Great book based on research, rather than subjective intuition. Although I feel that most of his counterintuitive results were actually things that already feel like the best way of doing things. His initial (and he argues wrong) approach of letting students struggle and find things for themselves always seemed impractical to me and suited only for experts.
Here is my personal summary of things that I liked, which I wrote while reading:
-----
Diagnostic Questions:
• Use often to diagnose required knowledge for current class - can be very quick
• Should be <10 sec to think,
• Should test only one skill (not many),
• Clear & unambiguous
• Impossible to get right with misconceptions
• Each wrong answer must correspond to a certain misconception, so it gives us diagnostic information.
• This teachers uses fingers, one for A, two for B, .... leaves students the time to think and then goes one,two,three vote! Students vote in silence. Then in order asks those for A, then B, then C, etc to explain their reason. Then does a re-vote. Giving the correct answer only after the re-vote. Rub out incorrect answer to avoid misconceptions.
Example-Problem pairs. "Worked example" followed by "Your turn" (this teacher adds the extra of: "Supercharged" problem, if in the middle you ask students to reflect, forcing them to Self-Explain).
• Worked example done in silence first. Then narrating what I did. Students not copying so as to have full attention. Then let students copy it. Then their turn to do a sample problem IDENTICAL to the example (but changing numbers).
• When doing a sequence of examples, use MINIMAL VARIATION between one and the next. Like an experimental method, changing only one thing to see its effect. Also examples seem a coherent whole rather than a random selection of examples. Make slight change so that students predict result. If its wrong, they will remember it much better because they had predicted it.
• Make sure examples include boundary examples! Quirky or unusual examples that will help students eliminate under-generalization and overgeneralization. Throw students off autopilot. Forces students to self-explain to themselves.
• Make sure no ambiguity in the question. Make sure there are not different (wrong) ways to arrive at the correct answer.
• Make sure no ambiguity is allowed in student answers, e.g. draw parallelogram, what if they draw a square? Either they are masters or know nothing, it's difficult to tell!
Deliberate practice:
• Like musicians practising separate bits of piece rather than whole piece at a time, and footballers practising individual skills rather than whole matches. Identify substeps that a novice sees in a single expert step. Practice the sub-steps INDIVIDUALLY and isolated from each other, to master each and all of them. No risk in having questions look very different to final exams. Giving final exam style questions might overload students
• Analogy: struggling to solve a problem (eg. changing a tire for the first time) can be so slow, confusing, and demanding of short term memory, that even if finally done successfully, next time will still be difficult. Learning is much less efficient than if you had a proper led training with deliberate practice of the steps to automate them.
Purposeful Practice (how to avoid boring repetition of similar quick problems):
• Purposeful Practice: lies between a boring series of exercises and a complex problem whose choice of method is uncertain. It is something which requires students to perform the specific task to be practiced (no choice of method) many times, but with an ultimate goal or purpose which makes it feel different. E.g instead of giving a list of fractions to add, give a set of fractions and ask students to find ways to add them to get the closest possible to 1. Novice learners will just practice the procedure of adding fractions while experts will learn new things, new ways of thinking.
Problems with early gains (my label) tips:
• Make a problem in which students are asked to find out everything they can, rather than a final single aim. This helps students feel relaxed, gives early gains and motivation as things are being found out.
Surface vs Deep structure in problems:
• Problems have surface structure (e.g a school trip... or a triangle is...) and deep structure (eg averages, Pythagoras, areas, perimeters). It is most common to batch problems with different surface but same deep structure (DS-SD) but we should also introduce problems with same surface but different deep structures! (SS-DD) so that students learn to recognise the differences in deep structures! Eg present problems that are all on triangles but relying in completely different things: areas, perimeters, ratios, angles, Pythagoras, ... all looking identical.
Errors: hypercorrection effect
• Errors committed with high confidence are more likely to be reflected on and learnt. Realising errors committed with low confidence does not shake the mind as much and has less likely a learning effect. So before showing answers to students, ask them to rate their confidence!
Long term memory
• Students might show amazing performance on the day you teach. Next week it’s as if they knew nothing.
• Retrieval power (ease to remember right now) vs Storage power (long term ease of increasing retrieval power).
○ Studying increases both.
○ If retrieval power is too high, then the act of retrieving will not increase storage power.
○ The lower the retrieval power, the highest increase in storage power when successfully retrieved.
Therefore, students need time to begin forgetting before retrieving!
This has been identified and replicated, "one of the most general and robust findings of experimental research on memory"
The spacing effect
• Dedicate part of each lesson to review concepts from weeks earlier
• Homework should be mostly for earlier topics (author recommends 2/3) and only 1/3 for current topic!
• Teacher’s exams and quizzes should be cumulative
Interleaving of topics
• Having to resolve through interference of topics Forces learners to find similarities and differences and higher level learning.
• Try to have current questions include as much earlier material as possible