Bloom's Taxonomy: 6 Levels, Verbs and Examples

Go from storyboard to an interactive course in minutes

Start Free

Picture a community pharmacy (a hypothetical example) that trains its new pharmacy assistants with a slide deck on over-the-counter medicines and a recall quiz on the questions to ask before any sale. Every new assistant passes. A month later, a customer comes back to the counter for a second packet of sleep aids in the same week, and the assistant on duty rings up the sale. The assistant can list the questions, but the course never asked her to notice a pattern or to decide when a request goes to the pharmacist.

I'm an instructional designer at Coursebox, where we make an AI quiz generator that writes quiz questions from course content, so I have a stake in how assessments get written. Bloom's taxonomy is the tool I use to find gaps like the pharmacy's. It shows whether each objective and its assessment ask for the same kind of thinking, and the pharmacy course is my example for all six levels.

What is Bloom's taxonomy?

Bloom's taxonomy is a framework for classifying learning objectives into six levels of thinking, ordered from simple to complex, first published in 1956 by a committee of college and university examiners and revised in 2001, when its levels became remember, understand, apply, analyse, evaluate and create.

David Krathwohl, one of the original authors and co-editor of the revision, describes it as "a framework for classifying statements of what we expect or intend students to learn as a result of instruction" (Krathwohl, 2002, p. 212). In workplace training, those statements are your learning objectives.

There are two sets of names in circulation. The 1956 handbook used nouns: knowledge, comprehension, application, analysis, synthesis and evaluation, with evaluation at the top. The 2001 revision put the names in verb form, renamed three of them and swapped the last two, so synthesis became create and moved to the top, and evaluate became fifth. The revised names are the ones to use today.

The examples use a hypothetical course for new community-pharmacy assistants. They follow an approved question routine for medicine requests and refer decisions outside their role to the pharmacist. The course targets the thinking the job requires; creating a new routine is an optional extension for experienced staff, subject to pharmacist approval.

Case study: Auditing introductory biology assessments

Momsen and colleagues classified 9,713 assessment items from 77 introductory biology courses taught by 50 faculty. Ninety-three per cent tested knowledge or comprehension in the original taxonomy. The cognitive level of stated course goals did not predict the level of corresponding assessments. This was an assessment audit, not a trial showing that Bloom's improved learning. Its practical lesson is to classify actual questions and compare them with course goals before claiming alignment. Read the biology assessment study to see how the audit was conducted.

The descriptions are the revision's category definitions as Krathwohl summarises them (2002, p. 215).

Case study: Faculty learning to write assessment questions

Faculty development workshops for newly joined medical faculty in Riyadh taught staff to write multiple-choice questions at higher cognitive levels. The questions were sorted with a two-level version of Bloom's: K1 for recall and comprehension, K2 for application and analysis.

• Setting: the medical faculty's exam questions from three courses, 240 questions in the year before the workshops and 240 in the year after, 2012 to 2014.
• Measured: the exam questions the staff wrote, which is their work output.

The 2015 study compared 240 questions before and 240 after workshops. Higher-level items rose from 84 to 123, and item-writing flaws fell from 47 to 5. Read the item-writing workshop study for the methods and results.

A broader analysis of 2,207 questions across 2012 to 2015 found higher-level items at 26.7% initially and 31.2% in the final year. That change was not statistically significant (p = 0.086), although flaws declined. The two papers use different question sets, so these figures do not directly measure whether the earlier gain lasted. The longitudinal study supports continuing question review rather than treating one workshop as sufficient.

Who created Bloom's taxonomy, and which version to use?

Bloom's taxonomy was written by a committee, with Benjamin Bloom as editor. The 1956 title page (Open Library record) reads "By A Committee of College and University Examiners", and it names five authors: Benjamin S. Bloom, Max D. Engelhart, Edward J. Furst, Walker H. Hill and David R. Krathwohl. Each of them worked in examinations, testing or research.

The idea came from testing. In the handbook's words, "The idea for this classification system was formed at an informal meeting of college examiners attending the 1948 American Psychological Association Convention in Boston" (Bloom et al., 1956, p. 4). Krathwohl later explained the practical aim: to share test items between universities "in order to create banks of items, each measuring the same educational objective" (Krathwohl, 2002, p. 212). The group met from 1949, and Handbook I: Cognitive Domain came out in 1956.

The authors also saw a risk that still applies to training teams. They were concerned about teachers who "merely selected what they believed to be desirable objectives from the list provided in the taxonomy" (1956, p. 5). The taxonomy sorts objectives you have already written from the job.

The revision was published in 2001 as A Taxonomy for Learning, Teaching, and Assessing, edited by Lorin Anderson and David Krathwohl with six co-authors, eight in all. Krathwohl says it was developed "in much the same manner 45 years later", and that it "gives much greater weight to teacher usage" (2002, pp. 212 and 215). The original was eventually translated into 22 languages (p. 213).

• Start with Krathwohl's 2002 article, "A Revision of Bloom's Taxonomy: An Overview", in Theory Into Practice 41(4), pages 212 to 218 (DOI). It is seven pages long and written by the co-editor. It covers the two dimensions, lists all 19 cognitive processes and shows the Taxonomy Table.
• Use the 2001 book (Longman) when you need the full detail. It is the current version, sold as a complete hardcover edition and a shorter paperback.
• Read the 1956 handbook for history and for its sample test items. It is in libraries but not legitimately free to read online.

Is Bloom's taxonomy a pyramid or a hierarchy?

Bloom's taxonomy is a hierarchy of complexity, and the pyramid came from somewhere else. The 1956 handbook presents the six classes as a numbered list and outline, and the 2001 revision uses a two-dimensional table. Researchers who traced the pyramid image found that "its origin remains a mystery: it was not part of Bloom's et al (1956) original framework or the later revision" (Mikhaylova and Pettersson, ECER, 2024). They also point out that the pyramid draws the simplest level as the widest and the most complex as the narrowest.

The original authors did order the classes as a ladder. They wrote that "the objectives in one class are likely to make use of and be built on the behaviors found in the preceding classes in this list" (1956, p. 18). On the next page they added, "Our evidence on this is not entirely satisfactory" (p. 19).

The revision loosened the ladder. Krathwohl writes that the categories still run from less to more complex, but "the requirement of a strict hierarchy has been relaxed to allow the categories to overlap one another" (2002, p. 215). His example is that explaining, a process under understand, can be harder than executing, a process under apply. In the pharmacy course, explaining to a worried customer why you ask about their other medicines can be harder than running the question routine on a routine cold request.

So you do not need to teach every fact before you ask people to think. In studies with middle school and college students, Pooja Agarwal found that "higher order and mixed quizzes improved higher order test performance, but fact quizzes did not" (Agarwal, Journal of Educational Psychology, 2019). In two sections of an introductory biology course, students given high-level exams did better on both the low-level and high-level questions in the final exam (Jensen et al., Educational Psychology Review, 2014). The authors read that result as support for the hierarchy, and it also suggests that working at the harder levels can help with the easier ones. For a course, that points to opening with a realistic counter scenario and teaching the facts inside it.

What is the knowledge dimension in revised Bloom's taxonomy?

The knowledge dimension is the second axis that the 2001 revision added. Krathwohl points out that objectives "typically consist of a noun or noun phrase", the content, and "a verb or verb phrase", the cognitive process (2002, p. 213). The original knowledge level mixed the two. The revision split them, so every objective now has a level of thinking and a type of knowledge.

The four types of knowledge, with Krathwohl's definitions (p. 214):

• Factual: "The basic elements that students must know to be acquainted with a discipline or solve problems in it." In the pharmacy course, the questions on the pharmacy's checklist.
• Conceptual: "The interrelationships among the basic elements within a larger structure that enable them to function together." Why each question is asked, and what turns a detail into a warning sign.
• Procedural: "How to do something", including "criteria for using skills, algorithms, techniques, and methods." The question routine, and the rules for when to refer to the pharmacist.
• Metacognitive: "Knowledge of cognition in general as well as awareness and knowledge of one's own cognition." An assistant knowing when she is out of her depth.

Put the two dimensions together and you get what the authors called the Taxonomy Table, with the types of knowledge down the side and the six processes across the top (p. 215). Krathwohl shows how to use it by placing one state standard in two cells at once, "B4, Analyze Conceptual Knowledge, and B6, Create [based on] Conceptual Knowledge" (p. 216).

The pharmacy course's final objective works the same way: when a customer buys a second sleep-aid packet in the same week, spot the warning sign and decide whether to refer them to the pharmacist. It sits in two cells. Spotting the sign is B4, analyse conceptual knowledge, because the assistant has to see why a repeat purchase is different from a routine one. Deciding is C5, evaluate procedural knowledge, because the decision follows the pharmacy's criteria for when to refer.

Mapping all six objectives shows the whole course at once:

The Taxonomy Table helps you inspect the kinds of knowledge and thinking your objectives target. Its blank cells invite reflection, not automatic additions: every course does not need every cell. If judging one's own limits is an intended outcome, add a task in which assistants explain what they know, what is missing and why the approved protocol requires referral. Assess that reasoning against criteria. A one-line self-check alone does not establish metacognitive evaluation.

The six levels of Bloom's taxonomy, worked through one course

Return to the hypothetical community-pharmacy course. New assistants need to handle routine requests within the pharmacy's approved protocol, recognise when information is insufficient and refer to the pharmacist. The walkthrough shows how the taxonomy changes the evidence you ask for. It does not require new starters to progress through every level; the Create example is an extension for experienced staff.

1. Remember

Remember is retrieving what you need from memory, and the revision gives it two processes: recognising and recalling. At work, it covers the facts people must have without looking them up.

• Sample objective: List the questions the pharmacy asks before any over-the-counter sale.
• How to assess it: A short recall quiz or a fill-in checklist. This is the one level where a recall quiz is the right test.
• In the course: A one-page summary of the questions, followed by a recall check at the end of the first section.

2. Understand

Understand is "Determining the meaning of instructional messages", and it is the broadest level, with seven processes from interpreting to explaining. Understand covers more than recall: the learner has to make sense of the material and say it in a new way.

• Sample objective: Explain to a customer, in plain words, why the pharmacy asks about the other medicines they take.
• How to assess it: A short written or recorded explanation, checked against a list: the reason is stated, the words are plain and the tone is reassuring.
• In the course: Good and poor explanations side by side, then practise explaining two of the questions.

3. Apply

Apply is "Carrying out or using a procedure in a given situation". Its two processes are executing, on a familiar task, and implementing, on an unfamiliar one.

• Sample objective: Handle a routine cold-and-flu request using the pharmacy's question routine.
• How to assess it: A decision scenario with feedback. The customer's answers arrive one at a time, and the learner chooses the next step.
• In the course: A demonstration of the routine at the counter, then two practice requests.

4. Analyse

Analyse is breaking material into parts and seeing how they relate. Its first process, differentiating, means "Distinguishing relevant from irrelevant parts or important from unimportant parts of presented material". An assistant does this while a customer talks.

• Sample objective: Pick out the details in a customer's answers that take a request outside routine.
• How to assess it: A counter conversation in which the learner flags the details that change the request and says why.
• In the course: Annotated conversations, then sorting practice: routine detail or warning sign.

5. Evaluate

Evaluate is "Making judgments based on criteria and standards", through checking and critiquing. The criteria are what separate it from opinion. At the counter, they are the pharmacy's rules for selling and referring.

• Sample objective: Decide whether to sell, suggest another option or refer to the pharmacist, and justify the decision against the pharmacy's protocol.
• How to assess it: Decision scenarios with feedback on the reasoning. The final one combines analyse and evaluate: a customer buys a second sleep-aid packet in the same week.
• In the course: Scenarios in which more than one response sounds reasonable, with feedback that explains which one the protocol supports.

6. Create

Create is "Putting elements together to form a novel, coherent whole or make an original product", through generating, planning and producing. At work, it is usually a plan, a procedure or a document that others will use.

• Sample objective: Plan a question routine for a product range new to the pharmacy, for the pharmacist to approve.
• How to assess it: The routine itself, reviewed by the pharmacist against a short rubric.
• In the course: I would keep this level for experienced assistants and leave it out of a new-starter course.

The alignment worksheet lays out the six levels side by side, with columns for the objective, the process verb, the type of knowledge, the assessment and the activity.

Bloom's taxonomy verbs: the 19 cognitive processes

Use the 19 cognitive processes in the 2001 revision as your Bloom's verb list. Krathwohl notes that in the revision these processes "receive the major emphasis", while the original gave far more attention to the six headings (2002, p. 214). The 1956 handbook gave each class "a definition of the category", "illustrative objectives" and "examples of items testing objectives in the category" (p. 4). It gave no verb lists. Those came from later writers on behavioural objectives, such as Metfessel, Michael and Kirsner, who in 1969 added "behaviorally oriented infinitives" to each class (ERIC, 1969), and Gronlund in 1970.

Alternative names in brackets are Anderson and Krathwohl's, spelt the Australian way, and the workplace verbs are my own additions.

How to use Bloom's taxonomy to design a training course

Bloom's taxonomy works best as a check on alignment between objectives, assessments and activities. Bloom himself hoped it would offer a "means for determining the congruence of educational objectives, activities, and assessments in a unit, course, or curriculum" (Krathwohl, 2002, p. 212). This is the order I work in.

1. Start from the job, and list the decisions and tasks people get wrong.
2. Write each objective as a process plus content, such as "pick out the warning signs in a customer's answers", in SMART and achievable wording.
3. Place each objective in the Taxonomy Table: its level of thinking and its type of knowledge.
4. Look at the spread. If every objective sits at remember and understand, add the higher levels the job needs, and check the empty cells.
5. Write the assessment for each objective at the same level, which is what makes the outcome measurable. Writing the assessment before the content is the core of backward design.
6. Plan activities that practise the same process the assessment asks for, with feedback, and use quick checks along the way for the formative work.
7. Classify your existing quiz questions by level and compare them with the objectives.

Case study: A nurse empowerment programme in Türkiye

Researchers from Koç University School of Nursing ran a randomised controlled trial of an empowerment programme for new graduate nurses at a private university hospital in Turkey. In their words, "the educational objectives and content of the research were prepared by considering Bloom's taxonomy".

• Setting: 64 new graduate nurses, 32 in the programme and 32 in a control group, June 2021 to September 2023, with no loss of participants. The programme mixed online modules with two face-to-face days.
• Measured: self-reported empowerment on two scales at three months.
• Result: The nurse empowerment trial reported structural empowerment rising from 52.53 to 67.81 in the programme group, compared with 50.91 to 56.22 in controls. Psychological empowerment rose from 33.47 to 43.09, compared with 33.25 to 37.38. These are self-reported outcomes from the full programme, not evidence that Bloom's alone improved clinical performance.

In professional education, a team in Quebec classified the objectives of 110 accredited continuing professional development activities for clinicians, run between November 2012 and March 2013. See the continuing education analysis.

The authors concluded that most of the activities "were generally not designed to promote clinical behavior change". In both studies the lower levels outweighed the higher ones, so classify your objectives and questions before you build anything.

Which Bloom's levels do course requests ask for?

To see which levels people ask for when they brief a course, we classified a sample of course requests typed into Coursebox. The sample came from the 90 days to 7 October 2026: requests from English-language browsers, one per person, with no uploaded file, and with pre-outline answers and template duplicates removed. A language model classified the 266 prompts that remained, 100 of them were course requests, and we spot-checked the results by hand.

The classifier counted generously, so analyse and create each came up in no more than about 1 in 4 requests, and only 1 of the 100 named Bloom's taxonomy at all. Within these sampled requests, judgement was the least often requested level. The classification does not establish what the resulting courses contained. If you brief a course builder, human or AI, write the evaluate objective and its scenario into the brief.

Our support inbox shows the same need from another angle. In a sweep of the inbox, no customer had written to us about Bloom's by name. The questions behind it do arrive: we confirmed 5 customers asking for deeper or scenario-based questions in place of recall, and 1 asking for easy, medium and hard tiers.

The affective and psychomotor domains

The 1956 group planned three domains. The handbook says, "Our original plans called for a complete taxonomy in three major parts", cognitive, affective and psychomotor (p. 7), and Bloom's taxonomy usually means the cognitive one.

• Affective domain: published in 1964 as Handbook II, by Krathwohl, Bloom and Masia, with Krathwohl as first author. It covers values and attitudes in five levels: receiving, responding, valuing, organisation, and characterisation by a value or value complex, where behaviour is consistently governed by a set of values. Consistently prioritising customer safety over sales could reflect that intended value. A short course or a correct scenario response cannot establish a stable pattern of values across working life.
• Psychomotor domain: never published by Bloom's group. The 1956 authors found "so little done about it in secondary schools or colleges" that they set it aside (pp. 7 to 8). Others filled the gap. Elizabeth Simpson proposed five tentative categories in 1966 (ERIC) and seven in 1972, ending with adaptation and origination (ERIC), and R. H. Dave (1970) and Anita Harrow (1972) published their own.

For training with attitude or physical skill goals, such as safe manual handling, plan those objectives with these domains, or with a model built around them.

Bloom's taxonomy vs Fink's taxonomy, SOLO and Webb's DOK

Bloom's sorts objectives by the thinking they ask for. Fink's taxonomy of significant learning adds kinds of learning that Bloom's cognitive domain leaves out, such as caring and learning how to learn. The SOLO taxonomy, from John Biggs and Kevin Collis (1982), grades the structure of a learner's answer, from prestructural to extended abstract. Norman Webb's Depth of Knowledge rates how complex an objective or test item is on four levels: recall, skill or concept, strategic thinking and extended thinking (Webb, 1999).

Webb's levels come with the same warning about verbs. His 1999 report notes that verbs such as "describe" and "explain" "could be classified at different levels depending on what is to be described and explained". In practice I write objectives with Bloom's, add Fink's categories when the course has to change attitudes, and use SOLO when I write a rubric for open answers.

Limitations of Bloom's taxonomy

• The ladder is weaker than it looks. Madaus, Nuttall and Woods tested the hierarchy with 1,128 students and found "a Y-shaped structure": knowledge, comprehension and application formed the stem, then analysis branched one way and synthesis and evaluation the other (ERIC, 1971). They also found that the synthesis and evaluation tests became "more a measure of general mental ability".
• People disagree on the level of a question. When faculty and students at a US medical school sorted multiple-choice questions into higher and lower order, agreement was low (Krippendorff's alpha 0.37 in each group), while the trained assessment team reached 0.90 (Monrad et al., Medical Teacher, 2021). If you classify a question bank, agree on a few examples first.
• It covers one domain at a time. The cognitive taxonomy has nothing to say about attitudes or motor skills, and the psychomotor domain never came from Bloom's group.
• It does not choose methods. Sequence, practice and delivery are decisions the taxonomy leaves to the designer.
• Workplace evidence is thin. I found no private-sector company case with measured results. The published cases come from health employers and staff development, and they measure exam questions written by staff, self-reported empowerment or knowledge. Only the exam-question case looked at work people produced in their jobs, so evidence at level 3 of the Kirkpatrick model, behaviour on the job, is thin.

Bloom's taxonomy with AI

Generative AI has changed what some levels prove, and Oregon State University Ecampus has published an update for it. Its "Bloom's Taxonomy Revisited" uses the taxonomy "as a touchstone for reconsidering course outcomes and student learning in the age of generative artificial intelligence (GenAI)" (Oregon State University Ecampus, version 2.0, 2024). For each level it sets "Distinctive Human Skills" against "How GenAI Can Supplement Learning Processes". Version 1.0, from 2023, flagged that activities "at the Remember and Analyze levels may be more likely to need amendment". The AI column reflects ChatGPT 4 as of October 2023, so check it against the tools you use now.

For the pharmacy course, I split each level two ways: what I would let AI help with when building or delivering the course, and what the assistant still has to do herself.

Three cautions apply when AI is part of the course. Keep real customer details out of prompts and practice scenarios. Check generated scenarios for bias, such as the same kind of customer always cast as the one misusing medicines. And watch for over-reliance: if the AI writes every justification, learners never practise the evaluate level the course is there to build.

Building a Bloom's course in Coursebox

On 7 October 2026 I gave Coursebox a brief for an over-the-counter medicine request course for new community-pharmacy assistants. It had five objectives, each tagged with a revised Bloom's level: remember, understand, apply, analyse and evaluate. The assessment rule was a recall quiz only for remember, and decision scenarios with feedback for apply, analyse and evaluate. The final scenario combined analyse and evaluate: a customer buying a second sleep-aid packet in the same week.

The first outline got the structure right. The sections followed the five levels in order, the apply, analyse and evaluate sections included scenarios, and the sleep-aid scenario came last.

The checks did not follow my requested formats. The remember section had no recall check, and the apply, analyse and evaluate sections ended in multiple-choice quizzes instead of the scenarios I had requested. Multiple-choice questions can test application and analysis, so their format alone did not establish the cognitive level. That judgement requires inspecting the actual question and what a learner must do to answer it.

One correction message fixed it. I asked for a recall check in the first section and for scenarios with feedback to test apply, analyse and evaluate. The revised outline also dropped two sections outside my objectives, and when I reloaded it, it ran about 1 hour 18 minutes across 6 sections and 28 pages. Counter Medicine Basics, the remember section, gained a recall page called Essential Facts Recall. Routine Request Workflow, the apply section, gained a Routine Request Scenario and kept a short 3-question check beside it. Recognising Red Flags and Choosing the Safest Action included decision-scenario pages intended to assess analyse and evaluate. The sixth section, Final Sleep Aid Scenario, held the combined analyse and evaluate case and closed the course. This test checked the outline and page labels. It did not establish the cognitive demand of the finished questions, the quality of feedback or learners' performance.

Putting the model to work

Choose the thinking the job requires, then inspect the task and evidence for each objective. Use both dimensions of the revised taxonomy: cognitive process and knowledge type. Keep useful factual checks, add decision tasks where judgement is required, and review the actual questions rather than their labels or format. You do not need every level or every grid cell. You need enough evidence to show that learners can do what the course promises.