[AlchemistCamp]

To be perfectly honest, I flagged it because of the knowledge gained graph. It's a wild extrapolation.

For context, I was a big fan of SRS and even contributed to Anki back in the day! I was really into foreign language learning, had majored in one language and was learning another language in a separate language family.

I built, ran and put my heart into brick and mortar language immersion school for years. Over time, I realized both from my learning experiences and those of my students that SRS fell far short of extensive reading.

It's tempting to break things down to "units of information", as you put it your assumptions document. SRS is great for decontextualized information (e.g., memorizing all the capital cities in the world), but that's not really how language works or how the brain works for most learning tasks. There are higher-level things your brain picks up, such as collocations, grammar and shared cultural beliefs.

Over the short term, SRS can be useful for building a scaffold to work from, but over the long term, Extensive Reading crushes it on pretty much every metric, including raw size of passive and active vocabulary.

[rsanek]

Extensive reading sounds compelling. Do you have recommendations for services that offer such content? In my own language learning, I have found a few websites here and there (eg Hola Qué Pasa [1]), but nothing that has a large database with varying levels of competence.

[1] https://holaquepasa.com/

[AlchemistCamp]

I'd recommend avoiding "services" and going for books, starting with graded readers. There's a wealth of options for Spanish learners.

If you absolutely hate books and want an online resource, then I'd suggest https://www.lingq.com. It has a lot of free content and lets you import your own. Their tech/design chops are meh, but it's run by true language learning enthusiasts and the founder dogfooded it for at least half a dozen languages.

[p-christ]

The simplifying assumption around units of information and the graphs are just to visualise the main point.

Effectiveness of spaced repetition scales really fast / exponentially whereas other learning methods don't scale like that - do you agree?

If so then over the long-run spaced repetition is always going to be extremely efficient relative to other learning methods.

[AlchemistCamp]

> Effectiveness of spaced repetition scales really fast / exponentially whereas other learning methods don't scale like that - do you agree?

Of course not. As I wrote, in the comment you’re replying to, I was a user and evangelist of SRS for years but eventually saw the reverse.

It’s a very useful tool for a certain narrow niche of memorization tasks. However, it’s an extremely inefficient learning method over the long run compared to reading and using information or skills in context.

[p-christ]

Do you agree with just this part?

> Effectiveness of spaced repetition scales exponentially

[AlchemistCamp]

Not really.

Given a fixed amount of review time per day, the largest deck you can maintain grows asymptotically over time, not exponentially.

[p-christ]

Do you agree with the assumptions of the Anki scheduler? That each time you review a card you'll remember it for incrementally longer and longer?

If so then a fixed amount of review time per day will let you remember things for longer & longer.

[copperx]

Is Extensive Reading just that? Reading a lot in general? Or is it reading a lot on the specific subject that you want to learn, taking all possible branches?

[AlchemistCamp]

It's about both the volume and the type of reading. See the 2nd page of this paper, under "What is extensive reading?": https://www.researchgate.net/publication/334535447_Extensive...

[copperx]

Thanks!

I'm not surprised. That's how I learned English.