Book Summary: Make it Stick

By Brown, Roediger & McDaniel

Learning is an essential part of being human. We benefit from it throughout our lives, not just in academic settings. However, much of what we have been taught about learning, or come to think as effective, can actually be impairing it. The authors (Brown, Roediger and McDaniel) use this book to articulate what the educational research actually tells us about effective learning. This is evidence-based in its nature, going beyond the many purely theoretical frameworks that can be found in the educational literature. As such, the guidance they give is truly life changing in its impact, turning learning into a skill you can do better. This is therefore one of the most important books that I have read and I hope you don't mind the extra detail that is included in this summary. 

The authors define learning as acquiring knowledge and skills so that they are easily accessible for future problem solving and opportunities. There is no denying that learning involves memory, but they are also clear that learning is much more than this (something that may be forgotten with the wrong focus). In addition, many of the skills used for effective learning have some counter intuitive aspects that dissuades learners from utilising them.

Their key claims are:

• Effective learning is effortful

• We are poor judges at what is effective- we often confuse what feels easy with efficacy

• Testing is an effective counter measure to such delusions 

• Retrieval practice, spacing and interleaving are effective techniques for making information stick

• Learning is constructive- you need to have foundational knowledge for more complex learning 

• Elaboration helps to expand learning potential because of making more connections 

• Rereading, highlighting, and cramming are not very effective learning tools

• The idea of learning styles is not supported by empirical evidence 

• Trying to solve a problem before learning about it helps improve understanding of the topic (termed generation)

• Learning is literally changing your brain. Genes do play some starting influence, but a lot is within our control 

Myths

As they note, there is a lot of misinformation throughout our culture about learning. Part of this comes from the system of knowledge conduction that we have, which has both aspects of tradition and dodgy theory. The empirical base is expanding but has been relatively slow to be utilised. A useful starting point is to deal with some of these myths and confusions.

Rereading 

This doesn't appear to be helpful for learning despite its popularity. This has been demonstrated well in several research studies. There has to be something more active about the engagement with material than simple exposure. It is much the same when students highlight text, either in a textbook or notes. A major problem here is that there develops a familiarity with the text which is mistaken for knowledge.

This is false mastery, or the illusion of knowledge. True knowledge is different from simple recognition. A major job of students is therefore to break through this illusion, such as with testing. The challenge here is the cognitive effort that this often entails. 

Creativity isn't special 

There is sometimes a dichotomy described with learning vs creativity. Some people think that true creativity is much more an innate talent that will just emerge spontaneously. However, the authors highlight that you can't be creative without some materials to work with. You need the building blocks of the skill and this comes from learning. In music, this may be the knowledge and proficiency with the different chords and scales, with painting it might be the hours that have been spent trying different techniques. Once you have earned these building blocks, creativity then arises.

How We Learn

As part of their explanation they do a summary of how we think learning happens. First, we take on board a lot of varied information that goes into short term memory. This is tenuous and so we encode it into longer term memory (or at least some of it). This becomes linked to things we know (that is, it is constructed) but is still fairly messy. As such, it undergoes some deepening and cleaning by the brain in a process called consolidation. This process strengthens the connections and makes them more durable. Sleep plays some part here, as may other subconscious processes, but either way, the process does happen over some time period - hours and days. 

However, it is not fixed like this and undergoes updating, in a process called re-consolidation. This is where new aspects of meaning are attached (such as new relevant knowledge) which makes the learning more solid. This is probably what happens with retrieval practice (as we shall see) as we reactivate the knowledge and restore it, sometimes adding new aspects and new connections. The capacity to store knowledge is probably essentially limitless, and actually may increase with added knowledge, as more information is available to make connections with (unlike the computer analogy). 

Retrieval is an aspect of learning that often gets overlooked, as we need to be able to effectively pull knowledge from our long term memory when needed. This involves the attachment of appropriate cues to the learned material. Some of this may be an active engagement/appreciation of how the learning will be utilised. This is separate from how well knowledge is stored. Aspects that affect retrieval include current context, recency of the learning, and the number and vividness of the cues attached to the learning. Indeed, we may want to suppress some retrieval if it gets in the way of current activity. Examples for this include learning a new foreign language, where other language words could interfere (you might keep remembering the German rather than the French). Interestingly, if the knowledge is deeply stored, this won’t have a major impact on future retrieval once the time is right. 

There is a clear link towards effort throughout this model. Easy practice, such as rereading or just doing the same exercise on repeat, is essentially just rerunning the activity through short term memory. It feels easy because it is, but the deeper consolidation is not happening. The activities that make learning harder (spacing, interleaving etc.) are all about bringing the information back out of long term memory and optimising it (adding extra information, removing extraneous connections and improving cues). This means that it the future these pathways will be richer and more stable.

The ultimate result of this is to create better mental models. These are the large scale connections of knowledge that allow expertise. We can only run a limited number of these at any one time so we need to create a library of effective and efficient ones. This is essentially what learning is all about. The process of driving a car is a good analogy. We can clearly understand how we go from poor models of separately using the clutch, gearstick, mirrors etc, into much more effective models e.g. reverse around a corner. The individual components are now integrated into a single better model, and this requires much less working memory space to utilise. 

Having this framework in mind, we can turn to look at the specific components of how we can learn effectively.

The Testing Effect

The testing effect is a fascinating aspect of how we learn and highlights the importance of testing as a component of learning rather than just an assessment of it.

The benefits are twofold:

1. Identifies gaps in knowledge 

2. Actually strengthens learning

Whilst the first point is probably obvious this last part is interesting as it is clear that difficult retrieval is helpful for learning. This actually improves the recall of what is being tested, seemingly by strengthening the connections of the material and by reinforcing the neural pathways involved in retrieval. Indeed, the act of retrieval seems to change the memory, rewriting it in a way that makes it stronger. This does need to be spaced and have some effort, rather than just a repetition of something by rote (which again is just rerunning information through superficial short term memory pathways).

There does still seem to be a lot of resistance to this, thinking that it is just memorisation. However, the authors continue to point out the need to have a store of knowledge to utilise higher functioning processes. Regardless, the evidence now clearly demonstrates that using testing practice helps overall retention. 

There are some nuances to this. One is that feedback on the quizzing is particularly beneficial to retention. Slightly surprisingly this is better if it is delayed. This may be because immediate answers become integrated with the problem and the correction is not fully retained (they use the analogy of stabilisation wheels on a bike- you become dependent on them without realising).

Also, a greater difficulty is usually more beneficial. This goes back to the term desirable difficulty that was first articulated by the Bjorks. As such, open questions or problems are usually better than multiple choice, but both are still better than just rereading. Again, this seems to arise from the neuronal activation that occurs when an answer is being looked for in the brain. Greater activation, in the case of searching for a difficult answer, can yield greater strength of those pathways.

Interestingly, quizzes before any learning are also very useful (within reason). These can flag what the student will need to know, and allow some scaffolding for future learning to hang on.

Mixing it up

The next observation is that mixing up practice is more effective than massed practice. Again, this goes against many intuitions as it feels like massed practice has better results e.g. cramming, repetition of the same skill. In some sense it does, as can be seen on a test the next day, but these are less durable and less applicable. Skills and knowledge learned this way decay faster and are less flexible.

Part of this will come from the spacing of practice here. Spacing allows time for processing of the information in a way that allows new connections and understanding to happen over time, sometimes subconsciously. It also allows some decay and forgetting, with the subsequent recall being more effortful and thus embedding the knowledge or skill more effectively (as per the testing effect above). 

Interleaving is another major factor here. Changing the topic of learning regularly, such as switching problem types, actually impedes initial learning because it may slow down the fluency that arises from repetition of the same task. However, research again shows that retention of the learning is much better after any reasonable delay. Again, the difficulty of this is much more apparent to students and teachers than the long term retention and, as such, it remains unpopular. 

This links into the concept of varied practice. By changing the skill type, habits are avoided and deeper mental models are created. This is laborious but produces better outcomes. It is somewhat against the tradition of repeating a task to get muscle memory, but the evidence is clear. One major advantage is the improved application of knowledge and skills that this gives. Real life doesn't appear with categories predefined so we have to be able to first identify the problem and then work out how to apply knowledge to solve it. With interleaving and variation this is constantly being done. As such, it provides a better development of conceptual knowledge than would be first thought. Indeed, it was thought that this level of knowledge would not be helped, unlike the more concrete factual knowledge. However, this may be because of the constant need to apply knowledge and identity differences, rather than just recognise similarities. Through this, concepts and classifications actually can be seen clearer.

Desirable Difficulty 

The observation through the research literature is that a suitable degree of difficulty in the learning creates better results. This has been termed ‘desirable difficulty’ by the Bjorks. This feels like it is slowing learning down, and it usually is, but it is also making it better. 

Errors

It is interesting to note here how useful making mistakes can be. As has been described, having a hard time with learning is valuable for engraining the learning material. As such, it seems to be the case that making mistakes can actually be helpful for learning, if appropriate correction is given. A caveat to this is if fear of failure (and the subsequent anxiety) takes up a lot of the students mental workspace. This may manifest if the act of failure seems to carry high stakes. However, with the appropriate perspective we can see how failure is simply a natural part of life. I find it well embodied in this quote from Thomas Edison - “I’ve not failed. I’ve just found 10,000 ways that don’t work.” Indeed, failure is an essential part of the scientific method, an approach to finding truth that has yielded much of our incredible world today.

Undesirable difficulties

Whilst there has been a lot of description about desirable difficulty in learning, it is clear that difficulty can actually hinder learning. This may be the case when learning is actually made impossible by the difficulty. An example of this could be trying to learn to read Japanese without any understanding of the different Japanese scripts. You simply can’t get anywhere. Also, as just described, it may be that some degree of difficulty is, whilst still possible, just too much for the learner. This is parallel to the anxiety problem described above - there can be too much of a barrier to overcome. As noted later, some barriers such as dyslexia increase the difficulty but don’t actually bringing learning benefit.

Cognitive Biases

We are naturally designed to make decisions and judgements. This has been a result of our evolutionary history, and we are generally good at it. However, as well documented by Kahneman, there are limits and errors that can occur. We need to be aware of how some biases can creep into our learning so as to avoid making mistakes. As the authors note, we have an inbuilt 'hunger for narrative'. We have a need to make everything a story and dislike it when there isn't one. Randomness or arbitrariness are uncomfortable for us. We bring this into everything by creating narratives where they probably didn't exist. This colours a lot of what we do, and highly influences out mental models. 

This is almost impossible to avoid (I note that even Kahneman didn't feel we had much chance of changing aspects of our thinking). The challenge here is that such connections are essential to our learning and memory, whilst also being potentially misleading. The literature is full of examples of problems that may arise here. Suggestive questioning, for example, can lead to a different recall of events. Completely fictional events can be misremembered as real (and this can even be manipulated). Interference of learning can blend two separate things into one. Even just through repetition, a false concept can become to feel familiar, and thus be mistaken as actually true. This could be a problem if wrong answers to testing are not corrected. 

The fluency illusion is major concern for learners, where recognition of a text is mistaken for mastery, as described earlier. As described through the book, these strategies have been developed to cut through this illusion. As such, creating strategies that mitigate other psychological flaws is also an important part of learning. Having an active awareness of them is a good start.

Mental models 

As noted, mental models are a key component of learning. They are the cognitive constructions that can be wielded within our limited mental workspace and yet carry large amounts of information. A challenge with them can be deconstructing them once you have become a master. This is because you have included so much information in it, it can be difficult to remember what the smaller component parts were at earlier stages of learning. This can be a barrier to expert teachers training novices - being unable to understand why their tuition is not making sense.

Another challenge can be to recognise when the wrong mental model has been applied, and to be able to switch. This can occur because of similar starting points, and the cognitive biases that inhibit identification. This may lead on to the unconscious incompetence as described by Dunning and Kruger. This is essentially an observation that unskilled people in a domain can lack a lot of insight into their deficits. This is particularly true at the lower edge of the competence spectrum. This poses a major problem for progression in learning, as it means students may not enact effective strategies, thinking they are doing okay.

As such, a clear role for external, more objective input is important for valid guidance of learning. There are many ways that this can be done:

• Testing

• Peer teaching

• Summarising

These are all valuable forms of objective feedback. Social feedback is also incredibly useful, and can come from teachers or in some situations peers. This approach forms a central component of the apprenticeship model, and also of effective teams. Simulation also incorporates many of these principles, as long as there are efforts to move towards fidelity. 

Individual Differences

There has been a lot of focus in educational circle about learning styles (visual, auditory etc) but this seems poorly supported by evidence. Instead, individual differences of application of learning are very relevant. This refers to how we are converting our learning experiences into the relevant mental models that we will be utilising. Reading ability is a major factor here, as difficultly here is not desirable, as with some learning. Problems such as dyslexia can just slow learning down and force new strategies to develop 

Learning styles matching has no evidence currently. Indeed, some research is frankly contradictory. The concern is that this can negatively affect someone's approach to learning, by making them believe a certain approach is pointless e.g. Reading. What is clear is that matching the learning style to the learning material (not the learner) is relevant. For instance, practical skills are likely best learned practically rather than from a book.

Intelligence is another key relevant factor, and one which psychologists and educators are still trying to understand. A useful differentiation is between fluid and crystallised intelligence. Fluid refers to the general problem solving ability and processing information. Crystallised refers to the stored knowledge and developed mental models of the individual. 

The multiple intelligences theory (by Gardner) is another useful analysis, describing how intelligence may actually be quite domain relevant. Challenges here remain the lack of empirical evidence, but there seem to be fewer negative impacts from looking at learning through this lens (in contrast to some other theories around intelligence). 

Sternberg has described another model of categorising intelligence: analytical, creative, and practical. This does have some more empirical support but is still early on. Sternberg is one who therefore proposes the concept of dynamic testing. This idea is to regularly assess and use it to identify weaknesses that can be focused on. This is in contrast to much other testing where the goal is to document strengths and thus is much more like the learning we do regularly in daily life. 

Structure building 

This is one proposed reason for differences in learning capacity. Some people are better at building the cognitive structures (mental models) for the desired topic. This is probably some skill itself as it involves distinguishing the relevant information from the irrelevant and being able to create the overall story. This seems to be about asking the right questions and reflection on learned material. There may be some link here with those that can identify key rules in things rather than just collating examples. Again, the idea is by exposure to the learning strategies described these skills will also be honed.

Improve Yourself

It is clear that we are a lot more plastic than we previously thought. Intelligence is much less fixed than we used to think, although there clearly is some genetic and environmental influence. This means that we have quite a lot of ability to improve and upgrade ourselves.

John T Breur provides a useful summary of the changes that we undergo in our brain development. Our synapse actually peak at age 1 to 2 and then undergo a pruning process to create our adult brain. This plasticity continues throughout life. This has been profoundly demonstrated in the ability of the brain to learn now sensation after certain injuries - entire sensory deficits can be partly replaced.

One question is therefore whether we can 'train' our brains in a general way. There is only one study looking at this brain training and it is not compelling. It doesn't seem that we can make the whole brain 'stronger' by exercises, like a muscle, as improvements are focal to the pathways utilised.

However, there is the observation that some small differences can compound. This has been noted before by Richard Nisbett in what he terms 'multipliers'. It seems that there are these cognitive multipliers available:

• Growth mindset

• Practice like an expert

• Constructing memory cues

Growth mindset

Carol Dweck has done a lot of work looking at the phenomenon of the growth mindset. This is the fact that the effort put in is a key variable in outcome. Thinking that a lot of achievement is innate can actually stymy both high achievers and low achievers. As such, a mindset that focuses on growing can transform the outcomes open to the individual, probably because it represents an active embrace of continuous learning.

Memory cues

Creating memory cues is another very effective technique. These are skills for retrieval, not for learning, and so are considerations after the material has been explored and learned. There are a few different ‘mnemonic devices’ for this. These are often based around the superior visuospacial capacities of the mind. We have evolved to have this system particularly effective (perhaps a throwback to our hunter-gatherer days) so converting other knowledge formats into it allows better retrieval. 

One example is the memory palace. This involves the imagination of a well known physical space which is then populated with highly memorable imagery that is related to the learning topic. When recall is desired you can then ‘walk around’ the memory palace in your head, which is surprisingly easy to remember. The imagery that you placed there will then be able to prompt you to remember the desired information.

Encoding systems are also used that provide a method for turning non visual info into visual. Memory champions have done this for things such as numbers and playing cards. For instance, each card in a pack of cards can be allocated a specific image. The act of memorising this is initially hard work but once the system is learned it can be relatively easily employed. In the cards example, the images can be placed around a memory palace and recalled in order, thus allowing memory champions to recall the order of entire decks of cards (sometimes more than one).

Summary

As a summary, we can see how such findings can be applied to students. First is to start with the reminder that effective learning can be hard, and that this is a sign of striving, not of failure. Striving is the path to excellence and should not be feared. Importantly, there is a huge capacity for self development through learning. Most of the limitation comes from how you limit your learning. 

They highlight the key strategies of:

• Testing

• Spacing

• Interleaving

They also promote:

• Elaboration 

• Generation 

• Reflection 

• Calibration 

• Mnemonic devices 

Similar advice is dispensed for higher learners and teachers. Ultimately, these strategies have a strong evidence base and fit into a larger model of learning that we are increasingly understanding. Some of them may initially seem counterintuitive, and at times even counterproductive, but this does not offset their effectiveness.