When to drop an item

Dropping an item can mean ending practice for the current session, suspending it for repair, or deleting it permanently. Those decisions should not share one automatic rule. The useful judgement is to treat repeated failure as a signal to inspect the item’s value, wording, supporting knowledge, and possible interference.

Experiments with undergraduates learning Swahili–English pairs found that dropping items after they met a preset success requirement could produce similar final performance with fewer trials. Efficiency depended strongly on the spacing between presentations.¹ In two other experiments with 218 college students learning geography facts or word definitions, extra same-session practice after one perfect trial produced a large advantage after one week, but that advantage declined sharply at longer delays.²

Later relearning changes the decision again. Experiments with students learning conceptual definitions found that both the initial success requirement and later relearning affected durability and study cost.³ Experiments with Swahili–English pairs also found that repeated weekly relearning could erase longer-term differences produced by different initial success requirements.⁴

The diagram shows how item value, clarity, and interference can guide responses to repeated failure, while continued practice remains a practical option.

flowchart TD
    A["Repeated failure on an item"] --> B{"Is the item important and well formed?"}
    B -->|"Important, cue unclear<br/>Manual guidance"| C["Edit the prompt or add supporting knowledge"]
    B -->|"Important, similar item competes<br/>Manual guidance"| D["Separate or suspend one item"]
    B -->|"Low value<br/>Manual guidance"| E["Delete or suspend it"]
    B -->|"Important and well formed<br/>Continued practice is a practical option"| F["Continue, then reassess cost and progress<br/>Later relearning affected durability and cost<br/>for the studied materials and delays"]
    C --> G["Return the revised item"]
    D --> G
    F --> G
    E --> H["Remove it from active review"]

Figure: Repeated failure calls for diagnosis: revise, separate, continue, suspend, or delete according to the item and evidence. Anki tags and suspends a review card at a configurable default of eight lapses. That is a product rule, not an experimentally validated optimum. Its manual suggests editing poor prompts, adding support, deleting low-value material, or separating interfering items.⁵ With twenty minutes available, investigate one repeatedly failed item instead of spending the whole session pressing the same answer button.

Boundary. Experiments show that stopping criteria and later relearning both affect cost and retention for particular materials and delays. They do not establish a universal lapse count, recall count, or interval for permanent retirement. Deleting a low-value item is a defensible practical choice, not a laboratory-derived threshold.