Newly acquired memories are initially labile and consolidate over time into longer-term traces. Although consolidated memories were traditionally considered resistant to interference, reactivation can render them transiently labile and susceptible to modification before reconsolidation. However, there is marked variability in whether a reactivated memory is disrupted, protected, or even strengthened by subsequent new information. The authors set out to systematically test which factors determine this outcome using a motor learning paradigm, testing the hypothesis that contextual inference and latent signals such as sensory prediction error guide whether a consolidated motor memory is modified or protected when opposing information is introduced.
The investigators used a classic A-B-A visuomotor rotation design. On Day 1 participants adapted to a 30-degree clockwise visuomotor rotation, referred to as perturbation A. On Day 2 participants experienced a 30-degree counterclockwise rotation (perturbation B) introduced under varied experimental conditions intended to manipulate reactivation and contextual signals. Relearning of A on Day 3 was used to assess whether the original memory was protected (savings) or disrupted.
Participants adapted to a 30° clockwise rotation (A) on Day 1. This served to establish a consolidated motor memory for the A perturbation prior to any interfering intervention.
On Day 2 the interfering 30° counterclockwise rotation (B) was introduced in several different conditions:
These manipulations were designed to vary the presence of explicit contextual cues, the magnitude of sensory prediction error at the A→B transition, and the state of reactivation of the original memory when interference began.
In a final experiment the authors added an explicit contextual cue — a secondary follow-through target — to distinguish A and B trials. This manipulation tested whether an explicit external cue alone could protect the original memory when a salient sensory prediction error was not present.
Reactivating the original A memory briefly before introducing interference protected that memory. Protection was demonstrated by significant savings during relearning of A on Day 3.
Direct introduction of interference (B) without prior reactivation disrupted the original A memory, as shown by impaired relearning.
The protective effect was consistent with a contextual-inference account: when the abrupt A→B transition generated a large sensory prediction error, that error acted as a latent contextual cue signaling a new context and thus shielded the original memory from being overwritten.
Eliminating the salient prediction error removed protection. Specifically, an immediate washout session that produced a similar error profile to the abrupt transition, or a gradual A→B transition that reduced abrupt prediction error, abolished the protective effect and led to disruption of the original memory.
When explicit contextual cues distinguished the two perturbations (the secondary follow-through target), memories were protected even in the absence of a salient sensory prediction error.
The authors interpret the pattern of results through a contextual inference framework: the brain uses available cues — both explicit contextual signals and latent indicators such as the magnitude of sensory prediction error — to infer whether incoming experience reflects the same context as a prior memory or a different context. If inference favors a new context (for example, prompted by a large unexpected error on abrupt transition), the new experience is stored separately, protecting the original memory. If inference favors a single shared context (for example, when prediction error is removed or transitions are gradual), interference is integrated and the original memory can be overwritten.
These results show that reactivation per se does not uniformly increase vulnerability to interference. Instead, whether a consolidated motor memory is modified, protected, or strengthened depends on contextual information available at the moment interference is introduced. Both latent signals like sensory prediction error and explicit contextual cues can determine the memory’s fate. This has implications for understanding variability in reconsolidation and for designing interventions that aim to modify or preserve motor memories.
The source article presents the experimental design, manipulations, and group-level outcomes summarized above. Specifics such as sample size, participant demographics, statistical values, trial counts, exact timing parameters, and detailed quantitative results were not reported in the text provided here. Where methodological or numerical details are required for replication or clinical translation, those details must be obtained from the full preprint.
The authors declared funding from the Department of Biotechnology (BT/PR51466/MED/122/377/2024). The authors declared no competing interests.