A Review of Models of the Teaching/Learning Process

The Teaching-Learning Process: A Discussion of Models
Deborah A. McIlrath and William G. Huitt

Citation: McIlrath, D., & Huitt, W. (1995, December). The teaching-learning process: A discussion of models. Educational Psychology Interactive. Valdosta State University. http://www.edpsycinteractive.org/papers/modeltch.html


Return to: |Readings in EdPsyc | EdPsyc Interactive: Courses |


Many researchers have tried to put together classroom- or school-based models that describe the teaching-learning process. A model is a visual aid or picture which highlights the main ideas and variables in a process or a system. The models presented in this paper include words or diagrams intended to give an understanding of the variables associated with school learning, especially as measured by scores on standardized tests of basic skills. The main models discussed and compared are by Carroll (1963), Proctor (1984), Cruickshank (1985), Gage and Berliner (1992) and Huitt (1995).

Two major questions are addressed in educational psychology: (1) "Why do some students learn required knowledge and skills taught in school, while others do not?" (a criterion-referenced evaluation question) and (2) "Why do some students learn more than other students?" (a norm-referenced evaluation question.) Unfortunately, the possible answers to these questions are enormous. Oftentimes research findings and theories of teaching and learning seem to contradict one another. What is an educator to do?

In this paper we will explore several models of teaching and learning. Gage & Berliner (1992) state that the use of models as learning aides have two primary benefits. First, models provide "accurate and useful representations of knowledge that is needed when solving problems in some particular domain" (p. 314). Second, a model makes the process of understanding a domain of knowledge easier because it is a visual expression of the topic. Gage and Berliner found that students who study models before a lecture may recall as much as 57% more on questions concerning conceptual information than students who receive instruction without the advantage of seeing and discussing models. Alesandrini (1981) came to similar conclusions when he studied different pictorial-verbal strategies for learning:

Research on the effectiveness of pictorial learning strategies indicates that learning is improved when pictures supplement verbal materials, when learners draw their own pictures while studying, and when learners are asked to generate mental pictures while reading or studying...the factor of sex was also included in the analysis due to its observed (although unexpected) effect (pp. 358, 363).

Interestingly, the females in this study had a tendency to benefit more than males if they related the specifics of their pictures to the whole concept.

Models have been used extensively in educational psychology to help clarify some of the answers researchers have found that might shed light on such questions as, "How do students learn effectively?" Or, "What is happening in this classroom that facilitates learning better than in another classroom?"

John Carroll's Model

Most current models that categorize the variables or explanations of the many influences on educational processes today stem from Carroll's (1963) seminal article defining the major variables related to school learning. Carroll specialized in language and learning, relating words and their meanings to the cognitive concepts and constructs which they create (Klausmeier & Goodwin, 1971). In his model, Carroll states that time is the most important variable to school learning. A simple equation for Carroll's model is:

School Learning = f(time spent/time needed).

Carroll explains that time spent is the result of opportunity and perseverance. Opportunity in Carroll's model is determined by the classroom teacher; the specific measure is called allotted or allocated time (i.e., time allocated for learning by classroom teachers.) Perseverance is the student's involvement with academic content during that allocated time. Carroll proposed that perseverance be measured as the percentage of the allocated time that students are actually involved in the learning process and was labeled engagement rate. Allocated time multiplied by engagement rate produced the variable Carroll proposed as a measure of time spent, which came to be called engaged time or time-on-task.

Carroll (1963) proposed that the time needed by students to learn academic content is contingent upon aptitude (the most often used measure is IQ), ability to understand the instruction presented (the extent to which they possessed prerequisite knowledge), and the quality of instruction students receive in the process of learning. Carroll proposed that these specific teacher and student behaviors and student characteristics where the only variables needed to predict school learning; he did not include the influences of family, community, society and the world that other authors discussed below have included.

The principles of this model can be seen in Bloom's (1976) Mastery Learning model. Bloom, a colleague of Carroll's, observed that in traditional schooling a student's aptitude for learning academic material (IQ) is one of the best predictor's of school achievement. His research demonstrated that if time is not held constant for all learners (as it is in traditional schooling) then a student's mastery of the prerequisite skills, rather than aptitude, is a better predictor of school learning. Mastery Learning's basic principle is that almost all students can earn A's if

1) students are given enough time to learn normal information taught in school, and
2) students are provided quality instruction.

By quality instruction Bloom meant that teachers should:

(1) organize subject matter into manageable learning units,
(2) develop specific learning objectives for each unit,
(3) develop appropriate formative and summative assessment measures, and
(4) plan and implement group teaching strategies, with sufficient time allocations, practice opportunities, and corrective reinstruction for all students to reach the desired level of mastery.

Proctor's Model

Prior to the sixties the research on important school- and classroom-related variables was directed toward the best traits or characteristics of teachers in an attempt to identify good teaching and the important characteristics of schools and communities that support good teaching. Proctor (1984) provides a model that updates this view by including important teacher and student behaviors as predictors of student achievement. It is derived from other teacher- and classroom-based models but is redesigned to emphasize teacher expectations. Proctor states that it is possible for a self-fulfilling prophesy (as researched by Rosenthal & Jacobson, 1968) to be an institutional phenomenon and the climate of a school can have an effect on the achievement of its learners. The attitudes, the norms, and the values of an educational faculty and staff can make a difference in achievement test scores. The paradigm most influencing Proctor's model is that of a social nature and not of a teacher/student one-on-one relationship. The other models include the variables that provide the focus for this model, but show these variables in a more subordinate manner.

Proctor's (1984) model begins with the factor of the School's Social Climate. Some of the variables included in this would be attitudes, norms, beliefs, and prejudices. This school climate is influenced by a number of factors, including such student characteristics as race, gender, economic level, and past academic performance.

The student characteristics also influence teacher attitudes and teacher efficacy. More recent studies support Proctor's (1984) position that student self-image and behavior are affected by teacher efficacy (e.g., Ashton, 1984; Woolfolk & Hoy, 1990).

The next category of variables is the interaction among the individuals involved in the schooling process. This includes the input of administrators as well as that of teachers and students. If expectations of learning are high (i.e., the school has good, qualified teachers and students who can learn) and there is high quality instructional input, corrective feedback, and good communication among students, parents, and educators, then the intermediate outcomes of student learning and student self-expectation goes up. On the other hand, adverse or negative attitudes on the part of instructors and administrators will cause student self-esteem, and consequently, student achievement to spiral downwards.

The interactions in Proctor's (1984) model include the school's overall policy on allowing time for children to learn or promoting other forms of student-based help when needed. This could include quality of instruction (as in Carroll's (1963) model above) or teacher classroom behaviors (as in Cruickshank's (1985) model below). These behaviors have an effect on student classroom performance (especially academic learning time and curriculum coverage) and self-expectations .

Finally, the student's achievement level in Proctor's (1984) model is an outcome of all previous factors and variables. It is hypothesized that there is a cyclical relationship among the variables. In Proctor's model, the main concept is that achievement in a specific classroom during a particular school year is not an end in itself. It is refiltered into the social climate of the school image and the entire process begins all over again. Proctor's model implies that change can be made at any point along the way. These changes will affect school achievement, which will continue to affect the social climate of the school.

Cruickshank's Model

The model by Cruickshank (1985) is more classroom- and teacher-based; he was heavily influenced by models created by Mitzel, Biddle, and Flanders. Mitzel contributed the concept of classifying variables as "product, process, or presage" (Cruickshank, p. 17). Product is learning on the part of the student (change in behavior or behavior potential) while process involves interaction between student and teacher. Presage is the teacher's intelligence, level of experience, success and other teacher characteristics. Presage is supposed to affect process and then, of course, process will affect the product.

Biddle (as cited in Biddle & Ellena, 1964) showed a relationship between specific learning activities and teacher effects. In his model, Biddle offers seven categories of variables related to schooling and student achievement: school and community contents, formative experiences, classroom situations, teacher properties, teacher behaviors, intermediate effects, and long-term consequences. This provides the foundation for Cruickshank's (1985) model.

Biddle also contributed a model of the transactional process of the classroom by analyzing the structure and function of the communication process. This is reflected in Cruickshank's model through the use of arrows depicting the interaction between teacher and pupil classroom behavior.

Biddle constructed his models to help answer questions he thought parents might ask, such as: "How often does my child get individual attention from the teacher?" Or, "Does the teacher really understand Junior's special problem?" (Adams & Biddle, 1970, p. 6). Biddle also helped define non-cognitive variables which contribute to the affective domain (i.e., self-concept and self-esteem of the students). An example of these variables would be teacher genuineness, "teacher-offered conditions of respect...and modification of low self-concept" (Good, Biddle & Brophy, 1975, p. 195).

Flanders (as cited in Cruickshank, 1985) offered the variables of teacher- and student-classroom-talk and devised an instrument which focused on this behavior. "His was the most frequently used instrument. It permitted observation of teachers' use of 'verbal influence,' defined as 'teacher talk' and 'pupil talk,' in a variety of classroom situations" (Cruickshank, p. 17). Cruickshank put them all together and added additional presage variables such as pupil characteristics, properties (abilities and attitudes) and school, community and classroom climate.

Gage and Berliner's Model

Gage and Berliner (1992) developed a model of the instructional process that focuses on those variables that must be considered by the classroom teacher as she designs and delivers instruction to students. This model attempts to define more precisely what is meant by "quality instruction" and presents five tasks associated with the instruction/learning process. The model is classroom- and teacher-based and centers around the question, "What does a teacher do?"

A teacher begins with objectives and ends with an evaluation. Instruction connects objectives and evaluations and is based on the teacher's knowledge of the students' characteristics and how best to motivate them. If the evaluations do not demonstrate that the desired results have been achieved, the teacher re-teaches the material and starts the process all over again. Classroom management is subsumed under the rubric of motivating students. Gage and Berliner suggest that the teacher should use research and principles from educational psychology to develop proper teaching procedures to obtain optimal results.

Huitt's Framework

The most recently developed framework to be discussed (Huitt, 1995) identifies the major categories of variables that have been related to school achievement. It is more a framework than a model because it does not identify specific variables and their relationships. The framework is not only school-, classroom-, teacher-, and student-based, but includes additional contextual influences as well. Huitt's framework attempts to categorize and organize all the variables that might be used to answer the question, "Why do some students learn more than other students?" This is a revision of a model by Squires, Huitt and Segars (1983) which focused only on those variables thought to be under the control of educators. This earlier model focused on school- and classroom-level processes that predicted school learning as measured on standardized tests of basic skills. One important addition in this framework is the redefinition of Academic Learning Time. It had long been recognized that Carroll's conceptualization of time spent measured the quantity of time engaged in academics, but was lacking in terms of the quality of that time. As discussed in Proctor's (1984) model, Fisher and his colleagues (1978) had added the concept of success as an important component of quality of time spent and coined the term Academic Learning Time (ALT) which they defined as "engaged in academic learning at a high success rate." Brady, Clinton, Sweeney, Peterson, & Poynor (1977) added another quality component--the extent to which content covered in the classroom overlaps to content tested--which they called content overlap. Squires et al. used the more inclusive definition of ALT proposed by Caldwell, Huitt & Graeber (1982)--"the amount of time students are successfully engaged on content that will be tested."

Huitt's (1995) framework considers variables related to context and student and teacher characteristics, some of which were the focus of the models by Proctor (1984) and Cruickshank (1985). It is an interactive framework along the lines of Biddle and Ellena (1964), Cruickshank, and Laosa (1982).

Huitt advocates that important context variables must be considered because our society is rapidly changing from an agricultural/industrial base to an information base. From this perspective, children are members of a multi-faceted society, which influences and modifies the way they process learning as well as defines the important knowledge and skills that must be acquired to be successful in that society. Huitt's framework shows a relationship among the categories of Context (family, home, school, and community environments), Input (what students and teachers bring to the classroom process), Classroom Processes (what is going on in the classroom),and Output (measures of learning done outside of the classroom). These categories appear superimposed in the framework since it is proposed they are essentially intertwined in the learning process.

This framework shows Input and Output as the beginning and end of the teaching/learning process. Huitt (1995) believes that educators must first identify or propose an end result (as stated by Gage & Berliner, 1992) because how you identify and measure the end product (Output) will influence the selection of important predictor variables (e.g., What You Measure Is What You Get, Hummel & Huitt, 1994). Until the outcome objectives are known, nothing else can be considered. Once outcome measures are selected, educators can begin to focus on those variables that can explain fluctuation or variability in those measures. Considering or changing specific goals or objectives may change the predictor variables from each of the other three categories. Thus, the Output or Outcome category is the most important and the focus of Huitt's framework.

In the United States, the most often cited Output measures are scores on standardized tests of basic skills such as reading, language arts, and mathematics as well as science and social studies. Since the United States is ranked 14th out of 15 countries in mathematics knowledge and 13th in science (Office of Policy and Planning, 1992), we need to take a very close look at how we can improve achievement on these measures. For example, the federal government focused on the task of increasing the Output measurements of students when it adopted Goals 2000 (Swanson, 1991).

However, student achievement in basic skills is not the only desired outcome of American education. The Secretary of Labor presented additional requirements in the report by the Secretary's Commission on Achieving Necessary Skills (SCANS; Whetzel, 1992). The SCANS report focuses on the skills necessary for students to find work in the information economy. It addresses two categories of skills: foundations (basic skills, thinking skills and personal qualities) that provide the platform on which the other skills will be built and competencies (handling resources, interpersonal skills, informational skills, system skills, and technology utilization skills) that more closely describe what workers will actually be doing. [Note: Huitt (1997) provides a critique of the SCAN report that addresses important outcomes that were omitted.]

The most direct impact on important measures of school learning are those variables related to Classroom Processes. This category includes two major subcategories (Teacher Behavior and Student Behavior), and an Other (or miscellaneous) subcategory that includes such variables as classroom climate and student leadership roles..

The category of Teacher Behavior includes the subcategories of planning (getting ready for classroom interaction), management (getting the class under control), and instruction (guiding the learning process). In general, planning activities have little predictable relationship to student achievement (Gage & Berliner, 1992). Both management and instructional variables are moderately related to achievement, but the lack of a strong relationship may be due to be a factor of teacher inconsistency (Rosenshine & Stevens, 1986). That is, teachers often change their management and instructional practices based on the time of day or the characteristics of a particular group of students. Three single variables, teachers providing corrective feedback (e.g., give an explanation of what is correct or incorrect and why), teachers' use of reinforcement, and level of student-teacher interaction (a variable developed from the work of Flanders, as cited in Cruickshank, 1985) seem to be the best single classroom predictors of student success (Rosenshine & Stevens). Direct or explicit instruction (Rosenshine, 1995) appears to be the best model of instruction when scores on standardized tests of basic skills is used as the outcome measure.

Huitt supports Proctor's (1984) position that intermediate outcomes, or more specifically Academic Learning Time (ALT) is one of the best Classroom Process predictors of student achievement. As stated above ALT is defined as "the amount of time students are successfully involved in the learning of content that will be tested." There are three components to ALT and each is as important as the other. The first is Content Overlap, defined as "the extent to which the content objectives covered on the standardized test overlaps with the content objectives covered in the classroom." This variable has also been labeled as "time-on-target." The idea is simple: if an objective or topic is not taught, it is not likely to be learned, and therefore we cannot expect students to do well on measures of that content. In fact, to the extent the content is not specifically taught, the test becomes an intelligence test rather than an achievement test. The fact that many educators do not connect instructional objectives to specific objectives that will be tested (Brady et al., 1977), is one reason that academic aptitude or IQ is such a good predictor of scores on standardized tests. Both tests measure the same construct: the amount of general knowledge an individual has obtained that is not necessarily taught in a structured learning setting.

The second component of ALT is Student Involvement, defined the same way that Carroll defined engaged time or time-on-task (allocated time X engagement rate). If the students are not provided enough time to learn material or are not actively involved while teachers are teaching they are not as likely to do well on measures of school achievement at the end of the year.

The last element is that of Success, defined as "the percentage of classwork that students complete with a high degree of accuracy." If a student is not successful throughout the year on classroom academic tasks, that student will likely not demonstrate success on the achievement measure at the end of the year.

Huitt proposes that these three components of Academic Learning Time should be considered as the "vital signs" of a classroom. Just as a physician looks at data regarding temperature, weight, and blood pressure before asking any further questions or gathering any other data, supervisors need to look at the content overlap, involvement, and success before collecting any other data or making suggestions about classroom modifications. Classrooms where students are involved and making adequate progress on important content are reasonably healthy and quite different from those classrooms where students are not.

In addition to the teacher's classroom behavior, other time components such as the number of days available for going to school (the school year), the number of days the student actually attends school (attendance year), and the number of hours the student has available to go to school each day (school day) can influence ALT (Caldwell et al., 1982). None of these additional time variables were included in Carroll's (1963) model.

What teachers and students do in the classroom will depend to some extent on the characteristics or qualities they bring to the teaching/learning process. In Huitt's (1995) framework these are labeled Input variables. The subcategory of Teacher Characteristics includes such variables as values and beliefs; knowledge of students and the teaching/learning process; thinking, communication and performance skills; and personality. While each of these is important to the classroom environment, teacher efficacy is one of the best predictors of student success from this subcategory (Proctor, 1984; Ashton, 1984). If a teacher believes that, in general, students can learn the knowledge or skills, and that, specifically, he can teach them, then that teacher is more likely to use the knowledge and skills he has and the students are more likely to learn.

A second subcategory of Input is Student Characteristics. This includes all of the descriptions of students that might have an influence on the teaching/learning process and student outcome. Study Habits; Learning Style; Age; Sex/Gender; Race/Ethnicity; Motivation; and Moral, Socioemotional, Cognitive, and Character Development all become important in the relationship of classroom processes/behavior and school achievement (Huitt, 1995). However, student aptitude and/or prerequisite skills are probably the best student characteristic predictors (Bloom, 1976). If time is held constant, then intelligence or ability to learn academic content will be a better predictor than prior knowledge because the amount of content learned in the classroom is allowed to vary. That is, if everyone has the same amount of time in which to learn, then the speed at which one learns (aptitude) will be the best predictor of achievement. However, if we vary the time students have to learn and keep the content to be learned constant (such as in Mastery Learning), then prior knowledge is more salient. Though we do not initially modify the student characteristics that each student brings to the classroom, as Proctor (1984) pointed out the teacher can arrange the teaching/learning process and modify each student's experience. This results in different Outcomes, which in turn becomes the Input for the next learning cycle.

Finally, Huitt (1995) includes the category of Context that includes such subcategories as School Processes and Characteristics, Family, Community, State and Federal Government, TV/Movies, and the Global Environment. For example, research shows that student achievement is impacted by class size (e.g., Bracey, 1995) and school size (e.g., Fowler, 1995; Howley, 1996). While all of the variables in these subcategories are important and influence variables in the other three major categories, probably the two most important are Family and the Global Environment. Mother's education and family expectations for student achievement have been shown to be excellent predictors of student achievement (e.g., Campbell, 1991; Voelkl, 1993; Zill, 1992) as well as the amount of technology in the home (Perelman, 1992). Perhaps even more significant is the movement from the industrial age to the information age (Perelman; Toffler & Toffler, 1995). This is because it is redefining the outcomes that ought to be the focus of schooling and is providing new technologies that can radically alter the teaching/learning process.

An simple example of how some of these variables might interact is shown in the following model. The size and region of the community combine with family characteristics and processes to impact teacher and student characteristics. School and state policies combine with teacher and student characteristics to impact teacher behavior, while student characteristics and teacher behavior influence student behavior. Student classroom behavior then influences teacher classroom behavior in an interactive pattern that eventually results in student achievement as measured by instruments influenced by state policies. Student achievement at the end of one school year then becomes a student characteristic at the beginning of the next.

Summary and Conclusions

Each of the above models and frameworks identifies important factors related to school learning and contributes important information as we attempt to answer the question "Why do some students learn desired knowledge and skills?" Over a period of years, these conceptualizations have been examined, reviewed, revised and edited to fit into today's modern society. Beginning with Carroll (1963) and ending (at least as far as this review is concerned) with Huitt (1995), we see teachers and school systems, families, communities and entire countries having an influence on students' school learning. None of the variables appears to be so influential that we need only pay attention to that particular factor in order to produce the kinds of educational changes society desires. For example, an individual teacher could project self-fulfilling prophesies on a student (as seen in Cruickshank's 1985 model), but so also could the institution itself (as seen in Proctor's 1984 model). Or the school may be successful in developing students' basic skills, but students could still be unsuccessful in life because other important outcomes were not developed (Whetzel, 1992).

Understanding all the variables and the relationships among each other and to student success may be more than we can expect of any educator. We may never fully grasp the significance of the entire process, but we can make every effort to understand as much as possible as we develop the teaching/learning processes appropriate for the information age. We can also identify the most important variables within a category or subcategory and make certain we attend to a wide variety of variables across the model.

Models are useful tools to better understand not only the learning processes of students, but ourselves as educators. At a glance the models might provide only more questions, but a careful study of the models can provide starting points to begin developing more appropriate educational experiences for our society's next generation.

References


Return to: