18.4 EXAMPLES OF CONDITIONS-BASED MODELS

Robert Gagné provided the intellectual leadership for a conditions-based theory of instruction. A number of scholars followed in his tradition by developing more detailed prescription of the external conditions that will support different types of learning.

The two texts edited by Reigeluth in the 1980s, Instructional Design Theories and Models (1983) and Instructional Theories in Action (1987), clearly delineate a number of models that we would describe as conditions-based models of design. (Some of the models in this text, such as those by Scandura, Collins & Keller, we would not describe as full conditions-based models, as they do not describe the cognitive and instructional conditions for more than one learning type.) It is not the purpose of this chapter to replicate the thorough discussions of the conditions-based models presented in Reigeluth (1983, 1987). However, we will briefly discuss and compare the models, because it is through the comparisons that many of the major issues regarding conditions-based models are revealed and exemplified. We will also briefly review research and evaluation studies that have examined the effectiveness of the model as a whole or individual features of the model. We have also included in our discussion some "models" not presented in Reigeluth's texts. Some examples provided are arguably not instructional design models at all (such as the work of Hom, 1976; Resnick, 1967; West, Farmer & Wolf, 1991), but all employ, reflect, or extend the conditions-based theory propositions listed in the introduction of this chapter in one important way or another.

18.4.1 Gagné and Gagné, Briggs, and Wager

We have thoroughly described Gagné's conditions-based theory of instruction elsewhere in this chapter. This theory was the basis of an instructional design model presented in Briggs's Instructional Design: Principles and Applications (1977) and Principles of Instructional Design (Gagné & Briggs, 1974, 1979; Gagné, Briggs & Wager, 1988, 1992).

Research examining the validity of Gagné's theory are of two types: those that have examined the validity of Gagné's instructional theory as a cluster of treatment variables and those that have examined the individual propositions of the theory as separate variables. Research of the latter type will be discussed later in this chapter. A few studies have attempted to evaluate the overall value of instruction based on Gagné's theory or portions of Gagné's theory that are not central to the conditions-based theory. We will describe several examples of studies of this first type. Goldberg (1987), Marshall (1987), Mengel (1986), and Stahl (1979) compared "traditional" textbook or teacher-led instruction to print-based or teacher-led instruction designed according to Gagné's principles. These studies were across age groups and subject matters. Mengal and Stahl found significant differences in learning effects for the versions developed according to Gagné's principles, and for Goldberg and Marshall no significant difference in treatments. Although we believe such gross comparison studies to be essential to the development of research in an area, they suffer from some of the same threats to- validity of conclusions as other comparison studies. In particular, it is unclear that the "traditional" versions did not include some features of Gagné's principles and that the "Gagnétian" versions were fully consistent with these principles. Research that has examined the principles from Gagnd's instructional design models that are directly related to propositions of his theory will be discussed in a later section of this chapter.

18.4.2 Merrill: Component Display Theory

Merrill's (1983) Component Display Theory (CDT), an extension of GagnCs theory, is a conditions-based theory of instructional design, as he prescribed instructional conditions based on types of learning outcomes desired.

18.4.2.1. TYpes of Learned Capabilities. Merrill classified leaming objectives (or capabilities) along two dimensions: performance level (remember, use, or find) and content type (facts, concepts, principles, or procedures). So, there are conceivably 12 distinct categories of objectives that his model addresses. Instead of having a declarative knowledge category, as Gagné does, which would include remembering facts, concept definitions, rule statements, and procedural steps, Merrill makes separate categories for each of these types of declarative knowledge. Similarly, instead of having a single cognitive strategies category as Gagné does, through his intersection of the two dimensions, Merrill proposes "find" operations for each of the content types: Find a fact, find a concept, find a rule, and find a procedure.

Merrill provided a rationale for his categorization scheme based on "some assumptions about the nature of subject matter" (p. 298). The rationale for content type is based on five operations that he proposes can be conducted on subject matter: identity' (facts), inclusion and intersection (concepts), order (procedures), and causal operations (principles). He derived his performance levels from assumptions regarding differences in four memory structures: associative, episodic, image, and algorithmic. His performance levels derive from the associative (remember: verbatim and paraphrased) and algorithmic (use and find) memory structures. Merrill does not explicitly address the internal processes that accompany the acquisition of each of these categories of leaming types.

18.4.2.2. External Conditions of Learning. Merrill described instructional conditions as "presentation forms" and classified these forms as primary and secondary. Primary presentation forms have two dimensions: content (generality or instance) and approach (expository or inquisitory). Secondary presentation forms are types of elaborations that may extend the primary presentations: context, prerequisite, mnemonic, mathemagenic help, representation or alternative representation, and feedback. Merrill's model then further describes for each category of capability" a unique combination of primary and secondary presentation forms that will most effectively promote acquisition of that type of objective" (p. 283).

18.4.2.3. Research on Component Display Theory. Researchers have examined component display theory in two ways: evaluation in comparison to "traditional" approaches and examination of individual strategy variations within component display theory. We briefly describe examples of both types of research.

In research across a range of content, age groups, and learning tasks, researchers have examined the effectiveness of instruction following design principles proposed by component display theory to existing or "traditional" instruction. For example, Keller (1982) compared more conventional mathematics instruction in both expository and discovery formats to instruction following a "modified discovery" approach suggested by CDT. Keller found no significant effects on acquisition of set theory concepts, concluding that is was important to learning that the generality be presented explicitly, but less important whether this generality was presented prior to or following presentation of examples. In contrast, Stein (1982) found a superiority of CDT for concept learning among eighth-grade learners, comparing four treatments: expository prose, expository prose plus adjunct questions, CDT with only primary presentation forms, CDT with both primary and secondary presentation forms. She found that both CDT versions were significantly more effective in promoting students' ability to recognize previously presented instances of these concepts and to generalize the concept to previously unencountered instances. In addition, she found that this effect was more pronounced for the more difficult concepts. In a similar prose study, Robinson (1984) found a CDT version of a lesson on text editing to be significantly superior (on recall of the procedure, marginally on use of the procedure [p = A I]) to two other versions of prose instruction: one version with summarizing examples, one version with inserted questions. Van Hurst (1984) found a similar positive effect of materials revised using CDT principles when compared with the existing instructional materials in Japanese language learning. The CDT version was found to promote significantly greater achievement and more positive affect and confidence than the original version.

Researchers have also examined individual variables in component display theory. For example, Keller (1985) examined the relative benefits of generality alone, best example alone, or both generality and best example on learning graphing concepts and procedures. She found that the combined treatment was superior for remembering the steps in the procedure. None of the treatments was superior for using the procedure (only practice seemed to be critical). Further, the combined condition was superior for promoting finding a new procedural generality. Chao (1983) also examined the benefits of two expository versions of CDT (generality, example, practice, generality/generality, example, practice) and two discovery treatments (examples, practice/examples, practice, generality) on application and transfer of concepts and principles of plate tectonics. Unlike Chao and similar to her earlier comparisons of expository and discovery sequences (1982), Keller found no statistically significant difference in the participants' performance on application or transfer measures. Although order of generality, example, and practice may not be found to affect performance, Sasayama (1984) found that for a procedure-using learning task, a rule-example-practice treatment had superior effects on learning than a rule-only, example-only, or rule-example treatment.

Many of the weaknesses of Merrill's model are similar to those of Gagné's, such as lack of an explicit and empirically validated tie between internal processes and external events. However, Merrill's model conjectures even less on internal processes. It is also less complete, as his model addresses only the cognitive domain, does not fully delineate the instructional conditions for the "find" (cognitive strategies) category, and does not have a category for complex learning reflected in what is often called "problem solving." A strength of CDT may be its evolution to fit with the demands of designing intelligent CAI systems, as noted by Wilson (1987).

18.4.3 Reigeluth: Elaboration Theory

Reigeluth and his associates (Reigeluth, Merrill & Wilson, 1978; Reigeluth & Darwazeh, 1982; Reigeluth & Rogers, 1980; Reigeluth & Stein, 1983) developed the elaboration theory as a guide for developing macrostrategies for large segments of instruction, such as courses and units. The elaboration theory is conditions based in nature as it describes (a) "three models of instruction and (b) a system for prescribing those models on the basis of the goals for a whole course of instruction" (p. 340). His theory specifies a general model of selecting, sequencing, synthesizing, and summarizing content in a simple to more complex structure. The major features of the general model are an epitome at the beginning of the instruction, levels of elaboration of this epitome, learning-prerequisite sequence within level of elaboration, a leamer-control format, and use of analogies, summarizers, and synthesizers.

The conditions-based nature of the model is obtained from Reigeluth's specification of three differing structures--conceptual, procedural, or theoretical-which are selected based on the goals of the course. Reigeluth further suggested that conceptual structures. are of three types: parts, kinds, or matrices (combinations of two or more conceptual structures). He described two different kinds of procedural structures: procedural order and procedural decision. Finally, he subdivided theoretical structures into two types: those that describe natural phenomena (descriptive structures) and those that affect a desired outcome (prescriptive structures).

The nature of the epitome, sequence, summarizers, perquisites, synthesizers, content of elaborations will vary depending on the type of knowledge structure chosen, which was based on the goals of the course. For example, if the knowledge structure is conceptual, the epitome will contain a presentation of the most fundamental concepts for the entire course. If the structure is procedural, the epitome should present the most fundamental or "shortest path" procedure. Reigeluth recommended use of Merrill's component display theory as the guideline for designing at the micro or lesson level within each elaboration cycle.

In recent years, Reigeluth (1992) has placed more emphasis on the importance of using a simplifying conditions method (SCM) of sequencing instruction than on the sequencing and structuring of instruction based on one of the major knowledge structures. The simplifying conditions method suggests that designers "work with experts to identify a simple case that is as representative as possible of the task as a whole" (p. 8 1). This task should serve as the epitome of the course with succeeding levels of elaborating "relaxing" the simplifying conditions so that the task becomes more and more complex. The theory still retains some of its conditions-based orientation, though, as Reigeluth has suggested, different simplifying conditions structures need to be developed for each of the kinds of knowledge structures he described (Reigeluth & Rogers, 1980; Reigeluth, 1987).

18.4.3.1. Research on Elaboration Theory. As with the previous models, some research has evaluated the effectiveness of instruction based on the principles of elaboration theory in comparison to instruction designed based on other models. Examples of this type of research are Beukhof (1986), who found that instructional text designed following elaboration theory prescriptions were more effective than "traditional" text for learners or learners with low prior knowledge. In contrast, Wagner (1994) compared instruction on handling hazardous materials designed using the elaboration theory to materials designed using structural learning theory (Scandura, 1983). She found that although it took longer for learners to reach criterion performance with the structured learning materials, they performed significantly better on the delayed posttest than learners in the elaboration theory group. Wedman and Smith (1989) compared text designed according to Gagné's prescriptions and following a strictly hierarchical sequence to text designed according to the elaboration theory. They found no significant differences in either immediate or delayed principle application (photography principles). Nor did they find any interactions with a learner characteristic, field independence/dependence. In another study using the same materials, Smith and Wedman (1988) found some subtle differences between the readthink-aloud protocols of participants from the same population who were interacting with the two versions of the materials. They found that participants interacting with the elaborated version (a) required less time per page than the hierarchical version, (b) made more references to their own prior knowledge, (c) made fewer summarizing statements, (d) used mnemonics less often, and (e) made about the same types of markings and nonverbal actions as participants interacting with the hierarchical version. They concluded that although instruction designed following the two different approaches may evoke subtle processing differences, these differences are not translated into differences in immediate and delayed principle application, at least within the 2 hours of instruction that this study encompassed. As Reigeluth proposes that the elaboration theory is a macrostrategies theory, effective for the design of units and courses, and recommends component display theory as a niicrodesign strategy for lessons, it is perhaps not surprising that researchers have not uniformly found positive effects of the elaboration theory designs with their shorter instruction.

Researchers have also examined design questions regarding individual variables within elaboration theory, such as synthesizers, summarizers, nonexamples in learning procedures, and sequencing. Table 18-5 below summarizes the findings of several of these studies.

18.4.3.2. Evaluation of Elaboration Theory. Elaboration theory is a macrostrategy design theory that was much needed in the field of instructional design. Throughout the evolution of elaboration theory, Reigeluth has proposed design principles that maintained a conditions-based orientation. Due to the strong emphasis on learning hierarchy analysis, until Reigeluth's work many designers had assumed that instruction should proceed from one enabling objective to another from the beginning to the end of a course. Reigeluth suggested a theoretically sound alternative for designing large segments of instruction. It is unfortunate that researchers in the field have not found it pragmatically possible to evaluate the theory in comparison to alternatives with course-level instruction.

In light of advances in cognitive theory, Wilson and Cole (1992) suggested a number of recommendations for revising elaboration theory. These suggestions include (a) deproceduralizing the theory; (b) removing unnecessary design constraints (including the use of primary structures, which form the basis of much of the conditions-based aspect of elaboration theory); (c) basing organization and sequencing decisions on what is known by the learners as well as content structure; and (d) assurning a more "constructivist stance" toward content structure and sequencing (p. 76). Reigeluth (1992) responded to these recommendations in an admirable way: Regarding the deproceduralization of the elaboration theory, he pointed out that he agreed that the theory itself should not be proceduralized, but that he has always included in his discussions of ET ways to operationalize it. Reigeluth proposes that he has already removed "unnecessary design constraints" (the second Wilson & Cole recommendation) by replacing the "content structure" approach by the simplifying conditions method. 'fbis-approach may more nearly reflect Reigeluth's original intentions for the elaboration theory. However, it does not eliminate the underlying conditions-based principle (which we interpret Wilson & Cole to be recommending), as the method for identifying simplified conditions seems to vary according to whether the instructional goal is conceptual, theoretical, or procedural. Reigeluth concurs with Wilson and Cole's recommendation to take the learners' existing knowledge into account in the elaboration theory, although beyond some revision in the sequencing of conceptual layers (from the middle out, rather than top down), he does not propose that this will be formalized in his theory. Regarding the reconimendation that he assume a more "constructivist stance," Reigeluth concurs that this may be important in ill-structured domains, which the elaboration theory does not currently address. However, he insightfully suggests, "People individually construct their own meanings, but the purpose of instruction-and indeed of language and communication itself-is to help people to arrive at shared meanings" (p. 81).

18.4.4 Landa

In terms of learning outcome types, Landa's (1983) algoheuristic theory of instruction, or "Landainatics," makes a distinction between knowledge and skills (ability to apply knowledge): categories that seem to be equivalent to declarative and procedural knowledge. According to Landa, learners acquire knowledge about objects and operations. Objects are known as a perceptive image, i.e., as a mental image or as a concept. A concept can be expressed as a proposition, but it is not necessary that a concept be expressed in order to be known. There are other kinds of propositions such as definitions, axioms, postulates, theorems, laws, and rules which can form, a part of knowledge. Operations (action on an object) are transformations of either real material objects or their mental representations (images, concepts, propositions). A skill is the ability to perform operations. Operations that transform material objects are motor operations. Operations that transform materials objects are cognitive operations. Operations can be algorithmic, "a series of relatively elementary operations that are performed in some regular and uniforin way under defined conditions to solve all problems of a certain class" (p. 175), or heuristic, operations for which a series of steps can be identified but are not so singular, regularized, and predictable as algorithms. Algorithmic operations appear similar to Merrill's conception of procedures, and the heuristic operations appear similar to Smith and Ragan's treatment of procedural rules and Gagné's problem solving (higherorder rule). A critical aspect of Landa's model is the importance that he ascribes to the verification of hypothetical description of algorithmic or heuristic process through observation, computer simulation, or error analysis. Such empirical validation is present in specifics of design models in task analysis but is generally missing in conditions-based models with regard to a generalized hypothetical cognitive task analysis for each class of outcomes that can be directly related to prescriptions for external conditions of learning.

Landa's theory suggests how to support processes that .turn knowledge into skills and abilities, a transition that provides much of the substance of Anderson's (1985) ACT* theory. He suggests the following conditions for teaching individual operations:

1. Check to make sure that the learners understand the meaning of the procedure.
2. Present problem that requires application of the procedure.
3. Have the student name the operation or preview what should be done to execute the operation.
4. Present next problem.
5. Practice until mastery.

Although he suggests a procedure for teaching students to discover procedures (algorithms), he points out that this process is difficult and time consuming.

18.4.4.1. Research and Evaluation. Research on Landa's model is not as readily available in the literature as the previously reviewed models. However, Landa has reported some evaluation of his model in comparison with more "conventional" training. Landa (1993) estimated that he has saved Allstate $35 million in costs due to (a) many times fewer errors (up to 40 times fewer), (b) tasks performed up to 2 times faster, and (c) workers' confidence level several times higher.

18.4.5 Smith-Ragan

Rather than developing a new conditions-based model, Smith and Ragan (1993) sought to exemplify and elaborate Gagné's theory. To address what they perceived to be limitations in most conditions-based models, they postulated a generalized cognitive process necessary for the acquisition of each of the different learning capabilities. With regard to the external conditions of learning, Smith and Ragan suggested that events of instruction as Gagné portrayed them insufficiently considered learner-generated and learner-initiated learning. Smith and Ragan restated the events so that they could be perceived as either learner-supplied, in the form of learning strategies, or instruction-supported, in the form of instructional strategies.

Smith and Ragan also proposed a model for determining the balance between instructional strategies (instruction-supplied events) and learning strategies (learner-supplied events) based on context, learner, and task variables. They also proposed that there is a "middle ground" between instruction supplied, supplantive (also know as mathemagenic) events (see 30.6), and learner-initiated events (see 33.2), in which the instruction facilitates or prompts the learner to provide the cognitive processing necessary to an instructional event.

18.4.5.1. Research and Evaluation. Smith (1992) cited theoretical and empirical bases for some of the learner-taskcontext-strategy relationships proposed in the COGSS model, which forms the basis of the balance between instruction-supplied and learner-generated events. In this presentation she proposed an agenda for validation of the model.

18.4.6 Tennyson and Rasch

Tennyson and Rasch (1988) described a model of how instructional prescriptions might be tied to cognitive learning theory. This work was preceded by a short paper by Tennyson (1987) which contained the key elements of the model. In this paper, part of a symposium on Clark's "media as mere vehicles" assertions, Tennyson discussed how one might "trace" the links between different treatments that media might supply and different learning processes. He described six learning processes (three storage processes: declarative knowledge, procedural knowledge, and conceptual knowledge; and three retrieval processes: differentiating, integrating, and creating) which he paired with types of learning objectives, types of knowledge bases, instructional variables, instructional strategies, and computer-based enhancements.

Tennyson and Rasch (1988a, b) and Tennyson (1990) suggested that kinds of learning should refer to types of "memory systems." As with the previous conditions-based models, Tennyson and Rasch employed an informationprocessing model as their foundation and suggested the main types of knowledge to be: (a) declarative, which is stored as associative networks or schemata and relates to verbal information objectives; (b) procedural, which relates to intellectual skills objectives; and (c) contextual, which relates to problem-solving objectives and knowing when and why to employ intellectual skills. Five forms of objectives are described as requiring distinct cognitive activity (verbal information, intellectual skills, conditions information, thinking strategies, and creativity). In discussing the relationships among the types of knowledge, Tennyson and Rasch noted that contextual knowledge is based on "standards, values, and situational appropriateness.... Whereas both declarative and procedural knowledge form the amount of information in a knowledge base, contextual knowledge forms its organization and accessibility" (Tennyson & Rasch, 1988a, p. 372).

In terms of instructional conditions, for declarative knowledge they recommended expository strategies, such as worked examples, which provide information in statement form on both the context and structure of information and question/problem repetition, which presents selected information repeatedly until the student answers or solves all items at some predetermined level of proficiency. For procedural knowledge, they recommended practice strategies in which learners apply knowledge to unencountered situations and some monitoring in terms of evaluation of learner responses and advisement. To teach contextual knowledge, they suggested problem-oriented simulation techniques. And for complex problem situations, they recommended a simulation in which the consequences of decisions update the situational conditions and proceed to make the next iteration more complex.

An interesting element is a prescription of learning time for the different types of learning: 10% for verbal information, 20% for intellectual skills, 25% for conditional information, 30% for thinking strategies, and 15% for creativity. One intent of this distribution was to reflect Goodlad's (1984) prescription of a reversal of traditional classroom practice from 70% of instructional time devoted to declarative and procedural knowledge and only 30% to conceptual knowledge and cognitive abilities. Although such general proportions may serve to illuminate general curriculum issues, specification of percentages of time to types of learning, regardless of consideration of other factors in a particular learning situation, may find limited applicability to instructional design.

18.4.6.1. Research and Evaluation. Tennyson and Rasch's model has not yet been subjected to evaluation and research. In terms of the extension of conditions-based models, some issues do emerge. Although other theorists propose this conditional knowledge, it is unclear whether the addition of a contextual type of learning will enhance the validity of the model. It is possible that such knowledge is stored as declarative knowledge that is in some way associated with procedural knowledge, such as in a mental model or problem schema. The suggestion of time that should be allocated to each type of learning is intriguing, as it attempts to point out the necessity of emphasis on higher-order learning. However, the basis for determination of the proportion of time that should be spent on each type of outcome remains unclear.

18.4.7 Merrill, Li, and Jones-ID2

In reaction to a number of limitations that they perceived in existing instructional design theories and models (including Merrill's own), Merrill, Li, and Jones (1990a, 1990b) have set about to construct a "second-generation theory of instructional design." One of the specific goals of its developers is to expedite design of an automated ID system, "ID Expert," and thereby expedite the instructional design process itself Ultimately, the developers hope that the system will possess both authoring and delivery environments that grow from a knowledge and rule base. Of all the models described in this chapter, ID2 is the most ambitious in its goal to thoroughly prescribe the instructional conditions for each type of learning. The ID2 model is being developed to (a) analyze, represent, and guide instruction to teach integrated sets of knowledge and skill; (b) produce pedagogic prescriptions about selection and sequence; and (c) be an open system that can respond to new theory. This model has retained its conditions-based orientation. Indeed, Merrill and his associates (p. 8) have elaborated on the relationships between outcomes and internal/external conditions:

(a) A given learned performance results from a given organized and elaborated cognitive structure, which we will call a mental model. Different learning outcomes require different types of mental models; (b) the construction of a mental model by a learner is facilitated by instruction that explicitly organizes and elaborates the knowledge being taught, during the instruction; (c) there are different organizations and elaborations of knowledge required to promote different learning outcomes.

Within ID2, outcomes of instruction are considered to be enterprises composed of entities, activities, or processes, which might loosely be interpreted as concepts, procedures, and principles, respectively. Merrill and his associates have spent a vast amount of effort describing the structure of knowledge relating to these types of knowledge and how these types of knowledge relate to each other.

Merrill and associates have described a number of conditions (external conditions or instructional methods) that can be placed either under either system or learner control. These conditions are described as "transactions" of various classes. Evidence of Merrill's component display theory can be found in the prescriptions for these transactions. To create this system based on his ID2, Merrill and his colleagues (1991, 1992a, b, c) have attempted to identify the decisions that designers must make regarding the types of information to build into the system and the methods by which this information can be made available to learners. Ibis analysis is incredibly detailed in and of itself. For example, Table 18-6 summarizes the "responsibilities" of the transactions that may be made available in instruction, the "methods" that make up these responsibilities, and the range (or parameters) of these methods. Merrill et a]. have made similar analyses of information that may be made available to learners when learning entities, activities, or processes. In addition to detailing the options of pedagogy and information that can be made available in instruction, the developers of a system may also establish the "rules" by which system choices may be made as to which of these options to present to learners.

18.4.7.1. Research and Evaluation. Parts of the system been evaluated by Spector and Muraida (1991) and by Canfield and Spector (1991). For example, one of the major evaluation questions has been: "Can the target audience of novice designers use the system, and can this system expedite instructional design activities?" In Spector and Muraida's study, investigating the utility of the system to expedite design, eight subjects participated in 30 hours of instruction in which they learned to use the system and, developed 1 hour of instruction. The results indicated that all subjects who remained in the study were able to complete a computerbased lesson using the support of a portion of the system.

As yet there are no comparison data with more conventional design processes. In their effort to carefully explicate necessary knowledge for learning and instruction, as well as the means by which these interact with each other, the developers have created a model that is quite complex. One benefit of the model is that its complexity reflects and makes concrete much of the complexity of the instructional design process. Unfortunately, it seems that terminology has shifted during development. ID2 is not without its critics. Among criticisms frequently leveled are its utility when used by novices, the lack of evidence of theory base, issues regarding sufficient agreement to generate strategies, and the likelihood of sameness of results in multiple applications.

18.4.8 Other Applications of Conditions-Based Theory

Although they have not sought to develop complete instructional design models, a number of notable scholars within and outside the instructional design field have utilized a conditions-based theory as a basis for much of their work. We will briefly describe four of these examples; they illustrate how pervasive and influential the conditions-based theory has been.

18.4.8.1. Jonassen, Grabinger, and Harris. Jonassen, Grabinger, and Harris (1991) developed a decision model for selecting strategies and tactics of instruction based on three levels of decisions: (a) scope (macro/micro); (b) instructional event (prepare learner, present inform-ation, clarify ideas, provide practice, and assess knowledge); and (c) learning outcome. Levels (b) and (c) are similar to the decisions patterns suggested by Gagné and Gagné and Briggs.

Jonassen, Grabinger, and Harris recommended making decisions regarding instructional tactics based on three major categories of learning outcomes: intellectual skills (concept or rule), verbal information, or cognitive strategy (iconic, verbal/digital). They suggested prescriptions for instructional events based on the learning outcome-for example, for the event of preparing the learning by supporting recall of prior knowledge of intellectual skills through presenting a verbal/oral comparative advance organizer, adapting content of instruction to learners' prior knowledge, and reviewing prerequisite skills and knowledge. This work has been elaborated into a more complete outcomes-based model for formatively evaluating instruction (Jonassen & Tessmer, 1996).

18.4.8.2. Horn. Horn's approach to text design has many elements of a design model and clearly employs a conditions-based set of assumptions. Horn's work, called structured writing, presents a highly prescriptive approach to the design of instructional and informative text. In addition to format concerns, Horn proposed different treatments for different types of learning. The types of learning he identified are procedures (which explain how to do something and in what order to do it); structure (about physical things, objects that have identifiable boundaries); classification (which shows how a set of concepts is organized); process (which explains how a process or operation works, how changes take place in time); concepts (which define and give examples and nonexamples of new aspects of the subject matter), and facts (which give results of observations or measurements without supporting evidence) (Horn, 1976, p. 17). Horn described differential conditions for text presentation by identifying what elements (or "blocks") each presentation relating to a particular type of learning (or "map") must have. Horn differentiated between necessary and optional elements for each type of learning.

18.4.8.3. West, Farmer, and Wolt. West, Fanner, and Wolf (1991) referred to three kinds of knowledge: (a) declarative, which is stored in propositional networks that may be ' semantic or episodic and may structured as data or state schemata; (b) procedural knowledge that is order-specific and time-dependent (p. 16), and (c) conditional knowledge, which is knowing when and why to use a procedure (similar to Tennyson & Rasch's "contextual knowledge"). They describe "cognitive strategies" that can support the acquisition of each of these learning types, which the instructional designer plans instruction to activate. In contrast to Gagné, who typically portrays cognitive strategies as instructional strategies, supplied by instruction, and in contrast to Smith and Ragan (1993), who portray the primary load of information processing as something that should shift between learner and instruction depending on circumstances, West, Fanner, and Wolf imply that strategies are always provided by the learner. These cognitive strategies are chunking, frames (graphic organizers, concept mapping, advance organizer, metaphor, rehearsal, imagery, and mnemonics). In terms of prescriptive or conditions-based models, West, Farmer, and Wolf prescribe the strategies as effective to support acquisition of all types of knowledge. However, they also use Gagné's five domains as types of outcomes for prescribing the appropriateness of each strategy, which is somewhat confusing, as procedural knowledge and intellectual skills, which are usually considered to refer to the same capabilities, are not given the same prescriptions for strategies. Our evaluation is that their prescriptions are for the declarative portion of higher-order knowledge types.

18.4.8.4. E.Gagné. Unlike most instructional models, E. Gagné's work (1993) is primarily descriptive, rather than prescriptive. E. Gagné based her conditions-based propositions on J. Anderson's (1990) cognitive theories of learning, and placed her theory base within the information-processing theories. She subscribed to Anderson's types of knowledge: declarative and procedural. Gagné described the representations of declarative knowledge as propositions, images, linear orderings, and schemas (which can be composed of propositions, images~ and linear orderings). Procedural knowledge is represented as a production system, which can result in domain-specific skills, doinainspecific strategies, and, to a limited degree, domain-general strategies.

Although the majority of her text is more descriptive than prescriptive, E. Gagné utilized the conditions-based theory as she discussed the internal processes required in the acquisition of each of the types of knowledge and the instructional support that can promote this acquisition. She described instructional support as increasing the probability that required processes will occur, or making learning easier or faster. A strength of E. Gagnés formulation is her description of internal processes. In addition, she provides empirical evidence of effectiveness of instructional support conditions.