Instructional Design in Practice: ADDIE's Disputed Origins, Bloom's Taxonomy Then and Now, and the Discipline of Writing a Testable Learning Objective

A training needs assessment, the subject of the previous lesson in this module, produces a list of gaps between what a role requires and what the people in it currently demonstrate. Instructional design is the discipline that turns that list into something a learner can actually study, practice, and be assessed against. The two are not the same activity, and treating them as interchangeable is a common practical mistake: a well-documented gap analysis can still produce weak training if nobody does the separate, deliberate work of deciding what the learner should be able to do afterward, in what sequence, through what materials, and how anyone would know it worked.

The acronym most training literature reaches for at this point is ADDIE, Analysis, Design, Development, Implementation, Evaluation, presented as a five-phase sequence a designer moves through. Before using that acronym the way most corporate training content does, it is worth being precise about what it actually refers to, because the precision changes how much authority a designer should lend it.

Analysis is the phase that absorbs the previous lesson's output directly: the documented performance gap, who the learners are, what constraints (time, budget, delivery channel, regulatory requirement) bound the solution, and whether training is even the right response at all, since some performance gaps trace back to unclear expectations, broken incentives, or missing tools rather than missing skill. Design translates that analysis into a blueprint: learning objectives, content sequence, assessment strategy, and delivery format, before any actual course material gets built. Development is where the blueprint becomes real material, slides, scenarios, job aids, e-learning modules, facilitator guides. Implementation is the rollout itself, delivering the training to real learners, which Lesson 44 covers in depth. Evaluation asks whether the design actually worked, at a level ranging from whether learners reacted well to whether the organization's performance gap actually closed, which is also where Lesson 44's treatment of the Kirkpatrick model picks up.

As a description of what has to happen somewhere in any serious instructional design effort, this is a reasonable, serviceable breakdown. The confusion begins with what training materials often imply next: that this five-phase sequence is itself a single, named, historically traceable methodology, invented by a specific person or team, validated as a system, and taught today largely unchanged from its original form. That claim does not hold up well against the documented record.

The documented institutional starting point is real and traceable, even if the acronym itself is not. In 1975, the Center for Educational Technology at Florida State University, working under contract with the United States Army, produced a five-volume report titled Interservice Procedures for Instructional Systems Development, usually abbreviated IPISD, authored by Robert Branson together with Gail Rayner, John Cox, James Furman, Floyd King, and Wallace Hannum, and published through the Army Training and Doctrine Command at Fort Monroe, Virginia. It was built for use across the Army, Navy, Air Force, and Marine Corps, and its influence was considerable: use of some version of instructional systems development, often simply called ISD, was eventually mandated not only within the US armed services but among the private contractors that trained alongside them.

Branson's own account of the project, published in 1978 in the journal Educational Technology, includes a graphic overview of the IPISD structure with five top-level headings. According to a direct reading of that account, those headings are Analyze, Design, Develop, Implement, and Control, not Evaluate. The report and its authors refer to the model only as IPISD throughout; nobody involved in writing or using it called it ADDIE. An earlier, related handbook by Leslie Briggs, also affiliated with Florida State University, published in 1970, contains similar ideas without using anything resembling the ADDIE heading set either. The acronym that training professionals now use constantly has no clean paper trail connecting it to this founding document, despite how often the two are presented as the same thing.

Michael Molenda, a professor of instructional technology at Indiana University, set out to resolve this directly after being asked to write an encyclopedia entry on "The ADDIE Model" and realizing he could not locate an original source to cite. His account of that search, published in 2003 in Performance Improvement, is unusually candid for an academic article, and it is worth following in some detail because it is the strongest documented evidence for just how imprecise the common framing of ADDIE actually is.

Molenda checked the major dictionaries and encyclopedias of instructional technology and education: no mention of ADDIE in any of them. He checked the standard histories of the field, both the long ones and the shorter overviews: nothing. He checked the leading instructional design textbooks of the period, across both older and newer editions: ADDIE appeared in none of them. He checked the published surveys that compare different instructional design models to one another: still nothing. By this point, as he writes, he had begun forming the theory that ADDIE existed more as a label than as an actual, specified model, so he contacted roughly twenty professors and practitioners whose institutional memory reached back to the 1960s. None of them could name an original source. The typical answer, when asked when they first heard the term, was "around the late 1980s." Several people he suspected might be the originator told him, in his own words, that they had assumed he was the one who invented it.

What he did find were sources that use the underlying five-step structure, analyze, design, develop, implement, evaluate, without ever calling it ADDIE. An American Society for Training and Development monograph from 1988 and a 1987 training-needs-assessment text both show diagrams with these five steps but label them simply "the ISD model." One of the earliest places anyone explicitly names it ADDIE in print is a 1996 book chapter Molenda co-authored himself, where he and his co-authors state plainly that they adopted the term because it was the one they kept hearing in informal conversation, not because they were citing or reviving a specific prior source. His own conclusion, stated directly: the label "seems not to have a single author, but rather to have evolved informally through oral tradition," and "there is no original, fully elaborated model, just an umbrella term that refers to a family of models that share a common underlying structure."

None of this means the five activities ADDIE names are fictional or useless; analyzing before designing, and designing before building, is sound practice under almost any label. What the history undermines is a specific, common rhetorical move: citing "the ADDIE model" as though it were a single peer-reviewed, empirically validated system, the way a vendor proposal or a training certification course often does, to lend borrowed authority to whatever five-step process that particular vendor happens to be selling. Molenda's research shows the opposite: anyone invoking ADDIE today is, in his words, "essentially creating and disseminating their own model," because no authoritative original exists to be faithfully applied or departed from. That is a genuinely different claim from "ADDIE is a proven methodology," and the difference matters directly for a training manager being asked to adopt one vendor's specific ADDIE-branded toolkit over another's on the strength of the acronym alone.

A separate, more practical critique concerns how ADDIE tends to get applied rather than how it originated. Many training organizations run the five phases as a strict waterfall: finish the full analysis, sign it off, finish the full design document, sign that off, then build the entire course before anyone outside the design team sees a working version of it. Applied this literally, the approach has a specific, well-documented failure pattern: stakeholders do not see a tangible product until development is substantially complete, by which point a wrong assumption made back in the design phase is expensive to fix, since fixing it may mean redoing work several phases downstream. The two phases meant to connect training most directly to business results, Analysis and Evaluation, are also, in practice, the two most commonly shortened or skipped altogether under time pressure, which is a specific, identifiable failure mode rather than a vague complaint about rigor.

Michael Allen, an instructional technology practitioner, made this critique the explicit basis of an alternative. In Leaving ADDIE for SAM, published with Richard Sites by ASTD Press in 2012, Allen proposed the Successive Approximation Model as a direct answer to the waterfall problem rather than a rejection of analysis or design as activities. SAM front-loads a short, intensive kickoff, often called a Savvy Start, lasting a day or two, where designers, subject-matter experts, and stakeholders sketch rough prototypes together instead of writing a design document alone. From there, the model runs two deliberately repeating loops, Iterative Design and Iterative Development, cycling through design, prototype, and review rather than finishing each stage once and moving on, so a rough but functioning version of the training exists within days rather than appearing, fully formed, only near the end of the project.

It would overstate the case to treat SAM as simply superior. Allen's argument is specifically against running ADDIE's five activities as a rigid, one-pass sequence, not against doing careful upfront analysis at all; a compliance-driven, highly regulated training program, where every objective has to map to a specific legal requirement and be defensible later, still benefits from the kind of documented, front-loaded analysis ADDIE's discipline provides, even if the later phases are run iteratively rather than as a single pass. SAM's own iterative loops still require someone to decide, early, what the training is actually trying to achieve; without that, an iterative process risks producing a polished but aimless prototype just as easily as a linear one risks producing an overdesigned document nobody tests until it is too late to change cheaply. The real design decision a training manager faces is not "ADDIE or SAM" as competing brands, but how much of the process to run as a sequence of signed-off stages versus a sequence of quick, testable loops, given the specific project's stakes, regulatory exposure, and tolerance for building something rough first.

The second framework this lesson owns has a cleaner origin story, though it is also commonly cited with a precision problem of its own. Benjamin Bloom, then Associate Director of the Board of Examinations at the University of Chicago, initiated the project that became Bloom's Taxonomy to solve a specific administrative problem: faculty across different universities kept writing similar comprehensive examination items from scratch every year, and Bloom hoped a shared classification of educational goals would let them pool and exchange test items that measured the same underlying objective. He recruited a group of measurement specialists from across the country who met roughly twice a year starting in 1949 to develop and revise the framework. Their final draft was published in 1956 as Taxonomy of Educational Objectives: The Classification of Educational Goals, Handbook I: Cognitive Domain, credited to Bloom, Max Engelhart, Edward Furst, Walker Hill, and David Krathwohl.

The cognitive domain they defined contained six major categories, ordered from simple to complex and from concrete to abstract: Knowledge, Comprehension, Application, Analysis, Synthesis, and Evaluation. The framework assumed what is usually called a cumulative hierarchy: mastery of a simpler category was treated as a prerequisite for mastery of the next one up. Reception was slow at first, partly because the word "taxonomy" was itself an unfamiliar term in education, but the framework eventually spread widely and was translated into twenty-two languages. One of its most common uses, classifying the objectives and test items in an actual curriculum, repeatedly turned up the same pattern: a heavy concentration of objectives at the Knowledge level, simple recall, even though objectives from Comprehension through Synthesis were usually considered the more important goals of education. That finding, reported directly by David Krathwohl himself decades later, is a genuinely useful piece of evidence for this lesson's own argument: a taxonomy is only as useful as the discipline applied to actually writing objectives across its full range, not just its lowest level, a point this lesson returns to when it reaches Mager's work.

Anderson, Krathwohl, and their collaborators published A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives in 2001. Krathwohl's own account of the revision, published the following year, explains the central structural problem the team set out to fix. A learning objective is ordinarily a sentence built from a noun, the subject-matter content, and a verb, the cognitive process to be applied to it, as in "remember the law of supply and demand." In the 1956 original, the Knowledge category alone carried both roles at once: its many subcategories specified the subject-matter content, while the definition of Knowledge itself implicitly specified the verb, to recall or recognize. That made Knowledge structurally different from the other five categories, which described only a cognitive process, not subject matter.

The 2001 revision resolved this by formally separating the noun and the verb into two independent dimensions. The Knowledge dimension, the noun axis, was expanded from the original's three implicit subcategories into four explicit ones: Factual Knowledge, Conceptual Knowledge, Procedural Knowledge, and a genuinely new category, Metacognitive Knowledge, knowledge about cognition in general and awareness of one's own thinking, a distinction the 1956 team did not have the cognitive-psychology vocabulary to name. The Cognitive Process dimension, the verb axis, kept six major categories but changed their form and order: Knowledge's implicit verb became Remember; Comprehension was renamed Understand, specifically because teachers already used that word constantly and its absence had been a frequent criticism of the original, even though the original committee had considered and rejected it, worried that "understand" meant different things to different teachers; Application, Analysis, and Evaluation were kept but converted to verb form as Apply, Analyze, and Evaluate; and Synthesis was renamed Create and swapped to the top position, above Evaluate, on the judgment that generating something new is more cognitively demanding than judging something that already exists.

This is not a historical curiosity confined to an academic footnote. A designer who picks a verb off a generic "Bloom's Taxonomy verb list" without knowing which taxonomy it comes from risks a specific, avoidable error: using "understand" itself as the objective's action verb. In the 2001 revision, Understand is a category label, not one of the specific, observable cognitive processes nested inside it, interpreting, exemplifying, classifying, summarizing, inferring, comparing, explaining. "Understand the bank's risk criteria" is not a testable objective in the sense the next section defines, because nobody can directly observe understanding; "explain the bank's risk criteria in terms a new loan applicant would follow" is testable, because explaining is something a learner can actually be asked to do and a trainer can actually watch or grade. Knowing that the taxonomy's category names are not themselves meant to be objective verbs, only the more specific processes nested under them are, is exactly the kind of precision the revision was built to support, and exactly the kind of precision a generic verb list stripped of its source tends to lose.

Robert Mager's Preparing Objectives for Programmed Instruction, published in 1962, is the foundational text most of today's objective-writing guidance still traces back to, whether or not the guidance says so. Mager's argument was specific and, for its time, genuinely disruptive: an instructional objective is only useful if it describes something observable enough that two different people, looking at the same learner's performance, would agree on whether the objective had been met. A goal stated as "appreciate good customer service" or "understand loan risk" fails this test immediately, not because those are unimportant outcomes, but because no two observers can reliably agree on what counts as evidence that appreciation or understanding has occurred.

Mager's original formulation specified three components a complete objective should contain. Performance describes exactly what the learner should be able to do, stated as an observable action rather than an internal state. Condition describes the circumstances under which that performance will occur, what tools, information, or constraints will or will not be available. Criterion describes how well the performance has to be done for it to count as acceptable, a standard, a tolerance, a time limit, or a comparison to an existing benchmark. An objective missing any of the three is, in Mager's terms, incomplete, even if it sounds reasonably clear on the surface. The book's influence went well beyond instructional design circles; California's 1972 Stull Act, which required teachers statewide to write measurable behavioral objectives as part of performance evaluation, is one concrete sign of how far Mager's framework traveled into actual policy, not just training-department practice.

Much current training material presents a four-part "ABCD" format, Audience, Behavior, Condition, Degree, and attributes it to Mager directly. That attribution is imprecise in a specific, checkable way. Mager's own 1962 formulation had three components, Performance, Condition, and Criterion, not four, and did not include Audience as a separate labeled element at all, on the reasonable assumption that an objective is always written for a specific, already-identified group of learners. The four-letter ABCD version is more accurately traced to a later instructional media textbook by Robert Heinich, Michael Molenda, James Russell, and Sharon Smaldino, who took Mager's existing Behavior, Condition, and Criterion (relabeled Degree) and explicitly added Audience as a fourth element, reasoning that stating who the objective applies to, even when it seems obvious, keeps a shared objective from drifting as it gets reused across different courses or audiences.

The distinction is not merely pedantic. A designer who believes the full four-letter structure was Mager's own original idea may assume, incorrectly, that naming an audience is itself the rigorous part of the exercise; Mager's actual contribution, and the harder discipline to apply consistently, is the insistence on an observable Behavior and a specific Criterion, the two elements most often left vague in practice even when an Audience and a Condition are stated clearly. Correctly crediting the ABCD mnemonic as a later elaboration, built on top of Mager's narrower but more demanding original formula, keeps the attention where Mager actually wanted it.

A training manager who has correctly diagnosed where ADDIE's label actually comes from, chosen a reasonable balance between upfront documentation and early prototyping, and written objectives with a real Behavior, Condition, and Criterion has done the design-stage work this lesson covers. None of it guarantees the training will transfer back to the job or move the organizational metric the original needs assessment was built to fix; that is a separate, empirical question, and it is the specific subject Lesson 44 takes up through the Kirkpatrick model and the research on learning transfer. The honest, unresolved tension this lesson leaves standing is a genuine one: a defensible, well-documented instructional design process and training that actually changes on-the-job performance are related but not guaranteed to move together, and an organization that only ever audits the design documents, never the downstream performance, can end up confident in a process it has never actually verified works.

Select a documented or plausible training need from your own organization, a role you have observed closely, or a disguised real case. The scenario should involve a specific performance gap and at least one real constraint, such as a tight timeline, a compliance requirement, or a limited budget, comparable to the Projonmo Microfinance case in this lesson.

Prepare a report of 2,000 to 2,500 words that:

1. states explicitly how much of your design process will run as a sequential, documented pass versus iterative prototyping loops, justifying the balance with reference to the specific stakes and constraints of your chosen scenario rather than treating ADDIE or SAM as a default choice;
2. writes at least four learning objectives for the scenario, each containing Mager's three required components (Performance, Condition, Criterion), and states which Knowledge-dimension category and which Cognitive-Process-dimension category (per Anderson and Krathwohl's 2001 revision) each objective belongs to;
3. identifies at least one objective in your set that uses a category label rather than a specific observable verb (for example, "understand") and revises it correctly;
4. explains, in your own words, the specific historical distinction between Branson's 1975 IPISD report and the ADDIE acronym now commonly attributed to it, and what that distinction should change about how you present your own design process to stakeholders; and
5. states explicitly where your proposed training blueprint's eventual success, in terms of actual on-the-job performance change rather than course completion, remains an open question this design-stage work alone cannot answer.

Use at least five academic or documented sources, including at least three from this article's reference list.

Assessment rubric (100 marks)

1. Allen, M. W., & Sites, R. (2012). Leaving ADDIE for SAM: An agile model for developing the best learning experiences. Alexandria, VA: ASTD Press.
2. Anderson, L. W., Krathwohl, D. R., et al. (Eds.). (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom's taxonomy of educational objectives. New York: Longman.
3. Bloom, B. S., Engelhart, M. D., Furst, E. J., Hill, W. H., & Krathwohl, D. R. (1956). Taxonomy of educational objectives: The classification of educational goals, Handbook I: Cognitive domain. New York: David McKay.
4. Branson, R. K. (1978). The interservice procedures for instructional systems development. Educational Technology, March, 11–14.
5. Branson, R. K., Rayner, G. T., Cox, J. L., Furman, J. P., King, F. J., & Hannum, W. J. (1975). Interservice procedures for instructional systems development (5 vols.). Fort Monroe, VA: U.S. Army Training and Doctrine Command.
6. Heinich, R., Molenda, M., Russell, J. D., & Smaldino, S. E. (2002). Instructional media and technologies for learning (7th ed.). Columbus, OH: Merrill/Prentice Hall.
7. Krathwohl, D. R. (2002). A revision of Bloom's taxonomy: An overview. Theory Into Practice, 41(4), 212–218.
8. Mager, R. F. (1962). Preparing objectives for programmed instruction. San Francisco: Fearon Publishers.
9. Molenda, M. (2003). In search of the elusive ADDIE model. Performance Improvement, 42(5), 34–36.

Verification notes: Molenda's (2003) "In Search of the Elusive ADDIE Model" and Krathwohl's (2002) "A Revision of Bloom's Taxonomy: An Overview" were both fetched and read in full via direct text extraction of their own published text, not a secondary summary, consistent with this series' standard for significant, specific claims. The IPISD report's full five-author byline, Branson's (1978) account of it, the 1956 and 2001 taxonomy publication details, Mager's (1962) original title and three-part objective format, the Heinich/Molenda/Russell/Smaldino ABCD attribution, and Allen and Sites's (2012) SAM publication details were checked against publisher and library-catalog records and multiple independent secondary summaries via web search. No statistic, date, or finding beyond what these sources report is claimed. The Projonmo Microfinance case, the decision scenario, and the closing synthesis are original teaching material for this series.

Lesson 44, next in Module 09, picks up directly from this lesson's Implementation and Evaluation phases: how training is actually delivered, what the research says about whether and when learning transfers back to the job, and the Kirkpatrick model's four levels for evaluating whether any of this design work produced a measurable result.