Is it Better to Review Back or Learn Ahead?

I have a lot of open questions about learning. One of those is whether it is better to review back or learn ahead to maintain knowledge.

Reviewing back would be going through material you’ve already completed and testing yourself on it again. This is the principle through which spaced repetition systems work. I’m currently reviewing back words and sentences I learned in China through Anki.

Learning ahead means learning the next level as a means of strengthening knowledge. I took an introductory class in artificial intelligence during the MIT Challenge. Learning ahead would mean taking a more advanced AI course that builds on those concepts.

Advantages of Reviewing Back

The main advantage of review is that you get complete coverage. Learning ahead to more advanced material usually only builds on a subset of prior material.

During the MIT Challenge I often did series of prerequisites: Single Variable Calculus -> Multivariable Calculus -> Differential Equations -> Signals and Systems. Each class used only some of the concepts from the prior class. Multivariable calculus introduced new concepts, but it didn’t stress some of the trigonometry identities of SVC. Signals and systems pretty much only considered transforms from differential equations.

Going back and reviewing means you cover all of those concepts, not just the ones used in a particular advanced course.

Advantages of Learning Ahead

Sometimes incomplete coverage is an advantage. The ideas that get reinforced in more advanced courses or projects tend to be the more important ones. If an idea is never reinforced in the future, it probably isn’t terribly important.

The main advantage of learning ahead, however, is that you learn new things and that can be more interesting than review. Learning new things is more motivating, and you’re expanding your knowledge rather than just reinforcing it. The idea of taking a more advanced course in AI is more appealing to me than the thought of going back and redoing old tests to make sure I remembered everything from three years ago.

The Inconsistency of Learning Ahead

When I did the MIT Challenge, my original plan was to use the learning ahead method as the primary way I’d maintain my knowledge. I still think there are some obvious benefits of that approach, but I’m starting to see some drawbacks.

The biggest drawback, for me, is that learning ahead is less consistent than review. Review, whether it’s done through software like Anki, or through some kind of scheduled re-testing, can be set up to maintain the information consistently. That way, if you get busy and don’t have the time to push your knowledge in a particular area, you can at least be confident it won’t slip back.

I mentioned that, with this language challenge, I’m being more proactive about fighting forgetting. I’m scheduling regular review habits in advance, independent of any future study I do. That way I can maintain my ability even if I have long gaps where I don’t actively improve a language.

Although I’m a bit late, I’m also looking to do something similar for the MIT Challenge and my university education.

Obviously, maintaining all of the knowledge through review is unnecessary and time consuming. Academic knowledge is delivered in large swaths, not all of which will be particularly useful or interesting to you. However, going back and selecting out the specific concepts and skills I want to maintain might not involve considerable long-term review work to maintain. Perhaps several hours every few months?

Sometimes it’s Good to Forget

The truth is much of what we learn we’ll eventually forget at some point, or it will be buried so deep that it doesn’t come up easily in the right contexts. Learning is, even when done well, adding water to a leaky bucket.

However, if you develop systems for learning ahead or reviewing back to preserve the minority of knowledge that is important to you, the knowledge that stays in the bucket will be the kind that matters to you.

  • Madhur

    Hi Scott,

    I think what you talked about is mainly related to academic reviewing. But there are things says rules/laws related to business, name of the singer who’s song your are hearing, how something is calculated etc. All these doesn’t come to mind easily. May be not at all sometimes. And there are many such points to remember. What I find in my case is I don’t recall such things easily or not at all. And may be this would be the problem with many others. But I don’t thing intentional reviewing would be the way to go about it. Occasionally, it happens that you get to review the points by coincidence through newspapers, talking to friends etc.

    So I think there should be way to stick such points to mind at least for considerable time period if not for life. And there are many non-academic points/concepts which we need in our life and those doesn’t come in the right context. I see people who remember it at one reading and can recall it when needed.

    What do you think? Any suggestion…

  • John

    Hello Scott,

    In addition, to spaced repetition there is another way to dramatically decrease forgetting with minimal retrieval practice. This is the method that world-class experts use. It is more difficult to describe than space repetition, but I will give it a try.

    As you know, in order to be world-class expert in an intellectual pursuit you need to accumulate a lot of knowledge. That is why it generally takes at least 15-20 years before someone reaches world-class levels. So how do you make sure that all the work you did in your first 5 years does not gradually evaporate?

    You suggested space repetition which can effective for beginners in a domain, but that is not what actual real world experts use. I have never heard of a Nobel prize winning physicist that use to review his high-school tests. So what do they do instead?

    The best way to understand this by looking at the world’s best chess player, Magnus Carlsen. Magnus is only 23 years old and he has already become the highest-rated player of all time. He is the Usain Bolt of the chess world.

    In a recent interview, Magnus said that he has memorized over 10,000 games. Given that an average game is 40 moves, Magnus has memorized 400,000 different moves. Most importantly, he did not use a fancy mnemonic technique such as the memory palace to do it. Instead through repetition (retrieval practice in the beginning of his career), he started building complex mental meta-models of how to play chess. Now he has built an incredibly complex mental model of chess. Think of the chess games as 10,000 bricks and the mental models as Bill Gates’ 100 million dollar mansion. To use Daniel Willingham terminology, he has achieved a remarkable understanding of the deeper structure of chess. For reasons we still don’t quite understand fully when you build such complex mental models the “upkeep” is quite minimal. One the reasons might be that the information is so well connected to so many other things in your long-term memory that it makes it really hard to forget.

  • Darko

    Well, I think it’s best to do both and as you learn ahead, to schedule a regular ‘review time’ for your cards and delete those that integrate into some advanced knowledge you already know, as well as those who you’ve never used practically.

  • Chris Krohn

    Visit the lessons of the past, but always push the frontiers of the future by anticipating it and creating it!

  • Eric-Wubbo

    “I have never heard of a Nobel prize winning physicist that use to review his high-school tests”

    Well, actually, Enrico Fermi, Nobel prize winner in physics and known by his contemporaries for his encyclopedic knowledge of physics (as well as being someone who both made important contributions to both theoretical and experimental physics) had the habit of rehearsing physics formulas during train journeys. So indeed, not really high school tests, but he did review some of his knowledge occasionally.

    Scott, you may know that ‘learning ahead’ is one of the main methods that Singapore math uses – and Singapore math currently likely being the best way to learn math (up to high school level) in the world. However, Singapore math was especially designed by experts so to make ‘repeating knowledge by learning ahead’ incorporated into the method (they call it the ‘spiral method’ or such); regular university courses or self-study generally don’t have that amount of planning.

    On the chess-player example: I know that there is some debate on spaced repetition being less important (or necessary) for complex information. And anyway, it seems that people’s vocabulary in their native tongue only increases as they get older, even though many unusual words are not encountered on any regular schedule. So I guess that facts, when connected to sufficient other concepts, are far less forgettable than Ebbinghaus’ curves suggest – which may be understandable as Ebbinghaus based his curves on nonsense syllables. Besides, there are biological mechanisms solidifying facts after a decade or so, the only real problem is usually that facts are ‘overgrown’ by similar facts.

    Whether knowledge sticks however also depends on the nature of the knowledge; I know at least one adult, academically educated Chinese who after about a decade living in Western Europe ‘forgot how to write’ (which was annoying as she had not yet learned to express herself comfortably in any other language)

    What this means for learning languages? Basically, I guess that one should always be patient while learning a language; the more you know about it, the easier it is to learn new words and the more slowly forgetting of existing words is, though likely you always need at least some maintenance.

    Review or going ahead? Both, of course. But I think people really overestimate the time that one has to spend on reviews because it seems ‘chore-like’; if I’m a bit behind on reviewing my 9000+ Chinese cards, it only takes a few extra sessions of half an hour to get back on track – unless you had a course that was really short and intensive, one should always be able to learn at least a few new words per day. And in the long term, with a few new words a day you will get a pretty nice vocabulary in a year or three…

  • Jonathan

    Hi Scott! Regarding this discussion of reveiwing and learning ahead, what do you think about Michel Thomas’s method of teaching languages? He claims that the students should not even try to retain what was taught, just to let it stick naturally..

  • Sebastian Aiden Daniels

    Damnit Scott. I can’t believe I am going to say this, but your most is making me miss being in school. I love that feeling when you see your knowledge building and the things that were once hard are now easier. I think it is important to balance reviewing back and looking ahead. Practice, practice, practice is what matters, but you don’t want to get too far ahead because then you may just get frustrated.

  • Katelyn

    In the context of SET, I’d lean towards learning ahead, perhaps mixed with realistic project work. Unlike foreign language conversation or chess, there’s very little in the engineering world that has to be designed so urgently you don’t have time to review half-forgotten concepts on an as-needed basis. Plus, there are very good resources out there that are hardly any effort to use. I’m more than happy to consign Lagrange multipliers to my increasingly fuzzy memories of student life and focus on things that I use like xml instead. In the unlikely even that I ever need Lagrange multipliers, I know just where to look.

  • werena

    well, this is a very interesting topic. Something, people might not think about when learning something… indeed looking back might be the best way to do it. It allows you to make the mistake, correct the mistake, reason why you thought differently, and learn the right thing. Thanks for the good read

  • Scott Young


    Why go back to school when you can attend Ivy-league classes for free?


  • Eric-Wubbo

    “Why go back to school when you can attend Ivy-league classes for free?

    Interesting remark – but I wonder if there is any evidence that Ivy-league classes are better than regular classes – top universities tend to attract top researchers, but top researchers are almost by definition much more interested in doing research than in devising interesting lectures [there were of course exceptions like Feynman, but I bet you would not have enjoyed Einstein’s or Newton’s lectures]. Besides, they don’t have to give very good lectures anyway as their students are much brighter and/or more motivated than average. Sad as it may be, most university professors I’ve met in my life were significantly worse at teaching than the average Dutch highschool teacher. In any case: Ivy league does not mean teaching quality (but do go there if you want to become a great researcher!)