Instructional design

Instructional Design Models Compared: Rarely Tested, Widely Used

The field's own survey of instructional design models found that few have ever been rigorously tested. They earn their place by being useful, so choose one by what you're building, not by its reputation.

A man with a shaved head, short grey beard and glasses pins a card to a large corkboard covered in hand-drawn loops, arrows and flow diagrams in a bright studio with white brick walls, skylights and plants, a laptop open on the worktable behind him
Mapping a design process before choosing a model to run it.
The short answer

Instructional design models are step-by-step maps of how to plan and build instruction. The best known are ADDIE, Dick and Carey, Morrison, Ross and Kemp, SAM and ASSURE. Few have been rigorously tested, so none is proven best: choose by what you're building (a lesson, a product or a whole programme) and how certain the brief is.

Key takeaways

  • Most instructional design models are versions of the same five kinds of work, analyse, design, develop, implement and evaluate, arranged for a particular setting.
  • A standard survey of the field found a 'disturbingly small volume of literature' testing the models; they gain credibility because practitioners find them useful.
  • Process models (ADDIE, Dick and Carey, SAM) say how to run a project; frameworks such as Merrill's first principles and backward design say what good instruction contains, and you can use one inside the other.
  • Gustafson and Branch sort models by what they produce: a few hours of classroom teaching, a self-instructional product, or a whole course or curriculum.
  • When the brief is uncertain, an iterative model such as SAM or rapid prototyping fits better than a long sequential one.

Instructional design models are maps of how to plan and build instruction. The ones you’ll meet most often are ADDIE, the Dick and Carey model, the Morrison, Ross and Kemp model, SAM (the Successive Approximation Model) and, for classroom teaching, ASSURE. Most comparison pages stop there and list the steps. The more useful fact is the one the field’s own reference work states plainly: almost none of these models has been rigorously tested.

That comes from Kent Gustafson and Robert Branch’s Survey of Instructional Development Models, the standard catalogue of the models, which has run through several editions (Gustafson & Branch, 2002).

Instead, they write, models with wide distribution “gain their credibility by being found useful by practitioners, who frequently adapt and modify them to match specific conditions.” In their conclusion they go further: it appears many models “have never actually been applied, never mind rigorously evaluated.”

That changes how to choose one. If no model is proven better than the others, the question isn’t which is best. It’s which fits what you’re building. This guide compares the main models, separates them from the frameworks often listed alongside them, and gives a four-step way to pick.

Process models and design frameworks are different things

Search for instructional design models and you’ll find lists that mix two kinds of thing:

  • Process models describe how to run a design project from start to finish: ADDIE, Dick and Carey, Morrison, Ross and Kemp, SAM, ASSURE.
  • Design frameworks describe what good instruction contains, whatever process you use: Merrill’s first principles of instruction, backward design, action mapping. Some lists even include Bloom’s taxonomy, which classifies learning objectives and isn’t a model at all.

You don’t choose between the two kinds. You run a process model and use frameworks inside its phases: backward design while you plan assessments, Merrill’s principles while you design each lesson.

Most process models are also versions of the same work. Gustafson and Branch describe “five core elements—analyze, design, develop, implement, and evaluate (ADDIE)” that the models share, and the textbook Foundations of Learning and Instructional Design Technology calls ADDIE a process and each model “a specific instance of a process that can be imitated or emulated” (Dousay, 2018). So learning one model well teaches you most of the others.

The main models at a glance

ModelKindShapeFits bestWatch out for
ADDIEGeneric processFive phases, meant to loopAny project, as a checklist of the workBeing run as a one-way waterfall
Dick and CareySystems processDetailed sequence of stepsWhole courses where analysis and evaluation matterHeavy for a small project
Morrison, Ross and KempSystems processNine elements in a cycle; start anywhereCurriculum planning with revision throughoutLess guidance on order for beginners
SAM and rapid prototypingIterative processShort rounds of prototype and reviewUncertain briefs, eLearning built with a clientPrototypes that harden into the final product
ASSUREClassroom processSix stepsA teacher planning lessons with existing mediaAssumes you select materials more than build them
Backward designFrameworkThree stagesAligning objectives, assessment and lessonsSays little about running the project
Action mappingFrameworkGoal, actions, practice, minimum informationWorkplace training tied to a business problemLess suited to open-ended education
Merrill’s first principlesFrameworkFive principlesDesigning lessons around real tasksNot a project process on its own

The process models

ADDIE: the generic process

ADDIE’s five letters, analysis, design, development, implementation and evaluation, name the work every course needs. It has no single author, and the strictly linear version critics attack is a reading of it rather than the model itself. We’ve covered it in depth in our guide to the ADDIE model, including what each phase should produce.

Dick and Carey: the detailed systems model

Gustafson and Branch call the Dick, Carey and Carey model “the standard” against which other models are compared. It starts with a needs assessment to set instructional goals, runs instructional analysis and learner and context analysis side by side, then moves through performance objectives, assessment instruments, instructional strategy, materials, and formative and summative evaluation. The survey classes it as a systems model and notes how detailed it is in analysis and evaluation.

Its strength is that order: assessments are written before materials, and the course is tried out and revised before release. Its cost is effort. For a three-lesson course it’s more procedure than you need.

Morrison, Ross and Kemp: the flexible cycle

Jerrold Kemp’s model, later developed with Gary Morrison and Steven Ross, is organised around nine elements, from identifying the instructional problem and learner characteristics to analysing content, writing objectives and choosing resources. Revision runs alongside every element, and the survey notes that a designer “can start anywhere and proceed in any order.” It suits experienced designers planning a curriculum who want a checklist rather than a sequence.

SAM and rapid prototyping: build early, test often

Rapid prototyping, which the survey traces to Tripp and Bichelmeyer (1990), “emphasizes early development of a simple and incomplete prototype” that evolves as the client’s needs become clearer. SAM, developed by Michael Allen’s company Allen Interactions, is the best-known version for eLearning. It runs in three phases: preparation, iterative design and iterative development. Early prototypes are meant to be thrown away: “the less investment in initial prototypes, the better” (Manning, 2016).

Gustafson and Branch describe rapid prototyping as most useful when the client isn’t sure what they want or the project needs a creative solution. That’s most eLearning briefs.

ASSURE: the classroom model

ASSURE, from Heinich, Molenda, Russell and Smaldino, has six steps: analyse learners, state objectives, select media and materials, use them, require learner participation, and evaluate and revise. It’s built for a teacher planning their own lessons, who typically selects and adapts existing materials rather than producing new ones. The survey praises it as practical and easy for teachers to apply.

The frameworks you use inside a model

Backward design

Backward design, from Grant Wiggins and Jay McTighe, runs in three stages: identify the results you want, decide what evidence will show learners have reached them, then plan the lessons. It’s the best habit to bring into any model’s design phase. Our guide to backward design works through it on an online course.

Action mapping

Cathy Moore’s action mapping starts from a measurable business goal, lists what people need to do on the job, and checks whether training is even the answer before designing anything (Moore, 2014). Where it is, learners practise realistic decisions and get only the information each activity needs. Her advice to “resist the temptation to present and then quiz” is worth taping to the monitor.

Merrill’s first principles: the one with some evidence

David Merrill reviewed instructional design theories to find the principles they share (Merrill, 2002). He arrived at five, each stated as “Learning is promoted when…”: learners are engaged in solving real-world problems, existing knowledge is activated, new knowledge is demonstrated, learners apply it, and they integrate it into their own world. Merrill called the paper “a preliminary report.”

Unlike the process models, the principles have been studied in real courses. Frick and colleagues surveyed 464 students in 12 undergraduate courses at one US university, then had instructors rate each student’s mastery of the course objectives about a month later (Frick et al., 2010). Students who agreed that both the first principles and their own successful engagement with tasks had occurred were about five times as likely to be rated high masters as students who did not. That’s some evidence, not proof: the study was correlational, used student ratings of the teaching, and covered face-to-face courses at a single university. The principles also overlap with better-tested findings, such as retrieval practice for the application step.

How to choose a model: four questions

1. What are you building?

Gustafson and Branch sort models by what they produce:

ClassroomProductSystem
Typical outputOne or a few hours of instructionA self-instructional productA course or whole curriculum
Who does itUsually one personA teamA team
Front-end analysisLowLow to mediumVery high
Tryout and revisionLow to mediumVery highMedium to high

A teacher planning a unit is in the classroom column: ASSURE, or a light ADDIE pass. A creator building a self-paced course that strangers will take alone is in the product column, where testing with real learners matters most. An L&D team redesigning a company’s whole onboarding programme is building a system: Dick and Carey or Morrison, Ross and Kemp earn their extra procedure there.

2. How certain is the brief?

If you know exactly what learners need to do, a sequential model is efficient. If the client keeps saying “I’ll know it when I see it”, or nobody has built this kind of course before, start prototyping. For example, a small team asked to build a customer-service simulation would mock up one scenario in a day, put it in front of a few staff and the manager, and revise before designing the rest. That’s SAM in practice, and it’s also what our guide to piloting a course recommends for any build.

3. Is training the answer?

Before any model’s design phase, ask what problem the course is meant to fix. If sales staff aren’t logging calls because the software is slow, a course won’t help. Action mapping’s flowchart is the quickest check, and it fits inside ADDIE’s analysis phase or SAM’s preparation phase.

4. Which framework will shape the lessons?

The process model gets the course built. What makes each lesson work comes from frameworks with evidence behind them: design the assessment first with backward design, build lessons around real tasks with Merrill’s principles, and keep each screen within what learners can process (cognitive load theory).

In higher education the unit outline and learning outcomes usually arrive already approved, so I start from them and work backwards: our learning model sits in the background as a checklist, but where I’m building a new unit in Canvas and showing it to the academic beforehand, we do reports on what has changed, and assemble student feedback or additional research, if possible, to justify changes and help us all save time in the sign off process.

What the comparisons miss

Most pages comparing instructional design models present them as rival products, each with pros and cons, as if picking the right one guarantees a good course. The evidence doesn’t support that, because it barely exists. Gustafson and Branch advise designers to “acquire a working knowledge of several models”, one from each of their three categories, and to choose for the job in front of them.

In practice, we’d keep it simple. Use ADDIE as the checklist of work, run it in short loops with a prototype early, and put your effort into the parts with the best evidence: clear objectives, assessments that match them, and plenty of practice with feedback.

What to do next

Before your next project starts, answer the first question in one line: are you building a lesson, a product or a system? Write it at the top of your plan, and choose the model that fits that column.

Reader poll

Two quick questions about how you work.

Answer to see the results

1. Which design model do you work from most often?
2. What do you mostly build?
0 of 2 answered

Questions people ask

What are the four main instructional design models?

There's no official list. The four most often named are ADDIE, SAM (the Successive Approximation Model), the Dick and Carey model and the Morrison, Ross and Kemp model. ADDIE is the generic process the others build on. Dick and Carey is the most detailed systems model, Kemp's version can start at any element, and SAM replaces long phases with quick rounds of prototyping and review.

What is the difference between ADDIE and SAM?

ADDIE names five kinds of work: analysis, design, development, implementation and evaluation. SAM, from Allen Interactions, runs the work as three phases, preparation, iterative design and iterative development, with cheap, disposable prototypes reviewed early and often. ADDIE describes what needs doing; SAM is one way to schedule it so problems show up before the expensive build.

Is Bloom's taxonomy an instructional design model?

No. Bloom's taxonomy classifies learning objectives by the kind of thinking they ask for, from remembering to creating. It's a tool you use inside the design phase of a model, when you write objectives and match assessments to them. Many lists of instructional design models include it, which is part of why the term gets confusing.

Which instructional design model is best for eLearning?

No model has been shown to produce better courses than the others, so pick by situation. For a self-paced course built by a small team for learners you won't meet, a product-oriented approach with plenty of tryout and revision fits. If the brief is still uncertain, an iterative model such as SAM helps you test ideas with prototypes before building the full course.

Do instructional design models actually work?

There's little direct evidence either way. Gustafson and Branch's survey of the models found very few published tests of them, and the models gain credibility because practitioners find them useful. There is some evidence for what good instruction contains: a study of 12 university courses linked Merrill's first principles to higher instructor-rated mastery, though the design was correlational.

Sources

  1. Gustafson, K. L., & Branch, R. M. (2002). Survey of Instructional Development Models (4th ed.). ERIC Clearinghouse on Information & Technology, Syracuse University. https://people.potsdam.edu/betrusak/621/survey_of_intructional_development_models.pdf
  2. Merrill, M. D. (2002). First principles of instruction. Educational Technology Research and Development, 50(3), 43–59. https://doi.org/10.1007/BF02505024
  3. Frick, T. W., Chadha, R., Watson, C., & Zlatkovska, E. (2010). Improving course evaluations to improve instruction and complex learning in higher education. Educational Technology Research and Development, 58(1), 115–136. https://doi.org/10.1007/s11423-009-9131-z
  4. Manning, E. (Allen Interactions) (2016). Using an iterative process to bake successful e-learning results. Allen Interactions blog. https://blog.alleninteractions.com/using-an-iterative-process-to-bake-successful-e-learning-results
  5. Moore, C. (2014). Action mapping on one page. Cathy Moore (blog). https://blog.cathy-moore.com/online-learning-conference-anti-handout/
  6. Dousay, T. A. (2018). Instructional design models. In R. E. West (Ed.), Foundations of Learning and Instructional Design Technology. EdTech Books. https://edtechbooks.org/lidtfoundations/instructional_design_models
Jim Lounsbury

Learning designer, documentary filmmaker and founder of The Learning Curve. Jim designs microcredentials and online courses for universities and businesses. More about Jim

Read next