From atomic cards to integrated recall: a new scheduling method for cloze deletion

Anki Suggestions
1

In previous posts (Improving FSRS for List Memorization and its revision, Extending Cloze Deletion for Sequential Item Rating), I proposed a method in which users review all items of a list on the same cloze note, with each cloze being a different item. These individual ratings are then combined to compute a single interval, which is applied to the entire list.

This new proposal does not replace the previous one, instead it offers an additional implementation based on the same underlying principles. The key difference is that here the interval is computed so that the average retrievability of the clozes—rather than the product of their retrievabilities, as in the earlier proposal—matches the desired retention. As a result, the calculated intervals will be longer and the retrievability of individual items lower.

The structure of this note is similar to a standard cloze note, and could follow this workflow:

Field 1(question): the user attempts to recall all clozes

Field 2 (text with hidden clozes): the user reveals cloze 1 → rates → reveals cloze 2 → rates → … → the algorithm computes a single interval and applies it to the whole cloze note

I propose to use the average retrievability because this approach is intended for notes with multiple clozes where it’s acceptable not to recall every item correctly at the same time, but where the user still prefers to review them together.

This is particularly useful for exam modalities that ask broad questions (e.g. “explain the pathophysiology of heart failure”) and that reward fluent recall of multiple related concepts. It also fits situations where several pieces of information form a natural chunk, making it more effective and more realistic to recall them together rather than splitting them into separate cards.

ADVANTAGES:

-better simulates broad exam questions

-minimizes context and interference

-trains fluency in recalling multiple notions together

-reduces the risk of losing the “big picture”

-allows different ratings for individual clozes while still producing a single interval, which is a new feature that would make Anki flashcards even more flexible and customizable.

DRAWBACKS:

-The scheduling logic and the optimal way to use this feature may not be immediately intuitive to all users

-retention of individual clozes may not always approximate the desired retention.

2

I think your remembering of the facts will not be as good as having the various Cloze options scattered throughout the other cards of a deck. If you just cycle through each field of a multi-field Cloze card, you are providing yourself with too many contextual hints. These hints may not be present in the real world, when you need to remember and use the information.

2 Likes
3

Thank you for your reply.

I think you are not considering one important detail that maybe I should have emphasized more: during the recall phase, the user only sees the question (for example, “Explain the pathophysiology of heart failure”), not the text containing the hidden clozes. As described in the scheme, the second field—the text with clozes—is revealed only after the user has made their initial recall attempt.

It is true that the user might remember an item when they see the text with hidden clozes, even if they initially forgot it while looking only at the question. This effectively gives them a kind of “second chance” to retrieve the information. However, it is the first recall attempt that really matters: if the user knows he did not recall an item during the initial response to the broad question, they can still choose to mark that cloze as “again.”

Even with this secondary cue, the primary goal of the workflow remains the user’s ability to answer the broad question itself. The clozes serve as a structured way to rate the individual components.

4

i would also add that the user can in principle press good even if he recalled the cloze on the second attempt. This reduces some of the benefit of minimizing context, but it’s fine a long as the user stays consistent.