Definition

Memory Stability ($S$) is a quantitative metric representing the durability of a memory trace in the brain. Specifically, within the FSRS Algorithm, Stability is defined as the number of days required for the probability of recall to drop from 100% to 90%.

The Mechanism: Storage Strength vs. Retrieval Strength

According to the New Theory of Disuse proposed by Bjork and Bjork (1992), human memory is not a single variable. It consists of two distinct components:

1. Retrieval Strength: How accessible a memory is right now (e.g., immediately after reading a flashcard). This decays rapidly.

2. Storage Strength (Stability): How entrenched the memory is. This determines how slowly Retrieval Strength decays over time.

The "Cramming" Paradox

Bjork’s research explains why cramming fails: passive study increases Retrieval Strength (you feel like you know it) but fails to increase Stability. Cognitive Load Theory suggests that true learning (increasing Stability) only occurs when a retrieval attempt is successful despite low Retrieval Strength (i.e., when it is hard to remember).


The Mathematical Model

Modern algorithms like FSRS operationalize Bjork's theory by calculating Stability ($S$) using a decay formula. As detailed in our internal analysis of learning metrics, the probability of recall ($P(R)$) is modeled as:

$$P(R) \approx 0.9^{\frac{t}{S}}$$

Where $t$ is the time elapsed and $S$ is the stability.


Application in Our Software

We do not track "Streaks" or "XP" as primary metrics; we track Stability. When you review a word, our goal is to increase its $S$-value.

  • Novice Word: $S = 1$ (Recall probability drops to 90% in 1 day).
  • Mastered Word: $S = 365$ (Recall probability drops to 90% in 1 year).This ensures you stop reviewing words you have already stabilized, focusing your efforts only on unstable items.

Works Cited