Minimal Guidance

This is effectively the second in a series on the forgotten history of educational research; the first being my post on Project Follow Through. I suggest that it is a forgotten history because many teachers and teacher educators are unaware of the debate or think it’s a false debate. I am not convinced by those who claim that there is no debate to be had and I will return to this point later.

I am talking about the publication in 2006 of a paper by Paul Kirschner, John Sweller and Richard Clark called, “Why Minimal Guidance During Instruction Does Not Work: An Analysis of the Failure of Constructivist, Discovery, Problem-Based, Experiential, and Inquiry-Based Teaching”. It led to three rebuttals in the same journal, a response by the authors, a conference and then finally a book, “Constructivist Instruction – Success or Failure.” So it is kind of a big deal.

The paper makes the case for fully guided instruction. In fact, a follow-up piece for American Educator magazine is called ‘The case for fully guided instruction’. The authors argue that information should not be withheld from students. Instead, concepts and procedures should be fully explained to them prior to guided practice. The authors draw on evidence from cognitive science and from educational research that shows that there is nothing to be gained and much to lose by not providing explicit guidance.

Two of the rebuttals, those by Schmidt et. al. and by Hmelo-Silver et. al. basically accept Kirschner et. al.’s points about cognitive load but claim that problem-based learning and inquiry learning do not fit the description of ‘minimal guidance’. This is a little semantic. The purpose of both approaches is for students to try to find out the answers to problems themselves without being explicitly instructed in how to do this. Whatever you call that thing, that’s what Kirschner et. al. are arguing against.

Hmelo-Silver et. al. state that problem-based learning is highly scaffolded. This is what Pea (2004) has to say about scaffolding;

“A theory of scaffolding should successfully predict for any given learner and any given task what forms of support provided by what agent(s) and designed artifacts would suffice for enabling that learner to perform at a desirable level of proficiency on that task, which is known to be unachievable without such scaffolding. Thus, one needs independent evidence that the learner cannot do the task or goal unaided.”

So we are to provide guidance to a learner only when we have evidence that he or she cannot complete the task unaided. That sounds pretty minimal to me, especially when you consider that a learner may be able to complete a task unaided but this may take him or her a long time, might be frustrating and might result in a less that optimal solution.

The third rebuttal is by Deanna Kuhn. She first questions whether knowledge should be taught at all before accepting that, “Of course we want children to acquire some rudimentary understanding of the physical and biological world around them” [my emphasis]. To Kuhn, rapid advances in the sciences mean that we cannot hope to teach all of the science that is out there.

Firstly, as a physics teacher, I wish to impart far more than a rudimentary understanding of the physical world. Secondly, although the frontiers of science are constantly extending, key principles such as Newton’s laws, evolution by natural selection or even the particle theory of matter have changed little in hundreds of years and are certainly worth understanding. Thirdly, Kuhn’s preference for emphasising the teaching of transferable skills  such as ‘argument’ or ‘inquiry’ ignore the fact that such skills are highly dependent on possessing a body of relevant knowledge. Inquiry, for instance, involves formulating a question and an hypothesis. How may we do this in a sophisticated way with only a rudimentary knowledge of the area in question?

People often point me towards these rebuttals. They less often mention Kirschner et. al.’s reply to these rebuttals which deals with the points that I have made above.

There are a few misconceptions that have developed around the case for explicit instruction epitomised by the Kirschner, Sweller, Clark paper. It’s worth clearing these up.

The evidence is broad

The evidence for explicit instruction does not just come from cognitive science and the sorts of experiments conducted by John Sweller on worked examples. It does not just come from Project Follow Through. In fact, the finding that explicit instruction is superior to less explicit forms of instruction has been replicated many times in process-product research and in expert/novice teacher studies.

Barak Rosenshine mentions three types of research from; cognitive science, process-product studies and studies on the use of cognitive supports to learn complex tasks. He states, “Even though these are three very different bodies of research, there is no conflict at all between the instructional suggestions that come from each of these three sources… The most effective teachers ensured that their students efficiently acquired, rehearsed, and connected background knowledge by providing a good deal of instructional support. They provided this support by teaching new material in manageable amounts, modeling, guiding student practice, helping students when they made errors, and providing for sufficient practice and review. Many of these teachers also went on to experiential, hands on activities, but they always did the experiential activities after, not before, the basic material was learned.” [Original Emphasis].

The evidence is not limited to basic rote memorisation

The evidence for explicit instruction is not limited to objectives such as the ‘rote’ memorisation of information. I doubt that it is easy to remember very much by rote – that is, without understanding – because we tend to remember things through their meanings. That’s why memory champions have tricks for assigning meaning to random events – for instance, they might imagine a walk through a house to remember a sequence of cards.

In Project Follow Through, students in the Direct Instruction condition outperformed the other groups in basic skills but also in reading passages full of inferences and in mathematical problem solving. Indeed, the research on complex tasks noted by Rosenshine above shows that explicit instruction in strategies improves performance in these ill-defined areas.

I suspect that if we could state an objective of education and agree that this was trainable i.e. not a fairly stable aspect of personality, and that it wasn’t something we all develop naturally such as walking, then any test we could devise would show that explicit instruction was superior in reaching this objective. There is nothing to be gained by failing to be explicit. As Gregory Yates of the University of South Australia notes, “The following misconception is noted: Knowledge is acquired superficially through direct instruction, but acquired more meaningfully though personal discovery… this idea did not find support in the myriad experiments conducted using Piagetian tasks. Instead, the added burden of attempting to discover or “problem-solve” new knowledge actively can increase cognitive load in such a way as to interfere with learning in novices.”

This is not simply obvious 

Once availed of the facts about explicit instruction, some may claim that this is just obvious; all teachers use it anyway so why go on about it? This is not quite right. Some teacher education texts advise providing explicit guidance only after a period of problem solving. This, in the least, represents a significant disagreement with the conclusions of Rosenshine.

Noted influential writers such as John Dewey and Paolo Freire have argued against explicit instruction; Freire devotes an entire chapter of the influential “Pedagogy of the Oppressed,” to criticising the ‘banking’ model of education which closely resembles explicit instruction. And these writers’ works continue to be popular.

In his influential paper on discovery learning, Mayer notes repeated attempts over many decades to introduce discovery approaches despite little evidence that they work. If the effectiveness of explicit instruction is obvious then why would this be?

Yates confirms this:

“It is notable that data from two empirical surveys challenge directly the notion that research into teaching has produced obvious findings. In separate studies, Townsend (1995) and Wong (1995) presented statements derived from process– product studies to teachers and student teachers, and requested them to rate how obvious the statements were. But in both studies, there was a twist. The item statements were presented in either of two ways, specifically either with the correct finding, or with its opposite. For example, Wong’s ninth item was expressed in two ways: Students were found to get better scores on achievement tests in classes with either (a) more teacher control and less student freedom to select learning experiences, or (b) less teacher control and more student freedom to select learning experiences. In both studies, the overall finding was that participants rated the stated propositions (both the correct and incorrect versions) as equally obvious. In fact for Wong’s item above, the correct form (a) was rated as less obvious that the incorrect form (b).”

So not obvious at all then.

Times are changing

However, acceptance of the evidence for explicit instruction does now seem be becoming more mainstream. In recent months, we have had the release of the Sutton Trust report in the UK and just recently the CESE report in New South Wales. Hopefully, the value of explicit instruction will now start to be appreciated.