Skip to content

Delayed Retention, Transfer, and Interference Battery

Delayed test battery — instantiates Reopened Malleability Window

Tests at a delay whether the installed change actually held — whether it survived over time, transferred to ordinary contexts, and resisted the return of the old pattern.

A response spike at the end of an induced session proves nothing about whether the reopening worked. Delayed Retention, Transfer, and Interference Battery is the out-of-session verdict: run at spaced delays after the window has reclosed, it asks the three questions that separate real change from a temporary one — did the new configuration persist, did it transfer to ordinary contexts beyond the trained material, and did it survive the return of the old pattern under load. Its distinguishing move is that it tests long after the window is shut and away from the training setting, and it specifically hunts for the pre-existing configuration reasserting itself — the failure the immediate, in-context measures are blind to.

Example

An adult learner of a second language has never been able to hear or produce a non-native contrast — the sound is fossilized into their native mapping. An intensive program reopened phonological plasticity and installed the new contrast; at the program's end their accuracy looked excellent. The battery then runs at one, three, and six months. Retention: can they still discriminate the contrast cold? Transfer: does it hold in untrained words and in spontaneous conversation, not just drilled pairs? Interference: does the old native mapping return under fast, distracted speech? The readout is sobering and useful — retention holds at one month, transfer to conversation is only partial, and by three months the old pattern spontaneously reasserts itself the moment they stop attending to it. The end-of-program score said success; the battery says durable but not selective under load — and that later verdict is the one that actually decides whether the reopening earned its risk.

How it works

The battery is defined by when and against what it tests, not just what:

  • Spaced delayed testing — probe at intervals well after reclosure, so decay and delayed relapse have time to appear, rather than reading the flattering moment right after training.
  • Three separate axes — retention (does it persist), transfer (does it generalize to ordinary, untrained contexts), and interference (does the prior pattern return under speed, stress, or competing cues) — passing one does not imply the others.
  • Comparator-referenced — the change is credited only where it exceeds what an ordinary-input arm achieved, so ordinary practice effects are not mistaken for the reopening's work.

Tuning parameters

  • Delay schedule — how far out and how many timepoints. Longer, denser schedules catch slow decay and late relapse but cost follow-up and lose participants.
  • Transfer distance — how far from trained items the probes reach. Far transfer is the honest test of generalization but is harder to pass and to interpret.
  • Interference load — how hard the old pattern is provoked. More load (speed, distraction, competing cues) surfaces fragile change; too much can fail even robust change.
  • Adverse sensitivity — how actively the battery looks for a returned prior pattern or a new distortion, versus only scoring target gain.

When it helps, and when it misleads

Its strength is enforcing the archetype's real success bar — selective, durable change that exceeds ordinary input — instead of the seductive end-of-session number. It is the mechanism most likely to reveal that a celebrated result quietly decayed, failed to generalize, or was overwritten by the old configuration coming back.

Its failure modes are all ways of testing too gently. Under-delayed or too-easy probes declare a durability that a later, harder test would break; reusing practiced items masks lost transfer; and dropping the comparator credits the reopening for gains ordinary practice would also have produced. The classic misuse is scoring only trained items at a short delay to draw a flattering retention curve. The discipline is spaced delays, far-transfer and load-stressed probes, active search for the returning pattern,[1] and every result read against the ordinary-input comparator.

How it implements the components

The battery realizes the archetype's longitudinal-outcome component — the delayed evidence that decides whether the change was real:

  • longitudinal_retention_transfer_and_adverse_change_signal — its entire function: the delayed signal of whether the target held, transferred to ordinary contexts, resisted the old pattern's return, and produced any adverse change.

It does NOT test immediate re-stabilization at reclosure (Selective Re-stabilization Challenge), aggregate outcomes across cases and sites (Longitudinal Adverse-Plasticity Registry), or run the ordinary-input arm it references (Ordinary-Training Comparator Protocol).

Notes

This battery reads one case over time; a run of delayed failures only becomes a pattern when aggregated across cases and sites, which is the Longitudinal Adverse-Plasticity Registry's job. Keeping the per-case delayed test separate from the cross-case record is what lets a single disappointing follow-up inform the registry without being mistaken, on its own, for a verdict on the whole method.

References

[1] Spontaneous recovery — a previously extinguished or overwritten response can return after a delay even without retraining. It is the standard learning-theory name for the old pattern reasserting itself, and the reason interference must be probed late and under load rather than assumed gone at session's end. ↩