In 2006 Paul Kirschner published, with John Sweller and Richard E Clark, a now-seminal piece of research that threatened to blow the doors off an often-accepted orthodoxy in teaching: that students learn best when they discover things by themselves. They proposed that not only was this not the case, but that the best learning frequently took place when guided direct instruction by an expert was the main strategy.
Decades of research demonstrates that for novices (the state of most students), direct explicit instruction is more effective and efficient – and in the long run enjoyable – than minimal guidance. So, when teaching new content and skills to novices, teachers are more effective when they provide explicit support and guidance. Direct, explicit instruction fully explains the concepts and skills that students are required to learn. It can be provided through all types of media and pedagogies (e.g., lectures, modelling, videos, computer-based presentations, demonstrations, class discussions, hands-on activities etc.) as long as the teacher ensures that the relevant information is explicitly provided and practised. Minimal instructional guidance, on the other hand, expects students to discover on their own most, if not all, of the concepts and skills they are supposed to learn. This approach has been given various names such as discovery learning, problem-based learning, inquiry learning, experiential learning, and constructivist learning.
Rich Mayer examined studies conducted from 1950 to the late 1980s that compared discovery learning (defined as unguided, problem-based instruction) with guided forms of instruction. In his famous three-strikes paper,2 he suggested that in each decade since the mid-1950s, after empirical studies provided solid evidence that the then-popular form of unguided approach did not work, a similar approach soon popped up under a different name with the cycle then repeating itself. This pattern produced discovery learning, then experiential learning, then problem-based and inquiry learning, then constructivist pedagogies, ad infinitum. He concluded that the ‘debate about discovery has been replayed many times in education but each time, the evidence has favored a guided approach to learning’ (p. 18).
Evidence from well-designed, properly controlled experimental studies as well as classroom studies from the 1980s to today also supports direct instructional guidance. The research has shown that when students try to learn with discovery methods or with minimal feedback, they often become lost and frustrated, and their confusion can lead to misconceptions: That because false starts (where students pursue misguided hypotheses) are common, unguided discovery is also inefficient. In a very important study,3 researchers not only tested whether science learners learned more via a discovery versus direct- instruction route but also, once learning had occurred, whether the quality of learning differed. The findings were unambiguous. Direct instruction involving considerable guidance, including examples, resulted in vastly more learning than discovery. Those relatively few students who learned via discovery showed no signs of superior quality of learning or superior transfer. Also, even if a problem or project is devised that all students succeed in completing, minimally guided instruction is much less efficient than explicit guidance. What can be taught directly in a 25-minute demonstration and discussion followed by 15 minutes of independent practice with good teacher feedback may take several class periods to learn via minimally guided projects and/or problem solving. And finally, minimally guided instruction can increase the achievement gap. A review of approximately 70 studies4 found not only that higher skilled learners tend to learn more with less guided instruction, while lower skilled learners tend to learn more with more guided instruction, but that lower skilled students who used less guided instruction received significantly lower scores on post-tests than on pre-test measures. For these relatively weak students, the failure to provide strong instructional support and guidance produced a measurable loss of learning.
Now let’s look at how we learn. There are two essential components that influence how we learn: long-term memory (LTM) and working memory (WM; often called short-term memory). LTM is a big mental warehouse of things while WM is a limited mental ‘space’ in which we think. However, to dispel a common misconception, LTM is not a passive repository of discrete, isolated fragments of information that permit us to repeat what we have learned, having only peripheral influence on complex cognitive processes such as critical thinking and problem solving. It is, rather, the central, dominant structure of human cognition. Everything we see, hear, and think about depends on and is influenced by our LTM. Expert problem solvers, for example, derive their skill by drawing on the extensive experience stored in their LTM in the form of concepts and procedures, known as mental schemas. They retrieve memories of past procedures and solutions, and then quickly select and apply the best ones for solving problems. We are skilful in an area if our LTM contains huge amounts of information concerning the area. That information permits us to quickly recognise the characteristics of a situation and indicates to us, often immediately and unconsciously, what to do and when to do it. And what are the instructional consequences of LTM? First and foremost, LTM provides us with the ultimate justification for instruction: the aim of all instruction is to add knowledge and skills to LTM. If nothing has been added to LTM, nothing has been learned.
WM, in contrast, is the cognitive structure in which conscious processing occurs. We are only conscious of the information currently being processed in WM and are more or less oblivious to the far larger amount of information stored in LTM. When processing novel information, WM is very limited in duration and capacity. We have known at least since the 1950s that almost all information stored in WM is lost within 30 seconds if it is not rehearsed and that the capacity of WM is limited to only a very small number of elements, estimated at about 7, but may be as low as 4±1.
For instruction, the interactions between WM and LTM may be even more important than the processing limitations. The limitations of WM only apply to new, to-be-learned information (i.e., information that has not yet been stored in LTM). When dealing with previously learned information stored in LTM, these limitations disappear. Since information can be brought back from LTM to WM as needed, the 30-second limit of WM becomes irrelevant. Similarly, there are no known limits to the amount of such information that can be brought into WM from LTM.
These two facts – that WM is very limited when dealing with novel information, but is not limited when dealing with information stored in LTM – explain why minimally guided instruction typically is ineffective for novices, but can be effective for experts. When given a problem to solve, novices’ only resource is their very constrained WM while experts have both their WM and all the relevant knowledge and skill stored in LTM.
One of the best examples of an instructional approach that takes into account how our working and long-term memories interact is the ‘worked example effect’. Solving a problem requires searching for a solution, which must occur using our limited WM. If the learner has no relevant concepts or procedures in LTM, the only thing they can do is blindly search for possible solution steps that bridge the gap between the problem and its solution. This process places a great burden on WM capacity because the problem solver has to continually hold and process the current problem state in WM (e.g., Where am I right now in the problem solving process? How far have I come towards finding a solution?) along with the goal state (e.g., Where do I have to go? What is the solution?), the relations between the goal state and the problem state (e.g., Is this a good step toward solving the problem? Has what I’ve done helped me get nearer to where I need to go?), the solution steps that could further reduce the differences between the two states (e.g., What should the next step be? Will that step bring me closer to the solution? Is there another solution strategy that I can use that might be better?), and any sub goals along the way. Thus, searching for a solution overburdens limited WM and diverts working-memory resources away from storing information in LTM. As a consequence, novices can engage in problem-solving activities for extended periods and learn almost nothing.
In contrast, studying worked examples reduces the burden on WM (because the solution only has to be comprehended, not discovered) and directs attention (i.e., directs WM resources) toward storing the essential relations between problem-solving moves in LTM. Students learn to recognise which moves are required for particular problems, which is the basis for developing knowledge and skill as a problem solver. As the learner progresses, various steps can be faded away so that the learner needs to think up and complete those steps themself (partially worked examples).
It is important to note that this discussion of worked examples applies to novices – not experts. In fact, the worked-example effect first disappears and then reverses as the learners’ expertise increases. That is, for experts with lots of knowledge in the LTM, solving a problem can be more effective than studying a worked example.
Why then, with all of this proof, do people continue to think that inquiry-based learning works? Turning back to Mayer’s review of the literature, educators seem to confuse constructivism as a theory of how one learns and sees the world, and constructivism as a prescription for how to teach. In cognitive science, ‘constructivism’ is a widely accepted theory of learning; it claims that learners must construct mental representations of the world by engaging in active cognitive processing (i.e., schema construction). Many educators (unfortunately including professors in colleges of education) have latched on to this notion of students having to ‘construct’ their own knowledge and assume that the best way to promote such construction is to have students discover new knowledge or solve new problems without much guidance from the teacher. Unfortunately, this assumption is both widespread and incorrect. Mayer calls it the ‘constructivist teaching fallacy’. Simply put, cognitive activity can happen with or without behavioural activity, and behavioural activity does not in any way guarantee cognitive activity. In fact, the type of active cognitive processing that students need to engage in to ‘construct’ knowledge can happen through reading a book, listening to a lecture, watching a teacher conduct an experiment while simultaneously describing what he or she is doing, etc. Learning requires the construction of knowledge. Construction is not facilitated by withholding information from students.
After a half-century of advocacy associated with instruction using minimal guidance, it appears that there is no body of sound research that supports using the technique with anyone other than the most expert students. Evidence from controlled, experimental (AKA ‘gold standard’) studies almost uniformly supports direct instructional guidance rather than minimal guidance for novice to intermediate learners. These findings and their associated theories suggest teachers should provide their students with clear, explicit instruction rather than merely assisting students in attempting to discover knowledge themselves.
Download a PDF version of this issue.
This is a condensed version of the article ‘The case for direct, explicit instruction’ written for American Educator by the original authors which itself summarised parts of the original article ‘Why minimal guidance during instruction does not work: an analysis of the failure of constructivist, discovery, problem-based, experiential, and inquiry-based teaching’ by Kirschner, P. A., Sweller, J. and Clark, R. E., originally published in Educational Psychologist 41 (2) pp. 75–86.
Mayer, R. (2004) ‘Should there be a three-strikes rule against pure discovery learning? The case for guided methods of instruction’, American Psychologist, 59 (1) pp. 14–19.
Klahr, D. and Nigam, M. (2004) ‘The equivalence of learning paths in early science instruction: effects of direct instruction and discovery learning’, Psychological Science 15 (10) pp. 661–667.
Clark, R. E. (1989) ‘When teaching kills learning: research on mathemathantics’ in Mandl, H., Bennett, N., De Corte, E. and Friedrich, H. (eds) Learning and instruction: European research in an international context, volume 2. London, UK: Pergamon, pp. 1–22.