In the previous section, I explained that the equation for the forgetting curve allows one to find the retrievability of a given memory. Also recall that retrievability is the probability of being able to successfully recall a memory at a specific point in time. The key word here is “probability.” Because the human brain is a biological organ, not a digital computer, it’s impossible to know exactly when you are going to forget something. We can make quite educated guesses, but there is always going to be a certain amount of randomness that has to be accounted for: sometimes you just randomly forget something, even though you previously knew it extremely well; other times you randomly remember something, despite only hearing it once several years ago.
As time passes without reviewing a memory, the probability that you still remember it (retrievability) decreases. The basic idea with spaced repetition is that if you always review memories before you are likely to have forgotten them, then you will almost never forget. The idea is simple yet immensely powerful. The problem is that because we only know the probability of how long we will be able to remember something, not the exact amount of time, we have to decide how low are we going to let that probability drop before we review.
Theoretically, immediately after reviewing a piece of information, the probability that you would be able to recall it if quizzed is 100%. But from that point onwards, the percentage begins to drop with each second that passes. If you decided to review memories each time retrievability dropped to 99%, then on average, at the time of review, you would remember 99% of your memories, and forget 1%. If you decided to review memories each time retrievability dropped to 85%, then on average, at the time of review, you would remember 85% of your memories, and forget 15%. At first glance, the first scenario might sound more desirable than the second, but this isn’t necessarily the case. Assuming that all other factors are the same, it’s going to take longer for retrievability to drop from 100% to 85% than to drop from 100% to 99%. This means that the person who remembers 99% of their material must review much more often than the person who is remembering 85%. In this way, we can see that how much you remember, and how often you review, are correlated.
The more often you’re reviewing, the more time you’re spending on reviews. So really, the cost of remembering is time. This leaves us with the question of what balance of time and retention is best? By default, Anki is calibrated in such a way that “for moderately difficult material, the average user should find they remember approximately 90% of mature cards that come up for review” (from the Anki manual). I assume that 90% was chosen as well-balanced middle ground, providing relatively high retention with a manageable amount of reviews. I will talk about whether I think this number is ideal or not in a different post.