22.3 Macro-adaptive instructional models

22.3. 1 A Taxonomy of Macro-Adaptive Instruction

Como and Snow (1983) developed a taxonomy of adaptive instruction to provide systematic guidance in selecting instructional mediation (i.e., activities) depending on the objectives of adaptive instruction and student aptitudes. Como and Snow distinguished two different objectives of adaptive instruction: (a) aptitude development necessary for further instruction such as cognitive skills and strategies useful in later problem solving and effective decision making, and (b) circumvention or compensation for existing sources of inaptitude needed to proceed with instruction. They categorized aptitudes related to learning into three types: (a) intellectual abilities and prior achievement, (b) cognitive and learning styles, and (c) academic motivation and related personality characteristics. (For in-depth discussions about aptitudes in relation to adaptive instruction, see Federico, 1980; Cronbach & Snow, 1977; Snow, 1986; Snow & Swanson, 1992; Tobias, 1987.) Como and Snow categorized instructional mediation into four types, from the least intrusive form of mediation to the most intrusive one: (a) activating, which mostly calls forth students' capabilities and capitalizes on learner aptitudes as in discovery learning; (b) modeling; (c) participant modeling; and (d) short-circuiting, which requires step-by-step direct instruction. This taxonomy gives a general idea of how to adapt instructional mediation for the given instructional objective and student aptitude. According to Como and Snow (1983), this taxonomy can be applied to both levels of adaptive instruction (macro and micro). For example, the activating mediation may be more beneficial for more intellectually able and motivated students, while the short-circuiting mediation may be better for the intellectually low-end students. However, this level of guidance does not provide specific information about how to develop and implement an adaptive instruction. More specifically, it does not suggest how to perform ongoing learning diagnosis and instructional prescriptions during the instructional process.

22.3.2 Macro-Adaptive Instructional Models

While Como and Snow's taxonomy represents possible ranges o f adaptation of instructional activities for the given instructional objective and student aptitudes, Glaser's (1977) five models provide specific alternatives for the design of adaptive instruction.

Glaser's first model is an instructional environment that provides limited alternatives. In this model, the instructional objective and activity to achieve the objective are fixed. Thus, if the student does not have appropriate initial competence to achieve the objective with the given activity, he or she is designated as a poor learner and is dropped out. Only students who demonstrate the appropriate initial state of competence are allowed to participate in the instructional activity. If the student does not demonstrate the achievement of the objective after the activity, the student is allowed to repeat the same activity or is dropped out.

The second model provides an opportunity to develop the appropriate initial competence for students who do not have it. However, no alternative activities are available. Thus, students who do not achieve the objective after the activity should repeat the same activity or drop out.

. The third model accommodates different styles of learning. In this model, alternative instructional activities are available, and students are assessed whether they have the appropriate initial competence for achieving the objective through one of the alternatives. However, there are no remedial activities for the development of the appropriate initial competence. Thus, if the student does not have initial competence appropriate for any of the alternative activities, she or he is designated as a poor learner. Once an instructional activity is selected based on the student's initial competence, the student should repeat the activity until achieving the objective or drop out.

The fourth model provides an opportunity to develop the appropriate initial competence and accommodate different styles of learning. If the student does not have the appropriate initial competence to achieve the objective through any of the alternative instructional activities, a remedial instructional activity is provided to develop the initial competence. If the student has developed the competence, an appropriate instructional activity is selected based on the nature of the initial competence. The student should repeat the selected instructional activity until achieving the objective or drop out.

The last model allows students to achieve different types of instructional objectives or different levels of the same objective depending on their individual needs or ability. The basic process is the same as the fourth model, except that the student's achievement is considered successful if any of the alternative instructional objectives (e.g., different type or different level of the same objective) are achieved.

Glaser (1977) described six conditions necessary for instantiating adaptive instructional systems: (a) the human and mental resources of the school should be flexibly employed to assist in the adaptive process; (b) curricula should be designed to provide realistic sequencing and multiple options for learning; (c) open display and access to information and instructional materials should be provided; (d) testing and monitoring procedures should be designed to provide information to teachers and students for decision making; (e) emphasis should be placed on developing abilities in children that assist them in guiding their own teaming; and (f) the role of teachers and other school personnel should be the guidance of individual students.

Glaser's conditions suggest that the development and implementation of an adaptive instructional program in an existing system are complex and difficult. This might be the primary reason why most macro-adaptive instructional systems have not been used as successfully and widely as hoped. However, computer technology provides a powerful means to overcome at least some of the problems encountered in the planning and implementing of adaptive instructional systems.

22.3.3 Aptitude-treatment interaction models

To provide systematic guidelines in selecting instructional strategies for individually different students, Carrier and Jonassen (1988) proposed four different types of matches based on Salomon's (1972) work: (a) remedial for providing supplementary instruction to learners who are deficient in a particular aptitude or characteristic, (b) capitalization/preferential for providing instruction in a manner that is consistent with a learner's preferred mode of perceiving or reasoning, (c) compensatory for supplanting some processing requirements of the task for which the learner may have a deficiency, and (d) challenge for stimulating learners to use and develop new modes of processing.

22.3.4 Aptitude Variables and Instructional Implications

To find linkages between different aptitude variables and learning, numerous studies have been conducted (see Cronbach & Snow, 1977; Gagne, 1967; Gallangher, 1994; Snow, 1986; Snow & Swanson, 1992; Tobias, 1983, 1989, 1994). Since the detailed review of ATI research findings is beyond the scope of this chapter, a few representative aptitude variables showing relatively important implications for adaptive instruction are briefly presented below.

22.3.4.1. Intellectual Ability. General intellectual ability consisting of various types of cognitive abilities (e.g., crystallized intelligence such as verbal ability, fluid intelligence such as deductive and logical reasoning, and visual perception such as spatial relations) (see Snow, 1986) is suggested to have interaction effects with instructional supports. For example, more structured and less complex instruction (e.g., expository method) may be more beneficial for students with low intellectual ability, while less-structured and more complex instruction (e.g., discovery method) may be better for students with high intellectual ability (Snow & Lohman, 1984). More specifically, Como and Snow (1986) suggested that crystallized ability may relate to, and benefit in interaction with, familiar and similar instructional methods and content, whereas fluid ability may relate to and benefit from learning under conditions of new or unusual methods or content.

22.3.4.2. Cognitive Styles. Cognitive styles are characteristic modes of perceiving, remembering, thinking, problem solving, and decision making. They do not reflect competence (i.e., ability) per se but, rather, the utilization (i.e., style) of competence (Messick, 1994). Among many different dimensions of cognitive style (e.g., field dependence versus field independence, reflectivity versus impulsivity, haptic versus visual, leveling versus sharpening, cognitive complexity versus simplicity, constricted versus flexible control, scanning, breadth of categorization, tolerance of unrealistic experiences, etc.), field-dependent versus field-independent and impulsive versus reflective styles have been considered to be most useful in adapting instruction. The following are instructional implications of these two cognitive styles that have been considered in ATI studies.

Field-independent persons are more likely to be: (a) self-motivated and influenced by internal reinforcement, and (b) better at analyzing features and dimensions of information and for conceptually restructuring it. In contrast, field-dependent persons are more likely to be: (a) concerned with what others think and affected by external reinforcement, and (b) accepting of given information as it stands and more attracted to salient cues within a defined learning situation. These comparisons imply some ATI research. For example, studies showing significant interactions revealed that field-independent students achieved best with deductive instruction, and field-dependent students performed best in instruction based on examples (Davis, 1991; Messick, 1994).

Reflective persons are likely to: (a) take more time to examine problem situations and make fewer errors in their performance, (b) exhibit more anxiety over making mistakes on intellectual tasks, and (c) separate patterns into different features. In contrast, impulsive persons have a tendency to: (a) show greater concern about appearing incompetent due to slow responses and take less time examining problem situations, and (b) view the stimulus or information as a single, global unit.

As some of the instructional implications described above suggest, these two cognitive styles are not completely independent of each other (Vernon, 1973).

22.3.4.3. Learning Styles. Efforts for matching instructional presentation and materials with the student's preferences and needs have produced a number of different learning styles (Schmeck, 1988). For example, Pask (1976, 1988) identified two learning styles: (a) a holist, who prefers a global task approach, a wide range of attention, reliance on analogies and illustrations, and construction of an overall concept before filling in details, and (b) a serialist, who prefers a linear task approach focusing on operational details and sequential procedures. Students who are flexible employ both strategies and are called versatile learners (Messick, 1994). Marton (1988) distinguished between students who are conclusion oriented and take a deep-processing approach to learning and students who are description oriented and take a shallow-processing approach. French (1975) identified seven perception styles (print oriented, aural, oral-interactive, visual, tactile, motor, and olfactory) and five concept formation approaches (sequential, logical, intuitive, spontaneous, and open). Dunn and Dunn (1978) classified learning stimuli into four categories (environmental, emotional, sociological, and physical) and identified several different learning styles within each category. The student's preference in environmental stimuli can be quiet or noisy sound, bright or dim illumination, cool or warm temperature, and formal or informal design. For emotional stimuli, students may be motivated by self, peer, or adult (parent or teacher), and more or less persistent, and more or less responsible. For sociological stimuli, students may prefer learning alone, with peers, with adults, or through a variety of ways. Preferences in physical stimuli can be auditory, visual, or tactile/kinesthetic. Kolb (1971, 1977) identified four learning styles and a desirable learning experience for each style: (a) Feeling or enthusiastic students may benefit more from concrete experiences; (b) watching or imaginative students prefer reflective observations; (c) thinking or logical students are strong in abstract conceptualizations; and (d) doing or practical students like active experimentation. Hagberg and Leider (1978) also developed a model for identifying learning styles, which is similar to Kolb's.

Each of the learning styles reviewed above provides some practical implications for designing adaptive instruction. However, there is not yet sufficient empirical evidence to support the value of learning styles, and no reliable methods for assessing the different learning styles developed.

22.3.4.4. Prior Knowledge. Glaser and Nitko (1971) suggested that the behaviors that need to be measured in adaptive instruction are those that are predictive of immediate learning success with a particular instructional technique. Since prior achievement measures relate directly to the instructional task, they should therefore provide a more valid and reliable basis for determining adaptations than other aptitude variables.

The value of prior knowledge in predicting the student's achievement and needs of instructional supports has been demonstrated in many studies (e.g., Ross & Morrison, 1988). Research findings showed that the higher the level of prior achievement, the less the instructional support required to accomplish the given task (e.g., Abramson & Kagen, 1975; Salomon, 1974; Tobias 1973; Tobias & Federico, 1984; Tobias & Ingber, 1976). Furthermore, prior knowledge has a substantial linear relationship with interest in the subject (Tobias, 1994).

22.3.4.5. Anxiety. Many studies showed that students with high test anxiety performed poorly on tests in comparison to students with low test anxiety (see Sieber, O'Neil & Tobias, 1977; Tobias, 1987). Since research findings suggest that high anxiety interferes with the cognitive processes that control learning, procedures for reducing the anxiety level have been investigated. For example, Deutsch and Tobias (1980) found that highly anxious students who had options to review study materials (e.g., videotaped lessons) during learning showed a higher achievement than other highly anxious students who did not have the review option. Under an assumption that anxiety and study skills have complementary effects, Tobias (1987) proposed a research hypothesis in an ATI paradigm: "Test-anxious students with poor study skills would learn optimally from a program addressing both anxiety reduction and study skills training. On the other hand, test-anxious students with effective study skills would profit optimally from programs emphasizing anxiety reduction without the additional study skill training" (p. 223). However, more studies are needed to investigate specific procedures or methods for reducing anxiety before guidelines for adaptive instructional design can be made.

22.3.4.6. Achievement Motivation. Motivation is an associative network of affectively toned personality characteristics such as self-perceived competence, locus of control, anxiety, etc. (McClelland, 1965). Thus, understanding and incorporating the interactive roles of motivation with cognitive process variables during instruction is important. However, little research evidence is available for understanding the interactions between the affective and cognitive variables, particularly individual differences in the interactions.

Although motivation as the psychological determinant of learning achievement has been emphasized by many researchers, research evidence suggests that it has to be activated for each task (Weiner, 1990). According to Snow (1986), students achieve their optimal level of performance when they have an intermediate level of motivation to achieve success and to avoid failure. Nicholla, Jagacinski, and Miller (1986) suggested that intrinsically motivated students engage in the task more intensively and show better performance than extrinsically motivated students. However, some studies showed opposite results (e.g., Frase, Patrick & Schumer, 1970). The contradictory findings suggest possible interaction effects of different types of motivation with different students. For example, the intrinsic motivation may be more effective for students who are strongly goal oriented, like adult learners, while extrinsic motivation may be better for students who study because they have to, like many young children.

Entwistle's (1981) classification of student-motivation orientation provides more hints for adapting instruction to the student's motivation state. He identified three types of students based on motivation-orientation styles: (a) meaning-oriented students, who are internally motivated by academic interest; (b) reproducing-oriented students, who are extrinsically motivated by fear of failure; and (c) achieving-oriented students, who are primarily motivated by hope for success. The meaning-oriented students are more likely to adopt a holist learning strategy that requires deep cognitive processing, while the reproduction-oriented students tend to adopt a serialist strategy that requires relatively shallow cognitive processing (Schmeck, 1988). The achieving-oriented students are likely to adopt either type of learning strategy depending on the given learning content and situation.

However, the specific roles of motivation in learning have not been well understood, particularly in relation to the interactions with the student's other characteristics, task, and other learning conditions. Without understanding the interactions between motivation and other variables, including instructional strategies, simply adapting instruction to the student's motivation may not be useful.

Recently, Tobias (1994) examined student interest in a specific subject and its relations with prior knowledge and learning. Interest, however, is not clearly distinguishable from motivation because interest seems to originate or stimulate intrinsic motivation, and external motivators (e.g., reward) may stimulate interest.

22.3.4.7. Self-Efficacy. Self-efficacy influences people's intellectual and social behaviors, including academic achievement (Bandura, 1982). Since self-efficacy is a student's evaluation of his or her own ability to perform a given task, the student may maintain widely varying senses of self-efficacy, depending on the context (Gallagher, 1994). According to Schunk (1991), self-efficacy changes with experiences of success or failure in certain tasks. A study by Hoge, Smith, and Hanson (1990) showed that feedback from teachers and grades received in specific subjects were important factors for the student's academic self-efficacy. Although many positive aspects of high self-esteem have been discussed, few studies have been conducted to investigate the instructional effect of self-efficacy in the ATI paradigm. Zimmerman and Martinez-Pons (1990) suggested that students with high verbal and mathematical self-efficacy used more self-regulatory and metacognitive strategies in learning the subject. Although it is clear that self-regulatory and metacognitive learning strategies have a positive relationship with students' achievement, this study seems to suggest that the intellectual ability is a more primary factor than self-esteem in the selection of learning strategies. More research is needed to find factors contributing to the formation of self-esteem, relationships between self-efficacy and other motivational and cognitive variables influencing learning processes, and strategies for modifying self-efficacy. Before studying these questions, investigating specific instructional strategies for low and high self-efficacy students in an ATI paradigm may not be fruitful.

In addition to variables discussed above, many other individual difference variables (e.g., locus of control, cognitive development stages, cerebral activities and topological localization of brain hemisphere, personality variables, etc.) have been studied in relation to learning and instruction. Few studies, however, provided feasible suggestions for adapting instruction to individual differences in these variables.

22.3.5 A Taxonomy of Instructional Strategies

Although numerous teaming and instructional strategies have been studied (e.g., O'Neil, 1978; Weinstein, Goetz & Alexander, 1988), selecting a specific strategy for a given instructional situation is difficult because its effect may be different for different instructional contexts. It is particularly true for adaptive instruction. Thus, instructional strategies should be selected and designed with the consideration of many variables uniquely involved in a given context. To provide a general guideline for selecting instructional strategies, Jonassen (1988) proposed a taxonomy of instructional strategies corresponding to different processes of cognitive learning. After identifying four stages of the learning process (recall, integration, organization, and elaboration) and related learning strategies for each stage, he identified specific instructional activities for facilitating the learning process. Also, he identified different strategies for monitoring different types of cognitive operations (i.e., planning, attending, encoding, reviewing, and evaluating).

Park (1983) also proposed a taxonomy of instructional strategies (Table 22-1) for different instructional stages or activities (i.e., pre-instructional strategies, knowledge presentation strategies, interaction strategies, instructional control strategies, and post-instructional strategies). However, these taxonomies are identified from the author's subjective analysis of learning/instructional processes and do not provide direct or indirect suggestions for selecting instructional strategies in ATI research or adaptive instructional development.

22.3.6 Limitations of Aptitude Treatment Interactions

In the 3 decades since Cronbach (1957) made his proposal, relatively few studies have found consistent results to support the paradigm and made little contribution to either instructional theory or practice. As several reviews of ATI research (Berlinger & Cohen, 1983; Cronbach & Snow, 1977; Tobias, 1976) have pointed out, the measures of intellectual abilities and other aptitude variables were used in a large number of studies to investigate their interactions with a variety of instructional treatments. However, no convincing evidence was found to suggest that such individual differences were useful variables for differentiating alternative treatments for subjects in a homogeneous age group, although it was believed that the individual difference measures were correlated substantially with achievement in most school-related tasks (Glaser & Resnick, 1972; Tobias, 1987).

The unsatisfactory results of ATI research have prompted researchers to reexamine the paradigm and assess its effectiveness. A number of difficulties in the ATI approach are viewed by Tobias (1976, 1987, 1989) as a function of past reliance on what he terms the alternative abilities concept. Under this concept, it is assumed that instruction is divided into input, processing, and output variables. The instruction methods, which form the input of the model, are hypothesized to interact with different psychological abilities (processing variables), resulting in certain levels of performance (or outcomes) on criterion tests. According to Tobias, however, several serious limitations of the model often prevent the occurrence of the hypothesized relations. The limitations are:

1. The abilities assumed to be most effective for a particular treatment may not be exclusive; consequently, one ability may be used as effectively as another ability for instruction by a certain method (see Cronbach & Snow, 1977).
2. Abilities required by a treatment may shift as the task progresses so that the ability becomes more or less important for one unit (or lesson) than for another (see Bums, 1980; Federico, 1983).
3. ATIs validated for a particular task and subject area may not be generalizable to other areas. Research has suggested that ATIs may well be highly specific and vary for different kinds of content (see Peterson, 1977; Peterson & Janicki, 1979; Peterson, Janicki & Swing, 1981).
4. ATIs validated in laboratory experiments may not be applicable to actual classroom situations.

22.3.7 Achievement-Treatment Interactions

To reduce some of the difficulties in the All approach, Tobias (1976) proposed an alternative model, achievement-treatment interactions (see 33.9. 1). While the ATI approach stresses relatively permanent dispositions for learning as assessed by measures of aptitudes (e.g., intelligence, personality, and cognitive styles), achievement-treatment interactions represent a distinctly different orientation, emphasizing task-specific variables relating to prior achievement and subject-matter familiarity. This approach stresses the need to consider interactions between prior achievement and performance on the instructional task to be learned. Prior achievement can be assessed rather easily and conveniently through administration of pretests or through analysis of students' previous performance on related tasks. Thus, it eliminates many potential sources of measurement error, which has been a problem in ATI research, since the type of abilities to be assessed would be, for the most part, clear and unambiguous.

TABLE 22-1. A TAXONOMY OF INSTRUCTIONAL STRATEGIES (PARK, 1984; SEIDEL, PARK & PEREZ, 1988). (THE LISTING OF INSTRUCTIONAL STRATEGIES IN THIS TABLE IS NOT EXHAUSTIVE, AND THE CLASSIFICATIONS ARE ARBITRARILY MADE.)

1. Preinstructional Strategies 1. Instructional objective Terminal objectives and enabling objectives Cognitive objectives vs. behavioral objectives Performance criterion and condition specifications 2. Advance organizer Expository organizer vs. comparative organizer Verbal organizer vs. pictorial organizer 3. Overview Narrative overview Topic listing Orienting questions 4. Pretest Types of test (e.g., objective: true-false, multiple choice, matching, etc. vs. subjective: short answer, essay, etc.) Order of test item presentation (e.g., random, sequence, response-sensitive, etc.) Item replacement (e.g., with or without replacement of presented items) Timing (e.g., limited vs. unlimited) Reference (e.g., criterion-reference vs. norm-reference)   2. Knowledge Presentation Strategies 1. Types of knowledge presentation Generality (e.g., definition, rules, principles, etc.) Instance: diversity and complexity (e.g., example and nonexample problems) Generality help (e.g., analytical explanation of generality) Instance help (e.g., analytical explanation of instance) 2. Formats of knowledge presentation Enactive, concrete physical representation Iconic, pictorial/graphic representation Symbolic, abstract verbal, or notational representation 3. Forms of knowledge presentation Expository, statement form Interrogatory, question form 4. Techniques for knowledge acquisition Mnemonic Metaphors and analogies Attribute isolations (e.g., coloring, underlining, etc.) Verbal articulation Observation and emulation

3. Interaction Strategies 1. Questions Level of questions (e.g., understanding/idea vs. factual information) Time of questioning (e.g., before or after instruction) Response mode required (e.g., selective vs. constructive; overt vs. covert) 2. Hints and prompts Formal, thematic, algorithmic, etc. Scaffolding (e.g., gradual withdraw of instructor supports) Reminder and refreshment 3. Feedback Amount of information (e.g., knowledge of results, analytical explanation, algorithmic feedback, reflective comparison, etc.) Time of feedback (e.g., immediate vs. delayed feedback) Type of feedback (e.g., cognitive/informative feedback vs. psychological reinforcing) 4. Instructional Control Strategies 1. Sequence Linear Branching Response-sensitive Response-sensitive plus aptitude-matched 2. Control options Program control Learner control Learner control with advice Condition-dependent mixed control 5. Postinstructional Strategies 1. Summary Narrative review Topic-fisting Review questions 2. Postorganizer Conceptual mapping Synthesizing 3. Posttest Types of test (e.g., objective: true-false, multiple choice, matching, etc. vs. subjective: short answer, essay, etc.) Order of test item presentation (e.g., random, sequence, response-sensitive, etc.) Item replacement (e.g., with or without replacement of presented items) Timing (e.g., limited vs. unlimited) Reference (e.g., criterion-reference vs. norm-reference)

Many studies (e.g., see Tobias 1973, 1976; Tobias & Federico, 1984) confirmed the hypothesis that the lower the level of prior achievement is, the more the instructional support is required to accomplish the given task, and vice versa. However, a major problem in the ATI approach, that learner abilities and characteristics fluctuate during instruction, is still unsolved in the achievement-treatment interaction. The treatments investigated in the studies of this approach were not generated by systematic analysis of the kind of psychological processes called upon in particular instructional methods, and individual differences were not assessed in terms of these processes (Glaser, 1972). In addition to the inability to accommodate shifts in the psychological processes active during or required by a given task, the achievement-treatment interaction has another problem: In this model, some useful information might be lost by discounting possible contribution of factors such as intellectual ability, cognitive style, anxiety, motivation, etc.

22.3.8 Cognitive Processes and ATI Research

The limitation of aptitudes measured prior to instruction in predicting the student's learning needs suggests that the cognitive processes intrinsic to learning should be paramount considerations in adapting instructional techniques to individual differences. However, psychological testing developed to measure and classify people according to abilities and aptitudes has neglected to identify the internal processes that underlie such classifications (Federico, 1980).

According to Tobias (1982, 1987), learning involves two types of cognitive processes: (a) macroprocesses that are relatively molar processes, such as mental tactics (Derry & Murphy, 1986), and deployed under student's volitional control; and (b) microprocesses that are relatively molecular processes, such as the manipulation of information in short-term memory, and are less readily altered by students. Tobias (1989) assumed that unless the instructional methods examined in ATI research induce students with different aptitudes to use different types of macroprocesses, the expected interactions would not occur. To validate this assumption, Tobias (1987, 1988) conducted a series of experiments in rereading comprehension using CBI. In the experiments, students were given various options to employ different macroprocesses through the presentation of different 'instructional activities (e.g., adjunct questions, feedback, various review requirements, instructions to dunk of the adjunct question while reviewing, rereading with external support, etc.). In summarizing the findings from the experiments, Tobias (1989) concluded that varying instructional methods does not lead to the use of different macrocognitive processes or to changes in the frequency with which different processes are used. Also, the findings showed little evidence that voluntary use of macrocognitive processes are meaningfully related to student characteristics such as anxiety, domain-specific knowledge, or reading ability. Although some of these findings are not consistent with previous studies that showed a high correlation between prior knowledge and the outcome of learning, they explain the reasons for the inconsistent findings in ATI research.

Based on the results of the experiments and the review of relevant studies, Tobias (1989) suggested that

Researchers should not assume student use of cognitive processes, no matter how clearly these appear to be required or stimulated by the instructional method. Instead, some students should be trained or at least prompted to use the cognitive processes expected to be evoked by instructional methods, whereas such intervention should be omitted for others (p. 220).

This suggestion requires a new paradigm for ATI research that specifies not only student characteristics and alternative instructional methods for teaching students with different characteristics but also strategies for prompting the student to use the cognitive processes required in the instructional methods. This suggestion, however, would make ATI research more complex without being able to produce consistent findings. For example, if an experiment did not produce the expected interaction, it would be virtually impossible to find out whether the result came from the ineffectiveness of the instructional method or the failure of the prompting strategy to use the instructional method.

22.3.9 Learner Control

An alternative approach to adaptive instruction is learner control (see 33.2) that gives learners full or partial control over the process or style of instruction they receive (Snow, 1980). Individual students are different in their abilities for assessing the learning requirements of a given task, their own learning abilities, and instructional options available to learn the given task. Therefore, it can be considered within the ATI framework, although the decision-making authority required for the learning assessment and instructional prescription is changed to the student from the instructional agent (human teacher or media-based tutor).

Snow (1980) divided the degree of learner control into three levels depending on the imposed and elected educational goals and treatments: (a) complete independence, self-direction, and self-evaluation; (b) imposed tasks, but with learner control of sequence, scheduling, and pace of learning; and (c) fixed tasks, with learner control of pace. Numerous studies have been conducted to test the instructional effects of learner control and specific instructional strategies that can be effectively used in learner-control environments (see 33.2) The results have provided some important implications for developing adaptive systems: (a) Individual differences play an important role in the success of learner control strategy; (b) some learning activities performed during the instruction are closely related to the effectiveness of learner control; and (c) the learning activities and effects of learner control can be predicted from the .premeasured aptitude variables (Snow, 1980). For example, a study by Shin, Schallert, and Savenye (1994) showed that limited learner control and advisement during instruction were more effective for low-prior-knowledge students, while high-prior-knowledge students did equally well in both full or limited learner-control environments with or without advisement. These results suggest that learner control should be considered both a dimension along which instructional treatments differ and a dimension characteristic of individual differences among learners (Snow, 1980). However, research findings in learner control are not consistent (see 33.5.4), and many questions remain to be answered in terms of the learner-control activities and metacognitive processes. For example, more research is needed in terms of learner-control strategies related to assessment of knowledge about the domain content, ability to learn, selection and processing of learning strategies, etc.

22.3. 10 An Eight-Step Model for Designing ATI Courseware

As reviewed above, findings in ATI research suggest that it is premature or impossible to assign students with one set of characteristics to one instructional method and those with different characteristics to another (Tobias, 1987). However, faith in adaptive instruction using the ATI model is still alive because of the theoretical and practical implications of ATI research.

In spite of the inconclusive research evidence and many unresolved issues in the ATI approach, Carrier and Jonassen (1988) proposed an eight-step model to provide practical guidance for applying the ATI model to the design of computer-based instructional (CBI) courseware. The eight steps are: (1) Identify objectives for the courseware; (2) specify task characteristics; (3) identify an initial pool of learner characteristics; (4) select the most relevant learner characteristics; (5) analyze learners in the target population; (6) select final differences (in the learner characteristics); (7) determine how to adapt instruction; and (8) design alternative treatments. This model is basically a modified systems approach to instructional development (Gagne & Briggs, 1979; Dick & Carey, 1985). This model proposes to identify specific learner characteristics of the individual student for the given task, in addition to their general characteristics. For the use of this model, Carrier and Jonassen (1988) listed important individual variables that influence learning. They are (a) aptitude variables, including intelligence and academic achievement; (b) prior knowledge; (c) cognitive styles; and (d) personality variables, including intrinsic and extrinsic motivation, locus of control, anxiety, etc. (see p. 205 in Carrier & Jonassen, 1988). For instructional adaptation, they recommended several types of instructional matches: (a) remedial, (b) capitalization/preferential, (c) compensatory, and (d) challenge.

This model seemingly has practical value. Without theoretically coherent and empirically traceable matrices that link the different learner variables, the different types and levels of learning requirements in different tasks, and different instructional strategies, however, the mere application of this model may not produce results much different from that of nonadaptive instructional systems. ATI research findings suggest that varying instructional methods does not necessarily invoke different types or frequencies of cognitive processing required in learning the given task, nor are individual difference measures consistently related to such processing (Tobias, 1989). Furthermore, the application of Carrier and Jonassen's (1988) model in the development and implementation of courseware would be very difficult because of the amount of work required in identifying, measuring, and analyzing the appropriate learner characteristics and in developing alternative instructional strategies.