Arthur Bakker has written a commentary for a new special edition of the journal, Instructional Science. The special edition focuses on discovery learning and Bakker asks the question: Is discovery learning a zombie, a phoenix or an elephant? I would offer that it may be none of these things. Perhaps another possibility is that it is a rabbit hole, down which things becomes curiouser and curiouser.
Using a common line of argument, Bakker laments the ‘dichotomy’ between discovery learning and direct instruction. Bakker wants to move on from such a simplistic framing. And yet this doesn’t quite work because he is clearly on one side of the debate.
For instance, when Bakker refers to Mayer’s seminal paper on pure discovery learning and Kirschner, Sweller and Clark’s paper on minimally guided instruction, he suggests, “These papers on minimal guidance have received considerable criticism and yet have high citation numbers—a phenomenon worth being studied by sociologists, philosophers, and historians of science.” It’s as if the influence of these papers is utterly mystifying. Bakker is right, of course, that these papers have been criticised. I don’t know about Mayer’s paper, but Kirschner et. al. responded to their critics, quite convincingly. In fact, the debate that the Kirschner et. al. paper provoked led to a conference and then an academic book with many contributors. So the citation numbers are hardly surprising.
Most of those who take issue with Kirschner et. al. take issue with the term ‘minimally guided’. Proponents of discovery learning often claim their procedures provide lots of guidance and ‘scaffolding’. Bakker discusses the way that discovery systems can be designed to limit the opportunity to follow unproductive pathways, for instance. However, a key distinction remains. Advocates of explicit teaching attempt to fully explain new concepts to novice learners. Indeed, if they follow the model of Carnine and Engelmann, they intentionally provide layers of explanation in addition to what might initially seem necessary. No matter how you cut it, advocates of discovery want to leave something out. This is where the paths diverge. And this leads to the fundamental question: If adding guidance to discovery enhances the effectiveness of discovery learning, why not go all the way and use fully guided instruction?
Bakker deploys a number of common arguments as to why discovery learning may be beneficial. The evidence is clear that it generally doesn’t lead to greater learning gains so proponents usually suggest it is motivating. Bakker adds the development of a ‘discovery attitude’ and ‘discovery skills’. He also suggests the following about discovery learning:
“…there is a strong link with constructivism: certain things cannot be transmitted; they have to be constructed or experienced. For those things that cannot be told, it makes sense to think through how educators create opportunities for construction or new experiences.”
I would like to know what these ‘things that cannot be told’ are, because I wonder if they are educational goals.
Bakker gives a hint of what he believes to be the way forward when he dives into ‘inferentialism’, which might just be the latest philosophical idea yoked to the discovery learning wagon. As I understand it, concepts are defined through the inferences you can make with them and this means that everything is connected. It also means that we gradually grasp concepts at deeper levels. As we become more expert with the mathematical concept of standard deviation, for example, we can make inferences we could not previously make. This means we cannot, as Kirschner et. al. claim, ‘fully explain’ concepts and procedures. Perhaps not, but this seems an uncharitable interpretation of what they are suggesting.
Bakker illustrates this by giving an example of how he arrived at a deeper understanding of standard deviation himself. By playing around with data from one of the test questions in one of the papers in the special edition, and after a hint from a colleague, he learnt something new about skewed distributions. Yet I don’t see why the discovery he made is something that could not be taught, or that is better discovered.
In a sense, some of the discussion around discovery is semantic. For instance, I cannot transmit to you the feeling of solving a complex maths problem, even if I may deploy descriptive and poetic language that gives you a sense of what this is like. So perhaps you have to discover this feeling for yourself. But the best bet for discovering this feeling is by being explicitly taught how to solve such maths problems. And when you have been taught such problems, you need plenty of practice to become proficient. Is practising problems you have been taught how to solve, noticing similarities and differences and apprehending the deep structure, discovery learning? If so, it’s not how discovery learning is commonly, if ever, interpreted by classroom teachers. But you could make that case, I suppose.
The key question is this: What educationally desirable goals are best taught through procedures that would be generally accepted as exemplifying discovery learning? To advocate for discovery, it is not enough to show that discovery and explicit teaching can work in tandem. We need to show that a discovery learning procedure is more effective for achieving certain goals than an explicit alternative. Moreover, the experiments that demonstrate this effect must not be confounded and the explicit control condition must teach the same content as the discovery condition. This is not too much to ask and such experiments should be relatively simply to run.
Maybe the papers in the special issue will provide such evidence.