicate an ordinal interaction (non-parallel, non-crossing). The I- predic- tions, however, are indicative of a disordinal interaction (regression lines cross within the range of the predictor measure). The fact that the lecture line is falling in the I+ figure and rising in the I- graph while the reverse is Achievement Score 100r 604 ' 50' Regression Lines for I 169 Ach a .42 1* + 63.7 -— Lecture --- CAI 5 10 15 17 1+ Subscore + ubscale Composite Sample Achievement Score 170 100V 90' Ach = .33 I" + 73.6 80 5.. I... “Ilse. 7O ' ..~...-~~ Ach = -.77 I' + 79.8 -' 60 r -—— Lecture --- CAI 50 I 5 10 15 17 I- Subscore Figure 4 Regression Lines for I- Subscale Composite Sample 171 true for the CAI lines explains the reason for nearly paral- lel lines for the IAR total score. Figure 3 indicates that lecture students scoring low on the I+ scale tend to gain higher scores than their CAI coun- terparts. Since the regression lines cross at I+ - 19.3 which is outside the range of the 1+ scale, students scoring high on this scale showed no apparent achievement differen— tial due to treatments. The opposite tends to hold using the I- scores as the independent measure. Students scoring higher on the 1- scale seem to benefit from the lecture while all low scoring students perform equally well in either treatment. Whether these trends are indeed statistically signifi- cant effects is determined by a test of parallelism. The following discussion is patterned after a procedure pre- sented by Johnson and Jackson (1959). A complete formula- tion of these results is found in Appendix E. The calcula— tions were performed for the I+ and I- scales on the compos- ite sample data since these were the only two cases where regression slopes had opposite signs. The linear regression equations for each treatment can be written as: N) = A1 + BlX (Lecture) and Na = A2 + BZX (CAI) 172 where 2 represents the predicted achievement score and X is the locus of control measure of interest. The A's and B's are the regression parameters presented earlier. To test whether the regression lines are parallel the null and alternate hypotheses can be written as: HA: (A1 — A2) + (B1 - B2)X # O The procedure, in outline form, is to demonstrate equival- ence of treatment variances with respect to the E distribu- tions, test whether B1 - B2, and then see if A1 - A2. The F-ratios associated with these three steps for each of the 1+ and I- calculations are presented in Table 13. None of the F-ratios are significant at the 0‘: .05 level. It is apparent that the regression lines are statistically paral- lel. The trends indicated in Figures 3 and 4 are not sta- tistically significant. The quantitative nature of the regression analyses confirm the ANOVA and ANCOVA results that the treatment groups are equivalent for all levels of locus of control--subscales as well as the total score. These analyses give insufficient grounds for rejecting the null hypothesis for the total IAR score or the two sub— scales; the regression slopes are equal. Indeed, since A1 a A 2 the lines themselves can be considered statistically equivalent. 173 Table 13 Test of Parallelism 1+ and 1- Regression Equations Composite Sample A Predictor 6; F Constant F Slope F Lecture 18.34 94.0 —1.15 I+ 1.21 2.91 12.47 CAI 16.70 63.7 .42 Lecture 18.47 73.6 .33 I- 1.23 2.72 27093 Note. F-ratios for significance are 1.94, 4.04, and 252. Johnson-Neyman Procedure The last issue centers on the hypothesis: 8. There will be no extreme locus of control condi- tions where a differential application of instruc- tional methodology is warranted based on signifi- cant differences in achievement scores. The equivalence of regression equations for all of the IAR scales precludes the need for conducting any analysis designed to resolve this hypothesis. The Johnson—Neyman procedure is one such technique. The lack of any signifi— cant ATI forces the decision to accept the null hypothesis that no region exists whereby one treatment is preferred on a statistical basis. In view of the acceptance of the equivalence of the treatment regression lines for both 1+ and I- as predictors, the present null hypothesis must be 174 accepted; no extreme regions of the locus of control measures predict significantly different achievement levels between the two treatments. In view of the decision to accept the null hypothesis, a presentation of the Johnson-Neyman procedure is inappro- priate at this point. This procedure is used to determine regions of statistical significance in the event that sig- nificant interactions occur. For those readers interested in the concept and underlying philosophy of this procedure, Appendix E contains a discussion and the detailed mathemati- cal calculations of the technique as it applies to the 1+ and I- composite data. StabilitygAcross Samples The results of this study are consistent for the two trials. The fact that no significant differences in main or interaction effects with and without covariate adjustments occurred for either the Winter or Spring samples emphasizes the acceptance of every null hypothesis. The regression equations do differ for the two separate trials. Table 12 presents the parameters for the linear regression lines. The slopes for the Winter and Spring trials are of opposite signs on the I+ and I- subscales. The I+ scale has negative slopes for the Winter trial but positive slopes for the Spring sample. The opposite trend occurs in the 1- scale. The slopes for the two treatments 175 within each of the samples, however, are the same. More- over, the slopes of the treatment lines using the IAR total score as the independent variable have the same sign for both trials. None of these trends, unfortunately, have any statistical significance. The question of whether those few individuals selected for the CA1 treatment but electing to remain in lecture influenced the outcome of the study was raised in the pre- vious chapter. Table 14 presents the mean achievement scores of those students who elected to remain in the lec- ture. As much as can be expected, the division is an even split between external and internal students; these students were not all external or internal in their locus of control classification. Their achievement scores paralleled those of the entire groups. Externals scored higher than inter- nals in the Winter and composite samples (see Tables 4 and 8) while the internal student out performed the external student in the Spring trial (see Table 6). Assuming these students had participated in the CA1 treatment and that they received the same achievement scores, the net effect would have been to improve the lecture group performance while reducing the CAI scores (Winter) or leaving them relatively unchanged (Spring and composite). The number of students involved and the parallel trends with the full sample indi- cate that, if these few students had worked with the CAI treatment rather than the lecture, the results would 176 strengthen the argument for lecture superiority over CAI. The results would not alter the locus of control factor nor the interaction effects. Table 14 Mean Achievement Scores Students Selected for CAI Electing Lecture Method External Internal Total Winter Trial i 92.5 56.0 70.6 n 2 3 5 Spring Trial 2 54.0 63.0 58.5 n 1 1 2 Composite Sample i 79.3 57.8 67.1 n 3 4 7 The overall absence of any statistically significant findings within both trials and the composite sample lend credence to the stability of the research results. The lone alternate hypothesis that was accepted occurred in the com- posite sample ANCOVA with the MSU math placement score as the covariate. There was a significant difference between instructional treatments. Any adjustments made for those seven students remaining in the lecture mode would only have strengthened lecture superiority over CAI. 177 Summary This chapter has provided the empirical results of the study. The four analysis procedures of ANOVA, ANCOVA, linear regression, and an abbreviated application of the Johnson-Neyman procedure were presented. The data was presented for each trial separately as well as in composite form. The reason for the separate sample analysis is based on a statistically significant difference in achievement scores between the two trials. The analysis of variance yielded no statistically significant difference in instructional treatments or locus of control classifications for either trial or for the composite sample. Interaction effects were also non-existent. The analyses of covariance were conducted using eight different covariates: four primary and four secondary. The SAT score was dropped due to large numbers of students lack- ing this measure. Of the 24 ANCOVA calculations, only one main effect F-ratio was in the region of significance. Instructional mode constituted a significant main effect for the composite sample when the MSU math placement score was controlled. The only other F-ratio to even approach signi- ficance was for the study time covariate analysis in the composite sample. Here the instructional mode main effect fell short of reaching significance, thus the null hypothe- sis of no significant main effect was accepted. There were no locus of control or interaction F-ratios that approached significance for either trial or for the composite sample. 178 Linear regression analysis in which the IAR total score served as the predictor variable produced practically paral- lel lines with lecture superior to CAI. The ANOVA results showed, however, that the differences were not statistically significant. Predictions based on the I+ and I- subscores revealed interesting trends. Graphs cf the composite sample regression data gave the appearance of an ordinal interac- tion for the I+ scale and a disordinal interaction for the 1- scale. Even though the regression lines using the I+ and I- scores as the independent predictors presented the pros- pect of interesting interactions, the statistical results indicated that the lines were parallel and, therefore, interaction effects were non-existent. The analysis concluded with a very brief discussion of the Johnson-Neyman procedure for determining regions of sig- nificance. Even though the Johnson-Neyman procedure is unwarranted for the data obtained in this study, this author considers the presentation of the method beneficial and of service to those readers interested in ATI research. Appen- dix E presents a more comprehensive description of the pro- cedure along with mathematical computations using the I+ and I- composite data. This technique was included, despite the non-significant interactions, for illustrative purposes. Many ATI studies "do not fully develop the analysis pro- ficedures to the point of obtaining regression equations and very few actually identify regions of significance when interactions are found. 179 The final section speculated on the possible impact of those students selected as part of the CA1 treatment but opting to remain in lecture. Their achievement scores paralleled those of the larger groups, and simply shifting them from the lecture group to the CAI group would have strengthen the trend for lecture to be a superior, though not significantly, method of instruction. Such a shift would not have affected the locus of control or interaction results. The overall analysis indicated that the two instruc- tional treatments were equivalent, the two locus of control groups were equivalent, and that no ATI's existed. These conclusions apply for both trials and the composite sample. These results were consistent across trials and inclusion of covariates had little impact on the main or interaction effects. The single exception, however, is within that expected by chance alone. Regression analysis, which treated locus of control in a quantitative rather than qualitative sense, also showed an equivalence of treatments. Regression analysis revealed no significant dependency of achievement on the 1+ or I- scales. There are several issues pertinent to these findings which increase the risk of committing Type II errors. Results using the composite sample, although consistent with the separate Winter and Spring trial analyses, lacks sta- tistical power due to a combination of different achievement measures. These findings are all the more tentative since 180 the two exams had statistically different means. The Winter trial and the composite samples suffer from unbalanced cell sizes. Even though each cell in the design matrix is filled, the cell sizes are not proportional. The seven students who remained in the lecture contributed to this situation thereby reducing the power of the analysis pro- cedures. Finally, the reliability indices for the IAR Questionnaire are not sufficiently high to make group deci- sions. The low reliability indices for the independent variable does not provide for as robust a study as desired. These are three major factors which might explain why no significant differences were found. CHAPTER V SUMMARY AND CONCLUSIONS Introduction This final chapter presents an in depth discussion of the research results, the degree of agreement with the stated predictions, and implications for continued research. The approach will be to first review the rationale for con- ducting the study and then to proceed with a discussion of the original research questions. This chapter is intended to aggregate the literature, the design and methodology of this study, and the experimental findings. The chapter con- cludes with suggestions for future research and personal reflections on this research effort. Overview of the Study The influence of computers, and more recently microcom- puters, for direct instructional purposes presents instruc- tors and administrators with a myriad of questions and deci- sions. The foremost question may center on the effective- ness of computer based instruction. A second concern is the cost, while a third question may focus on procedures for the efficient utilization of limited computer resources. This study investigated whether an aptitude-treatment interaction between student type, as measured by locus of control, and 181 182 instructional methodology existed. The effective and effi- cient utilization of educational materials and resources requires knowledge on how to do so. The underlying premise of this work was that knowledge of an interaction between student locus of control and the educational methods of lecture and computer-assisted instruction might permit the pairing of students with treatments so as to optimize learning as well as management of often scarce resources. Computer-assisted instruction has proven efficient in the reduction of direct instructional time. While the results show no detriment to achievement, scores still fall in a normal distribution-~some students do extremely well, others do very poorly. The same applies for most of the other instructional techniques. The problem remains one of identifying which student characteristics differentiate the low and high achievers. If a set of characteristics inter- acting with instructional methods, the so called aptitude- treatment interactions, could be found, then differential assignments of students to the appropriate treatment would be of benefit. This study focused on determining the influence of student locus of control on achievement within the learning environments of lecture and computer-assisted instruction. The Intellectual Achievement Responsibility (IAR) Question- naire was used to assess student locus of control. This questionnaire, specifically designed for academic settings, 183 contains two separate subscales: responsibility for succes- ses and responsibility for failures. Students were split into internal and external groups with the mean of the total IAR score as the dividing point. These two groups along with the two treatments composed a 2-by-2 completely crossed experimental design. The materials were drawn from a two week segment in a College Algebra course. There was a maximum of six lectures compared to three one-hour scheduled CAI sessions. The achievement measure was a teacher made test covering the materials presented during the experiment. The CAI system and courseware were designed and written by this author. The lectures were conducted by one of his colleagues. The experiment was performed during the Winter 1983 term with 32 students and again in the Spring 1983 quarter with 19 students. The subjects were predominantly Black, freshmen engineering students. A comparison group of mostly white students was given the IAR Questionnaire, thereby serving as an indicator for extrapolating the experimental results to a larger, more heterogeneous population. Analysis techniques consisted of analysis of variance and covariance, linear regression analysis, and an illustra- tive application of the Johnson-Neyman procedure for deter- mining regions of significance in ATI studies. The ANCOVA procedures investigated eight possible contributing factors. The four primary covariates of instructional time, test averages, MSU math placement score, and previous course 184 grade were predicted to be principle factors reflective of prior levels of achievement and ability. The four secondary covariates of study time, quiz averages, home work grade, and ACT math score were also considered as measures of abil- ity but not contributory in a substantial way. The ANOVA and ANCOVA procedures were performed on each trial separ- ately and on a combined sample. The emphasis of ATI research is not to identify which instructional method is the best on average but to develop regression equations which can be used in predicting student performance and possibly assigning students to treatments on the basis of one or more independent measures. This study considered the locus of control scales as the predictors while the lecture and the CAI methods represented the treat- ments. The Johnson-Neyman procedure defines those regions where a differential assignment of treatments is of statis- tical significance and benefit to the student. The Johnson- Neyman procedure was performed for the two IAR subscales on the composite sample and is included as an appendix for illustration purposes only. Discussion of Research Questions The study was designed and conducted to gain insight into four general research questions. The following discus- sion presents the original questions with the experimental results and conclusions. 185 1. Will the distinctly different strategies of tradi- tional lecture and computer-assisted instruction differentially affect academic achievement? This question was addressed by two research hypotheses, and in the 2-by-2 design of the study, is resolved by an analysis of treatment marginals. Analysis of achievement scores without covariate adjustments resulted in no signifi- cant differences on the basis of treatment assignments. Even though the F-ratio probability for the composite sample was within .03 of significance, the hypothesis that differ- ences in achievement are attributable to instructional mode was rejected. When covariates were included in the analysis only one of the eight variables produced a significant result. When both trials were combined to form the composite sample, the main effect due to treatments reached significance when the MSU math placement score was controlled. This particular score is used to initially place students in their beginning math class at MSU. This measure was considered as one of the primary covariates since it serves as a local measure of entry level ability. This factor reduced the F-ratio proba- bilities for both trials as well as for the composite sam- ple. It is not surprising, therefore, that this covariate produced a significant effect due to instructional method- ology. The lecture method produced higher achievement scores for both samples. The only other covariate producing a treatment main effect approaching significance was the amount of study 186 time. Although the F-ratio of 3.336 was not large enough to be accepted as significant, the study time factor appears to be one of some importance and should not be ignored. Sev— eral other covariate measures reduced F-ratio probabilities but none to a level of statistical significance. The only one to consistently reduce the F-ratio probabilities was the MSU math placement score. Results of this study indicate that the answer to this research question is a definite no. Even with the lone covariate factor reaching significance, the F-ratio proba- bility (.044) was just barely within the region to reject the null hypothesis. To the extent that this study was able to determine, the method of instruction, lecture or CAI, produced no significant differences in achievement. One could interpret this conclusion as an equivalence of methods and that CAI does just as well as the traditional lecture. As with most CAI studies, students in the CAI treatment spent less time in formal instruction than lecture students. For the composite sample, the lecture group spent an average of 4.57 hours in formal instruction as opposed to 2.67 hours for the CAI group. Even though the CAI group received 1.9 hours less instructional time--equivalent to two full class periods--their overall performance was no worse than their lecture counterparts. This difference in formal instruction time is highly significant at the °<- .05 level (t - 7.82). This study therefore, joins many others by reporting no 187 improvement or detriment in achievement for CAI but a signi- ficant reduction in formal instructional time (a 42% reduc- tion here). The amount of time spent studying outside of class was not statistically significant (t - 1.20) between the lecture and CAI groups although the CA1 students reported an average of two hours less study time. 2. Can levels of academic achievement be predicted for the different instructional techniques based on student locus of control measures? This study approached the locus of control issue in two different ways. The first, and simplest, was to set up internal and external locus of control groups on the basis of the IAR total score. A straight forward ANOVA procedure provided information on whether achievement differences between these two groups existed. Tables 5, 7, and 9 pre- sent the results of this analysis. The conclusion reached was that internal and external students performed equally well. The graphs in Figure 2 indicate that achievement scores for each treatment are parallel to a great extent. Contrary to the predictions and results noted in the review of the literature, the trend in this study was for inter- nally oriented students to perform less well than the exter- nal students. Although the differences between groups are not statistically significant, this trend is not easily explained. One possible explanation, however, may be related to the arbitrary classification of internal and external locus of control groups. Assuming a normal distribution of LOC 188 scores, approximately 68% of the sample will have IAR total scores within one standard deviation of the mean. Thirty- four percent of the internally (externally) classified stu- dents have IAR total scores within one standard deviation above (below) the sample mean. Cronbach and Snow (1977) contend that treating a continuous measure in a dichotomus manner, although not completely discouraged, tends to obfus- cate pertinent results. If the extreme internal and exter- nal students of the composite sample (having IAR total scores more than one standard deviation from the mean) are analyzed, the predicted trend emerges. The mean extreme external achievement score is 68.7 (n - 7) while the extreme internal average is 75.6 (n a 9). Such statistical manipu- lation itself is of little service in interpreting the results. However, it does hint that the expected trend is likely an artifact of the dichotomization of a continuous measure. Analysis of covariance, the second approach, resulted in no significant differences between internal and external classifications regardless of the covariate. The MSU math placement score analyses yielded increased locus of control F-ratios for each trial and for the composite sample. These increases, however, did not reduce the probabilities to a point of significance. This lone covariate consistantly increased the F-ratios for both the treatment and locus of control main effects. The MSU math placement score may possibly have more influence on achievement than locus of 189 control. As with the previous research question, this one must be answered in the negative—-academic achievement is insensitive to a dichotomized student locus of control scale. 3. Does there exist an interaction between instruc- tional methodology and student locus of control? This question is the type on which ATI research is based. ATI studies are not concerned with whether CAI is better or worse than lecture or with determining the rela- tionship between internal and external students. ATI research efforts attempt to identify stable and, hence, pre- dictable interactions between instructional treatment and student characteristics. This research question was the focal point of this study--does there exist an aptitude- treatment interaction between student locus of control and the instructional paradigms of lecture and CAI? Four research hypotheses were formulated to address this particular question. The analysis of variance calcula- tion produced a highly non-significant F-ratio for the 2-way interaction between instructional treatment and locus of control classifications. The F-ratios approached zero indi- cating a complete absence of any interaction whatsoever. Analysis of covariance generally increased the F-ratios regardless of the covariate used. This is expected with F— ratios close to zero--almost anything would help. However, none of the 2-way interactions with or without covariate adjustments reached significance. 190 As mentioned before, the locus of control measure is a continuous scale with a maximum value of 34--not simply external and internal. The ANOVA and ANCOVA procedures fail to fully utilize the continuous nature of this personality construct. A bilevel blocking on locus of control may, and in this case did, result in a non—significant ANOVA while regression slopes may indeed be significantly different. The ANOVA and ANCOVA procedures also prohibited separate analyses for the I+ and I- subscales of the IAR. Linear regression analysis provided the necessary tool for address- ing both issues. This analysis procedure considers the locus of control measure in a quantitative rather than qual- itative sense. Despite this quantitative approach, linear regression analysis confirmed the ANOVA and ANCOVA results. The regression slopes for the CAI and lecture regression lines were not significantly different--they were statisti- cally parallel. Parallelism of regression lines occurred when the IAR total score and the I+ and I- subscales were used as the independent variable. Moreover, a test to determine whether the regression slopes were different from zero failed to reach significance (see Appendix E). The student locus of control construct in total or as subscales failed to have any influence on achievement within the confines of this study. No ATI exists for the variables investigated. The predictions that CAI would benefit the more exter- nal student was not realized. The experimental results 191 indicate that the lecture method tended to be better than CAI for all levels of locus of control. CAI students who had extremely external scores on the I- subscale had higher achievement scores than their lecture counterparts. It must be reiterated that achievement scores were not statistically different. The point where the 1- scores begin to favor the CA1 method is at 5.6 which is below the minimum value of seven observed in the sample. The expected increase in achievement scores with increased internality did occur for the CA1 sections when 1+ was the predictor and for the lecture group when I- was the independent measure. Lecture students who attribute success (1+) to their own actions did not perform as well as those expressing the belief that success was mainly due to luck. The nature of the test (problem solving) might have influ- enced performance when viewed as a function of the 1+ sub- scale. Research has shown that the issue of congruence appears to have some impact on performance. Perhaps the problem solving type of test is more congrucus with the CA1 treatment. Both require logical, formalized operations, and contain a high degree of structure. The test structure may be incongruous with the lecture style or the manner in which it was conducted. A possible incongruity between expecta- tions in lecture and on the achievement test may explain the declining scores. Lecture students with external beliefs are probably not as affected by any such incongruity with respect to their responsibility for successes. 192 The trends are reversed when the responsibility for failure scale (1-) is used. Here it is the CA1 group that has declining scores for higher levels of personal accep— tance of failure. This trend may indicate an incongruity between the student and the CA1 instructional method. CAI is a novel use of computers, unfamiliar to most college stu— dents. Willing to accept a failure or a substandard per- formance, these students may have perceived the microcom- puter system as threatening or intimidating. CAI students with high 1- scores may have also viewed participation in the CA1 treatment as a means of taking a short break prior to the final exam. As in the previous discussion on the 1+ tendencies, the external 1- students are probably unaf- fected by the means of instruction--failure or success is largely due to luck or chance. 4. Are the results generalizable across student sam- ples? Since the study consisted of two samples, an indication of the stability and generalizability of the experimental results is available. The results were consistent--no sig- nificant treatment effects, locus of control effects, or interactions. Regression analysis of treatments for each trial produced the result of parallel regression lines with slopes statistically equal to zero. Regression lines based on the IAR total score had slopes of the same sign with the lecture method having a larger negative value than the CA1 group. 193 Regression analysis using the 1+ and I- subscales was reversed between the two trials. The signs of the lecture and CAI treatments were always the same regardless of trial or subscale. The signs became mixed only in the composite sample. The fact that the 1+ scale predicted declining achievement scores in the Winter trial and increasing scores for the Spring trial may be due to a difference in the over- all difficulty of the achievement measure. The Winter term average achievement score was significantly higher than that of the Spring trial. All discrimination indices were posi- tive and, except for three items (one on the Winter test and two on the Spring exam), had values of .30 or better. The difficulty indices for the Spring term exam, however, were in a range desirable for good classroom tests. Ebel (1979) states that items with discrimination indices in excess of .30 are reasonably good and may need some improvement. The difficulty of test items in a "good" classroom exam should be in the mid-range of 35-65%. Only one item on the Winter term exam met this recommendation while five of the nine Spring term questions fell in this range with one other within three percentage points. Refer to Appendix D for detailed statistics. The Winter term test, therefore, appears to be less than ideal for the adequate assessment of student ability with the material. The exam was probably too easy, despite the high reliability index and content validity judgements. The Spring term exam in contrast has a 194 better "statistical profile" and may be a better instrument for measuring the effects of the instructional techniques. The inconsistencies in the 1+ and I- regression equa- tions may have resulted in the difference in achievement measures. There may be an exam difficulty factor that influences the 1+ and 1- results. If the test is initially perceived as difficult, those students who take personal responsibility for success may work at the task in earnest while externally oriented students may work at less than their capacity because the "teacher made a hard test." This logic would explain rising achievement scores and regression lines as represented in the Spring term trial. The I- regression predictions may be interpreted as due to a heightened level of test anxiety arising from what is perceived as a difficult exam. Students expressing personal responsibility for failure may react in accordance with a perceived exam difficulty level. In an effort to overcome failure a higher level of test anxiety may occur which mani- fests itself in a debilitating manner, thus causing lower scores. Externally oriented students with respect to the 1- scale may not experience any increase in test anxiety and, therefore, perform at higher levels. This logic is again represented by the Spring trial data. Winter term results for the I- regressions could be due to increased test anxiety manifested as facilitating anxiety because of the relatively easier exam. The regression line slopes, albeit positive, are very small (see Table 12). 195 The 1+ Winter term regression slopes, which are nega- tive and rather large, are more difficult to explain. It would appear that the more internal students didn't really care about their performance. Since the exam was given at the end of the term, internal students may have perceived the importance of the "last exam" before the course final of little consequence. Internal students may have taken a "mental break" because they have enough confidence in them- selves to make up for a poor exam score on the final. Regardless of the above speculations, the dynamics occurring during each trial affected both treatment groups in a consistent manner. If the Spring trial consisted of students who had poorer mathematics backgrounds than the Winter term students, as might be expected, then covariate analyses, especially those using the MSU math placement score and previous math course grade, failed to expose any differential in ability. The fact remains that none of the main effects or interactions were statistically significant for either sample. The results, from a statistical vantage point, are generalizable and stable across student samples. Reflections and Observations Many issues pertaining to the conduct and ideas for the improvement of any study emerge as the investigation comes to a close. This section is included to address those issues and concerns, particularly as they relate to the limitations and the experimental procedures. 196 The study was conducted under a few severe constraints. There was only one microcomputer station available. Even though scheduling students for access to the system afforded all CAI students the time they required, there was very little room for adjustments or rescheduling. If the experi- ment was of a longer duration or the sample was larger, scheduling regarding microcomputer access would have become problematic. In fact, the system was dedicated to the study at the expense of other office and program concerns. To have extended the duration or increased the sample size would have seriously affected system access for both stu- dents and professional staff. This author was only able to prepare courseware for a fraction of the course material. The preparation time and expense involved with such courseware development is an intensive "front end" investment.‘ Although no accurate records were kept for the time required for courseware pre— paration, debugging, and student testing, an estimate of 100 hours for each of the three lessons used in this study is reasonable. The cost for preparing courseware, operating the microcomputer system over the expected lifetime of both the hardware and software, and the initial cost and mainten- ance of the hardware system itself are major concerns for administrators. The initial costs can be substantial and may represent a major expenditure. The unknown cost, of course, is that associated with courseware development and 197 maintenance. The courseware used for this study was spe- cifically designed and written for the subject matter of interest and the available hardware system. Commercially available courseware is scarce and often incompatable with existing course structures or hardware systems. Much of the available courseware, including materials developed for this study, cannot be easily modified or restructured. The duration of the experiment was limited by the number of microcomputer stations and available courseware. (The equipment demands during the study were almost double the recommended six students to one station ratio.) Even though the literature reports CAI and ATI studies lasting as little as one hour, the three one—hour sessions involved here may have been insufficient for either main or inter- action effects to emerge. The duration of the experimental treatments is related to locus of control in a subtle way. Locus of control, as a measure of generalized expectancy, depends upon the novelty or level of familiarity with a given situation. According to the theory promoted by Rotter (1954), the expected level of achievement is a function of both the specific situation and a generalized expectancy based on the similarity of the given situation to previous situations. Moreover, the impact of the general expectancy is reduced by the frequency with which the student has been in the same or similar sit- uations. The duration of this study, then, focused on the initial experience with a new instructional technique. 198 Extending the study to cover the entire course would provide information on the more steady state level of expectancy due to the increased familiariy of the CA1 system. The intended use of the IAR measure was one of prediction. Data was not collected on student perceptions during various stages of the study. The dynamics involved as students adjusted to the novel CAI method and the correlation of their percep— tions with their locus of control scores would have provided a test of the steady state hypothesis. Of those students electing to remain in lecture rather than participate in the CA1 treatment, one observation is noteworthy. These students stated that they felt comfort- able with the teacher and were apprehensive about trying something unknown. This reasoning was characteristic of the internal as well as the external students, although the external students were concerned about doing poorly on the exam as a consequence of the CA1 technique. Given the Option, these students preferred the familiar lecture method rather than an unknown situation. Conversely, there were just as many lecture students, both internal and external, who wanted to work on the computer system. These casual observations tend to reinforce the fact that locus of con- trol is not a factor in pairing instructional methods with students. The students involved in this study constituted a nar- rowly defined population. Their IAR total score and sub- scores were more internal than the more heterogeneous MMM 199 160 sample. Since the MMM 160 sample consisted of engi- neering majors but from all levels, freshman through senior, their more external orientation may be due to their college experience. Freshmen tend to be eager and enthusiastic about beginning their college career--it's a novel situation with probably a high level of generalized expectancies. For the more experienced students the college routine is no longer novel but in fact highly structured and regulated. The routinization of college activities may indeed promote a degree of cynicism, leading to the emergence of defensive externals who use the "situation" as a means of protection. One last observation needs comment. The fact that the students involved in this study were predominantly Black may limit the generalizability of the experimental findings. These students are not simply experiencing a faster, more demanding academic schedule as compared to high school, but they are adjusting to a completely new ond often foreign learning-living environment. The socio-cultural transitions minority students undergo when coming to MSU impact on their academic performance. The extent to which the learning, academic environment intermixes with the new living situa- tion is unknown. While the duration of the study may not have been long enough for the treatments to "take hold", a protracted experiment may have been influenced by transient but significant external factors that impact upon minority freshmen students. 200 Implications for Continued Research Like most research studies, this one has raised more questions than it has answered. The obvious questions con- cerning replication of the results using longer treatment durations, different samples (composition and size), and the choice of subject matter need to be addressed in any future studies. Extending the study to encompass an entire course or course sequence should also be accompanied by methods for determining changes in student locus of control. A detailed study of locus of control changes would provide information pertaining to Rotter's time dependence hypothesis. Using larger, more heterogeneous samples or samples drawn from the extreme locus of control regions would increase the power of the statistical procedures and the validity of the results. Implementation of the above recommendations would necessi- tate a complete removal of the software and hardware limita- tions that so constrained this study. Although limited in scope, this study has indicated that the personality construct of locus of control does not influence differences in academic achievement between instructional methods. Inclusion of more methods and/or affective characteristics may produce the aptitude treatment interactions sought here. Regardless of instructional method, some students did very well and others did very poorly. Locus of control was not a useful measure in explaining academic performance. Measuring test anxiety, 201 experimenting with different test formats, and using inter- active computer generated tests for the CA1 group compared to pencil and paper tests for the lecture students are other issues requiring attention. As colleges and universities move toward the next cen- tury the influx and proliferation of microcomputers will cause considerable reorganization in the conduct of the educational enterprise. The use and expense of microcompu- ter systems are major issues confronting administrators and curriculum planners. By 1989, the Massachusetts Institute of Technology and Brown University will invest $70 million each for CAI systems (Ploch, 1984). The University of Mich- igan is charging each engineering student $100 per term for access to one of four advanced microcomputer systems. (The revenue is allocated for the purchase of newer, more sophis- ticated systems as they become available.) Stevens Insti- tute of Technology requires each student to purchase micro- computer work stations. Microcomputer companies, who have traditionally focused on the business and home markets, are now heavily engaged in cultivating the educational market. The leaders currently appear to be IBM and Digital Equipment Corporation followed by Apple, Apollo, and Zenith. These companies are working with college and university administrators to provide hard- ware components at discounted rates. The financial stakes for hardware systems alone are very high. The major obstacle, however, is the quantity and 202 quality of the educational courseware. Whether it is of a tutorial, simulation, or artificial intelligence nature, the know-how of CA1 courseware development is prerequisite for successful implementation of hardware systems. It was the intent of this study to investigate one aspect of curriculum development as it pertains to CAI material. The development of effective courseware is sure to lag the appearance of CA1 work stations, even at the smallest of institutions, for some time to come. Robinson (1979) contrasts the "traditional" and "modern" educational methods. The traditional method attempts to control all aspects of the learning environment. This system produces "The Child" by requiring students to learn in a uniform, controlled situation, processing infor- mation at the same pace and style. The modern method attempts to remove all constraints and open the system up so as to be individualized. From a cybernetic point of view, the modern system is impossible because of the almost infi- nite variety of student-subject-environment configurations. The result is a mixed system whereby individual needs are addressed within the limitations of a controlled environ— ment. The mixed method of education can be described as a deregulated traditional system or an over regulated modern system. Even though the mixed method cannot theoretically exist because it violates several cybernetic principles, it does exist in a pragmatic sense. This investigation focused on one method for individualizing the educational process by 203 determining which instructional technique promotes higher levels of achievement for particular types of students. Despite the non-significant results, ATI research attempts to remove some of the classroom constraints by optimizing the match between student characteristics and instructional techniques. There is almost uniform agreement that computer tech- nology has the potential to restructure the educational pro- cess. Consequently, the manner in which educational insti- tutions, especially higher education, provide instructional services will also undergo radical change. The manner in which these changes take place over the next few years should be by design and not for convenience. This author does not foresee any hardware limitations. New advances in video capabilities, mass storage devices, and speech synthe- sis will only promote data, graphics, and audio information processing. The questions facing administrators, curriculum developers, and instructors will center on the effective usage of the advanced technology. These questions can be answered and appropriate decisions made only if adequate knowledge and information exists. Further research into the parameterization of instructional materials, modeling of procedures for the effective diffusion and implementation of the computer into existing educational structures, and the formulation of data driven models of higher education are required. 204 This research effort in particular and the trend for computerization in general prompt this author to propose the following six research areas. 1. What are the physical and operational features (lesson length, degree of student interactivity, instructional paradigm, cn-line questioning proto- col, feedback mechanisms, use of color, graphics, speech, etc.) of educational materials that opti- mize achievement? What student characteristics and instructional methods interact? Locus of control does not appear to be one such parameter, but variables like field dependence-independence, introversion-extroversion, or student attitudes and preferences may be useful in determining appropriate student-methodology pairings. Although not subject to ATI regression techniques, are simple nominal measures such as sex and major preference stronger predictors of achievement than the more quantitative variables? Educational reform is imminent, but at what cost? Institutions of higher education, already operating under often severe financial constraints, must develop sound policies for capital investments per- taining to computer systems, their maintenance, and their access. What is an optimal machine-user ratio? What funding strategies are available for 205 hardware, software, and courseware purchases? What support systems, personnel, networking capabili- ties, repair and/or replacement options, software/ courseware package requirements are needed? Can cost models be developed for these issues so as to permit comparison and optimization? What control strategies can or should be employed that would maximize long term retention of the sub— ject matter while minimizing cost? When should computers be used and when should they not? Are computer systems going to supplement the tradi- tional forms of instruction or are these powerful machines going to supplant them? Are artificial intelligence systems a viable approach to general undergraduate coursework and, if so, under what conditions should they be used in the classroom? The ultimate question to consider is whether class- rooms in the year 2000 will resemble those of today or whether classrooms will indeed exist at all? The growth of the microcomputer industry, while it has possibly saturated the current market, has, nonetheless, been phenomenal. The development of networking systems, satellite communications, and electronic libraries will facilitate long distance or in—home education. Institutions of higher edu- cation may become the realm for specialty subjects such as medicine or for advanced graduate research. 206 Universities may indeed face a difficult transi— tion period with the advent of electronic under- graduate programs. How are colleges and universi- ties planning to minimize the impact of a computer- ized education? What decisions need to be made so that computers can be incorporated into higher edu- cation with as little disruption as possible? One of the problems, of course, is that technological development is moving at an ever increasing rate. The latest hardware system today is obsolete within six months. Even though powerful hardware systems exist at reasonable prices, the software necessary to control and operate them, as well as quality courseware, is just now becoming availa- ble. Software and courseware development is currently a nascent business. The intent of this study was to determine whether locus of control could be one of the determining factors in the adoption and use of computer based educational systems. As the diffusion of computers into college classrooms con- tinues, research efforts directed toward better understand- ing their impact on the learning processes, faculty and students, and the educational institution itself must be studied. These research efforts are requisite for planning and controlling the fourth educational revolution. APPENDICES APPENDIX A COMPUTER—ASSISTED INSTRUCTION SYSTEM DESCRIPTION OVERVIEW SOFTWARE PROGRAMS DATA FILE CONTENTS ORGANIZATIONAL CHART FLOWCHART AND DATA LINKAGES TEXT PROCESSING COMMANDS EXAMPLES APPENDIX A COMPUTER—ASSISTED INSTRUCTION SYSTEM DESCRIPTION OVERVIEW This appendix provides a brief overview of the CA1 sys- tem developed for this study. This discussion also high- lights the distinction between the definitions of software and courseware. The design of the CA1 system is one of a data processing model. The software (computer programs) were written specifically to create, edit, and process the courseware which serves as the data. The software system is actually an integrated network of six programs. These pro- grams, while performing specific functions, contribute to the overall development, processing, and storage of informa- tion related to the instructional materials. The data processing model uses files which serve as either sources of information or for data storage. Six data files are used in the operation of the CA1 system. The contents of these files is presented in this appendix. Also presented are charts depicting the organizational structure and logical processes for the preparation and presentation of instructional materials. Although not used in this study, the CA1 system can perform a pre—processing configu- ration procedure whereby the instructional unit is tailored to the difficulty level of the student currently using the system. 207 208 There are a number of special processing commands available for the preparation of instructional materials. These commands constitute an authoring language and a brief description of each command is included. A final section presents two examples of text preparation and processing. SOFTWARE PROGRAMS CAI Processing Program Text Entry Program Curve Generator Character Generator File Development Program Summative Data Analysis Accessed by the students Processes instructional courseware Written in Z-80 assembly Provides authors a method for entering textual materials into the system Used to create instruc- tional units Written in Z-8O assembly Plots curves and graphs Permits experimentation with curve parameters before incorporation into instructional units Permits experimentation with special character and symbol development Allows permanent storage of character sets Used to create and update information on data files accessed by the CA1 program Provides analysis of the instructional units Performs item analysis Prepares student progress reports Lists student comments DATA User Verification File User Name User Identification Number User Password Assigned Instructional Unit Last Unit Completed Total Access Time Number of Accesses Last Access Date Instructional Mode Difficulty Level Counselor Code Non-Standard Symbol File Memory Storage Address Defining Bytes for Symbol Student Comment File Unit Name Comment Field 210 FILE CONTENTS Record Keeping File User Identification Number Access Date Unit Name Presentation Difficulty Level Counselor Code Unit Completion Time Student Responses Grading Algorithm Scores Grading and Scoring File Scoring Key Items #1-30 Primary Unit Assignment Alternate Unit Assignment Difficulty Level Decisions Unit Printing Decision Comment Field Instructional Units All instructional material and processing commands 211 ORGANIZATIONAL CHART Curve Character Generator Instruc- tional Units File Non- User Develop- Standard Verifi- ment Symbols cation CA1 Processing Program Student Grading Comments and 4 Scoring Student Records Circles-Programs Squares-Data Files Solid LineaDirect Linkage Dashed Line-Develop- mental Linkage Summative Data Analysis Generator File Develop- ment File Develop- ment 212 FLOWCHART AND DATA LINKAGES I User Sign-on Procedure User Verification , 1 Data File [g Instructional Unit Access Initialization of System Control User Verification Data File For Unit A ‘_l File Parameter Block I 1 Load Non-Standard Graphic Symbols File Parameter Block For Unit and Non- Standard Symbol File ficulty Index Blocks of Unit and User Verification File Pre-Processing Configuration of Instructional Unit Instructional Unit Processing and Display J 1 File Parameter and Dif- 1—1 F—_' Data Collection and Formative Analysis 1 Grading and Scoring, Decision-making and Record User Verification, and Keeping Student Records Data Student Input, Grading Scoring Data File T__' 1 Files Student Comments Student Comments Data File 10. TEXT PROCESSING COMMANDS Command Name Input Command Answer Command Spacing Command Question Command Random String Command Integer Number Generator Decimal Number Generator Text Phrase Generator Compute Command Compare Command Purpose To allow students to input answers to questions Provides immediate feedback indicating correctness of student input Used in conjunction with other special processing commands to ensure vertical alignment of characters Identifies the beginning of a question or problem Indicates number of random variables needed for insertion into a text line Generates a random integer number Generates a random decimal value Randomly selects text phrases from a list of possibilities Computes numerical values in a question using random variables that were pre- viously defined Used for answer matching when numerical values are input 11. 12. 13. 14. 15. 16. 17. 18. Foil Command Select Command Plot Command Graphic Characters Command On Command Calculate Command Assembly Language Routine Mapping Command 214 Calculates numerical multiple-choice foils for questions using randomly generated variables To select text phrases that are dependent on other randomly generated parts of a question, or to prepare textual multiple-choice foils Plots graphs of math functions and relations Loads a new set of graphic characters or symbols into computer memory Provides diagnostic feed- back to incorrect answers Provides temporary storage for numerical values used in complex calculations Loads and/or executes pro- grams written in assembly language Provides inverse mapping for multiple—choice foils which were ramdomized prior to presentation-—command is generated internally and is not for author use 215 EXAMPLES Instructional Unit Coding: (SP)QSN14. If (RV) molecules of fluorine react with (SP)COOO6*INT(SO*RND(1)+10) (SP) phOSphorus according to the equation P4 + 6F2 -> 4PF3 (SP) how many molecules of phosphorus trifluoride (SP) will be produced? (SP)FO 4*RVN(1) (SP)FO RVN(1)/6 (SP)FO 4*RVN(l)/6 (SP)FOFX4 (SP)INO4 Processed Version: 4. If 282 molecules of fluorine react with phosphorus according to the equation P4 + 6F2 -> 4PF3 how many molecules of phosphorus trifluoride will be produced? A. 1128 B. 47 C. 188 D. 4 What is your answer? __ * The (SP) and (RV) represent single graphic characters used for identifying a special processing command and indicat- ing the position for insertion of random variables. INT and RND represent the integer and random number func- tions respectively while RVN is a temporary storage array for numerical values randomly generated within a problem. 216 Instructional Unit Coding: (SP)QSSlThe amount of (RV) an object receives varies (SP)TX02|heatIlightI (SP) inversely as the square of the distance from the (SP)RSOlsource. How many times as much (RV) will an object (SP)SL RVN(1)IheatIlight| (SP)RSOlreceive if it is moved to a point (RV) as far away? (SP)TXO4|three timeslone-fourthIS timesltwo—thirdsl (SP)SL RVN(3) one-ninthll6 timesll/25Inine-fourthsl (SP)SL RVN(3) 9 timeslone-sixteenthl25 timeslfour-ninthsl (SP)SL RVN(3) one-sixthl8 timeslone-tenthll 1/2 timesl (SP)SLO7RVN(3) 6 timeslone-halfllo timeslfour-thirdsl Processed Version: The amount of light an object receives varies inversely as the square of the distance from the source. How many times as much light will an object receive if it is moved to a point 5 times as far away? :: 10 times :: one-tenth :: 1/25 :: 25 times This is a concealed multiple-choice item. As each foil is presented, the student has to state whether the foil is the correct answer or is an incorrect solution. An incorrect response from the student or presentation of the correct answer terminates the question. Note that the fourth option will not be presented because the third choice is the right answer. The foils are automatically randomized for this type of item. APPENDIX B UNIVERSITY CORRESPONDENCE MEMORANDA TO UNIVERSITY COMMITTEE ON RESEARCH INVOLVING HUMAN SUBJECTS (UCRIHS) UCRHIS RESPONSE MEMORANDUM TO COMMITTEE ON RELEASE OF CONFIDENTIAL INFORMATION COMMITTEE ON RELEASE OF CONFIDENTIAL INFORMATION RESPONSE APPENDIX B January 17, 1983 Memorandum To: University Committee for Research Involving Human Subjects From: Gregory C. Hamilto Re: Review of Doctoral Res arch Proposal Exemption is claimed as type 1 research project. As a graduate student in the College of Education, I am working on a dis- sertation for a degree in Administration and Curriculum. The dissertation focuses on comparing traditional lecture versus computer-assisted instructional methodologies for minority engineering students enrolled in one section of the College Algebra course MTH 108. I propose to investigate the existance of an aptitude-treatment interaction between instructional methodology and the affective variable of student locus of control. The class will be divided into two major groups on the basis of student locus of control (Intellectual Achievement Responsibility Questionnaire) and then randomly assigned to one of the two treatment methods. The dependent variable of the study is the score on an achievement test. Analysis will be performed by an analysis of covariance and multilinear regression techniques. The experiment will be performed in the Winter 1983 term and repeated with a second sample in the Spring 1983 term. For the past eight years I have been employed by the Office of Minority Student Education (OMSE) in the College of Engineering. During that time some of my responsibilities included course and curriculum development. The proposed research project is the culmination of an effort that began three years ago. The students for the project are, therefore, involved in our program and enrolled in a section of Math 108 offered through our office. The students are freshmen ethnic minority students pursuing an engineering degree. There is no potential risk to the students in any terms. All of the students have knowingly enrolled in the OMSE section of Math 108 for the purpose of obtaining academic instruction in College Algebra as part of the requirements for an engineering degree. All data collected will be coded to provide the striccest confidentiality. Additional data will be aggregated so that no individual student's identity will be revealed. 217 A 218 ‘ The benefits to be gained by the individual student is what we seek to determine. The advent of computer technology is rapidly entering into regular classroom activities even at the collegiate level. The effective usage of computers for instructional purposes depends not only on the computer system and the instructional materials but also on the character- istics of the students using the system. The potential benefits of the study are a better understanding of the interaction effects between affec- tive student characteristics and the method of instruction. Knowledge gained by this and similar studies can be incorporated into aptitude- treatment interaction regression models for the prescription of instruc— tional methodologies compatible with individual learning styles or prefer- ences. The benefits of using microcomputer technology as an instructional tool includes time reduction, cost/student decreases, and as this study intends to investigate, more effective learning for those students who are more responsive to this method of instruction. The consent procedure will be a verbal request by the instructor of the Math 108 class to the students for'voluntary participation in the IAR Questionnaire and the entire study. Since the proposed study dOes not involve any extraordinary classroom procedures, verbal rather than written consent will be used. Consent from the students will be obtained one week prior to the commencement of the experiment. The only formal consent form that may be used will be for the collection of ccvariant data. The form that will be used is attached. The data collection instruments consist of the IAR Questionnaire (attached), an achievement test, several quizzes, and homework assignments. All measures, except the questionnaire, are being prepared. The content and method of administering the in class measures is a standard practice in the Math 108 class. The questionnaire will be given as a written exercise one week prior to the beginning of the experiment. A verbal description of the study, the procedures to be followed, and the role of the students will be made prior to the administration of the questionnaire. 219 MICHIGAN STATE UNIVERSITY COLLEGE OF EDUCATION EAST LANSING ' MICHIGAN ° 48824 DEPARTMENT OF ADMINISTRATION AND CURRICULUM ERICKSON HALL January 12, 1983 Memorandum To: UCRIHS // From: Howard W. Hickey Dissertation Director . Subj: Research Porposal fo Gregory C. Hamilton I have reviewed the research proposal for Gregory C. Hamilton's doctoral thesis and state that it meets with my full approval. 220 MICHIGAN STATE UNIVERSITY UNIVERSITY costumer. on RESEARCH INVOLVING EAST LANSING - MICHIGAN - «324' HUMAN SUBJECTS (UCRIHS) ' ' 238 ADMINISTRATION BUILDING pnIMsnas February 8, I983 Mr. Gregory C. Hamilton Office of Minority Student Education 144 Engineering Building. Dear Mr. Hamiltonf Subject: Proposal Entitled, “The Impact of Student Locus of Control on Academic Achievement as a Function of Lecture Versus Computer-Assisted instruction" UCRIHS review of the above referenced project has now been ccmpieted. i am pleased to advise that the rights and welfare of the human subjects appear to be adequately protected and the Committee, therefore, approved this project at its meeting on February 7. 1933 . . You are reminded that UCRIHS approval is valid for one calendar year. if you pian to continue this project beyond one year, please make provisions for obtaining appropriate UCRIHS approval prior to the anniversary date noted above. Any changes in procedures involving human subjects must be reviewed by the UCRIHS prior to initiation of the change. UCRiHS must also be notified promptly of any problems (unexpected side effects, compiaints, etc.) involving homan subjects during the course of the work. Thank you for bringing this project to our attention. if we can be of any future help, please do not hesitate to let us know. Sincereiy, . Lo 7 7 r i not—we Henry E. Bredeck‘ Chairman, UCRIHS HEB/jms cc: Dr. Hickey 221 MEMORANDUM To: Dr. Lou Anna K. Simon From: Gregory C. Hamiito Date: June 15, 1983 Re: Release of Confidential Information Pursuant of successful completion of a doctorai degree in Education Administration and Curriculum, I am requesting the release of con- fidentiai information for the students on the attached list. Each of these students participated in a research study and signed a con- sent form. A copy of this form is attached and copies of the signed forms are available upon request. The purpose of the research study is to investigate the impact student locus of control has on academic achievement as a function of lecture versus computer-assisted instructional methodologies. To properly evaluate this interaction, information pertaining to the students' previous performance and abiiity is needed. This data is required for muitivariant regression anaiysis procedures. I, therefore, request release of the following data elements for the students listed. MSU Math Placement Score SAT Math Score ACT Math Score Math 082/104 Course Grade . Math 108 Course Grade (”thde I O O 0 Thank you for your cooperation. Enclosures 222 MICHIGAN STATE UNIVERSITY omc: or me novosr m1 unsmc - Mia-now - 43324 Abummnwarxmammumnc July 6, 1983 MEMORANDUM TO: Dr. Charles Eberly Dr. Horace King FROM: Lou Anna Kinsey Simon, Assistant Provost w SUBJECT: Release of Confidential Information Mr. Gregory C. Hamilton is conducting a doctoral research project to investigate the impact student locus of control has on academic achievement as a function of lecture versus computer-assisted instructional methodologies. This research project was reviewed and approved by the University Committee on Research Involving Human Subjects in February, 1983. Mr. Harrison has secured individual consent statements from each student in this study to permit the release of confidential information on test scores and grades in selected courses. Attached is a copy of the consent form signed by the students . Mr. Harrison has provided lists of the students in the study. He has provided to me copies of the signed consent forms. I have checked these forms against the lists and found the lists correspond exactly to the set of consent forms. Therefore, as chairperson of the Committee on Release of Information, I authorize the release of the following information for the students on the attached lists to Mr. Gregory C. Hamilton. 1. M30 Math Placement Score 2. SAT Math Score 3. ACT Math Score n. Math 082/10fl Course Grade 5. Math 108 Course Grade By copy of this memorandum, I am notifying Mr. Gregory C. Hamilton of this decision and am requesting that he contact Dr. Bberly about the test score information and Dr. King about the course grade information. Thank you for your support of this project. If you have any questions, please let me know. LAKS:Jm Attachment cc: Mr. Gregory C. Hamilton / Mr. Lynn Peltier Dr. Henry Hredeck MS U is en Affirmative Action/Emu] Courtenay Institution APPENDIX C RESEARCH FORMS AND QUESTIONNAIRES RESEARCH PROJECT CONSENT FORM INTELLECTUAL ACHIEVEMENT RESPONSIBILITY (IAR) QUESTIONNAIRE MATH 108 RESEARCH STUDY TIME SURVEY APPENDIX C Research Project Consent Form As a student in Math 108, I am freely participating in an educational research project investigating the effectiveness of formal lecture versus computer-assisted instructional techniques. I understand that the following information is needed for statistical analysis purposes. I also understand that this information will be held in the strictest confidence and will eventually be aggregated by the analysis proce- dures so as to obscure any connection between the data and myself. I, therefore, give consent that the registrar provide the following infor- mation to Gregory C. Hamilton: MSU math placement score, SAT math score, ACT math score, and Math 082 and Math 108 course grades. I also consent to the use of quiz, homework, and exam scores in the Math 108 course to be used in the analysis. Signed Student Number 223 224 Research Questionnaire Assigned Class Number As part of an educational research study, your cooperation and assistance is needed to obtain data for the purpose of determining several variables pertinent to the project. Please answer each of the following 34 questions by selecting one ando only one of the choices. For the results of this survey to be VETid and mean1ngful, each item must be answered this way. There is no “correct" answer and all of your responses will be confidential. Thank you for your time and assistance. 1. If a teacher gives you a good grade in a class, would it probably be a. because the teacher liked you, or b. because of the work you did? 2. When you do well on a test in school, is it more likely to be a. because you studied for it, or b. because the test was especially easy? 3. When you have trouble understanding something in school, is it usually a. because the teacher didn't explain it clearly, or b. because you didn't listen carefully? 4. When you read a story and can't remember much of it, is it usually a. because the story wasn't well written, or b. because you weren't interested in the story? 5. Suppose your parents say you are doing well in school. Is this likely to happen a. because your school work is good, or b. because they are in a good mood? 6. Suppose you did better than usual in a subject at school. Would it probably happen a. because you tried harder, or b. because someone helped you? 10. 11. 12. 13. 225 When you lose at a game of cards or checkers, does it usually happen a. because the other player is good at the game, or b. because you don't play well? Suppose a person doesn't think you are very bright or clever. a. can you make him change his mind if you try to, or b. are there some people who will think you're not very bright no matter what you do? If you solve a puzzle quickly, is it a. because it wasn't a very hard puzzle, or b. because you worked on it carefully? If someone tells you that you are dumb, is it more likely that they say that a. because they are mad at you, or b. because what you did really wasn't very bright? Suppose you study to become a teacher, scientist, or doctor and you fail. Do you think this would happen a. because you didn't work hard enough, or b. because you needed some help and other people didn't give it to you? When you learn something quickly in school, is it usually a. because you paid close attention, or b. because the teacher explained it clearly? If a teacher says to you, "Your work is fine," is it a. something teachers say to encourage students, or b. because you did a good job? 14. 15. 16. 17. 18. 19. 20. 21. 226 When you find it hard to work arithmetic or math problems at school, is it a. because you didn't study well enough before you tried them, or b. because the teacher gave problems that were too hard? When you forget something you heard in class, is it a. because the teacher didn't explain it very well, or b. because you didn't try very hard to remember? Suppose you weren't sure about the answer to a question your teacher asked you, but your answer turned out to be right. Is it likely to happen a. because she wasn't as particular as usual, or b. because you gave the best answer you could think of? When you read a story and remember most of it, is it usually a. because you were interested in the story, or b. because the story was well written? If your parents tell you you're acting silly or not thinking clearly, is it more likely to be a. because of something you did, or b. because they happen to be feeling cranky? When you don't do well on a test at school, is it a. because the test was especially hard, or b. because you didn't study for it? When you win at a game of cards or checkers, does it happen a. because you play real well, or b. because the other person doesn't play well? . If people think you're bright or clever, is it a. because they happen to like you, or b. because you usually act that way? 22. 23. 24. 25. 26. 27. 28. 29. 227 If a teacher fails you in a course, would it probably be a. because the teacher "had it in for you," or b. because your school work wasn't good enough? Suppose you don't do as well as usual in a subject at school. Would this probably happen a. because you weren't as careful as usual, or b. because somebody bothered you and kept you from working? If someone tells you that you are bright, is it usually 3. because you thought up a good idea, or b. because they like you? Suppose you become a famous teacher, scientist, or doctor. Do you think this would happen a. because other people helped you when you needed it, or b. because you worked very hard? Suppose your parents say you aren't doing well in your school work. Is this likely to happen more a. because your work isn't very good, or b. because they are feeling cranky? Suppose you are showing a friend how to play a game and he has trouble with it. Would that happen a. because he wasn't able to understand how to play, or b. because you couldn't explain it well? When you find it easy to work arithmetic or math problems at school, is it usually a. because the teacher gave you especially easy problems, or b. because you studied your book well before you tried them? When you remember something you heard in class, is it usually a. because you tried hard to remember, or b. because the teacher explained it well? 30. 31. 32. 33. 34. 228 If you can't work a puzzle, is it more likely to happen a. because you are not especially good at working puzzles, or b. because the instructions weren't written clearly enough? If your parents tell you that you are bright or clever, is it more likely a. because they are feeling good, or b. because of something you did? Suppose you are explaining how to play a game to a friend and he learns quickly. Would that happen more often a. because you explained it well, or b. because he was able to understand it? Suppose you're not sure about the answer to a question your teacher asks you and the answer you give turns out to be wrong. 15 it likely to happen a. because the teacher was more particular than usual, or b. because you answered.too quickly? If a teacher says to you, "Try to do better," would it be a. because this is something teachers might say to get students to try harder, or b. because your work wasn't as good as usual? 229 Math 108 Research Survey Assigned Class Number The amount of time you spend in a formal educational setting (lecture/ computer) is only a fraction of the total time spent studying the material. Consequently, much of the actual learning takes place out- side of the classroom. To finish up the research study would you kindly take a few minutes to complete the following form. Lecture Study Time Study Time Topic Section Date ' PRIOR to class AFTER Class The Complex Numbers 10.1 May 19 hr. hrs. Complex Roots of Equations 10.2 Poynomials, Remainder and 11.1 May 20 hr. hrs. Factor Theorems Synthetic Division 11.2 May 23 hr. hrs. Zeroes of Polynomials 11.3 May 24 * hr. hrs. The Rational Root 11.4 May 25 hr. hrs. Theorem Exam -—— May 27 hr. As always, results will be kept in the strictest confidence. APPENDIX D ACHIEVEMENT INSTRUMENTS INSTRUCTIONAL OBJECTIVES WINTER TRIAL EXAM AND STATISTICS SPRING TRIAL EXAM AND STATISTICS APPENDIX D INSTRUCTIONAL OBJECTIVES Lesson 1. Add, subtract, multiply, and divide complex numbers. Solve linear, quadratic, and cubic equations with complex and/or real coefficients. Lesson 2. Add, subtract, and multiply polynomials of varying degrees. Find the quotient and remainder functions Q(x) and R(x) when a given polynomial P(x) is divided by a given D(x) function. Use the Remainder Theorem to determine if (x — r) is a factor of a given P(x) polynomial. Use the Remainder Theorem to evaluate the polynomial P(x) for a given value of x. Use synthetic division of polynomials when the divisor, D(x), has the form (x - r). Lesson 3. Find a polynomial of lowest degree when given its roots and their multiplicities. Given a polynomial and some of its roots, find the other roots. Given a polynomial with integer or rational coeffi- cients, find all of the rational roots. 230 231 TUTOR NAME W '83 108 TEST 5 STUDENT NUMBER 1. (12 pts) Simplify, write the answers in the form a + bi 2. (12 pts) Solve for x and y real. 2x + 731-1 = 8 + (log3y)i 3. (12 pts) Find ALL solutions to x3 = 27 expressing the complex solutions in the a + bi form. Test 5 Page 2 232 A. (10 pts) Derive equations sufficient to determine A & B , such that o (x - 1) and (x + 2) are factors of f(x) = 2x3 + Ax” + Bx - 12 . Just derive the equations, DO NOT SOLVE. 5. ( 6 pts) What is the remainder when 2x100 - 3x5 + A is divided by (x + 1) . Show your method. 6. ( H)pts) Find the quotient q(x) and remainder r(x) if . 4 f(x) = 2x + 3x3 + 9x2 + 13x + 5 is divided by g(x) = x2 + 4 . Test 5 233 Page 3 7. (10 pts) Use synthetic division to find the quotient q(x) and remainder r when (5x + 3x4 — 7x3 + 1) is divided by (x - 2) 8. (10 pts) Find a polynomial of lowest degree with zeros 2, 1 + i , -i if a) the coefficients may be nonreal, and b) the coefficients meet be real. Leave in factored form. Test 5 234 Page 4 9. ( 18pts) Find ALL the zeros of f(x) = 2x3-+ x2 + 1 . Statistics: Test 5, Winter 1983 Exam Reliability Index 0.79 Item Value Average St. Dev. Disc Diff 1 12 8.75 2.36 0.53 73.33 2 12 10.63 3.26 0.58 76.67 3 12 8.66 3.91 0.68 70.00 4 10 6.22 2.56 0.45 57.00 5 6 5.66 1.13 0.38 85.00 6 10 8.41 2.68 0.42 80.00 7 10 9.63 0.96 0.00 100.00 8 10 7.47 2.61 0.38 75.00 9 18 15.13 4.11 0.50 76.11 Test Ave. = 80.53 Dev. = 15.51 Number of exams = 32 23:; NAME TUTOR - .3, 108 SP '83 STUDENT NUMBER TEST 5 VI l. (l2 pts.) Write each in simplest form. a) /:2 ~ JTS = b) 12" a c) The remainder when x11 - x2 + 3 is divided by x + l is 2. (12 pts.) Solve for z expressing the solution in the a + bi form. (2 + i): = 32 + 6 + 31 3. (l2 pts.) Find all real and complex solutions to x3 + 64 = 0 . Express the complex solutions in the a + bi form. ' Test 5 237 Page 2 ’ 4. ( 9 pts.) Solve for x and y real. . _ l . . 2y+21--4—+1log4x 3 2 5. (l0 pts.) Determine k such that x + 2 is a factor of x + kx - kx - l0 . 6. (l0 pts.) Find the quotient q(x) and remainder r(x) if f(x) = 8x3 - 6x2 - 4x is divided by g(x) = 2x2 - x + l . Test 5 238 Page 3 7. (l0 pts) Use synthetic division to find the quotient and remainder when 7x8 + 3x5 - x2 + 10 is divided by x — 1 . (l0 pts.) Find a polynomial of lowest degree with the given zeros if a) the coefficients may be nonreal b) the coefficients must be real. The zeros are 3, 3 + 21 , and 2 - 3i‘. Leave answers in factored form. I 1 Test 5 239 Page 4. 9. (15 pts.) Find all the zeros of 5x3 - x2 - l5x + 3 . 240 Statistics: Test 5, Spring 1983 Item Pt. Value Average St. Dev. Disc Diff 1 12 8.89 3.19 0.52 65.63 2 12 7.26 4.06 0.83 54.17 3 12 5.11 4.48 0.50 45.83 4 9 4.74 3.09 0.50 58.33 5 10 8.53 2.74 0.15 92.50 6 10 6.95 3.07 0.30 80.00 7 10 8.63 1.69 0.27 86.25 8 10 5.84 2.70 0.45 52.50 9 15 9.05 4.26 0.48 67.50 Test Ave. = 65.00 St. Dev. = 16.56 Number of exams = 19 Exam Reliability Index = 0.71 APPENDIX E JOHNSON-NEYMAN PROCEDURE INTRODUCTION AND DISCUSSION CALCULATIONS FOR I+ COMPOSITE DATA CALCULATIONS FOR I' COMPOSITE DATA APPENDIX E JOHNSON-NEYMAN PROCEDURE INTRODUCTION AND DISCUSSION This appendix provides a comprehensive discussion of the Johnson-Neyman procedure for determining regions of sig- nificance given the existence of significant interaction effects. Although no such interactions were found in this study, this appendix is provided as an illustration of the technique. Readers are cautioned that the discussion pro- vided herein are not to be considered as empirical findings of this research effort. The original formulation of this procedure (Johnson & Neyman, 1936) used two groups, two predictor measures, and one criterion variable. The theory has been extended to include any number of groups, predictors, and criterion measures (Abelson, 1953; Cahen & Linn, 1971; Carroll & Wilson, 1970; Johnson & Fay, 1950; Koenker & Hansen, 1942; Potthoff, 1964). The present application involves two groups, a single predictor, and one criterion variable. The applications presented in the latter sections of the appen- dix are patterned after the Johnson and Jackson (1959) examples, specifically those beginning on page 432. Setting aside the fact that no ATI exists and, for the purposes of illustration only, rejecting the hypothesis of equal 241 242 regression slopes, the following is a presentation of the Johnson-Neyman procedure. The crux of this procedure is to identify those values of the independent variable(s) which predict statistically significant differences in the dependent variable(s). Referring back to Figure 3, the treatment lines based on the I+ score would cross at 1+ = 19.3 if the scale extended that far. The extreme left region of the scale shows what appears to be a substantial difference in the predicted achievement scores; a difference of over 30 points exists at 1+ = 0. The task is to determine which value or values of 1+ (and I—) form the boundaries between equivalent and non- equivalent treatments. Putting this in the context of the I+ graph, does the value of 1+ = 10, for example, create one region, (I+ < 10), for which the lecture mode is signifi- cantly superior to CAI and another region (1+ > 10) where there is no statistical difference between them? For a single predictor variable the regions of signifi- cance are determined by a quadratic equation in one varia- ble. The solutions provide two values of the predictor which can be represented by vertical lines on a graphical display. If two predictors are used the problem usually becomes one of defining ellipses on two-dimensional pre- dictor planes (Cahen & Linn, 1971). As more variables are included, the complexity of the significant regions in the variable space becomes increasing complex as well. For the 243 present discussion the equations that the Johnson-Neyman procedure produce are presented in Table 15. Table 15 Illustration of the Johnson-Neyman Procedure: Equations and Solutions for Regions of Significance 1+ and I- Regressions Predictor Equations for Significance Solutions 1+ 24.301:2 — 659.30X + 4515.23 = O x .. 13.6 .t 1.331 I- 21.28x2 - 579.le + 3983.71 = 0 x .. 13.6 i 1.451 The imaginary components are a result of the large variance terms (constants) in the equations. This is pro- bably a reflection of the lack of any ATI which is prerequi- site for obtaining meaningful results. Although the issue of imaginary solutions is not addressed in the literature, and since this is presented for illustrative purposes only, this author will take the liberty of ignoring these terms. The regions of significance are graphed in Figures 5 and 6. Values of 1+ less than 13.6 would indicate that the lecture method is significantly better than CAI. Students having I+ scores greater than 13.6 would do equally well with either method. The I- graph shows that students having I- scores in excess of 13.6 would do better with the lecture method Achievement Score 100 70 60 50 244 " Ach = -1.15 1+ + 94.0 Region of Significant Difference r- "' I+ = 13.6 , Ach . .42 1* + 63.7 -—— Lecture --- CAI r 5 10 15 17 1+ Subscore Figure 5 Region of Significance for 1+ Subscale Composite Sample Illustration of Johnson-Neyman Procedure Achievement Score 100 90 80 70 60 50 245 Y Ach = .33 I” + 73.6 Region of/' Significant Difference I' = 13.6 . Ach . -.77 I" + 79.8 — Lecture --- CAI l . I .. 5 10 15 17 I- Subscore Figure 6 Region of Significance for I- Subscale Composite Sample Illustration of Johnson-Neyman Procedure 246 while students with low I- scores should perform equally well regardless of instructional method. The predicted dif— ferences in achievement scores at the cut-off points are 9.0 and 8.7 for the 1+ and I- predictors respectively. The last question is how many students fall into these regions of significance? Table 16 provides the number and percentage of students who, based on the preceeding analy- sis, would benefit from the lecture method. Table 16 Locus of Control Conditions Predicting Lecture Assignment Study Sample MMM 160 Sample Decision Rule N Z N Z 1. 1+ < 13.6 19 37.3 27 47.4 2. I- > 13.6 27 52.9 14 24.6 3. 1+ < 13.6 or I- > 13.6 36 70.6 37 64.9 4. 1+ < 13.6 and The table presents information for students in the composite study sample and the slightly larger MMM 160 group. The differential assignment of the lecture method is determined by the regions of significance. Assignments can be made according to one of four decision rules with the joint condition that I+ and I- both be in the regions of 247 significance being the most strict. Assuming this analysis is valid, these values show that only lO-ZOZ of the students would definitely benefit from the lecture treatment. The remaining 80-902 of the class would perform equally well with either treatment. Readers are again cautioned that this discussion has been for illustrative purposes only and the results are not the conclusions forwarded as part of the research. 248 CALCULATIONS FOR I+ COMPOSITE DATA Table 17 1+ Composite Data Work Sheet Lecture CAI External Internal External Internal I+ Ach I+ Ach I+ Ach I+ Ach 1 15 95 14 93 15 84 17 63 2 11 90 17 l4 14 80 17 92 3 15 98 17 89 12 73 16 76 4 13 55 14 72 13 76 16 65 5 11 78 16 81 11 84 16 57 6 ll 94 15 87 12 90 15 67 7 14 90 14 82 15 88 13 69 8 14 85 13 9O 12 61 16 45 9 10 71 14 85 12 72 17 61 10 13 95 16 84 11 23 11 11 54 17 78 15 68 12 13 98 15 61 13 15 64 15 61 14 ll 79 16 92 15 13 52 16 14 88 17 16 63 n 14 14 17 17 ll 11 9 9 fix: 177 1146 256 1272 142 799 143 595 i 12.6 81.86 15.1 74.82 12.9 72.64 15.9 66.11 [x2 2279 96922 3884 101592 1858 61519 2285 40679 6‘2 3.17 239.52 1.81 401.03 2.49 348.25 1.61 167.86 1 1:2 14584 19058 10437 9462 N 31 20 [x 433 285 I2 2418 1394 1x2 6163 4143 £22 198514 102198 Ixz 33642 19899 Sxx 114.97 81.75 Szz 9910.22 5036.27 249 2 2 Step 1’ Sz.xl = Szzl - Slelsxxl = 9910.22 - (132)2/114.97 = 9758.67 2 2 Sz.x2 = SzzZ - Szx2/Sxx2 = 5036.27 - (34.5)2/81.75 = 5021.71 F a (Nz’z) S:41 a (% %)(9758. 67) a 1 21 N -2 ' S2 5021. 71 ° z.x2 F20,29,.05 a 1.94 so result is not significant Therefore 61 = a} Equal variances of Z. 2 Step 2. NT =- 51 ixT - 10306 2 th 718 ZzT 300712 7:21, - 3812 ZxTzT - 53541 SzzT a 15783.4 SxxT = 197.69 szT a -125.98 Source of df SSz SSx Sum of Regression Variation Products Coef. Between 1 836.91 .97 -28.48 29.36 Within Lec. 30 9910.22 141.97 -132.00 -l.15 Within CA1 19 5036.27 81.75 34.50 .42 Total Within 49 14946.49 196.72 -97.50 -.50 Total 50 15783.40 197.69 -125.98 -.64 250 Step 3: Source of Variation df Sum of Squares Mean Square Total Within 48 14805.6 308.45 Group Residual Total Residual 49 15703.1 -— Difference 1 897.5 897.5 F = 897.5/308.45 = 2.91 ml = 1; r12 = 48 F needed for significance = 4.04 Therefore, A1 = A2 Step 43: Residual Sum of df Sum of Squares Mean Square Square of Z Within Lecture 29 9758.67 -- Within CAI 18 5021.71 -- Subtotal 47 14780.38 314.48 Total Within 48 14805.60 308.45 Difference Between 1 25.22 25.22 Regression Coef. F = 314.48/25.22 = 12.47 111 = 47; n = 1 F needed for significance = 252 Therefore, B1 = B2 b. Test for B = 0 2 (Sle + 82x2) = (:132.0 + 34.50)2 g 48 32 (Sxxl + Sxxz) (114.97 + 81.75) F = §%§f%% = .16 Not significant So, B1 = B2 = 0 251 Step 5: Assuming that A1 # A2 and B1 # B2 then: 22 = .42 X + 63.7 D = z1 - z2 = -1.57 x + 30.3 111 + n2 - 4 D2 F ‘ ( 1 ) ° (P+Q) s: F1,47,.05 = 4°°45 D2 > F(P+Q)( S: > '- 111 + n2 - 4 n +n (x-x )2 (x-x )2 1 2 1. 2. “Q” nn +—§— 1""?— l 2 xxl xx2 2 2 P+Q 3 31+20 + (3-13.97) + (x-l4.25) 31x20 114.97 81.75 P+Q = .021x2 - .592x + 4.264 32 - 52 + $2 = 9758 67 + 5021 71 = 14805 6 a z.xl z.x2 ' ' ' Substituting: 2 14805.6 (-1.57x + 30.3)2 2.(4.045)(.021x Rearranging: 0‘2 24.30x 47 2 - 659.30x + 4515.23 Using the quadratic formula to solve for x yields: x - 13.6 + 1.331 and x = 13.6 - 1.331 252 CALCULATIONS FOR I- COMPOSITE DATA Table 18 I- Composite Data Work Sheet Lecture CAI External Internal External Internal 1— Ach I- Ach I- Ach I- Ach 1 11 95 16 93 11 84 16 63 2 11 90 12 14 ll 80 15 92 3 12 98 15 89 7 73 15 76 4 14 55 16 72 14 76 13 65 5 16 78 14 81 14 84 14 57 6 14 94 13 87 15 90 13 67 7 11 90 16 82 12 88 15 69 8 11 85 16 90 13 61 15 45 9 11 71 15 85 10 72 16 61 10 12 95 15 84 13 23 11 12 54 15 78 10 68 12 8 98 12 61 13 11 64 13 61 14 15 79 15 92 15 14 52 16 15 88 17 14 63 n 14 14 17 17 11 11 9 9 2x 169 1146 246 1272 130 799 132 595 E 12.1 81.86 14.5 74.82 11.8 72.64 14.7 66.11 2x2 2095 96922 3588 101592 1590 61519 1946 40679 0.2 4.23 239.52 1.76 401.03 5.36 348.25 1.25 167.86 1x2 13712 18700 9453 8728 N 31 20 [x 415 262 {z 2418 1394 2x2 5683 3536 222 198514 102198 2x2 32412 18181 Sxx 127.35 103.80 253 2 Step 1' Sz.xl = Szzl — Szx1/Sxx1 = 9910.22 - (42.0)2/127.35 = 9896.37 2 2 Sz.x2 = SzzZ - Szx2/Sxx2 - 5036.27 - (80.4)2/103.8O = 4974.0 2 N 2-2 S z.xl 9896. 37 F = (N 2-2) ‘ ‘2— ‘ ("21" 9)(4974. 00) " 1°23 1 S z.x2 F20’29“05 = 1.94 so result is not Significant Therefore f3 - '2 Equal variances of Z. - 2 a Step 2. NT - 51 sz 9219 2 'L'xT - 677 ZzT - 300712 22.1. - 3812 szzT =- 50593 Source of df SSz SSx Sum of Regression Variation Products Coef. Between 1 836.91 1.01 28.17 27.89 Within Lec. 30 9910.22 127.35 42.00 .33 Within CAI 19 5036.27 103.80 -80.40 -.77 Total Within 49 14946.49 231.15 -37.60 -.16 Total 50 15783.40 232.16 -9.43 -.04 254 Step 3: Source of Variation df Sum of Squares Mean Square Total Within 48 14881.7 310.04 Group Residual Total Residual 49 15783.0 -- Difference 1 842.9 842.9 F - 842.9/310.04 = 2.72 111 = 1; n2 = 48 F needed for significance = 4.04 Therefore, A1 = A2 Step 43: Residual Sum of df Sum of Squares Mean Square Square of Z Within Lecture 29 9896.37 -- Within CAI 18 4974.00 -- Subtotal 47 14870.37 316.39 Total Within 48 14881.70 310.04 Difference Between 1 11.33 11.33 Regression Coef. F - 316.39/11.33 = 27.93 111 = 47; n2 = 1 F needed for significance = 252 Therefore, B1 = B2 b. Test for B a 0 2 . (Sle T SzxZ) _ (42.0 — 80.40)2 a 6 38 (Sxxl + Sxxz) (127.35 + 103.80) F a 3I0%%4 a .02 Not significant S0, B1 a B2 = 0 255 Step 5: Assuming that Al 4 A2 and B1 # B2 then: Z1 .33 X + 73.6 N) II D = 2 - 2 = 1.10 x — 6.2 1 2 111 + 112 - 4 D2 F ‘ ( 1 ) ° (P+Q) 8: F1,47,.05 = 4°°45 D2 > F(P+Q)( S: ) 111 + n — 4 n +n (x-x )2 (x-x )2 1 2 1. 2. P+Q= +—§‘—+—§— n1n2 xxl xx2 2 2 P+Q 8 31+20 + (x-13.39)_+ (x-13.10)g 31x20 127.35 103.80 P+Q = .018x2 - .463x + 3.143 2 2 2 S = S + S = 9896.37 + 4974.00 = 14870.4 a z.xl z.x2 Substituting: (1.10x — 6.2)2 2_(4.045)(.018x2 — .463x + 3.143)cl3§%%ei) Rearranging: 0.2 21.28.:2 — 579.01x + 3983.71 Using the quadratic formula to solve for x yields: x a 13.6 + 1.451 and x = 13.6 - 1.451 REFERENCE NOTES REFERENCE NOTES 1. Hestenes, M.D. Personal communication, April 1984. 2. Hill, R.0., Jr. Personal communication, March 1984. 256 BIBLIOGRAPHY BIBLIOGRAPHY Abelson, R.P. A note on the Neyman-Johnson technique. Psy- chometrika, 1953, 18, 213-218. Abrahamson, D., Schludermann, S., & Schludermann, E. Repli— cation of dimensions of locus of control. Journal gf Consulting and Clinical Psychology, 1973, 44, 320. Alker, H.A., & Wohl, J. Personality and achievement in a suburban and an inner city school. Journal 2; Social Issues, 1972, 28(4), 101-113. Alpert, D., & Bitzer, D.L. Advances in computer-based edu- cation. Science, 1970, 167, 1582-1590. Archer, R.P. Relationships between locus of control and anxiety. Journal gf Personality Assessment, 1979, 4;, Ashby, E. Machines, understanding, and learning: Reflec- tions on technology in education. The Graduate Journal, 19672.1(2)o 357-373. Atkinson, R.C. Ingredients for a theory of instruction. American Psycholggist, 1972, 21, 921-931. Avner, A., Moore, C., & Smith, S. Active external control: A basis for superiority of CBI. Journal .gf Computer- Based Instruction, 1980, 6, 115-118. Avoli, J.R. The ascent of computers. Microcomputing, 1981, '§(9), 108-114. Baker, F. Computer-managed instruction: A context for computer-based instruction. In H.F. O'Neil Jr. (Ed.), Computer-based instruction: ‘A state-of-the-art assess- ment. New York: Academic Press, 1981. Barnett, M.A., & Kaiser, D.L. The relationship between intellectual-achievement responsibility attributions and performance. Child Study Journal, 1978, 8, 209-215. Battista, M. The interaction between two instructional treatments of algebraic structures and spatial- visualization ability. Journal gf Educational Research, 1981, 14, 337-341. Bell, F. Classroom computers: Beyond the 3 R's. Creative Computing, 1979, 5(9), 68-70. 257 258 Berliner, D.C., & Cahen, L.S. Trait-treatment interaction and learning. In F.N. Kerlinger (Ed.), Review 2; research 13 education (Vol. 1). Itasca, IL: F.E. Peacock Publications, Inc., 1973. Bevan, N. Is there an optimum speed for presenting text on a VDU? International Journal .2: Man-Machine Studies, 1981, 14, 59—76. Bloom, B.S. Mastery learning. In J.H. Block (Ed.), Mastery learning; Theory and practice. New York: Holt, Rinehart, and Winston, Inc., 1971. Boettcher, E.G., Alderson, S.F., & Saccucci, M.S. A com- parison of the effects of computer—assisted instruction versus printed instruction on student learning in the cognitive categories of knowledge and application. Journal pf Computer-Based Instruction, 1981, 8, 13-17. Bork, A.M. Learning with computers. Bedford, MA: Digital Press, 1981. Bracht, G.H. Experimental factors related to aptitude- treatment interactions. Review ‘2: Educational Research, 1970, 49, 627-645. Brebner, A., Hallworth, H.J., Woetowich, H.M., Mah, S.N., & Huang, H.C. Results and implications of some recent research on CAI in mathematics. Proceedings pf 332 19th Annual Convention pf the Association 2: Educational Data Systems, 1981, 42-47. Brown, J.C., & Strickland, B.R. Belief in internal-external control of reinforcement and participation in college activities. Journal .2; Consulting gpd Clinical Psy- chology, 1972, 38, 148. Brown, J.S., & Burton, R.R. SOPHIE--A pragmatic use of artificial intelligence in CA1. Proceedings pf Egg 1974 Annual Conference .2: the Association ‘3; Computipg Machinery, 1974, 571-579. Bruner, J.S. Toward g theory pf instruction. Cambridge, MA: Harvard University Press, 1966. Bruner, J.S. The process of education. Cambridge, MA: Harvard University Press, 1977. Buck, M.R., & Austrin, B.R. Factors related to school achievement in an economically disadvantaged group. Child Development, 1971, 4;, 1813-1826. 259 Bunderson, C.V. The design and production of learner- controlled courseware for the TICCIT system: A progress report. International Journal .3; Man-Machine Studies, 1974, p, 479-491. Burke, R.L. CAI sourcebook. Englewood Cliffs, NJ: Prentice-Hall, Inc., 1982. Burns, P.K., & Bozeman, W.C. Computer-assisted instruction and mathematics achievement: Is there a relationship? Educational Technology, 1981, 21(10), 32-39. Cahen, L.S., & Linn, R.L. Regions of significant criterion differences in aptitude-treatment-interaction research. American Educational Research Journal, 1971, 8, 521-530. Campbell, D.T., & Stanley, J.C. Experimental and quasi- experimental designs for research. New York: Rand McNally, 1963. Carbonell, J.R. AI in CAI: An artificial-intelligence approach to computer-assisted instruction. IEEE Trans- actions gg Man-Machine Systems, 1970, MMS-ll, 190-202. Carnegie Commission on Higher Education. The fourth revolu- tion: Instructional technology $3 higher education. New York: McGraw-Hill, Inc., 1972. Carroll, J.B., & Wilson, G.F. An interactive-computer program for the Johnson—Neyman technique in the case of two groups, two predictor variables, and one criterion variable. Educational and Psychological Measurement, 1970, 30, 121-132. Cellini, J.V., & Kantorowski, L.A. Internal-external locus of control: New normative data. Psychological Reports, 1982, 51, 231—235. Chambers, J.A., & Sprecher, J.W. Computer assisted instruc- tion: Current trends and critical issues. Association for Computing Machineryl Communications, 1980,‘23, 332- 342. Clifford, M.M., & Cleary, T.A. The relationship between children's academic performance and achievement account- ability. Child Development, 1972, Q3, 647-655. Coleman, J.S., Campbell, E.Q., Hobson, C.J., McPartland, J., Mood, A.M., Weinfeld, F.D., & York, R.L. Eguality 2: Educational Opportunity. U.S. Department of Health, Education, and Welfare. Washington DC: U.S. Government Printing Office, 1966. 260 Corno, L., Mitman, A., & Hedges, L. The influence of direct instruction on student self-appraisals: A hierarchical analysis of treatment and aptitude-treatment interaction effects. American Educational Research Journal, 1981, 18, 39-61. Crandall, V.C., Katkovsky, W., & Crandall, V.J. Children's beliefs in their own control of reinforcements in intellectual-academic achievement situations. Child Development, 1965, 39, 91-109. Crandall, V.J., Katkovsky, W., & Preston, A. Motivational and ability determinants of young children's intellectual achievement behaviors. Child Development, 1962, 33, 643- 661. Cronbach, L.J. Coefficient alpha and the internal structure of tests. Psychometrika, 1951, pg, 297-334. Cronbach, L.J., & Snow, R.E. Aptitudes and instructional methods: A handbook for research pp interactions. New York: Irvington, 1977. Crowder, N.A. Automatic tutoring by means of intrinsic pro- gramming. In E. Galanter (Ed.), Automatic teaching: The state pf the art. New York: John Wiley & Sons, Inc., 1959. Crowder, N.A. Intrinsic and extrinsic programming. In J.E. Coulson (Ed.), Programmed Learning and computer-based instruction. New York: John Wiley & Sons, Inc., 1962. Davis, R.E., Alexander, L.T., & Yelon, S.L. Learning system design. New York: McGraw-Hill, Inc., 1974. deCharms, R. Personal causation: The internal affective deter-inants lg; behavior. New York: Academic Press, 1968. deCharms, R., & Carpenter, V. Measuring motivation in cul- turally disadvantaged school children. In H.J. Klaus- meier & G.T. O'Hearn (Eds.), Research and development toward the improvement 2f education. Madison, WS: Dembar Educational Research Services, Inc., 1968. Deigan, G.M., & Duncan, R.E. CAI in three medical training coures: It was effective! Behavior Research Methods g Instrumentation, 1978, IQ, 228-230. Domino, G. Interactive effects of achievement orientation and teaching style on academic achievement. Journal 2; Educational Psychology, 1971, 92, 427-431. 261 DuCette, J., Walk, 3., & Soucar, E. Atypical patterns in locus of control and nonadaptive behavior. Journal pf Personality, 1972, 50, 287-297. Duke, M.P., & Nowicki, S. Locus of control and achievement- The confirmation of a theoretical expectation. The Journal 3; Psychology, 1974, 81, 263-267. Dwyer, T.A. Heuristic strategies for using computers to enrich education. International Journal ‘gf Man-Machine Dwyer, T.A., & Critchfield, M. Multi-computer systems for the support of inventive learning. In P.R. Smith (Ed.), Computer-assisted learning. Oxford, England: Pergamon Press Ltd., 1981. Ebel, R.L. Essentials pf educational measurement (3rd ed.). Englewood Cliffs, NJ: Prentice—Hall, Inc., 1979. Edwards, J., Norton, 3., Taylor, S., Weiss, M., & Dussel- dorp, R. How effective is CAI? A review of the research. Educational Leadership, 1975,.33, 147-153. Ellis, A.B. The use and misuse pf computers .ifl education. New York: McGraw-Hill, Inc., 1974. Faust, G.W. Design strategy and the TICCIT system. View— points, 1974, 50(4), 91-101. Feather, N.T. Some personality correlates of external con- trol. Australian Journal .2: Psychology, 1967, 12, 253- 260. Fiegenbaum, E.A., & McCorduck, P. The fifth generation. Reading, MA: Addison-Wesley Publishing Co., 1983. Fendrich-Salowey, C., Buchanan, M., & Drew, C.J. Mathema- tics, quantitative and attitudinal measures for elemen— tary school boys and girls. Psychologipal Reports, 1982, 21, 155-162. Gable, A., & Page, C.V. The use of artificial intelligence techniques in computer-assisted instruction: An overview. International Journal pf Man-Machine Studies, 1980, 12, 259-282. Gagne, R.M. Essentials pf learning for instruction. New York: Holt, Rinehart and Winston, Inc., 1975. Gagne, R.M. The conditions pf learning. New York: Holt, Rinehart and Winston, Inc., 1977. 262 Gallagher, J.P. The effectiveness of man-machine tutorial dialogues for teaching attribute blocks problem-solving skills with an artificial intelligence CAI system. Instructional Science, 1981, 19, 297—332. Gaynor, P. The effect of feedback delay on retention of computer-based mathematical material. Journal pf Computer—Based Instruction, 1981, 8, 28-34. Gleason, G.T. Microcomputers in education: The state of the art. Educational Technolggy, 1981,.21(3), 7-18. Goldberg, L.R. Student personality characteristics and optimal college learning conditions: An extensive search for trait-by-treatment interaction effects. Instruc- tional Science, 1973, 1, 153-210. Gurin, P., Gurin, C., Lao, R.C., & Beattie, M. Internal- external control in the motivational dynamics of Negro youth. Journal pf Social Issues, 1969,.2§(3), 29-53. Hamilton, G.C. A microcomputer managed instructional system. Proceedings pf the 19th Annual Convention pf the Association pf Educational Data Systems, 1981, 112-118. Hamsher, J.H., Geller, J.D., & Rotter, J.B. Interpersonal trust, internal-external control, and the Warren commis- sion report. Journal ‘gf Personality and Social Psy- chology, 1968, g, 210-215. Hansen, J.B. Effects of feedback, learner control, and cognitive abilities on state anxiety and performance in a computer-assisted instruction task. Journal .gf Educa- tional Psychology, 1974, Q6, 247-254. Hays, W.L. Statistics for the social sciences (2nd ed.). New York: Holt, Rinehart and Winston, Inc., 1973. Heider, F. The psychology pf interpersonal relations. New York: John Wiley & Sons, Inc., 1958. Hestenes, M.D., & Hill, R.0., Jr. Algebra and trigonometry with calculators. Englewood Cliffs, NJ: Prentice-Hall, Inc., 1981. Hjelle, L.A. Internal-external control as a determinant of academic achievement. Psychological Reports, 1970, 2Q, 326. Hoffman, J.L., & Waters, K. Some effects of student person- ality on success with computer-assisted instruction. Educational Technology, 1982, 22(3), 20-21. 263 Hofmeister, A.M. Microcomputers in perspective. Excep- tional Children, 1982, 42(2), 115-121. Holmes, G. Computer-assisted instruction: A discussion of some of the issues for would-be implementors. Educa- tional Technology, 1982, 22(9), 7-13. Hunter, B., Kastner, C.S., Rubin, M.L., & Seidel, R.J. Learning alternatives lg U.S. Education: Where student and computer meet. New York: Educational Technical Pub- lications, 1975. James, W.H., & Rotter, J.B. Partial and 100% reinforcement under chance and skill conditions. Journal ‘3; Experi- mental Psychology, 1958, 55, 397—403. Jamison, D., Suppes, P., & Wells, S. The effectiveness of alternative instructional media: A survey. Review 2; Educational Research, 1974, £4, 1-67. Janicki, T.C., & Peterson, P.L. Aptitude-treatment inter- action effects of variations in direct instruction. American Educational Research Journal, 1981, l8, 63-82. Jelden, D.L. A CAI "Coursewriter" system for the microcom- puter. AEDS Journal, 1981, ii, 159-168. Jenkin, J.M. Some principles of screen design and software for their support. In P.R. Smith (Ed.), Computer- Assisted Learning. Oxford, England: Pergamon Press Ltd., 1981. Joe, V.C. Review of the internal-external control construct as a personality variable. Psychological Reports, 1971, 28, 619-640. Johnson, P.O., & Fay, L.C. The Johnson-Neyman technique, its theory and application. Psychometrika, 1950, 12, 349-367 0 Johnson, P.O., & Jackson, R.W.B. Modern statistical methods: Descriptive and inductive. Chicago, IL: Rand McNally & Co., 1959. Johnson, P.O., & Neyman, J. Tests of certain linear hypotheses and their application to some educational problems. Statistical Research Memoirs, 1936, 1, 57-93. Judd, W.A., Bunderson, C.V., & Bessent, E.W. quinvestiga- tion 2f the effects 2;, learner control 13 computer- assisted instruction prerequisite mathematics (MATHS). Austin, TX: The University of Texas at Austin, 1970. (ERIC Document Reproduction Service No. ED 053 532) 264 Kamm, S.D. Tutoring with microcomputers--A project report. Journal pf College Science Teaching, 1983, 12, 279-281. Karmos, J.S., Bryson, S., & Tracz, S. Equations for pre- dicting special students who will earn degrees. Journal p£_Instructional Psychology, 1982,‘g, 139-144. Katz, I. The socialization of academic motivation in minority group children. In D. Levine (Ed.), Nebraska Symposium pp Motivation (Vol. 15). Lincoln: University of Nebraska Press, 1967. Katz, I. A critique of personality approaches to Negro per- formace, with research suggestions. Journal .pf Social Issues, 1969,'2§(3), 13-27. Kearsley, 0., Hunter, B., & Seidel, R.J. Two decades of computer based instruction projects: What have we learned? T.H.E. Journal, 1983, 19(3), 90-94. Keller, F.S. "Good-bye, Teacher...". Journal ‘_f Applied Behavior Analysis, 1968,.l, 79-89. Kennelly, K., & Kinley, S. Perceived contingency of teacher administered reinforcements and academic performance of boys. Psychology ip the Schools, 1975, 12, 449-453. Kiehlbauch, J.B. Selected changes over time in internal- external control expectancies in a reformatory population (Doctoral dissertation, Kansas State University, 1968). Dissertation Abstracts, 1968,'22, 37lB-372B. (University Microfilms No. 68-9912) Koenker, R.H., & Hansen, C.W. Steps for the application of the Johnson-Neyman technique--A sample analysis. Journal Ipf Experimental Education, 1942:.19: 164-173. Koffman, E.B., & Blount, S.E. Artificial intelligence and automatic programming in CAI. Artificial Intelligence, 1975, 9, 215-234. ' Koffman, E.B., & Perry, J.M. A model for generative CAI and concept selection. International Journal 2; Man-Machine Studies, 1976,.8, 397-410. Kozma, R.B., Belle, L.W., & Williams, G.W. Instructional pechnigues_ip higher education. Englewood Cliffs, NJ: Educational Technology Publications, 1978. Kulhavy, R.W., & Anderson, R.C. Delay-retention effect with multiple-choice tests. Journal 1p: Educational Psy- chology, 1972, 9;, 505-512. 265 Kulik, J.A., & Jaska, P. PSI and other educational tech- nologies in college teaching. Educational Technology, 1977, 11(9), 12-19. Kulik, J.A., Kulik, G.C., & Cohen, P.A. Effectiveness of computer-based college teaching: A meta-analysis of findings. Review pf Educational Research, 1980, £0, 525- 544. Lahey, G.F. .122 effect .pf instructional presentation seguence ‘pp student performance $3 computer-based instruction. San Diego, CA: Navy Personnel Research and Development Center, 1979. (ERIC Document Reproduction Service No. ED 175 453.) Lahey, G.F. The effect of instructional sequence on per- formance in computer-based instruction. Journal pf Computer-Based Instruction, 1981, 1, 111-116. Lao, R.C. Internal-external control and competent and inno- vative behavior among Negro college students. Journal pf Personality and Social Psycholggy, 1970, 14, 263-270. Larson, I., Markosian, L.Z., Suppes, P. Performance models of undergraduate students on computer—assisted instruc- tion in elementary logic. Instructional Science, 1978, l, 15.35. Lefcourt, R.M. Internal-external control of reinforcement: A review. Psychological Bulletin, 1966, pp, 206-220. Lefcourt, H.M. Recent developments in the study of locus of control. In B.A. Maher (Ed.), Progress ip experimental personality research (Vol. 6). New York: Academic Press, 1972. Lefcourt, R.M. Locus pf control: Current trends .ip theory and research. Hillsdale, NJ: Lawrence Erlbaum Associ- ates, 1976. Lefcourt, H.M., & Ladwig, G.W. The American Negro: A prob- lem in expectancies. Journal '3: Personality and Social Psycholggy, 1965, 1, 377-380. Leherissey, B.L., O'Neil, H.F., Jr., & Hansen, D.L. Effects of memory support on state anxiety and performance in computer-assisted learning. Journal pf Educational Psy- chology, 1971,.92, 413-420. Leherissey, B.L., O'Neil, H.F., Jr., Heinrich, D.L., & Hansen, D.N. Effect of anxiety, response mode, subject matter familiarity, and program length on achievement in computer-assisted learning. Journal pf Educational Psy: chology, 1973,‘§4, 310-324. 266 Leiblum, M.D. Factors sometimes overlooked and underesti- mated in the selection and success of CAL as an instruc- tional medium. AEDS Journal, 1982, 15, 67-79. Lessing, E.E. Racial differences in indices of ego func- tioning relevant to academic achievement. The Journal_gf Genetic Psychology, 1969, 115, 153-167. Levien, R.E. The emerging technology: Instructional uses pf the computer 1y higher education. New York: McGraw-Hill, Inc., 1972. Lysiak, F., Wallace, S., & Evans, C.L. Computer-assisted instruction. 1975-76 evaluation report. Fort Worth, TX: Fort Worth Independent School District, 1976. (ERIC Document Reproduction Service No. ED 140 495.) Magidson, E.M. Issue overview: Trends in computer-assisted instruction. Educational Technology, 1978,.l§(4), 5-8. Malone, T.W., Macken, E., & Suppes, P. Toward optimal allo- cation of instructional resources: Dividing computer- assisted instruction time among students. Instructional Science, 1979, 8, 107-120. Markowitz, N., & Renner, K.E. Feedback and the delay- retention effect. Journal .2; Experimental Psychology, 1966, 12, 452-455. Masuo, D.M., & Furuta, B.A. Interactive effects of prior knowledge of "Credit Costs" concepts and math skills and mode of programmed instruction. Journal 2; Computer- Based Instruction, 1981, 8, 9-12. McClelland, D.C., Atkinson, J.W., Clark, R.W., & Lowell, E.L. The achievement motive. New York: Appleton- Century-Crofts, Inc., 1953. McCombs, B.L., & McDaniel, M.A. 0n the design of adaptive treatments for individualized instructional systems. Educational Psychologist, 1981, lg, 11-22. McGhee, F.E., & Crandall, V.C. Beliefs in internal-external control of reinforcements and academic performance. Child Development, 1968, 32, 93-102. Mehrens, W.A., & Lehmann, I.J. Measurement and evaluation .yg education and psychology. New York: Holt, Rinehart and Winston, Inc., 1973. Messer, S.B. The relation of internal-external control to academic performance. Child Development, 1972, 43, 1456- 1462. 267 Minton, H.L. Power as a personality construct. In B.A. Maher (Ed.), Progress .yp experimental pgrsonality research (Vol. 4). New York: Academic Press, 1967. Mitchell, P.D. Representation of knowledge in CAL course- work. In P.R. Smith (Ed.), Computer-assisted learning. Oxford, England: Pergamon Press Ltd., 1981. Mirels, H.L. Dimensions of internal versus external con- trol. Journal ‘2; Counsulting and Clinical Psychology, 1070, 35, 226-228. Mischel, W. 0n the future of personality measurement. American Psychologist, 1977,.32, 246-254. Moore, M.V., & Nawrocki, L.H. The educational effectiveness p£,grsphic displays for computer-assisted instruction: Final report. Alexandria, VA: Army Research Institute for the Behavorial and Social Sciences, 1978. (ERIC Document Reproduction Service No. ED 169 917.) Moore, M.V., Nawrocki, L.H., & Simutis, Z.M. The instruc- tional effectiveness pg three levels pf graphics displays for computer-assisted instruction. Arlington, VA: Army Research Institute for the Behavioral and Social Sciences, 1979. (ERIC Document Reproduction Service No. ED 178 057.) Morris, L.W., & Carden, R.L. Relationship between locus of control and extraversion-introversion in predicting aca- demic behavior. Psychological Reports, 1981, $8, 799- 806. Naisbitt, J. Megatrends. New York: Warner Books, Inc., 1982. Nie, N.H., Hull, C.H., Jenkins, J.C., Steinberger, K., & Bent, D.H. Statistical package for the social sciences (2nd ed.). New York: McGraw-Hill, Inc., 1975. Nowicki, S., Jr., & Segal, W. Perceived parental character- istics, locus of control orientation, and behavioral cor- relates of locus of control. Developmental Psychology, 1974, 12, 33-37. Nowicki, S., Jr., & Strickland, B.R. A locus of control scale for children. Journal 'pf Consulting and Clinical Psychology, 1973, go, 148-154. Oettinger, A.G., & Marks, S. Runp computer, run: The myth- ology pg educational innovation. Cambridge, MA: Harvard University Press, 1969. 268 Ohlsson, K., Nilsson, L.C., & Ronnberg, J. Speed and accur- acy in scanning as a function of combinations of text and background colors. International Journal pf Man-Machine Studies, 1981, 15, 215—222. Orlansky, J. & String, J. Computer-based instruction for military training. Defense Management Journal, 1981, .ll(2), 46-54. Papert, S. Mindstorms: Children, computers, and powerful ideas. New York: Basic Books, Inc., 1980. Park, 0., & Tennyson, R.D. Adaptive design strategies for selecting number and presentation order of examples in coordinate concept acquisition. Journal .2; Educational Psychology, 1980, 12, 362-370. Perlmuter, L.C., & Monty, R.A. The importance of perceived control: Fact or Fantasy? American Scientist, 1977,-§§, 759-765. Peterson, P.L., Janicki, T.C., & Swing, S.R. Ability x treatment interaction effects on children's learning in large-group and small group approaches. American Educa- tional Research Journal, 1981, 18, 453-473. Petzel, T.P., & Gynther, M.D. Effects of internal-external locus of control and skill or chance instructional sets on task performance. The Journal pf General Psychology, 1970, 82, 87-93. Phares, E.J. Expectancy changes in skill and chance situa- tions. Journal ‘pf Abnormal and Social Psychology, 1957, ii, 339-3420 Phares, E.J. Locus pf control: A personality determinant pi behavior. Morristown, NJ: General Learning Press, 1973. Phares, E.J. Locus pf control Lg personality. Morristown, NJ: General Learning Press, 1976. Ploch, M. Micros flood campuses. High Technology, 1984, Posner, M.I., & Osgood, G.W. Computers in the training of inquiry. Behavior Research Methods '8 Instrumentation, 1980, 12, 87-95. Potthoff, R.F. On the Johnson-Neyman technique and some extensions therof. Psychometrika, 1964, 22, 241-256. Powell, A., & Vega, M. Correlates of adult locus of con- trol. Psychological Reports, 1972, 30, 455-460. 269 Prawat, R.S. Mapping the affective domain in young adoles- cents. Journal of Educational Psychology, 1976, go, 566- 572. Prawat, R.S., Grissom, S., & Parish, T. Affective develop- ment in children, grades 3 through 12. The Journal of Genetic Psychology, 1979, 135, 37-49. Pressey, S.L. A simple apparatus which gives tests and scores-and teaches. School and Society, 1926, 23, 373- 376. Rankin, R.J., & Trepper, T. Retention and delay of feedback in a computer-assisted instructional task. Journal .of Experimental Education, 1978, §§(4), 67-70. Rapaport, P., & Savard, W.C. Computer-assisted instruction. Research pp school effectiveness project: Topic summary report. Portland, OR: Northwest Regional Educational Laboratory, 1980. (ERIC Document Reproduction Service No. ED 214 707) Reece, M.J., & Gable, R.K. The development and validation of a measure of general attitudes toward computers. Edu- cational and Psychological Measurement, 1982, .gg, 913- 916. Ripple, R.E., Millman, J., & Glock, M.D. Learner character- istics and instructional mode: A search for disordinal interactions. Journal ‘of Educational Psychology, 1969, pp, 113-120. Robinson, M. Classroom control: Some cybernetic comments on the possible and the impossible. Instructional Science, Roblyer, M.D. When is it "Good Courseware"? Problems in developing standards for microcomputer courseware. Edu- cational Technology, 1981, 21(10), 47-54. Rothen, W., & Tennyson, R.D. Application of Bayes' theory in designing computer-based adaptive instructional strategies. Educational Psychologist, 1978,'l2, 317-323. Rotter, J.B. Social learning and clinical psychology. Englewood Cliffs, NJ: Prentice—Hall, Inc., 1954. Rotter, J.B. Generalized expectancies for internal versus external control of reinforcement. Psychological Mono- graphs, 1966 80(1, Whole No. 609). 270 Rotter, J.B. Some problems and misconceptions related to the construct of internal versus external control of reinforcement. Journal ‘of Consultiog and Clinical Psy- chology, 19751.32: 56-67. Rotter, J.B., Chance, J., & Phares, E.J. Applications .of 3 social learning theory .2£ personality. New York: Holt, Rinehart and Winston, Inc., 1972. Rushby, N.J. Computers 13 the teaching process. New York: John Wiley & Sons, Inc., 1979. Rushby, N.J. (Ed.) Selected readings .ip computer-based learning. New York: Nichols Publishing Co., 1981. Salisbury, A.B. Computers and education: Toward agreement on terminology. Educational Technology, 1971, ll(9), 35- 40. Salomon, G. Heuristic models for the generation of aptitude-treatment interaction hypotheses. Review .of Educational Research, 1972, 3;, 327-343. Sarason, S.B., Davidson, K.S., Lighthall, F.F., Waite, R.R., & Ruebush, B.R. Anxiety ‘ip elementary school children. New York: John Wiley & Sons, Inc., 1960. Sassenrath, J.M., & Yonge, G.D. Delayed information feed- back, feedback cues, retention set, and delayed reten- tion. Journal _o£ Educational Psychology, 1968, 22, 69- 73. Schurdak, J.J. An-approach to the use of computers in the instructional process and an evaluation. American Educa- tional Research Journal, 1967, 4, 59-73. Shaw, R.L., & Uhl, N.P. Control of reinforcement and aca- demic achievement. The Journal pg Educational Research, 1971, pg, 226-228. ' Skinner, B.F. Teaching machines. Science, 1958, 12 , 969- 977. Smith, I.D. Impact of computer-assisted instruction on student attitudes. Journal .2£ Educational Psychology, 1973, pg, 366-372. Solomon, D., Houlihan, K.A., & Parelius, R.J. Intellectual achievement responsibility in Negro and white children. Psychological Reports, 1969,‘g£, 479-483. Spielberger, G.D. Theory and research on anxiety. In C.D. Spielberger (Ed.), Anxiety and behavior. New York: Academic Press, 1966. 271 Spielberger, C.D. Anxiety as an emotional state. In C.D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. 1). New York: Academic Press, 1972. Spielberger, C.D., O'Neil, H.F., Jr., & Hansen, D.N. Anxiety, drive theory, and computer-assisted learning. In B.A. Maher (Ed.), Progress i3 experimental personality research (Vol. 6). New York: Academic Press, 1972. Splittgerber, F.L. Computer-based instruction: A revolution in the making? Educational Technology, 1979, 12(1), 20- 26. Steinberg, E.R. Review of student control in computer- assisted instruction. Journal of Computer-Based Instruc- tion, 1977’ A, 84.90. Stipek, D.J., & Weisz, J.R. Perceived personal control and academic achievement. Review .of Educational Research, 1981, 51(1), 101-137. Strassberg, D.S. Relationships among locus of control, anx- iety, and valued-goal expectations. Journal ‘of Con- sultiog and Clinical Psychology, 1973, ii, 319. Sturges, P.T. Verbal retention as a function of the infor- mativeness and delay of informative feedback. Journal-of Educational Psychology, 1969, Q9, 11-14. Sturges, P.T. Information delay and retention: Effect of information in feedback and tests. Journal .of Educa- tional Psychology, 1972, 93, 32-43. Sturges, P.T. Immediate vs. delayed feedback yp'g computer- managed test: Effects op loog-term retention. Chico, CA: California State University, 1978. (ERIC Document Repro- duction Service No. ED 160 635) Sturges, P.T., Sarafino, R.P., & Donaldson, P.L. Delay- retention effect and informative feedback. Journal 2; Experimental Psychology, 1968, 18, 357-358. Suppes, P., Fletcher, J.D., & Zanotti, M. Models of indi- vidual trajectories in computer-assisted instruction for deaf students. Journal u2£ Educational Psychology, 1976, .631. 117-127. Surber, J.R., & Anderson, R.C. Delay-retention effect in natural classroom settings. Journal of Educational Psy- chology, 1975, Q1, 170-173. 272 Tennyson, C.L., Tennyson, R.D., & Rothen, W. Content struc- ture and instructional strategies as design variables in concept acquisition. Journal .of Educational Psychology, 1980, 12, 499-505. Tennyson, R.D. Adaptive instructional models for concept acquisition. Educational Technology, 1975, l§(4), 7-15. Tennyson, R.D. Instructional control strategies and content structure as design variables in concept acquisition using computer-based instruction. Journal oi Educational Psychology, 1980,.12, 525-532. Tennyson, R.D. Use of adaptive information for advisement in learning concepts and rules using computer-assisted instruction. American Educational Research Journal, 1981, 18, 425-438. Tennyson, R.D., & Rothen, W. Pretask and on-task adaptive design strategies for selecting number of instances in concept acquisition. Journal of Educational Psychology, 1977, pg, 586-592. Tennyson, R.D., & Rothen, W. Management of computer-based instruction: Design of an adaptive control strategy. Journal oi Computer-Based Instruction, 1979, 2, 63-71. Thibaut, J.W., & Kelley, H.H. The social psychology .of groups. New York: John Wiley & Sons, Inc., 1959. Thomas, D.B. Effects of alternative response mode require- ments go the accuraoy .2£ student response ip computer- based testing. Iowa City, IA: Iowa University, 1979. (ERIC Document Reproduction Service No. ED 183 130.) Throop, W.E., & MacDonald, A.P. Internal-external locus of control: A bibliography. Psychological Reports, 1971, .2§(1), 175-190. Tobias, S. Effect of creativity, response mode, and subject matter familiarity on achievement from programmed instruction. Journal 2; Educational Psychology, 1969, pp, 453-460. Tobias, S. Distraction, response mode, anxiety, and achievement in computer-assisred instruction. Journal of Educational Psychology, 1973a, o5, 23-237. Tobias, S. Sequence, familiarity, and attribute by treat- ment interactions in programmed instruction. Journal oi Educational Psychology, 1973b, Q4, 133-141. Tobias, S. Achievement treatment interactions. Review [of Educational Research, 1976, 39, 61-74. 273 Tobias, S., & Abramson, T. Interaction among anxiety, stress, response mode, and familiarity of subject matter on achievement from programmed instruction. Journal pg Educational Psychology, 1971, 9;, 357—364. Tobias, S., & Duchastel, P.C. Behavioral objectives, sequence, and anxiety in CAI. Instructional Science, 1973, 3, 231-242. Toong, H.D., & Gupta, A. Personal computers. Scientific American, 1982, 247(6), 86-107. Vinsonhaler, J.F., & Bass, R.K. A summary of ten major studies on CAI drill and practice. Educational Tech- nology, 1972, 12(7), 29-32. Watson, D. Relationship between locus of control and anxiety. Journal .of Personality and Social Psychology, 1967, p, 91—92. Watson, D., & Baumal, E. Effects of locus of control and expectations of future control upon present performance. Journal of Personality and Social Psychology, 1967, 9, 212-215. Weiner, B. A theory of motivation for some classroom experiences. Journal “of Educational Psychology, 1979, ll, 3-25. Weiner, B., Nierenberg, R., & Goldstein, M. Social learning (locus of control) versus attributional (causal stabil- ity) interpretations of expectancy of success. Journal .of Personality, 1976, 55, 53-68. Westrom, M. The instructional uses of microcomputers. AEDS Monitor, 1983, 21(7,8), 22-27. Wexler, J.D. Information networks in generative computer- assisted instruction. IEEE Transactions .23 Man-Machine Systems, 1970, MMS-ll, 181-190. "11111114411111117111111171.7115