Chapter 9 · The Levers That Turn the Resource

We said the direction of learning design is to reduce the share of resource that leaks without aiding learning. To make this direction more workable, we must divide the burden on working memory by its source. Even for burdens that drain the same resource, where they come from differs, and so does what can be done about them.

Two Sources of Load

One burden comes from the content itself. The more representations that must be related at once and the deeper the hierarchy, and the more the lower representations are not in place as units and must be built up on the spot, the greater the burden. This is the floor set by the content to be handled. To someone learning calculus for the first time, calculus is inherently heavy, and that weight cannot be reduced without changing the content. This burden inherent to the content is called intrinsic load.

The other burden, regardless of content, is added by clumsy presentation. The diagram and the explanation of a solution lie apart, so the eye must go back and forth matching them; or the information you seek is scattered in the wrong places, so you spend resource gathering it. This burden arises not because the content is hard but because the vessel holding it is bad, and the resource that leaks that way remains as no learning at all. This is called extraneous load.

The labor of synthesis itself—binding floating representations to build something new—is inseparable from handling the content's inherent weight. So it belongs to intrinsic load: the resource poured into handling intrinsic load is the resource of synthesis, and that synthesis remains as learning. So there are two burdens: the intrinsic load set by the content, and the extraneous load added by the vessel.

Divide it this way and the prescription gathers into one. Intrinsic load is the floor set by the content and cannot be reduced directly—only the lower representations can be put in place beforehand to raise comprehension. So what can be touched is extraneous load: blocking the resource the vessel makes leak, so that resource goes to handling intrinsic load, that is, to synthesis. Learning design is, in the end, this work. The techniques below can each be read as a lever that turns this allocation in a particular direction.

Levers That Gather Resource Toward Synthesis

In teaching math or physics solutions, there is a method of, instead of having learners solve a problem themselves, laying out a finished solution down to its steps for them to study. Made to solve it directly, the learner must each time find the next step from long-term memory (retrieval) and build it anew (formation). But to a novice who has no solution structure yet, that finding is groping from bare ground for a method that is not even there. There must be something to recall for retrieval to firm a path, but the novice has no path to firm yet. That empty groping and building take all the resource, leaving no capacity to hold the solution's structure in the head. Give the laid-out solution and the steps are already in place, so the learner is relieved of the burden of search and retrieval and can spend resource on synthesizing along the structure. The resource that was leaking extraneously has been gathered toward synthesis. But this help is tied to the learner's state. To an expert who already has that solution structure, reading every already-known step and matching it against one's own knowledge becomes excess instead, and what was a foothold flips into extraneous load.

Inserting questions here and there through the material also turns the resource's course. When a question asking "why was that just so" is inserted, the learner's attention turns back to that passage and deep processing that goes back over it happens there. But because working memory's resource is limited, the more the part the question points to is deeply processed, the shallower the part it does not point to becomes. An inserted question does not add new amount of learning but reallocates where the limited resource is spent. What you ask decides what is left deep and what is let slip.

Levers That Make You Spend More Resource on Purpose

If the previous two save resource and gather it toward synthesis, there are also levers that conversely make you spend more resource on purpose to change what remains. When solving problems of several types, the method of mixing the types rather than drilling one type at a time is such a one. Drill one type straight through and the type is already fixed for every problem, so you need only apply that type's procedure. Mix the types, by contrast, and at every problem you must first sort out "what type is this." The added synthesis of sorting the type—discrimination—is laid on. And this discrimination is, in the end, retrieval that pulls from long-term memory the method matching the cue that is the problem. Drill them massed and the method is already fixed and you need only set your hand to it, so that pulling is skipped; mix them and at every problem the method must be found anew from the cue. This finding eats more resource per problem and makes the moment taxing, and in return the problem's features are firmed as a cue that summons the representation corresponding to the solution method. Drill them massed and there is no such pulling, so the burden is small but the cue is not firmed; mix them and the burden is large but the cue is firmed. So if the aim is to tell apart, by cue, types that look similar, mixing is better, while if the aim is to master one procedure smoothly, massing is better. It is a choice of whether to spend resource on the added synthesis of discrimination.

The Feedback That Decides Where Resource Goes

Feedback given after a mistake, too, helps or cuts learning depending on where it sends resource. Give information about the task itself, like "you missed a sign here, so change it this way," and that information is used straight away in the next attempt's synthesis and retrieval and helps learning; resource flows to task processing. Give an evaluation about the self, like "well done" or "is this all you can do," and the learner's resource tilts toward defending and steadying the self, so the share left to process the task shrinks. Under the same name, feedback, one side sends resource to the task and the other to the self. The reason large analyses pooling feedback report that no small amount of feedback actually lowers performance is read as lying in this difference of direction.

All Support Is Relative to State

One thing cuts across the four levers: no support is good or bad in itself; it is relative to the learner's current state of representation. The laid-out solution was a foothold to the novice and excess to the expert. Mixing was beneficial when discrimination was the aim and a hindrance when mastering a single procedure was the aim. Because where resource leaks and where it should go depends on what the learner already holds as units, the same manipulation gathers resource toward synthesis for one person and scatters it in vain for another. So "this method is good" is true only ever with the condition attached, "for whom, in what state."

We have seen the narrowness of working memory and the resource competition, and the levers that turn that resource. But so far we have mostly seen what working memory blocks us from doing. The thing working memory actually gathers resource to accomplish—binding several representations to build something new, building meaning from a text—the product itself we have not yet properly looked at. Having seen the limits, we now look at what working memory actually makes.