Is Learning 2x in 2 Hours Possible?

How learning works the four steps

Recently, I decided to commit to building in the education space. Since then, I've been reading up on a new kind of school called Alpha School. They claim that their students learn twice as much in only 2 hours a day. My question was: is that marketing, or is that possible?

So in an attempt to both learn about learning, and understand what is possible, I've written the following piece:

When you're talking about history and someone quotes an author, or when you're cooking and need to do some math in your head, it becomes important to have actually learned. So how does information get into our brains? How does learning work?

Put simply, learning goes through a few steps:

Step 1: Information consumption. This is when you read a book or blog post, listen to a song, talk to a friend, or see an advertisement. When any of your senses tell you anything.

Step 2: Store in your working memory. Your working memory is like the RAM on a computer, or your basket at the grocery store. It can only store a little bit and for just a little while. Scientifically speaking, working memory only stores about "4 units." (Basically, you could remember F, C, N, M, or you could remember FBI, CIA, NBA, MLS — regardless, it's four chunks.)

Step 3: Encoding. Once information is in your working memory it can then be encoded/stored into your long term memory. This is like saving a document to your computer or cloud. It is also like the transfer from your grocery basket to your fridge or pantry.

Step 4: Retrieval. This is where the information stored in your long term memory is recalled from where it is stored to your consciousness so you can use it. Like looking at the document in your downloads folder or grabbing milk from the fridge.

Note: this retrieval creates a pathway from where it's stored, so generally the more you retrieve it the stronger that pathway is and the easier it is to remember — like painting. The more coats, the thicker the paint.

So, all of that to say, summarizing and pulling all of the steps above and the learning science below, from memory, is helping me better learn and know the concepts. A little learning and a little investigation into the frontier of education, what will we find?

Now that we know fundamentally how learning works and have it defined, let's look into the 8 most effective science-backed approaches to improving the quality and speed of the learning process, as well as 1 example of techniques bundled and 1 important theory.

1. Worked examples with faded support ~0.5–0.6 SD

What this means is, working on real problems, that in the beginning you couldn't do on your own, with help, and as you improve the help fades slowly.

Said in an acronym, the EDGE method. This is a technique I learned from my time in the Boy Scouts. EDGE stands for Explain, Do, Guide, Enable and is literal steps for a teacher, guide, coach to take. Here's an example from my life:

I run a pickup soccer league in New York City with anyone from professional players from Europe to guys that are just getting back into it. Sometimes there are neighborhood kids playing on the field or running around and for some reason many of them come up to me. Often, I'll invite them to play with us, and in doing so comes many opportunities for coaching.

My favorite opportunity is encouraging the kids to push my adult friends more. First, I will Explain that using your forearm against the opponent's hip is both legal and one of the best places to push them because it is a center of balance. Then I will Do it to the kid. After that I will grab their arm and Guide them in pushing me. But then the real smiles begin, and I enable them to push all the big kids.

In doing so I take them from having aggression and knowing it's important to being enabled to use it effectively, and push around adults :) Good fun.

(SD ~0.5–0.6 from these studies: Sweller & Cooper 1985; Renkl & Atkinson 2003 on fading)

2. The testing effect ~0.50 SD

The testing effect, but it's more aptly called a quizzing effect. This is low stakes, high frequency quizzing that practices step 4 of learning (retrieval) as opposed to going over what you read or wrote.

A good example of this is how I improved my CCAT (Cognitive Aptitude Test) score from the 85th percentile to the 98th percentile. The CCAT is supposed to be a test of how fast you process and think, but my problem was that I hadn't been in school doing math or vocabulary in over 5 years, and hadn't even learned my prime numbers or practiced my multiplication tables well enough when I was in school. I was missing fact fluency.

In order to increase my score, I needed to practice solving the kinds of problems that are on the test, quickly, and relearn my facts. Said differently, create and strengthen the pathways of retrieval for the information and techniques required on the test.

When I would finish one round of theory, i.e. how do I do percentages in my head, I would immediately give myself a problem to solve. This quickly devolved into my girlfriend quizzing me randomly, but I learned quick!

(SD ~0.50 from these studies: Rowland 2014, meta-analysis; Adesope, Trevisan & Sundararajan 2017)

3. Mastery learning ~0.5 SD

In my opinion the most intuitive of all of these techniques. It simply means that students move forward when they know the material, as opposed to the current model that's in majority, where students move forward when they are a certain amount of "older."

An example I like here is most, if not all, professional sports leagues. They don't care if you're 17 or if you're 40, they just care if you can compete. Personally, in cross country I raced kids that were 1-5 years older than me, and when I was playing soccer I loved to play with kids older than me, and nobody minded if I was good.

(SD ~0.5 from these studies: Kulik, Kulik & Bangert-Drowns 1990, meta-analysis)

4. Spacing ~0.4–0.6 SD

This talks specifically about the space in between lessons on a subject, skill, or concept. Basically, the harder it is to recall a piece of information the more it strengthens that pathway, unless you can't remember it of course.

An example of this effect working negatively against me is 'cramming' when studying for tests. In the traditional American public school I went to, I used to memorize all of the slides in a teacher's lecture and then regurgitate that information when I was tested. That or I would just google everything I needed to know as I was walking to class, recognize it on the multiple choice test and then forget all of it.

I could have spent the same amount of time I did cramming, but broken up into a little every day, and I would know that information better because it would have had time to fade some before the recall.

(SD ~0.4–0.6 from these studies: Cepeda, Pashler, Vul, Wixted & Rohrer 2006, meta-analysis)

5. Interleaving ~0.42 SD

This is the one to me that I grumble about. It means mixing the order up, and knowing them, not just back to front, but individually.

I love to have things in a nice neat order, step 1, step 2, step 3, I feel free! The problem is that learning things in order, forwards or backwards, is significantly worse for learning than interleaving.

A story about this comes from geography in 8th grade. I had the privilege of traveling a lot when I was young, so I loved to look at maps. When we did the 50 states challenge where we memorize the whole 50 states names and capitals, I would test myself by looking at flashcards of just the blank state shape and needing to remember where it was, the name, and the capital with no reference to any other state. This interleaving and randomly selecting states to study helped me to better know them rather than know them by order.

(SD ~0.42 from these studies: Brunmair & Richter 2019, meta-analysis; Rohrer & Taylor 2007)

6. Feedback ~0.4 SD

When and how do you learn you were right or wrong? Basically, in an optimized learning environment students learn their errors before they practice the errored concept and are not just told they're wrong, but also told a better way of doing it next time. The quality of the feedback, and how it is delivered, matters. (More on the how in my next write-up.)

A good example of this comes from my time coaching sales. We had a script, a simple one, 5 lines that, if followed, could get a sale 1/20 times. If they added in some better body language and talking through objections it could get down to 1/10 times.

One of my early hires and I were role playing their pitch, and they kept looking away from my eyes. In their culture I knew it was respectful, but in America it's suspicious. Immediately I told him what he was doing wrong, and that looking into the person's eyes at the end of sentences is most important if not even the whole time. Not only did I make sure to recognize his culture, but I also gave him a pathway to working up to full eye contact. He crushed his sales that very day.

(SD ~0.4 average from these studies: Kluger & DeNisi 1996, meta-analysis — the same analysis found roughly one third of feedback interventions made performance worse)

7. Tutoring ~0.37 SD

This one has some controversy. Although it's sure to help learning move faster, the question is, how much faster? There was a paper published in 1984 by a guy named Benjamin Bloom, and although him having a cool name is not up for debate, the findings of his paper are.

What his paper claimed was that the average tutored student performed in the 98th percentile, but the findings haven't been replicated in 40 years. Regardless, tutoring studies have consistently replicated around .4 SD above the average (or about 6 IQ points).

An example of this is my struggle in Pre-Calculus my junior year of high school. I had some compounding problems stemming from the change to Common Core, and never mastering some fundamentals before moving forward to these more advanced math courses. So, while I continued to be more confused by my Pre-Calc teacher and the way he would skip steps while teaching, I worked with a tutor after school set up by the family I was living with. I went from a C to an A in that class.

(SD ~0.37 from these studies: Nickow, Oreopoulos & Quan 2020, meta-analysis of 96 randomized trials — not the 2.0 claimed by Bloom 1984, which has never been replicated)

8. Self Explanation g ~0.55

This means having students explain why they got the answer, how they got the answer, why this material matters, and other such justifications.

My story for this takes me to an event I did recently called a Pithy Party. Much unlike a pity party, this was focused on exciting ideas and teaching them. Each person brought an idea, opinion, or finding and had to explain it in 5 minutes. Through doing this everyone better understood what it was they believed and why they believed it. Having a student explain themself forces them to create connections between different frameworks in their mind.

(g ~0.55 from these studies: Bisra, Liu, Nesbit, Salimi & Winne 2018, meta-analysis of 69 effect sizes)

The Bundle Example: Direct Instruction ~0.5–0.6 SD

This technique is actually a bundle. The reason I mention it is because of how effective it is at increasing learning quality and speed, and the 50 years of science behind it. It also shows that when bundling learning science strategies together, they can be very effective. It makes Alpha School look significantly more possible. The techniques it includes are the worked examples and feedback from above along with explicit instruction and high frequency response.

Explicit instruction is defining the concept specifically rather than working with the student for them to discover it on their own. High Frequency Response is essentially the amount students are responding to the teacher during teaching.

An example of explicit instruction and high frequency response is my physics class from high school. To me, it felt like the entire 50 minute lecture was a conversation. My teacher would define a concept and then ask us questions, have us explain it to our groups, then respond again to her prompts. This amount of interaction within the boundaries of the explicit definition helped walk the path of encoding and retrieval many times by the end of class.

(SD ~0.5–0.6 from these studies: Stockard, Wood, Coughlin & Rasplica Khoury 2018, meta-analysis covering 50 years of research; Project Follow Through)

The important concept: Cognitive Load Theory a framework

As I mentioned above, our working memory only holds a limited amount of information at once. Understanding this and designing around it helps to limit overload and therefore more efficiently move information from our working memory to our long term memory.

An example of this is the game of soccer. It's easy to notice who is experienced in the game by who is scanning the field with their eyes and who is not. This scanning takes information about where players and space are and stores it in working memory to use when making decisions. Players that haven't trained technique don't have space to receive the ball, pass it where someone else is on the field, defend the ball, keep balance, and remember where multiple other players on the pitch are.

An example of using this in design is putting problems on the same page as you want them to write the answer, taking text OFF of your slides, and putting labels on diagrams instead of in a key on the side. Each of these helps to streamline information and feel better to people's brains.

(No single effect size — cognitive load theory is a framework rather than one technique, and it explains why several of the above work. Specific predictions do have numbers, e.g. the split-attention effect around 0.7. See Sweller 1988; Cowan 2001 on the four chunk limit)

So, is 2x possible?

These techniques as a whole are why there is so much room to be improved in the education space. With all of these methods combined I would venture to say that students' learning can be improved close to ~1 SD a year if combining diverse techniques. I say this because if a clear studied example of 4 techniques hits ~.6 SD then adding in more techniques that work on different parts of learning, in my estimation, would get it to ~1 SD a year. If this is repeated every year, the improvement would compound. So even at the normal slowing speed of learning due to amount of content, a student who started in Kindergarten would be ~5 grade levels ahead by the time they were a 5th grader. This conclusion makes me believe the learning 2x in 2 hours a day possible.