Storage strength and retrieval strength are the two dimensions every memory trace possesses in Bjork and Bjork's 1992 framework. Storage strength reflects the item's integration into long-term memory—its richness of encoding, its connections to other knowledge, its resistance to interference. Retrieval strength reflects the item's current accessibility—how likely it is to come to mind when needed right now. The two are independent: a deeply stored item can have low retrieval strength (a childhood friend's name you struggle to recall), and a shallowly stored item can have high retrieval strength (a phone number you just looked up). The independence resolves the paradox of desirable difficulties: spacing, generation, and interleaving reduce current retrieval strength (making practice feel harder) while building storage strength (making retention stronger). AI maintains external retrieval strength permanently—users always have access—preventing the drop in retrieval strength that triggers the effortful re-retrieval that builds storage strength.
The framework's predictive power for the AI age lies in its specification of what builds storage strength versus what maintains retrieval strength. Storage strength increases through effortful retrieval—the harder the retrieval, the greater the increment—and through rich, varied encoding that connects the item to multiple retrieval cues. Retrieval strength increases through recent activation but decays rapidly without it. Massed practice maximizes retrieval strength temporarily (the item is highly accessible immediately after practice) while barely affecting storage strength (because effortful retrieval never occurred). Spaced practice allows retrieval strength to decay between sessions, so each session requires effortful retrieval that substantially increases storage strength.
AI tools create a historically unprecedented cognitive profile: users with comprehensive external retrieval strength (anything can be looked up) and minimal internal storage strength (because the conditions for building it—effortful retrieval after forgetting—never occur). The profile feels like mastery, because retrieval strength is high and humans assess their knowledge by accessibility. The profile lacks the internal architecture that genuine expertise requires: the richly interconnected web of knowledge that allows connection-making, pattern recognition, anomaly detection, and the flexible transfer that storage strength enables.
The theory predicted a specific failure mode now visible across AI-augmented workforces: practitioners who perform excellently on known-query retrieval (looking up what they know they need) and poorly on the cognitive operations storage strength enables (noticing what they did not know to look for, connecting information across domains, recognizing when something contradicts the broader pattern). The gap between query-based and connection-based cognition is the gap between maintained retrieval strength and absent storage strength—and it is a gap that widens with every AI interaction that substitutes external access for internal encoding.
The New Theory of Disuse emerged from Bjork's three-decade investigation of retrieval inhibition and interference—the mechanisms by which memory access is regulated. Earlier dual-process theories (short-term versus long-term memory, recall versus recognition) had proposed multiple memory systems; Bjork and Bjork proposed multiple strength dimensions within a unitary system. The innovation was treating storage and retrieval as independent rather than correlated variables, allowing the theory to explain phenomena (like the feeling of knowing something you cannot currently retrieve) that single-dimension models struggled with.
Elizabeth Ligon Bjork's research on metamemory provided the empirical foundation for understanding how the two strengths produce divergent subjective experiences. Learners assess their knowledge by retrieval strength—if information comes to mind easily, they judge it well-learned. The assessment is accurate for current performance (high retrieval strength does predict immediate recall) but inaccurate for future performance (storage strength, not retrieval strength, predicts delayed recall). The dissociation between what learners track (retrieval) and what matters (storage) became the explanatory mechanism for why students choose massed over spaced practice despite spacing's proven superiority.
Two strengths, one trace. Every item in memory has independent storage strength (depth and richness of encoding) and retrieval strength (current accessibility)—conflated by intuition into a single 'memory strength' whose separation explains the performance-learning paradox.
Storage builds through effortful retrieval. The primary mechanism for increasing storage strength is retrieving an item when retrieval is difficult—after spacing-induced forgetting, after interference from other items, under conditions requiring reconstructive effort rather than automatic access.
AI maintains external retrieval, prevents internal storage. When external systems (search engines, language models, knowledge bases) provide permanent retrieval strength, the drop in internal retrieval strength that would trigger effortful re-learning never occurs—users maintain access while storage strength remains permanently low.
Access is not understanding. The capacity to find information (maintained retrieval strength) is categorically different from the capacity to think with information (storage strength enabling connections, patterns, flexible application)—a difference invisible to systems measuring output rather than cognitive architecture.