The forgetting curve

The forgetting curve’s useful lesson is its shape, not a universal timetable. Retention falls with delay, often more sharply earlier and more slowly later. The measured points need not form a perfectly smooth line. In the original work and later replications, observations around one or two days sometimes sat above smooth fitted curves.¹²

The original experiment was much narrower than popular summaries imply. Ebbinghaus was the sole participant. He memorised ordered series of thirteen nonsense syllables and later relearned them to the same criterion. His outcome was savings, meaning the reduction in work required for relearning, rather than the proportion of material he could freely recall.¹ A later replication followed one Dutch-speaking man learning Dutch-compatible nonsense syllables and found a broadly similar savings curve. It also agreed with curves from two participants in an earlier German replication, apart from a lower observation at the longest delay.²

The figure illustrates declining savings in relearning an ordered series of thirteen nonsense syllables, emphasizing the recurring shape rather than a universal timetable.

<svg viewBox="0 0 590 300" xmlns="http://www.w3.org/2000/svg" role="img" aria-labelledby="curve-title" font-family="system-ui, sans-serif">
  <title id="curve-title">Qualitative illustration of savings in relearning ordered series of thirteen nonsense syllables</title>
  <g fill="none" stroke="currentColor" stroke-width="2">
    <line x1="78" y1="48" x2="78" y2="230" />
    <line x1="78" y1="230" x2="555" y2="230" />
    <path d="M 145 90 C 168 130 198 151 235 165 C 266 177 291 166 318 171 C 360 179 404 184 437 181 C 456 180 471 186 485 190" stroke-dasharray="9 8" />
    <path d="M 485 190 L 466 192 M 485 190 L 472 176" />
  </g>
  <g fill="currentColor">
    <text x="82" y="24" font-size="16">Savings in relearning ordered series of 13 nonsense syllables</text>
    <text x="12" y="60" font-size="14">More savings</text>
    <text x="12" y="224" font-size="14">Less savings</text>
    <text x="82" y="255" font-size="14">Shorter delay</text>
    <text x="448" y="255" font-size="14">Longer delay</text>
    <text x="215" y="125" font-size="14">Qualitative direction only</text>
  </g>
</svg>

Figure: Savings generally decline as delay increases in this qualitative illustration of relearning ordered nonsense-syllable series. With a deck and twenty minutes, treat the curve as a warning that delay matters, not as a review timer. A forgotten answer does not reveal a universal amount lost. Record what the prompt tested and whether success meant unaided recall, recognition, or faster relearning. Those outcomes are not interchangeable.

Boundary. These experiments support a recurring curve-shaped decline under their conditions. They do not establish fixed forgetting percentages for meaningful lessons, a universal mathematical law, or one mechanism that explains forgetting. Training materials that attach fixed percentages to the curve have extended a narrow savings experiment far beyond what it measured.¹³