Frank Noschese recently pointed out on the Modeling Physics listserve an article, Putting Students on the Path to Learning: The Case for Fully Guided Instruction”, by Clark, Kirschner, and Sweller in the Spring 2012 American Educator about why Guided Instruction is the best for most students. My favorite quotes:
A number of reviews of empirical studies on teaching novel information have established a solid research-based case against the use of instruction with minimal guidance. Although an extensive discussion of those studies is outside the scope of this article, one recent review is worth noting…
The article goes on to discuss a review of other studies comparing totally unguided learning with guided forms of learning, even though the authors initially concern themselves with teaching in which learners “must discover or construct some or all of the essential information for themselves” (emphasis mine). One review is the best evidence that we expect?! I had hoped to find some information to refine my model of how students learn, namely under what circumstances different forms of learning work best. To be fair, some of this is found in the article, but the descriptions of the research are lacking. I can only conclude that he means to test discovery learning by means of his example. I would rather they just tell me accurately what the research says than to go hunt through the footnotes, which I find to be an inefficient waste of my time. Well-played, sirs!
Before I read, I’m going to list my prior conceptions of learning. My current understanding of guided inquiry indicates the following features of learning.
- Students retain more when they are initially expected to work with what they know. Highly interconnected knowledge is more likely to be committed to long-term memory. When students engage a problem with their prior conceptions, they search their brains for existing mental coat hooks on which to hang new knowledge. Interconnected knowledge makes learning more coherent and helps students to organize their knowledge.
- Students inured to schooling do not venture willingly into the uncertain realm of inquiry. Students who have proven themselves with traditional modes of learning (i.e. lecture and reading) do not initially enjoy it, and students who are struggling with prerequisite concepts may not benefit from it.
- Students are more motivated to learn something abstract when they need to know. The rest of the teaching world insists that best practice is to pre-teach and front-load vocabulary. In Modeling, we do not apply the word to its meaning until there is a compelling reason to do so.
- Extending the lessons of vocabulary (a verbal representation), students need repeated exposure to multiple representations of a concept. In the act of translating between representations, students (to borrow Robert McDuff’s CIMM definition of learning) coordinate the activation of multiple parts of the brain. As with connectedness of knowledge, this ensures long-term retrievability.
- Learning via guided inquiry is effective but not efficient. Although the authors argue its efficacy, interactive engagement and Modeling methods have been shown to promote long-term retention, but instructors have to leave out much of the traditional curriculum. However, there are also positive externalities to learning this way, including promoting self-reliance and teaching process skills. Having my daughter figure out the details of how to perform a chore may not be the most efficient way to communicate the subtleties of a task, but people usually bristle at too much lecture on the finer points of sweeping with a broom. There is a balance here. As long as she has been shown the basic mechanics and understands the goal involved, she is ready to begin. There is a Gradual Release of Responsibility here, and I can view it as a mode of direct instruction or a mode of guided inquiry. There are things I want her to discover for herself, and things I want to show her first. By setting her loose, I am building her work-ethic and responsibility too.
- Sometimes breaking concepts into small chunks and teaching each works great. Jump Math is one of my favorite examples of this. They also do it in a way that is fun, by showing students what they can do with what they already know. The method does ask students to try things for themselves, but they are usually little things with which students are guaranteed to succeed.
- Sometimes one has to supply a few tools for thinking, and students cannot help but think. CIMM (Cognitive Instruction in Mathematical Modeling) is my favorite example of this.
- Whatever they are learning, students need feedback. In Modeling and many inquiry methods, the feedback comes from guiding questions, which place greater cognitive demands on students. Students also give feedback to each other, which means that they have to listen to each other. Depth of processing is key. One of the only ways to get students to listen to each other is for the teacher to stop talking, and while this is necessary, it is not sufficient. If students aren’t taught or guided how to listen, they will feel frustrated when the teacher doesn’t answer their questions directly.
- What students think about (depth of processing) while they read or learn is important. Classroom culture is a good way to build this expectation.
- Just because students do not become experts at acting like scientists without ten years of experience doesn’t mean we should wait to teach these skills for five years.
- Observation is theory-laden, that is, interpreted through preexisting schemata. Students will misremember or misobserve to fit their prior conceptions. Thus, teachers must engage with student prior conceptions either before or after teaching new concepts. Not addressing misconceptions is not an option.
Like my students learning something new, my learning about learning and teaching is fragmented. Right now it is a list, but as I go on, I must organize it into a more useful and connected form. I’ve been experimenting with concept-mapping in my Earth & Space Science class, so perhaps sometime soon I’ll build connected models of teaching, learning, and practicing through this medium.
Back to the article, which notes that transfer is a problem for minimally-guided methods. It is no wonder. Transfer requires some guidance. Students must be guided how to abstract away differences and build lasting edifices of knowledge. In some circumstances even, abstract reasoning should be taught before particular concrete examples (See http://www.sciencemag.org/content/320/5875/454.figures-only/reply and the interesting comments).
A second point concerns whether discovery-learning actually creates misconceptions. Although we want students to use the scientific method as a general tool for dispelling their own misconceptions, it would be helpful if students did not have so many. I tend to blame an incoherent teaching of elementary and middle school science, which in the U.S. has been neglected for over a decade. Teaching students concepts without application and for which they are not ready is the real harm. When students know the words for big scientific concepts or have memorized Newton’s Laws by heart but can’t explain them, we know that we have overreached student abilities. Teaching something poorly early is more damaging than having students practice observing and measuring in the early grades. To be sure, feedback must come early enough that students do not form lasting misconceptions before they can be corrected.
The achievement gap problem has a simple solution: Allow students who are frustrated to meet for quick direct instruction and practice. The long-term benefits of this type of metacognition are clear for self-regulation. Because it is clear that some struggling students will choose to work with their friends anyway, one can make it mandatory at first using formative assessment results to choose the target population.
I think I will have to read more about Mayer’s “constructivist teaching fallacy” (Note: Look up “Three-Strikes Rule” and “Constructivism as a Theory of Learning versus Constructivism as a Prescription for Instruction”). I would agree that a theory of learning cannot be immediately applied to teaching. I would not say that I am constructivist in my teaching but rather that I want students to know how to make science work for them. In the grand scheme of things, I have no doubt that I will teach my students wrong and/or useless ideas. However, if in doing so, I have taught students how to think and learn for themselves in even the most challenging situations, then I have succeeded. Might there be some connections to a growth mindset here?
The limitations of working memory make it all the more important to help students assemble their knowledge of the world into working-memory-friendly chunks. Since memory works by “forgetting” the useless details, is it possible with direct instruction to control what details students consolidate and forget? I like how the MIT RELATE project‘s MAPS (Modeling Applied to Problem Solving) SIM model and Model Hierarchy, which I think scaffolds student behavior in a good way and reduces demands on working memory.
I would also like to learn more about the worked-example effect (See the articles end notes, 29-32). It could be used within Modeling during the Model Deployment phase. To unburden working memory also seems to call for clever use of a problem-solving strategy and keeping track of steps on paper.
Update: The Twitterverse comes through with more details. @jybuell refers to a comment by Richard Hake on @ddmeyer‘s blog, which refers to a previous response to Kirschner and Swell by Hmelo-Silver, C.E., R.G. Duncan, and C.A. Chinn. They make many of the points I made, only much better. Conflating minimally guided instruction like discovery learning with more guided modes of inquiry is no good. IL (Inquiry Learning) and PBL (Problem-Based Learning) provide many forms of scaffolding for students that address the working memory concerns and offer positive empirical results.