Testing, Memory and Learning

Some Pain, More Gain

• The author says that memory research informs classroom teaching and ultimately makes them a better teacher.
• In January 2021, the author dreaded teaching virtually due to the pandemic.
• Traditional tests were rendered nearly impossible due to online instruction and cheating.
• Testing can drive learning, not just measure it.

Testing and Memory Retention

• Brain imaging studies show testing on recently viewed material increases information retention.
• Students had a three-day window to take an open-book quiz online every week.
• Weekly quizzes prompted critical thinking and provided opportunities for feedback.
• Quiz topics were revisited on midterms, sharpening knowledge.
• 85% of students "strongly agreed" that weekly quizzes helped them learn.
• Students said quizzes helped them prepare for the midterm and being tested on things that were just learned.
• We are wired to learn from mistakes and challenges. The phenomenon is called error-driven learning.

The Struggle is Real

• The author resolved to write a book and learn how to surf at fifty years old.
• Surfing is humbling and has a steep learning curve that demands core muscles.
• The pain from making mistakes is a big obstacle to learning.

Error-Driven Learning

• Error-driven learning uses errors as learning opportunities.
• Neuroscientists believe skilled movements are learned by observing the difference between what we intend and what we actually do.
• Error-driven learning explains benefits from actively learning by doing rather than passively learning by memorizing.
• Memory researchers know that learning under challenging conditions benefits.
• Training by testing yourself is better than studying.

Testing Effect

• Studying alone provides a false sense of confidence.
• Testing exposes weaknesses.
• Those who were tested had the humbling experience of struggling.
• Students who repeatedly studied retained only half of what they initially learned.
• Students who tested themselves retained over 85 percent.
• The testing effect is reliable.
• "Pre-testing" can be remarkably effective in learning.
• Giving the brain a chance to actively struggle seems to help us learn more and retain that information over time.
• Machines learn a lot when they have to struggle, and humans benefit from error-driven learning.

Hippocampus

• If the model could memorize new information, but it was poor at retaining what it had learned.
• Error-driven learning effectively turbocharged the hippocampus, allowing it to learn and retain much more in the face of interference.

Memory

• Memory does not benefit from mistakes if you have no idea what you did wrong.
• Expert instruction and feedback can be helpful.
• Benefits come from stress-testing connections in the brain by optimizing how and when you learn.
• The better way to learn is to break up your study time into several shorter sessions that are spaced out in time.
• The brain saves space and learns quickly by focusing on what we don't already know.
• Hippocampus model shows that the benefit you get from spacing out your learning comes from error-driven learning.
• When you test yourself, your brain will try to generate the right answer, but the result isn't quite perfect.
• This struggle provides a huge opportunity to learn.
• Memory updating is key.
• Hippocampus generates episodic memories by tying our experiences to a particular context.
• If you keep returning to the same information periodically, spacing out your study periods in between, the hippocampus can continually update those memories until they have no discernible context.
• It makes it easier to access them in any place at any time.
• Repetitively updated memories lose distinctive bits and become easier to access.

Brain Functions

• Error-driven learning explains the dynamic nature of memory.
• Error-driven learning extends into myriad activities in our brains while we are fast asleep.
• Brains are hard at work during sleep.
• Clearing out metabolic waste that accumulates during the day is part of the work that happens during sleep.
• The brain reorganizes memories by seeking out connections between different events we have experienced.
• The brain goes back and forth between at least five different states in a night's sleep.
• Levels of neuromodulators and hormones such as coristol change dramatically during these states.
• SWS (slow-wave sleep) and REM (rapid eye movement) sleep work hand in hand to transform our recent experiences into knowledge that we can use.

Sleep

• SWS, the deepest sleep stage, is most reliably linked to learning and memory.
• There is orchestrated interaction between the hippocampus and the neocortex during SWS.
• During REM sleep, the brain generates its own sensory input and tries to make sense of it in the form of dreams, constructing an alternate reality.
• SWS plays a critical role in solidifying memories.
• Sleep doesn't just strengthen memories.
• Sleep seems to play a specific role in undoing collateral damage that occurs when we retrieve particular events.
• After a night of sleep, people had better memory for events that were related to what they had previously been tested on, or retrieval-induced facilitation.
• During sleep, error-driven learning helps the brain use threads of disparate experiences to weave a tapestry of knowledge.
• Sleep helps us integrate what we have recently learned across different events so we can use that information more effectively.
• Memories for events sometimes become less context-dependent after sleep.
• After sleep, we can better see big-picture relationships.
• Sleep can allow us to convert memory into wisdom. 
• Memories are less context-dependent after sleep.
• Taking a brief daytime nap or rest can help in some cases.

Targeted Memory Reactivation

• Ken's team had people learn information while sounds were playing in the background.
• The subjects took a nap, and some of those sounds were played again while they were in deep sleep.
• The sounds affected their memories for what they had learned from the previous day.
• Targeted memory reactivation can improve learning a video game or a new language and can stimulate creative approaches to solving a problem.
• Reactivating memories can reduce implicit bias.

Spillover

• Memory is an ecosystem of interacting cell assemblies.
• A ripple effect can extend to other memories occupying a similar place in the ecosystem.
• Recalling one memory can make it harder to pull up related memories, a phenomenon called retrieval-induced forgetting.
• Repeatedly testing someone on hammer will strengthen the memory for that word, but conversely weaken the person's memory for screwdriver if it was not tested.
• Recalling real events from one's life can have a similar effect.
• Recalling something can spill over to other memories in a positive way, strengthening both the memory you pulled up and related memories.
• The benefits of the tests spilled over to related facts, retrieval-induced facilitation.
• Tests are incomplete because it's usually not feasible to test on every piece of material covered in a class. Then testing can sometimes be counterproductive.
• If students are encouraged to memorize a bunch of disconnected pieces of information testing on some of those facts could heat up the competition and hurt retention of other facts that were not on the test.

Reshaping Learning

• Error-driven learning can help students memorize information.
• Traditional tests are optimized for last-minute learning that enhances short-term performance at the expense of retention.
• Lasting benefits come from error-driven learning, during which we are not always going to be successful.
• Learn and retain more from the struggle of pushing ourselves to the edges of our knowledge.
• Normalize mistakes and failures and incentivize constant improvement.