Why have the results of the Follow Through evaluation failed to
impact the policies and practices of the educational community? Why
have the most effective approaches for educating children not been
widely disseminated? Why has the knowledge gained from the Follow
Through evaluation not been used to reform education in America? The
answers to these questions may be found in part by looking at how
instruction is viewed by the various elements of the educational
Follow Through provides an opportunity for such an analysis because it revealed how the educational industry collectively conceived of, planned, conducted, and interpreted a large scale educational experiment. When I was in graduate school, I wrote a lengthy paper in which I traced the history of Project Follow Through, looked at its implications for education, and analyzed the contingencies that determine educational practices. This article is condensed from the paper, which will be published this fall by the Cambridge Center for Behavioral Studies. I made a vow that I would tell the story of Project Follow Through to anyone who would listen.
Many people know the history of Project Follow Through far better
than I, because they lived it. As I understand it, it goes something
like this. In 1964 Congress passed the Economic Opportunity Act,
which initiated a range of programs intended to fight poverty. The
federal policy that emerged from the EOA was influenced by a growing
consensus that education would be the antidote to poverty by
providing skills necessary to break out of the exiting cycle. One of
the best known programs to develop from this rationale was Head
Head Start began in the summer of 1965. It was immediately popular and continues to enjoy tremendous public support today. It is commonly believed that Follow Through received impetus from the success of Head Start and from a study showing that gains made by Head Start children dissipated when they began school. In reality, the decision to initiate Follow Through was probably a function of both conviction and expediency. In any event, in February 1967, President Johnson requested that Congress establish a program to "follow through" on Head Start. The outcome was Public Law 90-92, authorizing Project Follow Through. Although it has been referred to as the largest and most expensive educational experiment, Follow Through was not initially conceived as an experiment, but as a comprehensive social services program. However, before the program got underway, budget cuts forced a reconceptualization and Follow Through was converted to a longitudinal experiment aimed at finding effective methods for teaching disadvantaged children. The Follow Through experiment involved close to 10,000 children from 120 communities each year from 1968 to 1976. Follow Through continued as a service program until funding was eliminated in 1995.
The design of the Follow Through experiment was called planned
variation. Based on the notion that a variety of curricula and
instructional methods could be designed, implemented and evaluated,
the planned variation approach was intended to reveal differences in
effectiveness among different teaching approaches.
A plan was devised that made it possible to implement a variety of educational models in local school districts, while avoiding the appearance of unwarranted federal intervention. The Office of Education contracted with developers of educational approaches who then acted as sponsors of their model and worked cooperatively with districts to implement the model in Follow Through classrooms.
Each sponsor was responsible for translating the model's particular approach to education into practice. This included selecting or developing instructional materials, and training teachers in the model's instructional method. The Follow Through sponsors' task of designing a complete curriculum for the entire school day, had never before been attempted in educational reform.
The selection of sponsors began in January of 1968. Individuals or groups who were involved in developing new approaches for teaching young children were invited to present information about their programs. Sixteen developers subsequently submitted formal proposals, twelve of which were accepted. The approaches represented the entire spectrum of assumptions about instruction, ranging from the carefully controlled approach of the Direct Instruction and Behavior Analysis models to child-centered approaches such as Bank Street and Open Education. Ten additional sponsors were added over the following three-year period, not because they offered unique approaches to compensatory education, but because they offered the possibility of enlarging the Follow Through constituency.
The selection of sites progressed synchronously with sponsor selection. From among a group of 225 nominated school districts, a total 51 were selected, based on their ability to begin a comprehensive services program before the start of the school year, their willingness to participate in the planned variation experiment and their working relationship with local community action agencies.
Sites and sponsored models were paired during a four day conference held in Kansas City in February, 1968. In an effort to increase cooperation in implementing the various models, local representatives were encouraged to choose the model they believed was most compatible with the goals and interests of their district. Each model was implemented in a variety of sites, where children received daily instruction in the model. Performance data were collected when children entered the program and at the end of each school year until they completed third grade.
The evaluation of this enormous project was complex and expensive.
The data were collected by Stanford Research Institute and analyzed
by Abt Associates.
Eleven outcome measures were included in the national evaluation. All sponsors agreed upon the outcome measures, which were intended to assess performance in different learning domains including basic academic skills, general problem-solving skills, and the development of self-concept.
For evaluation purposes Abt Associates divided models into three broad categories according to their areas of primary emphasis. The typology was determined based on the sponsor's own program description and stated goals and objectives of the models. The Basic Skills category included models that focused primarily on directly teaching fundamental skills in reading, arithmetic, spelling and language. The Cognitive-Conceptual category included that intended to develop "learning-to-learn" and problem solving skills. Models in the Affective-Cognitive category emphasized development of self-concept and positive attitudes toward learning, and secondarily, "learning-to-learn" skills. Nine of the major models included in the national evaluation are described by model type in Table 1.
For each outcome subtest, Abt evaluators compared the performance of a group of Follow Through children at a given site with a comparison group. This process resulted in more than 2,000 comparisons. The difference between a Follow Through group and the comparison group was used as the measure of effect. An effect was judged to be educationally meaningful if the difference 1) was statistically significant and 2) was at least one quarter standard deviation. When Follow Through scores exceeded non-Follow Through scores the outcome was considered positive. When non-Follow Through scores surpassed Follow Through scores, the outcome was considered negative. Average effects were computed for individual models, as well as for model types.
Abt Associates produced yearly reports, which were published in four volumes titled Education as Experimentation: A Planned Variation Model. Volume IV (Stebbins, et. al., 1977), provides the most comprehensive evaluation of the differential effectiveness of the models. The following findings of average effects for model types are paraphrased (in italics) from Volume IV-A (pp. 135-148).
Models that emphasized basic skills succeeded better than other models in helping children gain these skills. Groups of children in Basic Skills models performed significantly better on measures of academic skills than did non-Follow Through groups. Abt evaluators concluded that a Basic Skills model would be preferable if an educator was concerned with teaching skills such as spelling, math computation, language, and word knowledge. Note that the Abt report refers to the superiority of a model type. However, it is not inclusion in a category that leads to educational effectiveness, but the particular instructional materials and procedures used. The Direct Instruction model had an unequivocally higher average effect on scores in the basic skills domain than did any other model.
Where models emphasized other skills, the children they served tended to score lower on tests of basic skills than they would have done without Follow Through. With the exception of the Florida Parent Education model, all Cognitive-Conceptual and Affective-Cognitive models had more negative than positive outcomes on measures in the basic skills domain. That is, performance of students in the comparison group was superior to that of the Follow Through students in those models. At the end of third grade, children taught in these models had achievement scores that were, in fact, lower than would have been predicted in the absence of "compensatory" education. Thus, four years spent in the majority of models actually increased the educational deficits that Follow Through was intended to remediate.
No type of model was notably more successful than the others in raising scores on cognitive conceptual skills. No model type had an overall average positive effect on measures in this domain, which included reading comprehension and problem solving. One model that did have considerable impact on cognitive conceptual skills was the Direct Instruction model. Not one model in the Cognitive-Conceptual category obtained a positive average effect on these measures, despite the fact that their instructional programs emphasized development of these skills. Models that focused on cognitive-conceptual skills were incapable of influencing standardized measures of those skills after four years of instruction.
Models that emphasized basic skills produced better results on tests of self-concept than did other models. On the average, children in models the evaluators classified in the Basic Skills category, performed better on affective measures than did children in Cognitive-Conceptual or Affective models. All models in the Basic Skills category had positive average model effects. The only other model to demonstrate a positive average effect was the University Florida's Parent Education model. In every case, the models that focused on affective development had negative average effects on measures in this domain.
The Direct Instruction and Behavior Analysis models ranked first and second, respectively, in average effects on affective measures. Both of these approaches stress careful structuring and sequencing of curriculum materials that are designed to limit the number of errors and ensure successful performance. In addition, they both rely on frequent measurement of the child's progress in order to provide immediate remediation. These models view positive self-concept as an outcome of skill acquisition. In other words, rather than considering self-concept a necessary prerequisite for learning, they contend that instruction resulting in academic success leads to improved self-concept. The data uphold this view.
It would be a mistake, however, to claim that instruction in a Basic Skills model leads to academic success and improved self-concept. Significant differences on both categories of measures were observed for only two of the Basic Skills models, Direct Instruction and Behavior Analysis. In other words, describing the result as a "Basic Skills" effect does not identify the specific instructional variables that lead to significantly better performance in both outcome areas. The fact remains however, that no model classified as "Affective" had a positive average effect on affective measures.
The average effects for nine individual models are represented in Figure 1. The centerline of the figure indicates no difference between students in a Follow Through model and comparison students. Notice that the Direct instruction model is the only model to show sizable positive effects on all measures. The majority of models show considerable negative effects (performance below the level of the comparison group) on all measures. These findings clearly show the Direct instruction model to be superior on these measures compared with traditional programs and with other Follow Through models.
Figure 1: This figure shows the average effects of nine Follow Through models on measures of basic skills (word knowledge, spelling, language, and math computation), cognitive-conceptual skills (reading comprehension, math concepts, and math problem solving) and self-concept. This figure is adapted from Engelmann, S. and Carnine, D. (1982), Theory of Instruction: Principles and applications. New York: Irvington Press.
The evaluation was not only costly, but controversial. At least three other major reanalyses of the data were independently conducted. None of these analyses show significant disagreement with respect to achievement data. Results of the national evaluation and all subsequent analyses converge on the finding that the highest achievement scores were attained by students in the Direct Instruction model. The Follow Through experiment was intended to answer the question "what works" in educating disadvantaged children. If education is defined as the acquisition of academic skills, the results of the Follow Through experiment provide a clear answer to the question.
The purpose of the Follow Through planned variation experiment was
to identify effective educational methods. However, there is little
utility in identifying effective methods if they are not then made
accessible to school districts. The Joint Dissemination Review Panel
and the National Diffusion Network were created to validate and
disseminate effective educational programs. In 1977, Follow Through
sponsors submitted programs to the JDRP. "Effectiveness" was,
however, broadly interpreted. For example, according the JDRP, the
positive impact of a program need not be directly related to academic
achievement. In addition, a program could be judged effective if it
had a positive impact on individuals other than students. As a
result, programs that had failed to improve academic achievement in
Follow Through were rated as "exemplary and effective." And, once a
program was validated, it was packaged and disseminated to schools
through the National Diffusion Network.
The JDRP's validation practices did not go unchallenged. According to former Commissioner of Education, Ernest Boyer, "Since only one of the sponsors (Direct Instruction) was found to produce positive results more consistently than any of the others, it would be inappropriate and irresponsible to disseminate information on all the models..." (quoted in Carnine, 1984, p. 87). However, commissioner Boyer's concerns could not prevent the widespread dissemination of ineffective instructional approaches. The JDRP apparently felt that to be "fair" it had to represent the multiplicity of methods in education. Not only did this practice make it virtually impossible for school districts to distinguish between effective and ineffective programs, it defeated the very purpose for which the JDRP and NDN were established.
The effect of the Follow Through evaluation may also be measured
by the extent to which the findings influenced decisions about
funding. While all Follow Through models received budget cuts over
the years, the disbursement of available funds was not based on
effectiveness, but on a non-competitive continuation basis. In fiscal
year 1982, the funding formula was changed so that sponsors with
JDRP-validated programs received the lowest level of funding, while
the highest level of funding went to those sponsors that had not been
Not surprisingly, funding ineffective programs at a higher level did not make them effective. Not one additional program was validated during the following year. Yet the same funding policy continued to be implemented, favoring ineffective programs. It is clear that increased financial support by itself does not lead to increased performance by students. How children are taught is critically important.
The results of the Follow Through evaluation also failed to influence decisions about allocation of federal research funds. Planned variation makes it possible to identify the best performing programs and then subject them to further analyses. Instead the Office of Education and National Institute of Education agreed to spend 12 million dollars to develop and study new Follow Through approaches with the primary concern being "whether or not an approach can be put in place and maintained, not with the effectiveness of the approach in improving student outcomes" Proper and St. Pierre, 1980, p. 8) [emphasis added]. According to Stallings (1975), the Direct Instruction model was not only most effective, it and the Behavior Analysis models were the most easily implemented. If information about implementation was needed, these two models provided a good starting point. The plan that was pursued shows total neglect of the findings of the Follow Through evaluation and astonishing disregard for the academic achievement of students.
Perhaps even more disturbing is the fact that twenty years after the publication of the Follow Through evaluation, there is little evidence that the results have altered educational practices in American classrooms. The majority of schools today use methods that are not unlike the Follow Through models that were least effective (and in some cases were most detrimental). Barriers at all levels of the educational system preclude widespread adoption of the model that was most effective.
The Educational Establishment
The history of Follow Through and its effects constitute a case
study of how the educational establishment functions. As in other
bureaucracies, it is composed of parochial vested interests that work
to either maintain the status quo or to advance a self-serving
agenda. As a result, the largest educational experiment in history
(costing almost one billion tax payer dollars) has been effectively
prevented from having the impact on daily classroom practices that
its results clearly warranted. Let's look at some factors that
operate at each level of the educational establishment to influence
decisions about teaching methods and materials.
Policymakers. Follow Through demonstrated that public policy is based on public support, not on empirical evidence. Thus, the position that officials adopt with respect to teaching methods is most likely to be congruent with the position of the majority. Because the Direct Instruction model represents a minority view in education, it was not surprising that policymakers failed to take a strong position in support of the Follow Through results.
Although some policymakers may have some formal training in areas of education, they typically rely on input from education professionals when developing and supporting programs. The influence of stakeholders in traditional educational practices can be seen throughout the history of Project Follow Through. Planning committees, advisory boards, and task forces were composed of representatives of universities and research centers. These professionals usually represent educational philosophies that the Follow Through results suggest do not, and cannot, lead to the development of effective teaching methods. For example, the chairman of the Follow Through National Advisory Committee was the dean of the Bank Street College of Education, whose model was ineffective in improving academic achievement or affective measures.
Clearly some professionals with a self-interest have the power to influence educational policy in a direction that will not necessarily lead to improved education. In fact, some social policy analysts assert that in situations where administrators are strongly convinced of the effectiveness of a program, it is likely that an evaluation will be disregarded. This is tragically illustrated in California where policy makers enamored with Whole Language were seemingly incapable of attending to data showing serious declines in students' reading performance, including a national assessment on which California students placed last. By ignoring outcome data, policy makers continue to make educational decisions that negatively impact children. And the most vulnerable learners are those who are most seriously harmed.
An additional problem is that policymakers frequently rely on information that others provide them. Thus their decisions are often based on incomplete and inaccurate data that reflect not what research has revealed, but the biases of program administrators, and supporters. An Office of Education document that was read at an appropriations meeting claimed that "when contrasting all Follow Through children with their non-Follow Through comparisons... there emerge large differences in achievement, motivation, and intense effects" (U. S. Congress, 1974, p. 2361), a statement leading senators to believe that the Follow Through program as a whole was successful and should be continued. John Evans, OE's Acting Deputy Commissioner for Planning, Budgeting, and Evaluation, explained to Congress that:
...Follow Through is made up of a different set of alternative ways of approaching alternative education, different models, different programs. And the task and central purpose of that program...is to find out which of those methods or approaches are more or less effective. The evaluation evidence we have compiled indicates just what we would expect from that kind of experiment: namely, that some of those models and approaches are very reassuringly effective, and the kinds of things we would want to see disseminated and used more broadly...other models are not successful and not effective and not the kinds of things we would want to carry on or continue to fund or support indefinitely (U. S. Congress, 1974, p. 2360).
This example illustrates how reports and interpretation of results may serve as a source of confusion when decision makers are faced with the task of determining the fate of a program.
It is acknowledged that policy makers are more likely to be influenced by social and political contingencies than by empirical data. However, others may be expected to pay more heed to the findings of major research programs in their field.
Colleges of Education. Project Follow Through was unique because it examined not only instructional programs, but the educational philosophies from which they were developed. While the Follow Through models varied greatly in specific differences, they may generally be considered to represent one of two general philosophies of education. The majority of models were based on philosophies of "natural growth" (Becker and Carnine, 1981) or what Bijou (1977) referred to as "unfolding." According to these models, learning involves changes in cognitive structures that are believed to develop and mature in the same manner as biological organs. Whole Language is an example of instruction derived from this philosophy. It is based on the belief that reading develops naturally given sufficient exposure to a print-rich environment.
The second philosophical position is concerned with principles of learning or "changing behavior" (Becker and Carnine, 1981). From this perspective, teaching involves specifying what is to be taught and arranging the environment in such a way that the desired change in behavior results.
Although the data from Follow Through support the latter position, the majority of colleges of education espouse a philosophy of cognitive restructuring. Thus, the data from Follow Through fail to support the philosophy that dominates colleges of education. This obviously made it difficult for educators to accept the Follow Through findings and they responded by discrediting the evaluation as well as by voicing specific objections about the Direct Instruction model or questioning the values of the model. For example, educators are fond of accusing direct teaching approaches of ignoring the "whole child" by emphasizing academic achievement at the expense of affective development. The Follow Through data clearly show that no such trade-off occurs. The Direct Instruction model was more effective than any other model on measures of self-esteem. A second objection is that this Direct instruction is reductionistic and results in only rote learning of non-essential skills. Yet, the data show that students in the Direct Instruction model demonstrated superior performance on measures of complex cognitive skills. In contrast, not a single model that set out to improve these cognitive skills was able to do so.
Although effective methods may be rejected simply because of their philosophical underpinnings, it is possible that they are rejected for more practical reasons as well. If teachers are to become competent in the use of effective teaching methods, teacher training programs must be restructured and those who are responsible for teacher training must themselves become proficient in those methods. Effective restructuring will require changes not only in what is taught, but in how it is taught as well. The training paradigm underlying most teacher training programs has little to recommend it, with students spending the majority of their time listening to lectures about theory and method. Sponsors of Follow Through models found that lectures about teaching had little impact upon actual teaching practices. Training was most successful when it included modeling of the desired behaviors, opportunities for teachers to practice, and feedback about their performance (Bushell, 1978). This has important implications not only for preservice training of teachers, but for how schools provide inservice training.
Teachers. Probably the biggest obstacle is the fact that the instructional methods a teacher uses are most likely to be those taught during his or her own training. Although it is assumed that teachers have acquired the skills necessary to teach their students, in reality teachers are woefully unprepared. For example, there are currently thousands of teachers in classrooms who do not know how to teach beginning reading, because the professors who "taught" them adhered to a philosophy of "natural growth." One teacher confided to me, "I do not know how to teach reading to someone who doesn't already know how to read"! If our teachers do not, by their own admission, know how to teach, how will our children learn?
Teachers may not seek out empirically validated methods, such as Direct Instruction, because they fail to recognize that their current methods are not effective. Student failure is more likely to be attributed to deficits within the child or to external factors such as the child's home life, than to ineffective instruction. Furthermore, many teachers are not even aware that methods exist that would enable them to be more effective. In many instances, the only information teachers have about Direct Instruction is misinformation. And, even if teacher did know there was a better way to teach, how would they acquire the necessary skills? Surely not by returning to the schools where they received their initial teacher training.
Teachers who are motivated to look for and use effective methods, often run into opposition. For example, if Direct Instruction materials have not been approved for purchase by curriculum committees, teachers will, in effect, be unable to purchase those materials. Even if appropriate materials can be obtained, teachers may be forbidden to use them. In addition, districts often refuse to provide funds for teachers to attend Direct Instruction conferences and training sessions, preferring to send them to receive information about the most current fads.
School Districts. The fact that effective teaching methods are available does not mean that they will be adopted. According to Alan Cohen (personal communication, 1992), "We know how to teach kids, what we don't know is how to get the public schools to do it!" Because there are no incentives for adopting effective methods or penalties for adopting ineffective ones, the choice of instructional programs will be made based on other factors. One factor that determines whether a particular method will be adopted is how greatly it differs from existing practices. The best candidates for adoption are those most similar to ongoing practices, because they are least disruptive. Stallings and Kaskowitz (1974) described the behavior of teachers in Direct Instruction Follow Through classrooms as "quite atypical of generally practiced classroom behavior" (p. 220). This decreases the probability of adoption because it requires so much change.
Financial incentives may also influence adoption decisions. While funding may provide the inducement to adopt an innovation, monitoring is needed to ensure its continued implementation. One way that Follow Through differed from other federally funded programs was that in exchange for funding, particular instructional practices were specified and monitored. This system of supervision resulted in a higher degree of fidelity of implementation of the model than might otherwise be expected. However, schools are generally not organized to provide the level of supervision that Follow Through model sponsors found necessary to ensure fidelity of implementation.
Publishers. Much, perhaps most, of what a teacher does is determined by the materials he or she uses. Yet, those who develop instructional materials typically do not have the skills required to develop effective materials. Few educational programs are based on state-of-the-art programming principles. Worse yet, materials are not field tested to ensure their effectiveness with children. The publishing industry does not initiate the development of instructional materials, but instead reacts to the demands of the educational marketplace. California provides a good illustration of this dependent relationship. In California the state adopts an instructional framework. Criteria for instructional materials are then derived from the framework. Publishers are provided these criteria and busily get to work developing instructional materials that conform to them. They submit their materials during the textbook adoption process and panels evaluate the extent to which the materials correspond to the specified criteria. Noticeably absent from these criteria is any mention of measured effectiveness. Given this process, a program could meet every single criterion and be recommended for adoption, and not be effective in teaching a single child! But, field tests are expensive, and the prevailing contingencies provide absolutely no incentive for publishers to conduct them in order to provide learner verification data because such data are not considered in textbook selection and adoption. (See "Why I sued California, Engelmann, ADI News, Winter, 1991).
The Public. Although the public is not typically considered part of the educational establishment, it can be included in this discussion because it supports education. What the public has supported is a system which has continued to neglect effective methods of instruction. Of course, the public's support has been innocent because it is generally unaware of instructional options and their differential effectiveness. Parents and others have been led to accept that the failure of a great many students to learn is due to deficits in the children. The general public has no way of knowing that children's achievements are largely a function of how they are taught. However, this may be changing.
Toward the Future
The American public's dissatisfaction with public education is
becoming increasingly clear. The failures of public education have
been well publicized. Endless studies and reports call attention to
important factors such as improving curricula, increasing teacher
salaries, expanding the length of the school day and/or year, and a
variety of other changes. Although some of these changes may be
necessary, they will not be sufficient to produce the substantial
academic improvement that is possible. The critical factor that has
been historically ignored is instructional method. Our educational
problems will not be solved until it is recognized that how well
students learn is directly related to how well they are taught.
Is there any evidence that research is beginning to influence educational policy and practice? Recent events in California may point to progress in that direction. The Report of the California Reading Task Force (1995) stresses effective teaching and recommends that every school and district implement a "reading program that is research based" (p. 3). In February of this year, Assembly Bill 3075 (1996) was introduced in the California State legislature. This bill would amend the minimum requirements for a teaching credential to include satisfactory completion of comprehensive reading instruction "that is research-based" and includes "the study of direct, systematic, explicit phonics." In September of 1995, Governor Wilson signed Assembly Bill 170, referred to as the ABC bill, requiring the State Board of Education to "ensure that the basic instructional materials that it adopts for mathematics and reading...are based on the fundamental skills required by these subjects, including, but not limited to systematic, explicit phonics, spelling, and basic computational skills." It is possible that these documents offer the promise of hope for the future. I will close with the words of Leonardo da Vinci: "Tell me if anything ever was done."
Becker, W. C. & Carnine, D. (1981). Direct instruction: A
behavior theory model for comprehensive educational intervention with
the disadvantaged. In S. W. Bijou & R. Ruiz, (Eds.), Behavior
modification: Contributions to education (pp. 145-210).
Hillsdale, NJ: Erlbaum.
Bijou, S. W. (1977). Practical implications of an interactional model of child development. Exceptional Children, 14, 6-14.
Bushell, D. (1978). An engineering approach to the elementary classroom: The Behavior Analysis Follow-Through project. In A. C. Catania, & T. A. Brigham (Eds.), Handbook of Applied Behavior Analysis ( pp. 525-563). New York: Irvington Publishers.
Carnine, D. W. (1983). Government discrimination against effective educational practices. Proceedings of the Subcommittee on Human Resources Hearing on Follow Through Amendments of 198, 99-103. Wash. D. C.: U. S. Government Printing Office.
Carnine, D. W. (1984). The federal commitment to excellence: Do as I say, not as I do. Educational Leadership, 4, 87-88.
Engelmann, S. (1991). Why I sued California. ADI News, 10, 4-8.
Engelmann, S. & Carnine, D. (1982). Theory of instruction: Principles and applications. New York: Irvington Publishers.
Every child a reader: The report of the California reading task force. (1995) Sacramento: California Department of Education.
Proper, E. C., & St. Pierre, R. G. (1980). A search for potential new Follow Through approaches: Executive summary. Cambridge, Mass.: Abt Associates. (ERIC Document Reproduction Services No. ED 187 809)
Stallings, J. A., & Kaskowitz, D. H. (1974). Follow Through classroom observation evaluation (1972-1973). Menlo Park, CA: Stanford Research Institute.
Stebbins, L. B., St. Pierre, R. G., Proper, E. C., Anderson, R. B., and Cerva, T. R. (1977). Education as experimentation: A planned variation model (Volume IV-A: An evaluation of Follow Through). Cambridge, Mass.: Abt Associates.
U. S. Congress, Senate. (1974, May 29). Hearings before a subcommittee of the Committee on Appropriations.