Daniel T Willingham, Department of Psychology, University of Virginia, USA

Individuals vary in their views of what students should be taught. How should teachers discuss the misdeeds of a nation’s founders? What is the minimum accomplishment expected of each student in mathematics? But there is no disagreement on the importance of critical thinking skills. In free societies, the ability to think critically is viewed as a cornerstone of individual civic engagement and economic success.

But is there evidence that explicitly teaching critical thinking brings any benefit? There is, and such evidence is available for different subject matters. For example, in one experiment researchers taught college students principles for evaluating evidence in psychology studies – principles like the difference between correlational research and true experiments (Bensley and Spero, 2014). These principles were incorporated into regular instruction in a psychology class, and their application was practised in that context. Compared to a control group that learned principles of memory, students who learned the critical thinking principles performed better on a test that required evaluation of psychology evidence.

But perhaps we should not find this result terribly surprising. You tell students that this is a good strategy for this type of problem, and you have them practise that strategy, so later they use that strategy when they encounter the problem.

When we think of critical thinking, we think of something bigger. When I teach students how to evaluate the argument in a set of newspaper editorials, I am hoping that they will learn to evaluate arguments generally, not just those that they read, and not just those that they would find in other editorials. This aspect of critical thinking is called transfer, and the research literature evaluating how well critical thinking skills transfer to new problems is decidedly mixed.

It is no surprise that programmes in school meant to teach general critical thinking skills have had limited success. Such programmes are usually curricular add-ons, during which students engage in critical thinking activities for perhaps five hours each week over the course of a year or two. Unfortunately, the evaluations of these programmes seldom offer a rigorous test of transfer. If the critical thinking regimen entails argument and debate, the outcome measure is usually the ability to evaluate arguments or take both perspectives in debate (see Kuhn and Crowell, 2011; Reznitskaya et al., 2012). When investigators have tested for transfer in such curricular programmes, positive results have been absent or modest and quick to fade (Ritchart and Perkins, 2005).

It is not useful to think of critical thinking skills, once acquired, as broadly applicable. Wanting students to be able to ‘analyse, synthesise and evaluate’ information sounds like a reasonable goal. But analysis, synthesis and evaluation mean different things in different disciplines. Literary criticism has its own internal logic, its norms for what constitutes good evidence and a valid argument. These norms differ from those found in mathematics, for example. And indeed, different domains – science and history, say – have different definitions of what it means to ‘know’ something. Thus, our goals for student critical thinking must be domain-specific.

But wait. Surely there are some principles of thinking that apply across fields of study? There are indeed principles that carry across domains of study. The problem is that people who learn these broadly applicable principles in one situation often fail to apply them in a new situation. The law of large numbers provides an example. It states that a large sample will probably be closer to a ‘true’ estimate than a small sample – if you want to know whether a set of dice is loaded, you’re better off seeing the results of 20 throws than two throws. People readily understand this idea in the context of evaluating randomness, but they are less likely to see the need for a large sample when judging academic performance; they are ready to say that someone who received poor grades on two maths tests is simply bad at maths.

We know that a student has understood an idea like the law of large numbers. But understanding it offers no guarantee that the student will recognise new situations in which that idea will be useful.

Critical thinking as problem recognition

We seem to face a significant challenge: how can we improve student critical thinking if it is difficult for them to appreciate that some new problems are actually ones that they have solved in the past?

Richard Catrambone developed a different technique to address a slightly different transfer problem. He noted that in maths and science classes, students often learned to solve standard problems via a series of fixed, lock-step procedures. That meant that students were stumped when confronted with a problem requiring a slight revision of the steps, even if the goal of the steps was the same. For example, a student might learn a method for solving work problems like ‘Trisha can paint a house in 14 hours and Carole can do it in eight. How long would it take them to paint one house, working together?’ A student who learns a sequence of steps to solve that sort of problem is often thrown by a small change – the homeowner had already painted a quarter of the house before hiring Trisha and Carole.

Catrambone (Catrambone, 1998; Margulieux and Catrambone, 2016) showed that student knowledge will be more flexible if students are taught to label the sub-steps of the solution with the goal it serves. For example, work problems are typically solved by calculating how much of the job each worker can do in an hour. If, during learning, that step were labelled so that students understood that the calculation was part of deriving the solution, they would know how to solve the problem when a fraction of the house is to be painted.

Open-ended problems and knowledge

Students encounter standard problems that are best solved in a particular way, but many critical thinking situations are unique. Critical thinking is needed when playing chess, designing a product or planning strategy for a field hockey match. But there are no routine, reusable solutions for these problems. Nevertheless, just as with routine problems, critical thinking for open-ended problems is enabled by extensive stores of knowledge about the domain (North et al., 2011).

Knowledge aids critical thinking in three ways. First, the recognition process previously described (‘Oh, this is that sort of problem’) can still apply to sub-parts of a complex, open-ended problem. Complex critical thinking may entail multiple simpler solutions from memory that can be ‘snapped together’ when solving complex problems (Koedinger et al., 2012; Taatgen, 2013). Calculating the best value among several vacation packages may be a novel, open-ended problem, but if the method of comparison calls for long division, I don’t need to think through a method to execute that sub-step.

The second way that knowledge contributes to critical thinking in open-ended problems is through its impact on working memory. Working memory is where you hold information and manipulate it to carry out cognitive tasks. So, for example, if I asked ‘How is a scarecrow like a blueberry?’, you would retrieve information about scarecrows (not alive, protect crops, found in fields, birds think they are alive) and blueberries (purple, used in pies, small) from your memory and then you’d start comparing these features, looking for overlap.

An important feature of working memory is its limited size. Suppose that I said, ‘What do these objects have in common: a blueberry, a scarecrow, a flowerpot, a drumstick and a dishwasher?’ Working memory would be overwhelmed. There’s probably space for the five words, but not for the five words and a bunch of information about each word, plus leftover attention to compare them.

So how does a chess player think about all 32 pieces on the board and their relative positions, and have attention left over to contemplate effective moves? Knowledge allows the player to treat groups of pieces as a single unit. The king, a castle and three pawns in a corner of the board relate to one another in the defensive position, so the expert will treat them as a single unit. This ability to clump multiple entities into a single, meaningful unit has been observed in many domains of expertise, as varied as dance and computer programming. When experience allows you to unite many separate dance moves into a single unit, it saves working memory space. That allows more working memory space for the dancer to think about more subtle aspects of movement, rather than crowding working memory with ‘what I am to do next’.

The third way that knowledge may contribute to critical thinking is in enabling you to deploy thinking strategies. We can tell students that they should evaluate the logic of the author’s argument when they read an opinion piece. Students should have no trouble recognising ‘Oh, this is that sort of problem’, and they may have committed to memory the right thinking strategy. But they may not be able to use the strategy without the right domain knowledge. This point is rather obvious in the case of a critical thinking skill like evaluating an argument: abstract principles like ‘look for hidden assumptions’ won’t help much in sizing up an opinion piece about the war in Afghanistan if you know very little about the topic. And never mind evaluating the argument – if you lack background knowledge about the topic, ample evidence from the last 40 years indicates that you will not comprehend the author’s claims in the first place (Willingham, 2017).

How to teach students to think critically

So what does all this mean? Is there really no such thing as a ‘critical thinking skill’ if by ‘skill’ we mean something generalisable?

Maybe. But one fact ought to be salient. We are not sure that the general skills exist, but we are quite sure that there is no proven way to teach them directly. In contrast, we have a pretty good idea of how to teach students the more specific critical thinking skills. I suggest that we do so. Here is a four-step plan.

First, identify what is meant by critical thinking in each domain. Be specific. What tasks showing critical thinking should a secondary school graduate be able to do in mathematics, history and other subjects? It is not useful to set a goal for students to ‘think like historians’. If students are to read as historians do, they need to learn specific skills like interpreting documents in light of their sources, corroborating them and putting them in historical context. These skills should be explicitly taught and practised – there is evidence that simple exposure to this sort of work without explicit instruction is less effective (Abrami et al., 2008; Halpern, 1998; Heijltjes et al., 2014).

Second, identify the domain content that students must know. We have seen that domain knowledge is a crucial driver of thinking skills. For example, sourcing historical documents means interpreting their content in light of the author, the intended audience and the circumstances under which the author wrote. It is not enough to know that a letter was written by an army sergeant to his wife just before the Battle of Romani. The student must know enough about the historical context to understand how this sourcing information ought to influence his or her interpretation of the letter.

What knowledge is essential to the type of thinking that you want your students to be able to do? That of course depends on one’s educational goals. The prospect of someone deciding which knowledge students ought to learn – and what they won’t learn – sometimes makes people uneasy exactly because this decision depends on one’s goals for schooling, and goals depend on values. Selection of content is a critical way that values are expressed (Willingham, 2012); making that choice will lead to uncomfortable trade-offs. But not choosing is still making a choice. It is choosing not to plan, and to let random forces determine what students learn.

In the third step, educators must select the best sequence in which to learn the skills. It is obvious that skills and knowledge build on one another in mathematics or history, and what is true of maths and history is true of other domains of skill and knowledge; we interpret new information in light of what we already know. The right preparation makes new learning easier.

Fourth, educators must decide which skills should be revisited across years. Studies show that even if content is learned quite well over the course of half a school year, about half will be forgotten in three years (Pawl et al., 2012). That doesn’t mean that there’s no value in exposing students to content just once; most students will forget much but they’ll remember something and, for some students, an interest may be kindled. But when considering skills that we hope will stick with students for the long term, we should plan on at least three to five years of practice (Bahrick, 1984; Bahrick and Hall, 1991). Most of the time, this practice will look different – it will be embedded in new skills and content. But this revisiting should be assured and planned.

Some practical matters of teaching critical thinking

When should critical thinking instruction start? There is not a firm, research-based answer to this question. What children can and cannot do varies depending on the content. For example, in some circumstances, even toddlers can understand principles of conditional reasoning, and in other circumstances, conditional reasoning confuses adult physicians. It all depends on the content of the problem (Willingham, 2008).

Thus, research tells us that including critical thinking in the schooling of young children is likely to be perfectly appropriate. It does not, however, provide guidance into what types of critical thinking skills to start with. That is a matter to take up with experienced educators, coordinating with colleagues who teach older children in the interests of making the curriculum seamless.

Should everyone learn critical thinking skills? The question sounds like a set-up. But the truth is that, in many systems, less capable students are steered into less challenging coursework, with the hope that by reducing expectations, they will at least achieve mastery of the basics. Access to challenging content and continuing to tertiary education is, in nearly every country, associated with socioeconomic status (OECD, 2018). Children from high socioeconomic status families also have more opportunities to learn at home. If school is the chief or only venue through which low socioeconomic status students are exposed to advanced vocabulary, rich content knowledge and demands for high-level thinking, it is absolutely vital that those opportunities be enhanced, not reduced.

This is an edited extract from Willingham D (2019) How to teach critical thinking. Education Future Frontiers: Occasional Paper Series. education.nsw.gov.au

Copyright: Daniel T. Willingham and the NSW Department of Education, Education Futures and Governance, 2019


Abrami PC, Bernard RM, Borokhovski E et al. (2008) Instructional interventions affecting critical thinking skills and dispositions: A stage 1 meta-analysis. Review of Educational Research 78(4): 1102–1134.

Bahrick HP (1984) Semantic memory content in permastore: Fifty years of memory for Spanish learned in school. Journal of Experimental Psychology: General 113(1): 1–29.

Bahrick HP and Hall LK (1991) Lifetime maintenance of high school mathematics content. Journal of Experimental Psychology: General 120(1): 20–33.

Bensley DA and Spero RA (2014) Improving critical thinking skills and metacognitive monitoring through direct infusion. Thinking Skills and Creativity 12: 55–68.

Catrambone R (1998) The subgoal learning model: Creating better examples to improve transfer to novel problems. Journal of Experimental Psychology: General 127(4): 355–376.

Halpern DF (1998) Teaching critical thinking for transfer across domains: Disposition, skills, structure training, and metacognitive monitoring. American Psychologist 53(4): 449–455.

Heijltjes A, Van Gog T and Paas F (2014) Improving students’ critical thinking: Empirical support for explicit instructions combined with practice. Applied Cognitive Psychology 28(4): 518–530.

Koedinger KR, Corbett AT and Perfetti C (2012) The knowledge-learning-instruction framework: Bridging the science-practice chasm to enhance robust student learning. Cognitive Science 36(5): 757–798.

Kuhn D and Crowell A (2011) Dialogic argumentation as a vehicle for developing young adolescents’ thinking. Psychological Science 22(4): 545–552.

Margulieux LE and Catrambone R (2016) Improving problem solving with subgoal labels in expository text and worked examples. Learning and Instruction 42: 58–71.

North JS, Ward P, Ericsson A et al. (2011) Mechanisms underlying skilled anticipation and recognition in a dynamic and temporally constrained domain. Memory 19(2): 155–168.

OECD (2018) Education at a Glance 2018: OECD Indicators. Paris: OECD Publishing.

Pawl A, Barrantes A, Pritchard DE et al. (2012) What do seniors remember from freshman physics? Physical Review Special Topics – Physics Education Research 8(2): 020118.

Reznitskaya A, Glina M, Carolan B et al. (2012) Examining transfer effects from dialogic discussions to new tasks and contexts. Contemporary Educational Psychology 37(4): 288–306.

Ritchart R and Perkins DN (2005) Learning to think: The challenges of teaching thinking. In: Holyoak KJ and Morrison RG (eds) The Cambridge Handbook of Thinking and Reasoning. Cambridge: Cambridge University Press, pp. 775–802.

Taatgen NA (2013) The nature and transfer of cognitive skills. Psychological Review 120(3): 439–471.

Willingham DT (2008) What is developmentally appropriate practice? American Educator 32(2): 34–39.

Willingham DT (2012) When Can You Trust the Experts? How to Tell Good Science from Bad in Education. San Francisco, CA: Jossey-Bass.

Willingham DT (2017) The Reading Mind. San Francisco, CA: Jossey-Bass.