Recently I dropped into an online discussion of Ten Steps to Complex Learning, about which I wrote a series of posts some years ago. Among the four components of its learning blueprint were:
Procedural information, which helps when you’re dealing with skills that you use pretty much the same way each time. You can think of these as the routine parts of larger tasks, like knowing how to navigate Excel and create formulas, as opposed to figuring out what you need to solve.
Supportive information that helps when you’re working on skills you apply differently to different problems. This includes things like mental models and cognitive strategies for whatever domain you’re working in.
In a learn-and-perform context, you can make use of procedural information while carrying out some task. Think of how-to demonstrations, cognitive feedback, and even job aids. The supportive information that van Merriënboer and Kirschner talk about, though, isn’t something you can rely on in mid-task; you work with it beforehand, or afterward. And you build it up in part through the practical application–you can’t practice theory.
Speaking of practice, as someone who likes to cook but is by no means a chef, I rely on recipes. And I often rely on what I’ve learned through cookbooks and videos by Jacques Pépin. Renowned as a master chef but even more as a master teacher, Pépin combines “learning how to cook this dish” with a broader “learning how to cook.”
Here’s a video essay in which he talks about the paradox of recipes:
The recipe is a teaching tool, a guide, a point of departure. You have to follow it exactly the first time you make the dish. But as you make it again and again, you will change it, you will massage it to fit in your own taste, your own sense of esthetic.
I’ve had dinner many times at the home of friends who cook from one of my cookbooks, and I’ve often been amazed at how far away the dish has moved from the original recipe. But it is not necessarily a negative experience; in fact, it is sometimes better than the original, and I end up getting credit and thanks for a dish that had nothing to do with me anymore.
Learning how and when to massage the recipe is part of that bridge from the specific task to the larger construct of principles and relationships in cooking.
Even in his earliest cooking shows, Pépin would underscore that while he was a professional chef – and therefore his skills had been honed by years of practice – his viewers could follow a recipe and begin applying the specific steps to achieve a result.
Along the way, he’d point out variations and considerations. Those can come awfully quickly, but he’s not trying to get you to memorize them; they’re more like highlights you could turn to, or glimpses at the culinary cognitive map in his head.
See what you think. The video essay earlier was about the concept behind a certain dish. Here’s the recipe for braised pears in caramel sauce and (at the 11:00 mark in the embedded video) his own demonstration.
If my hunch is right, a couple of people reading this post will be buying pears next weekend.
As the learning designer, you’re not the one heading to the interview, but you do select the car and choose the route. You not only give directions, but also mark the lanes, erect street signs, and string up the traffic signals. Whatever roads and vehicles the driver encounters, you put there. And if you make mistakes, the driver won’t arrive on time or do well in the interview.
With that happy thought, let’s discuss how to manage cognitive load. In other words, how can you reduce demands on working memory and maximize a learner’s chances of success?”
Research suggests—with a very loud “ahem!”—that multimedia razzle-dazzle can actually work against effective learning. Even background music can interfere with success, the way sound from the car radio makes it harder for you to navigate through a work zone. In “Nine Ways to Reduce Cognitive Load in Multimedia Learning,” Richard E. Mayer and Roxana Moreno explain what they mean by cognitive load and offer a three-part theory for how to make information meaningful.
Meaningful learning: how it happens, how it doesn’t
Multimedia learning, according to Mayer and Moreno, involves delivering information through words (printed or spoken) and images (drawings, photos, animations, videos). By “meaningful learning,” they mean you’re able to apply that information to a new situation.
How that happens, they say, is affected by three factors:
The dual-channelassumption says that we handle incoming information through two channels: one for words and one for images.
The active processingassumption says that we need to do significant mental work in order to learn. We decide what to pay more attention to. And, for things that make the cut, we go on to figure out what they mean and how they interact. This processing is how we create a mental construct for what we’re learning, and connect that construct to existing knowledge.
The limited capacity assumption says that we can only work with so much at a time in a cognitive channel. We can only handle so many words or so many images at a time.
Assuming you’ve done some active processing with those three points, you can already see the implications. Learning is challenging enough; the way we present information through words and images can help or hinder.
Mayer and Moreno also identify five ways overload can happen, and they present strategies to overload. I’ll discuss three here, and two more in the next post.
Overload in a single channel
Imagine that a section of your multimedia lesson has most of its information in a single channel–say, a large block of text. Let’s say it’s all necessary information. In fact, because it’s necessary, you decide to include a voiceover to reinforce the print message.
You’re asking the learner to read and listen at the same time. The two streams of verbal information — printed text and spoken words — compete for working-memory resources and can overwhelm the verbal channel.
Assuming all the information truly is relevant, Mayer and Moreno suggest off-loading some content: move some from verbal to visual. Use images to anchor key concepts, reduce the printed text, and let the audio channel carry the message. “Students understand a multimedia explanation better when the words are presented as narration rather than as on screen text,” write Mayer and Moreno.
Remember our interview candidate? She navigates traffic more smoothly with a GPS that combines spoken directions with a graphic map—far more so than if she had highly detailed, text-only directions.
What happened in the researchers’ experiments? One way to express the strength of an outcome is through “effect size.” Using one common measure, Cohen’s d, an effect size of 0.1 – 0.3 would be small, 0.3 – 0.5 would be moderate, and greater than 0.5 would be significant.* In six experiments involving offloading, Mayer and Moreno report an effect size of 1.17.
* In Cohen’s terminology, a small effect size is one in which there is a real effect — i.e., something is really happening in the world — but which you can only see through careful study. A ‘large’ effect size is an effect which is big enough, and/or consistent enough, that you may be able to see it ‘with the naked eye’.
What if both channels, verbal and visual, have too much essential information? No matter how much you need to cover or how elegant the presentation, too much is too much. When the learner can’t process everything, she can’t organize the input into a useful mental model, let alone integrate it with what she already knows.
Again, our driver trying to make the interview can’t easily cope simultaneously with a nagging GPS, unfamiliar street signs, shifting traffic, and a message board displaying cryptic data about a detour—even though it’s all important.
Mayer and Moreno offer two solutions. One is to segment content; break material into smaller pieces, and allow the learner to decide when to move on. An experiment broke a three-minute segment into 16 segments, linked by CONTINUE buttons. Compared with a control group, students who could choose when to continue, thus taking the time they wanted with the current segment, performed substantially better.
When segmenting won’t work, a second solution is to offer pre-training, which means providing some information ahead of time, such as the names or functions of major parts. In order to build a mental model of what you’re learning, you need a component model (how each major part works), and a causal model (how the parts affect each other). Pre-training gets you to the component model faster so it’s easier to construct your causal model.
Suppose our interview candidate has traveled to Washington, D.C. Before she gets her car, she might learn the different names for the most important freeway (I-495, I-95, the Beltway) and the meaning of “Inner Loop” and “Outer Loop.” That could help her negotiate the trip from Dulles airport to Bethesda.
Part 1 (here) deals with how we process information through two channels (one for words, one for images), and how overload can occur in one channel or in both.
Overload from extraneous information
(Spoiler alert: “Nice to know” doesn’t mean “good to include.”)
Mayer and Moreno point out that “interesting but extraneous material” takes up cognitive capacity. The learner has to pay some attention—for instance, it’s hard to not listen to background music. Effort goes into deciding whether anything deserves further attention. The more this happens, the less capacity remains for learning what actually does matter.
You probably can guess what the researchers recommend: weeding. Remove the extraneous. What’s the bare minimum that people need to know in order to accomplish the skill or apply the knowledge? Force everything else to justify its inclusion.
In an animated sales-call lesson, for example, I don’t need to see the customer driving in. I don’t need an animated phone, virtual pens, and virtual paper clips. I do need a customer statement to respond to. I need time to analyze it. I need clear examples of responses and how effective they are in a situation like the one I’m seeing.
To me, the weeding of nonessential material is the difference between the rich but irrelevant detail of a war story and the crisp relevance of a pertinent example. Our interview candidate probably doesn’t need to know that there’s a library two blocks before she gets to Midcounty Highway; she does need to know when she gets there, the two right lanes are right-turn-only.
Granted, sometimes you can’t edit details out. Suppose you’re explaining how to operate packaging machinery in a pharmaceutical plant. Your learner will confront lots of equipment and lots of steps, along with potentially overwhelming detail in the video close-ups.
When weeding is not an option, Mayer and Moreno recommend is signaling—providing cues to the learner about how to organize the material. So, the lesson might start by breaking packaging into four stages: product into plastic blisters, blisters into cardboard wallets, wallets into carton packs, cartons into cases. In subsequent lessons, arrows or similar highlighting emphasize key components of each stage.
Overload from poor presentation
Sometimes overload results from the confusing presentation of essential information. Imagine an animation in one part of a screen and related text in another. The learner has to shift focus between the two areas, as well as figure out which parts are related to which.
Mayer and Moreno recommend closer alignment of words and pictures. Placing text inside a graphic, rather than alongside as a caption, aligns the explanation more closely with the visual for what’s being explained.
In a related situation, information arrives as animation, onscreen text, and audio narration. The simultaneous presentation of text and narration, which the researchers call redundant presentation, requires the learner to work at reconciling the two verbal forms while also dealing with the visual form. It’s as if our interview candidate were watching an animation of the route to follow and reading directional text while the person next to her recited those directions.
Mayer and Moreno cite studies with a significant shift as a result of reducing redundancy, such as dropping onscreen text and using only narration. An interesting twist they add is that if there’s no animation, students learn better from concurrent narration and on-screen text than from narration alone. The interpretation is that the on-screen text by itself doesn’t overload the visual channel the way it would with the animation there as well.
Overload from “Hold that thought!”
The final type of cognitive overload involves both essential processing and “representational holding.” Mayer and Moreno explain that as having to retain information in working memory. For example, if you read about the thermoforming process for drug packaging, and then watch a video showing the process, you have to keep elements of that text in memory during the video, which reduces your ability to select, organize, and integrate.
One way to avoid this overload is to synchronize—interweave text or audio with the video. Words about the sealing step should arrive as the visual does; a description of the check-weigher should come while the learner sees that device in action.
Researchers cite robust evidence that “students understand a multimedia presentation better when animation and narration are presented simultaneously rather than successively.” Meanwhile, Mayer and Moreno point out that if the non-synched elements are brief—a few seconds of narration followed by a few seconds of animation—there’s less overload, mostly likely because the learner has less representational holding to do: fewer things to keep in mind from the verbal information.
But what if you can’t synchronize? Then, the recommendation is individualization, or ensuring that you have learners skilled at holding things in memory. If for your work you’re able to match “high-quality multimedia design with high-spatial learners,” you’re all set. Personally, I’m rarely able to manage that.
Final thoughts
I started by comparing the multimedia learner to someone who has to drive through a strange city to make an interview. Mayer and Moreno highlight ways that your design decisions can make that trip far more difficult than necessary. Pick up some learning principles and lessons from this research—and take off a little cognitive load.
I’ve worked on a lot of technical training: how to forecast warehouse inventory orders, how to create mockups of proposed software, how to write flood insurance policies. Programs like these are a complex version of working with tools to produce specific results.
To navigate that complexity, to keep on track while designing training, I find it helpful to frame the purpose in a single phrase. And as a phrase, I think this:
Producing retirement estimates using the ViaComp system.
is better than this:
Using the ViaComp system to produce retirement estimates.
The distinction’s clearer if, when asked what the training’s about, you drop the last clause:
Producing retirement estimates. Using the ViaComp system.
What comes first in the longer versions is what’s assumed to be the most important. In English, word order matters. That’s less true in a language like Finnish, where word endings matter much more. Here are some examples, thanks to my friend Riitta Suominen:
Miehen karhu söi… (The bear ate the man [and not the berries].)
Miehen söi karhu… (The bear ate the man [while the wolf ate the rabbit].)
Emphasis in Finnish comes from word order and case endings. In English, to make the distictions clear in print, I have to add visual emphasis.
For those learning goals above, dropping the ending was a way for me to emphasize how distinct they are, the better to choose the right one for the training program. 80% of the time may be spent using the ViaComp system, but what we’re really doing is producing retirement estimates.
Would I ever choose “using ViaComp to produce retirement estimates” as the one-sentence summary? Sure — for example, if we were changing from the Acme Pension Suite to ViaComp. The emphasis would be on how to use ViaComp so you can accomplish the same work as in APS.
Even if the organization is switching systems, though, what matters to the learner (far more than the underlying system) is the work that gets done. And most of the time, you’re not switching. You’re expanding the skills of people who already handle other pension-related tasks, or you’re helping newcomers become productive.
Neither one of those groups cares much what the system’s called, what server it’s on, or similar infra-facts. They do care about their jobs, which means they care about immediate context. How you describe learning goals is an entry to that context, as well as a beacon during the design process.
I’ve been collaborating on a course with my colleague, Tanis. An unexpected benefit has been the ability to float an initial idea, talk about it, and have it improve from the discussion, from feedback, and from new ideas these things engender.
I want to talk about one of those engendered nuggets. I’m a bit hesitant, because when you get to the end, you may well think “Yeah, so?” For me the path was well worth following, and I might not have taken it without the back-and-forth with Tanis. So this is another form of working out loud.
The topic doesn’t actually matter much. If you’re curious, see the following aside; otherwise, just skip past.
About the topic:
Many pension plans allow purchase of service, a way for a person who hasn’t contributed to the plan (for example, during a leave of absence) to pay additional money into the plan. That payment is the purchase. The person then gains credit for the corresponding work time–that’s the service. Purchasing service can increase the amount of your eventual pension.
Different plans have different rules and coverage, and within a plan there are usually several types of purchase of service. You can see typical examples here (for an Ontario plan) and here (a Pennsylvania plan).
Some pension plans use other terms, but purchase of service is the one we use.
The nugget emerged as we juggled three goals for the first part of our course:
Introduce a new type of purchase
Connect this new type to what people already know
Provide a framework to show what the various types of purchase have in common
Employees taking our course would already have learned how to handle certain purchases, like the leave of absence mentioned above. In the new course, they’ll learn the details for purchasing arrears (payment for a period when contributions should have been made to the pension plan but were not–for example, because of clerical error).
Working with internal documents and with our subject-matter experts, we discovered a pattern that seemed to apply at a high level to all purchases:
Circumstances occur that make a purchase possible.
The plan receives an application for the purchase.
Plan staff analyze the application to see whether the purchase is permissible.
Plan staff calculate the cost of the purchase.
There’s a lot more to it, and there are nuances and conditions for each of those, but it didn’t seem like a bad framework. Having laid it out, we could ask participants how a leave-of-absence purchase would fit into this, since they’d already know how those purchases work. Then we could start talking about arrears purchases, to show how at this level they’re like other purchases the participants have worked with.
On second or third glance, though, this version felt abstract. Our plan staff don’t work directly with frameworks; they work with the specific purchases. And so we moved to…
Version two: known → framework → new
In the revision, we decided to start by describing out a leave-of-absence purchase according to our framework: a person goes on maternity leave; she later applies to purchase the service; the staff evaluate the application; we provide a quote for the cost. We’d make sure participants saw how at a high level thus was how the LOA purchase worked. Finally, we’d introduce arrears purchases using the same framework.
This felt better, in no small measure because we began with the specific and not the abstract. And we felt we were doing a better job of connecting to what people already knew.
As we worked on other parts of the course, we’d revisit the intro. Gradually we began to feel that we were explaining for the sake of explaining.
I’ve been in the instructional design field longer than Tanis has, and I feel as though I should have known better. It’s always tempting to try and make things clear. As we poked at this, though, we realized that the key point is not that a leave-of-absence purchase follows these four stages, and so does an arrears purchase.
What was important? Knowing about LOA helps you to learn about arrears.
Version three: known → new
Here’s the sequence we now have–and in the course, the sequence takes much less time than you’ve spent reading this post:
Ask participants to describe the phases of a LOA purchase, from the member’s point of view, in 25 words or less. (We don’t care about word count; brevity encourages big-picture summary.)
Show a diagram with LOA information illustrating our four phases: the maternity leave, the application, our research, the cost estimate. Discuss how the experience of the participants aligns with this pattern.
Redraw the diagram with an arrears purchase replacing the LOA one.
We really like having the participants start by sharing their own ideas about the processes involved in the purchases they already work with. We then show our summary (the LOA in the four phases) and make sure they see their own experience in that summary. Finally, we can start talking about arrears.
So now we don’t belabor the four phases; they’re just stepping stones between the familiar and the new. We’re inviting participants to build the connections that work for them.
What comes next? We use this intro as a springboard to what’s different about purchasing arrears. We ask participants what they think might trigger an arrears. If they already have an idea, great–we can reinforce that. If they don’t, that’s okay, too; their interest level is higher as we move into the explanation.
