Book Review: How Learning Works
As promised, I review and point-by-point summarize How Learning Works: 7 Research-Based Principles for Smart Teaching by Susan A. Ambrose, Michael W. Bridges, Michele DiPietro, Marsha C. Lovett, and Marie K. Norman (2010), hereafter HLW as I scratch in futility at the sprawling length of this post.
The authors aim to provide “a bridge between research and practice” for teaching and learning, very much in the spirit of Practical Advice Backed by Deep Theories. They concentrate on widely-supported results that are independent of subject matter and environment, so while the discussion is directed towards instructors in K-12 and college classrooms, there are also implications for essentially anyone in a teaching or learning role.
Let me restate that a little more strongly: any student, autodidact or not, would be well-served by internalizing the models and recommendations presented here. Teachers have even less of an excuse not to read the book, which is written very clearly and without sinking to punchy popularization. This is basic stuff, in the best possible way.
Sure, there are more sophisticated ideas out there; there exist subgenres of domain-specific research (especially for math and physics education); you can find diverse perspectives in homeschooling communities or in philosophy of education. There’s even some controversy in the depths of the research on some of the points in this book (though for the most part the scope of disagreements is still contained within the boundaries drawn by the authors). But as far as most people need concern themselves, HLW is an earnest and accurate if not quite comprehensive account of What We Know about learning.
[I do wish there were a similar account of And How We Think We Know It, looking into common research techniques, metrics of learning outcomes, systematic errors to guard against, reliability of longitudinal studies, statistics about replicability and retractions, and so on, but this isn’t it. The book lightly describes methods when it sees fit, and my scattered checks of unfamiliar studies leave me fairly confident that the research does in fact bear the claims the book makes.]
The book organizes research on teaching and learning into seven principles in order to “provide instructors with an understanding of student learning that can help them (a) see why certain teaching approaches are or are not supporting students ’ learning, (b) generate or refine teaching approaches and strategies that more effectively foster student learning in specific contexts, and (c) transfer and apply these principles to new courses.” The principles are
Students’ prior knowledge can help or hinder learning.
How students organize knowledge influences how they learn and how they apply what they know.
Students’ motivation determines, directs, and sustains what they do to learn.
To develop mastery, students must acquire component skills, practice integrating them, and know when to apply what they have learned.
Goal-directed practice coupled with targeted feedback enhances the quality of students’ learning.
Students’ current level of development interacts with the social, emotional, and intellectual climate of the course to impact learning.
To become self-directed learners, students must learn to monitor and adjust their approaches to learning.
Hopefully these ideas are not surprising to you. They are not meant to be; they stand mostly to organize diverse research findings into a coherent model (see principle #2). And if many of those research findings are old news to you as well, I also take that to be a point in favor of the book, and I trust that you will understand why.
Each chapter begins with two stories meant to illustrate the principle, a discussion of the principle itself, a discussion of the research related to that principle, and recommendations that take the principle into account. The chapters are interconnected but stand on their own. If you don’t plan to teach, you might get most of your value from Chapters 4, 5, and 7. There’s some fluff to the book, but not much. My summary, though long, leaves out the stories and examples, useful repetitions and rephrasings, detailed explanations, and specific recommendations, not to mention descriptions and citations of the relevant studies. I do not consider it a substitute for reading the book, which isn’t really that long to begin with.
Before I summarize HLW, I’ll make a couple brief comparisons. Why Don’t Students Like School: A Cognitive Scientist Answers Questions About How the Mind Works and What It Means for the Classroom by Daniel T. Willingham (2009) looks pretty similar, down to the format in which chapter titles ask questions which are then answered by Principles of Learning, followed by a discussion of the principle, followed by recommendations for the classroom. It’s written at a more popular level, with less discussion of actual research and lots more fluff. Only occasionally does it draw connections directly to a study, rather than use that as the chief mode of exposition (as in HLW). Each chapter does have a short annotated bibliography divided into less and more technical texts, which is nice. Willingham comes down strongly in favor of drilling and factual knowledge preceding skill. While that’s something I’ve approvingly polemicized about at some length, it needs a mountain of caveats. In general he optimizes (explicitly, in fact) for counterintuitive punchiness, and it’s not always clear how well-supported his advice really is. The organization and coverage feels haphazard to me, but where he hits on topics covered by HLW, he seems to agree.
The 25 Principles of Learning [pdf] from the University of Memphis learning group is a short document with a similar aim: a few sentences describing each principle, a couple sentences describing the implications, and a couple of references. It covers important points that HLW addresses only indirectly or that it inexplicably leaves out entirely (spaced repetition, testing, and generation effects, for example). It’s worth looking over to fill in those gaps. But it’s really “25 Important Findings on Learning”: it doesn’t give examples, offer very specific advice, or attempt to organize these principles into a causal model of learning. Consider them exercises for the reader.
1. How Does Students’ Prior Knowledge Affect Their Learning?
Students link new ideas and information to what they already know. This can hinder learning in the case of inactive, insufficient, inappropriate, or inaccurate knowledge, but it can also be harnessed to enhance learning.
In some ways this is common sense—for example, in the way a mathematics lecture directly relies on definitions and theorems. A student without sufficient background knowledge might still learn to manipulate the symbols, but with more effort, worse retention, worse transfer, and worse ability to explain. But there are also indirect, non-obvious mechanisms at work, in which background knowledge that is not explicitly prerequisite can help learning (as in general knowledge of soccer enhancing recall of arbitrary soccer-match scores).
Declarative knowledge (object-level concepts) and procedural knowledge (how and when to apply those concepts) do not always go hand in hand. One without the other is a knowledge gap that can be tricky to spot, especially in self-assessment.
Existing knowledge needs to be active to be effective; activation can be achieved with minor prompts and reminders, as well as questions designed to trigger recall.
Students may activate existing knowledge that’s inappropriate (e.g. the colloquial/intuitive meaning of “force” when learning Newtonian physics) or inaccurate. Such activation interferes with learning, leads to incorrect conclusions, and predisposes students to resist conflicting evidence.
Inaccurate isolated facts can be unlearned through empiricism and explicit refutation. Deeper misconceptions can be extremely persistent, but patient repetition and a long series of small inferential bridges can help.
Strategies for teachers:
Determine the extent, quality, and nature (e.g. declarative vs. procedural) of students’ prior knowledge:
Talk to previous instructors
Use diagnostic tests
Ask self-assessment questions
Use brainstorming or concept mapping
Look for patterns of error
Address gaps in prior knowledge:
Identify for yourself what knowledge is necessary
Remediate insufficient knowledge as determined above
Activate relevant prior knowledge
Explicitly point out connections
Use analogies and examples
Use exercises that explicitly ask students to use their prior knowledge
Avoid activating inappropriate prior knowledge:
Highlight the boundaries of what knowledge is applicable, either explicitly or with rules of thumb
Explicitly identify discipline-specific conventions
Show where analogies break down and examples don’t generalize
Help students revise inaccurate knowledge:
Ask students to make and test predictions
Ask students to justify their reasoning
Help students practice using knowledge meant to replace misconceptions
Allow sufficient time
2. How Does the Way Students Organize Knowledge Affect Their Learning?
Developing expertise requires rich connections between various facts, concepts, and procedures, organized around abstract principles and causal relationships. Although an expert does not necessarily build such knowledge networks explicitly or consciously, it is possible for a novice learner to deliberately organize knowledge into expert-style structures, improving learning, performance, and retention.
The optimal organization of knowledge depends on how that knowledge is to be used. Learning physics in a historical framework has advantages and disadvantages when compared with learning the same physics according to physical principles.
Students whose knowledge networks (graphs with “pieces of knowledge” as nodes linked by their relationships) are more densely connected will retrieve their knowledge faster and more reliably, and are more likely to notice inconsistencies and contradictions.
Experts, as a result of their densely connected knowledge networks, process information in coherent chunks where novices process individual bits of information (as for chess positions and circuit diagrams). Memorization of digit sequences can be greatly boosted by hierarchical chunking of subsequences. These facts are seen as related.
Expert knowledge networks have more meaningful connections and deeper organizing principles.
Students learn better when provided with a structure for organizing information. Causal relationships are especially effective organizing principles.
Studying worked examples, analogies, and contrasting cases helps students organize their knowledge meaningfully.
Organize the material:
Create a concept map for the material to be taught
Identify the knowledge organization best suited to the purpose of learning
Enhance students’ knowledge organization:
Explicitly describe the organization of material at each level in the hierarchy of presentation—subject, course, lecture, discussion
Use contrasting and boundary cases
Explicitly point out deep similarities and other connections
Use multiple organizing structures
Expose students’ knowledge organization
Ask them to draw a concept map
Use a sorting task
Look for patterns of mistakes
3. What Factors Motivate Students to Learn?
Students are motivated by the subjective value of a goal and by their expectancy of success. [You may be reminded of the Procrastination Equation, which also describes penalties for impulsiveness and delay.] Students may be guided by different goals, and recognizing this can help you foster their motivation.
Students who pursue learning goals, which emphasize the intrinsic or instrumental value of material, are generally the most motivated and have the best learning outcomes.
Students may also be guided by performance goals, related to their self-image and reputation. These may themselves be performance-approach or performance-avoidant; the former seems to entail a cognitive framework more conducive to learning.
Work-avoidant goals (“do as little work as possible”) can be directly at odds with learning, but are generally context dependent.
There are, broadly, three broad determinants of subjective value: attainment value (satisfaction from mastery or accomplishment), intrinsic value, and instrumental value. These may mutually reinforce each other.
To be motivated, a student should expect both their own ability to succeed and for success to bring about a desired outcome.
Expectancy of success is influenced by the student’s past success rate in similar situations, and even more strongly by the reasons the student identifies for their past success or failure. Specifically, attributing success to internal and controllable causes* and failure to controllable but temporary causes increases expectancy. Attributing success to luck and failure to personal inadequacy decreases expectancy. [*Interestingly, the authors make no real distinction here between internal and controllable causes for success, which is a fundamental distinction between the “fixed” vs. “malleable” (which you may know as “growth”) mindsets addressed in Chapter 7.]
Supportive environments also increase motivation.
Connect material to students’ interests
Provide authentic tasks
Show relevance to students’ academic lives
Show relevance of generalizable skills
Identify and reward what you (as the instructor) value
Give students opportunities to reflect on the value of their work
Clarify the course goals and your instruction and assessment strategies
Identify and set an appropriate level of challenge
Help students build success spirals with early challenges
Provide feedback on progress
Help students attribute success and failure appropriately
Discuss effective study strategies
Give students flexibility and control in course work to increase both value and expectancy
4. How Do Students Develop Mastery?
Consider a driver changing lanes, making many small motions, visual checks, and mental evaluations fluently and automatically. An expert performs complex tasks with little conscious awareness of the complexity involved. To approach that level of mastery, a novice must not only learn the component skills, but also integrate the skills and know when to apply them.
Experts do not necessarily make good teachers: they process information in chunks, they employ shortcuts and skip steps, they perform with automaticity, and they overestimate students’ competence. Their unconscious mastery leads to so-called expert blind spots.
Students will perform poorly if their component skills are weak.
Student performance is greatly improved when instructors identify component skills required for a complex task and target weak ones through practice. A small amount of focused practice on a component skill can have a large impact on performance of the complex task.
Multitasking degrades performance by way of excess information-processing demands or cognitive load. The same applies to combining skills for a complex task, but much more so for novices than for experts.
Cognitive load can be reduced when learning a complex task by allowing the student to focus on one component skill at a time. It may also be helpful for the instructor to support other aspects of the task while students do their focused practice. This is known as scaffolding.
Another instance of scaffolding effect appears when the instructor presents students with worked examples rather than problems, freeing up cognitive resources to think about principles and techniques.
Results on drilling component skills in isolation, as compared with practicing the overall task with focus on the components, are mixed. Some skills afford isolated practice better than others. A highly complex but easily divisible task can be learned more effectively by initially practicing the components in isolation, and then progressively combining them.
Mastery also involves knowing when to apply learned skills outside of the learning context. Doing so is referred to as transfer. Transfer occurs rarely and with difficulty, and is worse the more dissimilar the learning and transfer contexts.
Overspecificity and context-dependence of knowledge hurt transfer; deep understanding of principles and relationships helps transfer. The latter effect can be targeted with structured comparisons and analogical reasoning also help transfer.
Minor prompts and reminders facilitate transfer, much as they help activate appropriate knowledge (see Chapter 1).
Expose component skills:
Map out your own expert blind spot
Enlist help from those with mere conscious competence
Talk to others in your discipline
Talk to others outside your discipline
Explore educational materials
Reinforce component skills
Focus students’ attention on the key aspects of the task
Diagnose weak or missing component skills
Provide isolated practice of those skills.
Build fluency and facilitate integration of skills
Give students practice exercises explicitly to increase automaticity
Temporarily constrain the scope of the task
Explicitly include integration in performance criteria
Discuss conditions of applicability
Give exercises explicitly about conditions of applicability
Provide opportunities to practice in diverse contexts
Use hypothetical scenarios for practice questions
Ask students to generalize to abstract principles
Identify deep features using comparisons
Prompt students to retrieve relevant knowledge
5. What Kinds of Practice and Feedback Enhance Learning?
Practice is often misguided and feedback poorly timed, insufficient, or unfocused. To be effective, practice should be directed by goals and coupled with targeted feedback.
Learning can be predicted by time in deliberate practice, which is marked by being directed toward a specific goal and an appropriate level of challenge. [I’ve often heard deliberate practice described with an emphasis on mindful attention, in contrast with practice in a flow state (for example in an article by Ericsson himself—the last paragraph before “Future Directions”), but the authors questionably suggest that flow is a sign of appropriate challenge. For motivation, perhaps it is, but I would argue not so for deliberate practice.]
Clearly specified performance criteria can help direct students’ practice.
Learning is hampered by either insufficient or excessive challenge.
The success of individual tutoring is largely driven by the ability to tailor challenges to a level appropriate to deliberate practice.
An instructor can improve learning outcomes with difficult tasks by adding structure and support to bring it within the bounds of the student’s competence. This can consist of guidance by the instructor, or of written prompts and checklists. (C.f. “scaffolding” in Chapter 4.)
The benefits of deliberate practice accrue gradually with increasing time spent practicing; both students and teachers underestimate the time needed.
The effectiveness of feedback is determined by both content and timing. It should communicate progress and direct subsequent effort, and it should be supplied when students can best use it.
Feedback that identifies specific items that need improvement will aid learning more than will a mere indication of error.
Unfocused feedback can be counterproductive by overwhelming the student and failing to direct effort well.
Generally, more frequent and more rapid feedback is better for learning. Delayed feedback can be useful in helping students learn to recognize and correct their own errors.
Be explicit about course goals, and phrase them in terms of capabilities rather than knowledge
Use a rubric to communicate performance criteria
Give contrasting examples of high and low quality work
Progressively refine goals
Encourage deliberate practice:
Assess prior knowledge to set an appropriate challenge
Create many chances to practice
Build scaffolding into assignments
Set expectations about practice
Look for patterns of errors
Use prioritized feedback to direct student efforts
Give feedback on strengths and weaknesses
Allow frequent opportunities for feedback
Provide feedback at the group level, potentially in real-time
Require peer feedback on assignments
Require students to describe how they incorporated feedback
6. Why Do Student Development and Course Climate Matter for Student Learning?
People vary not just intellectually, but also socially and emotionally. Students’ identities may be entangled with the course material and environment in complicated ways that often go unrecognized. A student’s entire state—not just the intellect—interacts with the social, emotional, and intellectual climate of the course to impact learning, for better or for worse. [When I saw this chapter title, I had a vague worry that it would seem out of place, a perfunctory nod to diversity studies or something. I’m still not entirely comfortable with parts of the treatment here, but the above premise is sound.]
The research involved in this first section is of a different nature from the rest of the text. In the first part, the authors seek to describe student development, and cite a model which characterizes developmental changes into seven dimensions: developing competence, managing emotions, developing autonomy, establishing identity, freeing interpersonal relationships, developing purpose, and developing identity. They then cite research characterizing intellectual developments in terms of stages: duality, multiplicity, relativism, and commitment. Similarly, stages for social development. The point is that people can have a lot of different implicit and explicit beliefs, modes of communication, and ways of processing new information, which they can’t just switch off and homogenize when they enter a classroom, and that people have done a lot of work to attempt to enumerate and connect these things. [I think the discussion here is the weakest part of the book, and I’d be interested in better resources on the subject, if they exist.]
For course climate, they describe a classification in terms of whether an environment is marginalizing or centralizing (describing how the perspectives of groups might be discouraged or welcomed), and whether this occurs implicitly or explicitly. Implicitly marginalizing classrooms are the most common of the four quadrants.
In implicitly marginalizing environments (i.e. without overt exclusion or hostility towards outgroups), individuals may suffer an accumulation of micro-inequities that over time has a large impact on learning. A number of studies have found that perceptions of a marginalizing climate are negatively correlated with learning and career outcomes. The authors identify four important channels for marginalization: stereotypes, tone, faculty-student interactions, and content.
The activation (in the sense of Chapter 1) of stereotypes can influence learning, generally impairing performance; this effect is known as stereotype threat. The activation does not have to be a result of explicitly invoking the stereotype; implicit communication of assumptions or apparently innocuous comments also have effects.
The immediate mechanism for stereotype threat seems to be a disruptive emotional reaction; this as opposed to self-efficacy or self-esteem being depressed or otherwise brought in line with the stereotype. The effect does not require any belief in the stereotype. There are deeper nuances as well as strategies for mitigating the effect in the literature.
Perceived hostility or expectations of failure in stereotypes can decrease motivation and drive students from a discipline.
A positive, constructive, and encouraging tone in discussions and syllabi improves student motivation and behavior. (This in contrast to punitive, critical, or demeaning tone.)
Perceived positive faculty attitudes towards and interactions with undergrads are correlated with higher rates of graduate education and better self-reported learning outcomes. Faculty availability is a major factor in students’ academic decisions.
Course content itself in its orientation towards inclusiveness can have cognitive, motivational, and socio-emotional effects on learning.
Promote intellectual development:
Make uncertainty, ambiguity, and complexity safe
Resist a single right answer
Incorporate use of evidence into performance criteria
Promote social development:
Examine your assumptions about your students
Be mindful of accidental cues regarding stereotypes
Do not ask individuals to speak for an entire group
Recognize students as individuals.
Promote an inclusive climate:
Be a model for inclusive language, behavior, and attitudes
Use multiple and diverse examples
Establish and reinforce ground rules for interaction
Make sure course content does not marginalize students
Use the syllabus and first day of class to establish climate
Set up processes to get feedback on the climate
Anticipate and prepare for sensitive issues
Address tensions early
Turn discord and tension into a learning opportunity
Facilitate and model active listening.
7. How Do Students Become Self-Directed Learners?
As one progresses in academic and professional life, one takes progressively more responsibility for one’s own learning. The jump between high school and college can be especially jarring in this regard. Metacognition, “the process of reflecting on and directing one’s own thinking,” becomes increasingly important, but falls outside the scope of most instruction. Still, to effectively direct their own learning, students must learn and practice an array of metacognitive skills.
One model represents metacognition as a continuously looping cycle of task assessment, evaluation of strengths and weaknesses, planning, execution and simultaneous monitoring, and reflection; all of these five steps are informed by a student’s beliefs about intelligence and learning.
Assessing the task is not always natural or obvious to students (essay prompts are often ignored; learning goals are not always clear).
People are poor judges of their own knowledge and skills, tending to overestimate their abilities more the weaker they are.
Novices spend little time in the planning phase of the cycle relative to experts in physics, math, and writing. Novice plans are often poorly matched to the task.
Students who naturally and continuously monitor their performance and understanding learn better.
Students can be taught to self-monitor, and this also improves learning.
Monitoring alone is not sufficient; novice problem solvers will continue to use a strategy after it has failed (and certainly after it has proven modestly successful and familiar but not optimal).
Students who believe their intelligence is malleable rather than fixed are more likely to learn and perform well.
Moreover, the “malleable” perspective can be promoted by external influences, still leading to better performance.
Promote task assessment:
Be more explicit about assignments than you think is necessary
Tell students what you do not want
Check students’ understanding of the task in their own words
Provide a rubric
Give timely feedback
Provide opportunities for self-assessment.
Have students implement a plan you provide
Have students implement their own plan
Make planning the central goal of the assignment.
Provide simple heuristic questions for self-evaluation
Have students do guided self-assessments
Require students to reflect on and annotate their own work
Use peer review
Promote reflection and adjustment:
Prompt students to reflect on their performance
Prompt students to analyze effectiveness of study skills
Present multiple strategies
Create assignments that focus on strategizing
Promote useful beliefs about intelligence and learning:
Address these beliefs directly
Broaden students’ understanding of learning
Help students set realistic expectations
Model your metacognitive process for your students
Scaffold students in their metacognitive processes
Conclusion: Applying the Seven Principles to Ourselves
The authors turn their principles inward and discuss learning to teach. For the most part this is a restatement of the principles with no particularly new insights in their application to teaching, but there are interesting comments regarding the first few:
Many teachers were formerly atypically successful students, and their prior knowledge can lead to distorted expectations; accordingly, many of the recommendations involve gathering data about the students.
The organization of this book into principles is itself a deliberate application of the second principle.
For motivation, the authors try to connect the content of the course with what every teacher really cares about: efficiency. They also suggest focusing on one or two aspects of teaching in a given semester, in order to build up small successes in improving teaching.
In terms of mastery, practice and feedback, climate, and metacognition, teaching is not so different from any other skill.
HLW has eight appendices on tools mentioned throughout the book, with a reiteration of their nature and utility, and most importantly, example checklists and worksheets. These are
Ground rules [for discussion]
Exam wrappers [for promoting metacognition on graded exams]
Reader response/peer review
These alone would have been an improvement over most teaching materials I grew up with.