A self-consistent set of principles for thinking about learning, drawn from cognitive science and educational psychology literature. Two layers: theory (claims about how learning works) and practice (claims about how to affect learning). Specific advice on how to operationalize P-claims for engineering or application is out of scope and will be handled later in a separate document.
This framework was constructed using an engineering approach for creating domain-specific languages as a method to think, plan, and reflect. An observant reader would notice that each layer is its own domain language, following the tradition that a specialized language is better than a generic one.
The standard for inclusion in either language is twofold: whether the claim being introduced adds something the existing claims don't address, and whether the claim is correct in the only operational sense available — consistency with the latest literature on the topic and various audits that attempt to actively refute it using that literature.
The only legitimate criticism the framework accepts is "you have not validated this with studies specific to your synthesis" — a real constraint on evidence strength, as no studies have been conducted nor are any planned. All other legitimate criticisms where data is presented, the author will wrestle with and attempt to incorporate their data into the framework.
On rejection-claims (RT or RP). Most disagreements with literature can be handled by considering the rejection a matter of either situational interpretation or scale; there is no need to take a side in the debate as the framework's author takes all views as having an element of truth. However, sometimes the framework must pick a position; in such cases the rejected position is placed near the framework's chosen stance along with an explanation and citations showing where the rejection has grounds.
The framework started as a computer scientist's attempt to rectify folk science with empirical work, drawing from 91 cited works across cognitive science, educational psychology, behavioral genetics, and the learning sciences. Early versions had Broadwell's four stages of competence and the folk version of ZPD front and center, but as the framework evolved the author realized these are engineering shortcuts used to help laymen like himself simplify extremely complex science. This has since shaped the framework's view that some folk theories survive because they do real engineering work, simplifying decisions for non-experts, even when their mechanistic framing is imprecise. Decision-making support was given its own P-claim to emphasize the author's learning journey.
The framework's claims carry forward three risks. First, the author is trained in computer science rather than a relevant field. Second, the author made heavy use of generative AI to search for and summarize hundreds of papers — a method whose known hazard is fabricated citations or misrepresented findings; to reduce this, the framework was put through multiple audits run from different directions plus automated verification, with manual spot-checks where the AI's summaries did not add up, though the citations have not yet been systematically verified one-by-one against their primary sources. Third, none of the deductions made in this framework has a purpose-built study evaluating its claims. Some residual risk remains on all three.
This document is a thinking aid for navigating the literature, not professional educational, psychological, or medical advice; for decisions about a specific learner, consult a qualified professional.
The current document contains two layers: Theory (T-claims) and Practice (P-claims). Two additional layers naturally derive from the existing P-layer:
- A-layer (Application): teacher-facing folk guidance. A-layer entries derive from one or more P-claims and translate into guidance a teacher could use without first having to read hundreds of research papers or acquire a relevant degree. The A-layer relaxes orthogonality rules to favor accessibility; entries cross-reference each other and use the framework's binning vocabulary throughout.
- E-layer (Engineering): system architecture and design choices for a recommender (or recommender-like) system built on the framework. Each E-layer entry would survive P-scrutiny (architecturally derivable from or compatible with P-claims) but would be voiced for engineers. Since most framework claims concern activity type and difficulty, a recommender is the natural E-layer target, though variants like online tutoring (recommending problems instead of courses) could also be E-layer.
The full architecture is T → P → {A, E}: theory grounds practice, practice splits into teacher-facing translation and engineer-facing implementation. The acronym T.A.P.E. is a play on words: the framework tapes together several sciences for practical use. Each layer would evolve at its own rate, audience-driven, with the decoupling work in the current framework already structured to support this expansion. A-layer and E-layer construction is deferred for a future date, and the architecture serves as a useful tool to park topics the author would otherwise have engaged in.
The framework uses these prefixes for claims:
- T — theory claim (e.g., T1, T5). Located in Part I. Header format:
#### Tn. Title [provenance markers]. Heading level: 4 hashes. - Tn[a-z] — theory sub-claim, a refinement or specialization of a parent T-claim that is substantive enough to warrant its own naming and citation but not orthogonal to its parent claim. Sub-claims should be thought of as having significant overlap with their parent claim. Sub-claims with the same parent claim are orthogonal to each other (T3a is independent of T3b). Currently used: T3a (scaffolding-bound brittleness), T3b (format-bound brittleness), T7a (passive observation and mental practice), T9b (learning experience affects future motivation), T9c (belief-installation interventions produce smaller effects). Header format:
##### Tn[a-z]. Title [provenance markers]if grounding differs from parent, or##### Tn[a-z]. Titleif inheriting parent's grounding. Heading level: 5 hashes (one deeper than parent). - RT — rejected theory claim. The framework's chosen side in a live debate, with falsification grounded either in explicit conditions or in the empirical pattern that established the rejection (where additional studies finding the rejected phenomenon would weaken the rejection). Currently used: RT5. Located in Part I alongside the corresponding T-claim. Header format and heading level same as T-claims.
- P — practice claim (e.g., P1, P12). Located in Part II. Header format:
### Pn. Title → operationalizes Tx, Ty, Tz [type: planning|execution|judgment|motivation]with optional[scoped to ...]tag for procedures with restricted applicability. Heading level: 3 hashes. - Pn[a-z] — practice sub-claim, analogous to Tn[a-z]. Currently used: P1a (frustration response), P4a (use three high-element-interactivity contexts to verify generalization), P5a (route help through external attribution), P5b (develop the learner's help-seeking metacognition), P8a (manufactured whys have costs), P8b (confirm the learner's why under pressure).
- RP — rejected practice claim. Reserved for analogous use at the practice layer if a practice tradition needs explicit rejection. Currently unused.
Provenance markers attach to T-claims and theory sub-claims:
- [literature] — directly grounded in cited published work
- [synthesis] — reasoning from literature without direct empirical evidence
Refutation claims use the RT prefix (e.g., RT5). The RT prefix encodes that the claim is a refutation of a position in the cited literature; provenance markers on RT claims indicate the relationship to the literature being refuted ([literature] for literature-backed refutation, [synthesis] for reasoning-based refutation).
Type tags attach to P-claims with one of four allowed values:
- [type: planning] — applied at planning time before the activity
- [type: execution] — applied during the activity
- [type: judgment] — applied at decision points to inform planning or execution choices
- [type: motivation] — identifies where motivational work is needed and what dimension of motivation is involved, without prescribing the specific methodology used to do that work; practitioners may use any methodology that serves the dimension named
A P-claim may also carry a [scoped to ...] tag indicating restricted applicability conditions. Currently used: P12 (scoped to one-to-one mentorship with stage 4 + stage 5 in the specific procedure-novice combination).
Claims and references are in order of author's discovery; numbering gaps are remnants of deleted content.
Note on heading-level asymmetry: T-claims use 4 hashes and P-claims use 3 hashes, but this is not arbitrary inconsistency. Part I groups its T-claims under a "Theory claims" section header at heading level 3, with individual T-claims at heading level 4 nested inside. Part II lists P-claims directly under Part II at heading level 3 with no intermediate wrapper. The asymmetry reflects a real organizational difference rather than a structural error. Two ways to make them symmetric: add a "Practice claims" section header inside Part II (preserves Part I's wrapper, demotes P-claims to heading level 4), or remove Part I's "Theory claims" wrapper (promotes T-claims to heading level 3, matches Part II's flat structure). Currently unfixed; flagged for awareness.
Claims about how learning works. The standard here is defensibility against the literature.
Cognitive Load Theory [1], [2]: working memory imposes a binding constraint on learning. Material is intrinsically harder when more elements must be held and related simultaneously (element interactivity). Instructional design either focuses limited capacity on relevant elements or wastes it on irrelevant ones; the former supports learning, the latter hinders it.
Skill acquisition theory [3]: procedures become fluent and eventually automatized. Used as foundational reference for fluency and automatization claims. Folk-pedagogical phase models [4] are not used in the theory layer; they appear only in the practice layer (Stage binning) as decision-support vocabulary.
Memory and forgetting research [8], [9]: foundation for retrieval practice and spacing. Ebbinghaus (1885) [8] established the forgetting curve — steep initial forgetting that flattens over time. Cepeda et al. (2006) [9] meta-analyzed 254 effect sizes across distributed-practice studies and found robust retention gains for spaced over massed practice across delays from 14 days to 350 days.
Desirable difficulties [10], [11], [12]: effortful retrieval and varied practice produce more durable learning than fluent practice. The learning-versus-performance distinction [13] is foundational: performance during practice is a poor predictor of long-term retention.
Motivation research: directional finding that positive motivational orientations correlate with better learning outcomes; negative orientations with worse. Framework draws on the directional finding without committing to a specific theory of motivation. See T9 for the empirical case across multiple research traditions.
A learner's apparent state on any given skill can vary session-to-session in ways that are not noise around a stable mean. The underlying trajectory is non-monotonic: periods of apparent regression often reflect reorganization rather than measurement variability or loss. This pattern is documented across multiple developmental domains — motor skill acquisition, language development, theory of mind, reasoning, creativity, and face perception — under the term U-shaped development [22], [23].
The dynamic systems perspective [24] proposes a mechanism: behavior is "softly assembled" from multiple contributing components that self-organize differently as components and conditions change, so apparent regression reflects reconfiguration rather than true loss of capability.
Siegler's overlapping waves model [25] is the mainstream developmental position on cognitive change and is directly opposed to staircase models of stage progression. The empirical claim is that children use multiple strategies simultaneously and that within-child variability is substantial at all ages and is itself positively associated with subsequent learning.
In the practical section, the framework may use a staircase model as a binning approach to talk about broad parts of the spectrum. But even in such discussions, the framework holds the modern view that learned state is a continuum over a staircase to be the correct model.
Cognitive measurement, like all measurement, disturbs the state being measured. The methodological tradition tracing to Webb, Campbell, Schwartz & Sechrest (1966) [90] established this empirically: when the same behavior is measured both obtrusively and unobtrusively, the results often differ substantially. A complication specific to cognitive measurement: the state being measured isn't just information or skill but includes the learner's beliefs about themselves and the task. Measurement that asks "can you do this?" disturbs not just what the learner knows but what they believe about whether they can do it.
Disturbance can be reduced through unobtrusive measurement (physiological proxies, naturalistic observation, implicit tests, continuous monitoring), but every method involves a trade-off: lower disturbance generally means measuring correlates rather than the target state directly, or measuring under conditions that limit what can be inferred. Three established findings further constrain what any cognitive assessment can claim:
Practice effects [42]: any repeated formal assessment produces score improvements partly attributable to familiarity with the format rather than to true cognitive change. The same diagnostic given twice produces different results from the same underlying state. Formal assessments thus produce snapshots that are partly measurements and partly teaching, with the proportions varying by item difficulty and learner state.
Multidimensionality of competence [43]: cognitive states are not one-dimensional numerical quantities, despite psychometric conventions that treat them as such. A learner's competence on a skill is a configuration of related abilities — fluency, conceptual understanding, recognition of when the skill applies, ability to execute under variable conditions — that gets projected onto a single number for measurement convenience. The projection loses information that matters for instructional decisions.
Classroom assessment versus psychometric measurement [44]: these are distinct operations with different epistemic properties. Psychometric measurement uses standardized instruments to produce comparable scores across learners; classroom assessment uses contextual observation by an instructor who knows the learner. Tools from the psychometric regime can serve as instruments within classroom assessment but cannot substitute for it, because the contextual information is part of what makes classroom assessment work.
How to measure unobtrusively is outside the scope of this framework.
The conditions under which a skill is practiced shape the representation that forms. Practice in deliberately narrowed conditions — heavy scaffolding, repeated single format, decontextualized drill — produces representations that work within those conditions but fail when conditions change. This is the engineered-isolation effect: representations developed under diverse-input training pressure generalize better than representations developed under narrow-input training [30].
The phenomenon has been demonstrated in laboratory studies of skill isolation. The cognitivist tradition associated with Anderson [3] and ACT-R has shown, in work such as Singley and Anderson's text editor studies [37], that isolation can be deliberately engineered. The framework's position is that engineered isolation, while achievable, produces inferior outcomes than embedded learning, narrow, brittle, fast-decaying competence rather than robust capability.
Two mechanisms support the brittleness prediction. The unified-representation effect [30] establishes that diverse-input training produces representations connecting more broadly to the rest of cognition. Decay vulnerability (T6) compounds the problem: isolated skills, lacking anchors in the broader cognitive system, decay faster once instruction stops because no ambient encounter maintains them. The empirical finding that cognitive/artificial/accuracy-based tasks decay faster than physical/natural/speed-based tasks [28] is consistent with this — the more a skill is deliberately abstracted from its embedding, the faster it loses retrieval strength when not actively rehearsed.
The distinction between learning (durable change) and performance (current behavior under conditions) [13] is foundational here: behavioral output under narrowed conditions can be high-performance without being durable learning, and the discriminator is whether the output survives when conditions change.
This claim manifests in two practical patterns:
A learner who produces correct outputs only when prompted through each step is consistent with either acquisition having occurred or not having occurred; the prompted output does not discriminate between these states. As scaffolding increases, the practice context narrows toward engineered isolation along the active-execution dimension, and the resulting representation becomes increasingly bound to the scaffolded format.
The discriminator is delayed independent performance. If outputs survive removal of prompts (with appropriate delay), the representation extended beyond the scaffolded context. If outputs collapse, the representation was bound to the scaffolding.
A learner who produces correct outputs across multiple sessions in the same format may have acquired a verbal or procedural performance that reflects no underlying schema. Unlike scaffolding-bound brittleness, which self-corrects when prompts are removed, format-bound brittleness can persist indefinitely and resists detection by formal assessment that mirrors the training format. Repeated practice in a single format narrows the practice context toward engineered isolation along the surface-form dimension, and the resulting representation becomes bound to the format.
The discriminator is performance under surface-form variation. If outputs survive perturbation, schema is present. If outputs collapse, additional practice in the original format is contraindicated — it deepens the format-bound representation rather than producing the missing schema.
The claim concerns sustained engineered isolation across the learning trajectory, not within-session focus. A focused thirty-minute session on coin counting does not violate T3 — it is one embedded sub-activity within a varied life, not engineered isolation in the structural sense. Within a single session, focus is genuinely useful because of an unrelated mechanism: cognitive startup cost. Each task has a non-trivial entry cost (orienting to the problem, loading context, getting working memory into the right state), and distractions force the learner to pay this cost repeatedly. Within-session focus and across-session variety are operating at different timescales and do not conflict.
The error T3 warns against is sustained engineered isolation across the whole learning trajectory — drilling for weeks with carefully decontextualized worksheets, isolating a skill from any real-world context for the duration of its acquisition. That produces the brittle, fast-decaying competence the mechanisms above predict.
T4. Some skills require explicit instruction; they are not acquired through ambient encounter alone [literature]
For a small class of skills, no amount of ambient encounter — however rich, and whether active or passive — builds the skill; it must be constructed through deliberate explicit instruction, after which ambient encounter can maintain it.
There is evidence of at least two skills belonging to this class: systematic decoding (the letter-sound code underlying reading) and arithmetic-fact automaticity (instant retrieval of basic facts). For decoding, immersion in rich text without explicit instruction in the letter-sound code tends to produce context-and-picture guessing that can resemble reading until the context stops carrying the load; explicit phonics instruction produces measurable gains in decoding, word reading, and comprehension (Ehri, Nunes, Stahl & Willows 2001 [118], National Reading Panel meta-analysis, overall d = 0.41, larger when begun early). For arithmetic facts, ambient numerical activity builds number sense but not automaticity; automaticity is built specifically by retrieval practice, and it frees working memory for later mathematics (McNeil, Jordan, Viegut & Ansari 2025 [119]). Conceptual understanding and retrieval practice are complementary rather than substitutes — understanding does not by itself produce automatic retrieval, and retrieval practice is the component that builds it. In both cases the ambient activity remains valuable: it builds the scaffolding that makes the explicit instruction faster and more durable. But it is the explicit instruction that builds the skill.
Our examples are narrow and should not be taken to mean that all of language arts or all of math falls into this category. The framework has not surveyed for other members of this class; the two named are those with strong evidence, and other skills should not be assumed into the class without comparable evidence. The framework takes no position on why the class exists and therefore makes no attempt to predict which skills belong to this class and which don't.
Transfer is real. Where two contexts genuinely share schema components, competence in one supports competence in the other — but two conditions must both hold: the overlap must be real (shared components, not merely abstract resemblance), and the learner must recognize it. This position accepts the identical-elements tradition (Singley & Anderson 1989 [113], who demonstrated transfer between text editors sharing components) and the analogical-transfer tradition (Gick & Holyoak), rejecting only far transfer (RT5).
Recognition is the gate, not the existence of shared structure. Gick & Holyoak [108], [109] showed this directly: given a structurally analogous source story, only about 30% of learners spontaneously applied it to a new problem, but about 75% did once cued to recognize the connection — the shared structure was available the whole time; recognition determined whether it was used. Where overlap is obvious (two text editors with shared features), recognition is immediate; where it is less obvious (a story whose structure maps onto a problem), recognition must be prompted; where there is no real overlap, only abstract resemblance, there is nothing to recognize (RT5).
Regardless of the exact mechanism (reconstruction vs. direct use), breadth of knowledge provides more material for new learning (Bransford & Schwartz 1999 [110]).
Phillips [26] proposes a formal theory of relational schema induction as reconstruction by cross-task structural comparison. The dynamic systems perspective [24] treats behavior as softly assembled from contributing components rather than as the deployment of stored schemas. Situated cognition theory [31], [32] holds that knowledge is inseparable from the activity and context in which it is acquired — Lave's research on everyday cognition provides direct support, with children performing market arithmetic they cannot reproduce in school formats, showing that "the same" operation is in fact different cognitive operations in different contexts.
The reconstruction account has structural implications for sequencing. Generality emerges from triangulation across at least two instances of an underlying pattern; a generalizable schema cannot form from a single instance because there is nothing to compare against and the abstractions necessary for transfer are not visible from inside a single domain. This in turn means variation-based practice presupposes a base level of acquisition: there must be something against which variation can be detected. The practical sequencing follows from T5 combined with T1's non-monotonic trajectories and T6's decay dynamics; the practice section operationalizes this sequencing.
The decay of schemas through disuse is a well-established phenomenon (T6); decay matters for T5 because what is remembered of a prior schema is what serves as starting material for the next reconstruction. As decay proceeds, less prior structure is available to seed the next instance.
Path dependence falls out of T5 combined with T10 as a derived consequence rather than as a separate principle. Given that substrate must be present (T10) and new schemas are reconstructed using prior schemas as starting material (T5), when multiple acquisition orderings could satisfy substrate requirements, which prior schemas were available shapes what new schema gets constructed. Different orderings produce different sets of available priors at each construction step, producing different end-state schemas. This asymmetry is visible in second language acquisition: L1 effects on L2 processing persist into high proficiency [33] and L3 acquisition reflects both prior languages' structure [34], because reconstruction in those domains has rich prior-schema configurations that vary substantially across learners. For elementary skills with small prior-schema requirements (counting, basic literacy), reconstruction produces similar outputs across learners regardless of order, because the available priors are similar. The framework does not need a separate path-dependence claim; T5 + T10 produce the prediction.
T5a. Deriving facts new to oneself requires suspending plausibility checking [literature, synthesis]
Personal innovation, a term coined for this framework, is the ability for a person to derive new facts about the world through their own knowledge. Such facts may or may not be novel in the general sense; elevating personal innovation to general innovation is outside the scope of this framework, and is traditionally addressed through study for a doctorate or a business degree.
Personal innovation occurs when a person starts imagining the possible with no plausibility filter, followed by a filter to eliminate the nonsense. Even after the obvious nonsense has been eliminated, one still must validate that the surviving links are real with an external source like a teacher, book, or an experiment, if no expert exists to consult, because the feeling of understanding a self-generated conclusion systematically outruns the actual understanding (Rozenblit & Keil 2002 [117]).
Applying the plausibility filter during generation has two costs: it suppresses the candidates that would otherwise surface (Amabile 1979 [114]; Amabile, Goldfarb & Brackfield 1990 [115]), and it degrades recognition of the good candidates among those that do (Mueller, Melwani & Goncalo 2012 [116]).
Once candidate ideas have been surfaced, the literature on near-transfer takes over (T5), the only difference being that the bridge was self-made rather than supplied by an outside source.
Far transfer — the claim that learning in domain A produces benefit for learning in incompatible domain B by virtue of some abstract shared base — is rejected. Two domains can share an abstract structure (both involve sequential ordering, hierarchical decomposition, or the like), but for that shared structure to be usable by the learner, the learner needs a representational language that bridges the two domains. Until that language forms, the learner has no mechanism to access the shared abstract structure from either domain.
Far transfer (across incompatible domains) fails replication. The far transfer literature [14], [16], [17], [38] consistently produces negative evidence; Barnett & Ceci's taxonomy [15] frames the dimensions along which transfer distance is judged (and cautions that a single effect size for far transfer is misguided).
The framework chooses to believe that ensuring a related base schema is a prerequisite even at the theoretical level, because P5 (prefer offering help) has consistently shown practical value: when in doubt about the learner's substrate, providing more help rather than less reliably produces better outcomes. This practical evidence reinforces the theoretical commitment.
Narrow far-transfer successes can occur when a learner has built enough bridging language to manually transfer between two schemas. Such cases are accommodated by T5's reconstruction mechanism (the bridging language is itself an accumulated representational structure) rather than requiring a separate explanatory mechanism.
Anderson, Reder & Simon (1996) [91] critiqued the situated-cognition tradition's stronger anti-transfer readings, arguing that abstract symbolic representations produce transferable competence when representation and degree of practice are sufficient. This is not a contradiction of T5 but a case T5 already covers. When a schema is abstract enough, it can generalize to multiple areas by being usable as a sub-skill — and grabbing as many such sub-skills as fit a new domain is precisely what ARS call transfer. T5 adds one component ARS leave implicit: some construction in the new target domain also occurs, because the sub-skills don't compose themselves — assembling them into a target-domain schema is reconstruction work that happens within the target context.
Acquired skills and schemas decay over time when not retrieved or practiced. This is a long-established empirical finding. Arthur, Bennett, Stanush, and McNelly [28] meta-analyzed 189 data points from 53 studies and found effect sizes for skill loss ranging from approximately zero immediately post-training to d = -1.4 after more than 365 days of nonuse. The decay pattern is not uniform across skill types: cognitive, artificial, and accuracy-based tasks decay faster than physical, natural, and speed-based tasks.
The Bjork & Bjork "new theory of disuse" [29] supplies the mechanism. Memory has two separable properties:
- Retrieval strength: how accessible a skill is right now. Erodes through disuse.
- Storage strength: durable underlying memory. Persists longer than retrieval strength.
This separation predicts the observed pattern of decay and recovery. Skills become inaccessible (retrieval strength low) but recover faster than initial learning when re-encountered (storage strength persisted). It also explains why the spacing effect works: spaced practice allows retrieval strength to drop between sessions, which produces larger storage-strength gains when retrieval is rebuilt than massed practice does. Cepeda et al. [9] meta-analyzed the spacing effect across distributed practice studies and found robust retention gains for spaced over massed practice across multiple delay intervals.
Skill decay is also a recognized operational concern in safety-critical professions (aviation, medicine, military) where refresher training schedules are designed around documented decay curves.
The rate of decay also varies based on external conditions.
T7. Encounter type affects memory: active production exceeds passive exposure [literature, synthesis]
Independent of decay dynamics (T6), the type of encounter a learner has with a skill affects how memory is updated. The framework distinguishes encounter types along two axes:
| Active production | Passive exposure | |
|---|---|---|
| Deliberate (instructor-scheduled) | Retrieval practice [45] | Restudy |
| Incidental (environment-driven) | Active ambient encounter | Passive ambient encounter |
The testing effect literature establishes that the active-production cells (left column) produce substantially better retention than the passive-exposure cells (right column). Roediger and Karpicke's foundational study [45] found retention 50% higher for tested versus restudied material at one-week delay. The Rowland meta-analysis [46] found g = 0.50 across 61 studies; the Adesope meta-analysis [47] found g = 0.61 across additional studies. The effect generalizes from laboratory to classroom contexts and from simple to complex material. The mechanism, in Bjork & Bjork's terms (T6), is that active production strengthens both retrieval strength and storage strength, while passive exposure strengthens primarily storage strength to a lesser degree.
The framework also distinguishes whether skill recall happens because of explicit instruction or because the environment occasioned it:
Ambient skills (numerical reasoning, literacy, time, money, basic measurement, language, social rules, basic spatial reasoning) are continuously exercised across heterogeneous surface forms in daily life, predominantly through active ambient encounters, with some passive exposure. The diversity of contexts is what makes ambient encounter robust: training under diverse-input pressure produces representations connecting more broadly to the rest of cognition than narrow-input training does (the unified-representation result from machine learning [30]). This is also why ambiently-exercised skills are unusually robust against decay (T6) — they were never single-anchor schemas to begin with.
Explicit skills exist primarily within curricular contexts and receive no ambient encounter. For these skills the instructor controls the entire schedule, and the choice between retrieval practice and restudy has substantial consequences for durability.
At elementary age, almost all formally-instructed skills are also ambient. As specialization narrows the ambient field, typically in secondary education — explicit skills come to dominate, and the type of encounter in scheduled review becomes binding rather than incidental.
T7a. Passive observation and mental practice produce partial but distinct cognitive effects [literature]
Two encounter types short of active practice produce real cognitive effects: passive observation of skilled performance and mental practice (active cognitive rehearsal in the absence of physical execution). Both produce partial versions of what active practice produces, both are complementary with active practice, and combined approaches outperform either alone.
Passive observation. Surgical training studies [56] show measurable performance improvements from observation alone in subsequent active practice. Motor learning research confirms observational learning produces movement representation and coordination patterns through partial activation of execution-related neural circuits [57]. Surgical skill literature [58] establishes that combined observation-and-practice produces better outcomes than practice alone. What passive observation specifically produces: recognition of competent versus incompetent execution, familiarity with how the activity unfolds, vocabulary and conceptual landscape, error detection capability, awareness of strategic patterns, partial movement representations.
Mental practice. The mental practice literature is one of the most replicated findings in skill acquisition research. Feltz and Landers' meta-analysis of 60 studies found g = 0.48 [59]; Driskell, Copper, and Moran's more rigorous follow-up reported d = 0.527 [60], with a 2020 replication confirming the effect holds up. Recent sport-specific work [61] found g = 0.476 overall, with combined mental-plus-physical practice (g = 0.579) outperforming mental practice alone (g = 0.298), which itself outperforms no practice. Effects are larger for skills with substantial cognitive content (strategy, decision-making, sequencing) than for purely motor or strength tasks. Mental practice is particularly effective during periods when physical practice is impossible, which is why the phenomenon shows up clearly in athletes during injury recovery or in extreme cases like prisoners of war maintaining or improving skills they cannot physically perform.
What neither passive observation nor mental practice produces: the predictive model that drives execution under uncertainty in dynamic conditions. Both produce real cognitive products that accelerate active practice when it eventually happens, and both can substitute partially for active practice when active practice is impossible, but neither fully replaces the active prediction-and-consequence loops that build the predictive model. The practical implication: extended observation or mental rehearsal before active practice begins produces faster active-practice acquisition because the recognition layer (and for mental practice, partial cognitive-strategy layer) is already built when active practice begins; active practice can focus on building the predictive model rather than building both layers simultaneously.
All things being equal, positive motivation increases the rate of learning, and negative motivation decreases the rate of learning.
Linnenbrink-Garcia, Tyson, & Patall (2008) [72] reviewed 90+ peer-reviewed articles and found ~40% of effects positive for adaptive motivational orientations and consistently negative effects for avoidance-based orientations. Richardson, Abraham, & Bond (2012) [73] meta-analyzed motivational predictors of academic performance with correlations ranging from -0.14 (avoidance orientation) to 0.59 (positive self-efficacy). Van Yperen, Blaga, & Postmes (2014) [74] replicated the directional pattern across education, sport, and work domains. Ryan and Deci (2000) [41] and decades of subsequent work establish that autonomous motivation produces better learning outcomes than controlled motivation, with hundreds of replicated studies. Multon, Brown, & Lent (1991) [75] meta-analyzed self-efficacy and academic performance across 36 studies finding r = 0.38; Honicke & Broadbent (2016) [76] replicated with r = 0.33.
Effect sizes are modest (r ≈ 0.10 to 0.60) but reliable.
Pekrun's control-value theory [19] establishes that achievement emotions during an activity feed forward into motivation for re-engagement. Positive activating emotions promote re-engagement; negative deactivating emotions suppress it. Camacho-Morles et al. (2021) [77] meta-analyzed 68 studies (sample sizes up to N=31,868 for enjoyment) and found the predicted directional pattern: enjoyment correlates positively with academic performance (ρ = 0.27), anger negatively (ρ = -0.35), boredom negatively (ρ = -0.25). Talsma et al. (2018) [86] strengthens this directionally with longitudinal cross-lagged evidence: performance has a substantial net positive effect on subsequent self-efficacy (β = 0.205, p < 0.001), nearly three times larger than the reverse direction. Mastery experience — actual performance success — is the strongest single predictor of self-efficacy in the literature (Britner & Pajares 2006 [94], Usher & Pajares 2008 [95]).
T9c. Belief-installation interventions produce smaller effects than motivation-learning correlations suggest [literature]
Belief-based motivation interventions produce real but small effects on learning outcomes — smaller than the correlations in T9 would suggest. Growth mindset shows d = 0.03 to 0.10 across 273 studies (Sisk et al. 2018 [83]), confirmed at similar magnitude by the National Study of Learning Mindsets (Yeager et al. 2019 [84], n = 12,490). Grit shows a meta-analytic correlation of r = 0.18 with academic performance, largely overlapping with conscientiousness (Credé et al. 2017 [85], 88 samples, n = 66,807). When self-efficacy is installed in isolation, the effect on subsequent academic performance is small (β = 0.071, Talsma et al. 2018 [86] cross-lagged panel analysis of 11 longitudinal studies, n = 2,688 — the methodologically cleanest estimate available, controlling for the reciprocal structure that inflated earlier estimates). The pattern: motivation correlates with learning at modest magnitudes (T9), but trying to install motivation produces smaller effects than even those correlations suggest.
The Talsma effect is age-moderated. The reciprocal-effects model holds for adults but not for children: in the child subsample, performance uniquely impacts subsequent self-efficacy beliefs, while self-efficacy does not measurably impact subsequent performance (β near zero). For children, the direction is one-way — performance experiences build self-efficacy, but installed self-efficacy does not produce performance gains. This developmental moderation has theoretical grounding: self-efficacy intervention requires metacognitive reflection capacity (Flavell 1979 [96]) that children have not yet developed.
A learner without the prerequisite vocabulary or conceptual repertoire to attach a new schema cannot acquire the schema regardless of practice volume; practice in this condition deposits performance without acquisition. This follows from the literature on prior knowledge gating instructional effectiveness [21], [50]: instructional moves appropriate for one prior knowledge state are inappropriate for another, and when prior knowledge is sufficiently absent, instruction designed for higher prior knowledge fails entirely.
The empirical evidence comes from contrasts at extremes. The Tetzlaff et al. 2025 meta-analysis [50] (60 studies, 5924 participants) compared low prior knowledge learners benefiting from high-assistance instruction (d = 0.505) with high prior knowledge learners benefiting from low-assistance instruction (d = -0.428). These two values are contrasts at the extremes of the prior knowledge gradient; the underlying phenomenon is continuous. The current literature is increasingly framing this as a special case of element interactivity (Chen, Kalyuga & Sweller 2017 [97]): the same instructional content imposes different cognitive loads at different levels of prior knowledge, with the effect emerging from how prior schemas chunk content into manageable element groups. Looking at extremes informs how to think about the continuous gradient, not whether expertise comes in two discrete states.
Absence of substrate is a contraindication for the activity, not a condition that further instruction can overcome at the current session.
The substrate requirement applies not only to the skill itself but to communication about the skill. Goals, explanations, and feedback are themselves cognitive content the learner must parse, and parsing requires substrate from the learner's current language. A goal articulated in vocabulary the learner does not yet possess does not function as a goal regardless of how clearly stated. The practice section operationalizes this constraint, with demonstration as a substitute for verbal goal-setting when verbal substrate is absent.
T13. Biological substrate and accumulated environment interact dynamically across development [literature]
Beyond the prior-schema substrate addressed by T10, biological substrate — including genetic factors and non-learning environmental factors such as nutrition — affects acquisition curves under any given learning environment. Twin and adoption studies estimate heritability of cognitive ability and Big Five personality traits at 40-60% in adulthood [52]. However, heritability is not a fixed quantity. It varies systematically with age, environmental conditions, and developmental stage.
The Wilson Effect [53]: heritability of cognitive ability increases from approximately 41% in early childhood to 55% in adolescence to 66-80% in adulthood. The accepted mechanism is amplification through active gene-environment correlation [54]. As learners gain capacity to select their own environments, genetic predispositions shape environmental selection, and selected environments reinforce predispositions. The framework's existing T7 (encounter type effects) and T9 (autonomous motivation drives selection) jointly predict this dynamic, but what T7 and T9 in isolation do not surface is two consequences. First, non-learning factors (genetic predispositions, nutritional and developmental conditions) contribute as much to acquisition trajectories as any prior schema development through pure learning, which means one universal teaching technique is impossible: biology-environment interactions vary across learners, and the same instructional move operates differently against different substrates. Second, learner activity choice is biased data: voluntary choices reflect predispositions plus accumulated comfort, not just current capability. Following voluntary choice blindly amplifies the existing trajectory through the selection-reinforcement loop and widens gaps between preferred and avoided domains. Briley and Tucker-Drob's meta-analysis [53] (16 studies, 11,500 twin and sibling pairs) found that early genetic influences amplify over development rather than being innovated newly at later ages, consistent with this transactional account.
Heritability is also environment-dependent. A mouse sibling-intervention study [55] showed heritability of cognitive ability dropping from 55% in standard control conditions to non-significant 15% in enriched environments, same genetic differences, attenuated genetic effect when environmental range expanded. Human early childhood intervention research shows analogous patterns: effect sizes depend on initial environmental conditions, with greatest effects when initial deprivation was greatest and modest effects when initial conditions were already supportive.
The implication: what appears as a biological ceiling on acquisition is typically a ceiling under current conditions. Targeted intervention can shift what's achievable. The framework operates on the accumulation side of learning. Biological substrate affects acquisition curves but does not fix them. Trait measurements (Big Five, cognitive ability assessments) compress biological substrate, accumulated experience, and current environmental supports into single summaries; the framework treats them as inputs without claiming to disentangle the contributions.
The framework's claims are calibrated to typical development. Atypical development (neurodevelopmental conditions, specific learning disabilities, sensory processing differences) involves substrate considerations that require clinical assessment by trained professionals and is outside the framework's defended scope.
T14. Teaching a subject is a different but related knowledge type than the subject itself [literature]
Subject matter expertise and teaching capability are distinct kinds of knowledge that operate on the same domain. Practitioners who develop one do not automatically develop the other. The relationship is asymmetric: teaching capability requires subject knowledge as substrate (you cannot teach what you do not know), but subject knowledge does not produce teaching capability without separate development.
Empirical support comes from pedagogical content knowledge (PCK) research [63], which has documented across multiple subjects (chemistry, history, mathematics, medicine) that experienced teachers possess a kind of knowledge that subject matter experts who don't teach lack. PCK is topic-specific within a subject — teaching photosynthesis requires different PCK than teaching chemical bonding. The construct has been productive across 35+ years of research with continued refinement.
This claim has implications across the framework. The stage 5 binning category (in the practice section) operationalizes T14 for instructional decisions — it is the subset of teaching knowledge specifically about decomposing one's own integrated procedural knowledge for transmission. Specific procedures in the practice section that involve full-execution scaffolding require both subject expertise (stage 4) and teaching knowledge (stage 5) for the procedure-novice combination at hand. T13's claim that no universal teaching technique exists draws on T14 — variation in teaching knowledge across practitioners is one source of why the same instructional move works differently.
The doctorate-without-industry pattern (subject expertise without sustained teaching producing practitioners recognized as experts but limited as teachers) is observational evidence for the distinction. The reverse pattern (effective teaching without sufficient subject expertise) is rare because teaching requires the substrate that subject knowledge provides.
T15. Procedural knowledge becomes integrated through accumulated practice, shifting from effortful to automatic execution [literature]
With sufficient varied practice, procedural knowledge transitions from conscious effortful execution (requiring working memory and attention) to automatic execution (requiring minimal conscious attention). This transition has specific cognitive properties: the practitioner can perform the procedure under cognitive load, can perform secondary tasks while executing the procedure, and can no longer easily articulate why each step happens because the integration is below conscious access.
Empirical support is well-established across cognitive psychology and neuroscience. Fitts and Posner [64] articulated the three-stage skill acquisition model (cognitive, associative, autonomous) with the autonomous stage corresponding to integrated execution. Anderson's ACT-R [65] proposes specific computational mechanisms (proceduralization, composition) by which declarative knowledge becomes procedural and procedural knowledge integrates. Logan's instance theory of automaticity [66] proposes accumulated instances as the mechanism — performance shifts from algorithm-based to memory-retrieval-based as instances accumulate. Neuroimaging studies show shifts in brain region involvement as skills automate, with reduced prefrontal cortex activity and increased basal ganglia and motor region activity for motor skills [67].
This claim has implications across the framework. The stage 4 binning category in the practice section operationalizes T15 for instructional decisions — it is the framework's category for "learner has reached integrated execution for this procedure." T14 (teaching is a different knowledge type) connects to T15: integrated procedural knowledge (T15) is what subject matter expertise produces; teaching capability (T14) is a separate knowledge type that may require partial decomposition of the integration. The doctorate-without-industry pattern reflects practitioners who have T15 integration without T14 teaching capability.
Integration is also subject to T6 decay. Once-integrated knowledge can return to conscious effortful execution if disuse is sufficient — practiced procedures abandoned for years no longer execute automatically. Re-integration requires re-accumulation of practice. The rate of decay depends on integration depth and on whether ambient encounter has been sustaining the procedure (T7).
The transition is non-linear at the population level (different practitioners reach integration at different rates depending on substrate, practice quality, and accumulation conditions) but is the expected end-state of sufficient varied practice for typically-developing learners. The framework's stance is that T1's non-monotonic trajectories is the underlying reality, and stage 4 (operationalizing T15) is decision-support binning rather than a strict developmental stage.
Academic burnout — characterized by emotional exhaustion, cynicism, and reduced sense of accomplishment — produces measurable negative effects on academic achievement (Madigan & Curran 2021 [87] meta-analysis: total burnout × achievement r = -0.24 across 29 studies, N = 109,396; reduced-efficacy dimension r = -0.39). The dominant causal model is effort-reward imbalance (Siegrist 1996 [98]): chronic mismatch between effort invested and reward perceived produces the depleted state. Burnout is a depleted state, not merely an absence of engagement; meta-analyzed interventions (mindfulness, CBT, REBT) produce g = 0.90 reduction (Madigan, Kim & Glandorf 2023 [88], 17 studies, n = 2,462).
Declarative memory — memory for facts and events that can be consciously stated — is reconstructed at the moment of recall rather than played back from storage. Bartlett (1932) [71] established memory as reconstructive rather than reproductive. The fallibility this implies is a defining property, not a marginal failure: leading questions reshape what witnesses report (Loftus & Palmer 1974 [111]), and entire events that never happened can be implanted as confident, detailed memories (Loftus & Pickrell 1995 [112]). Even autobiographical memory — the closest thing a person has to a stored personal record — is reconstructed, and corruptible, on each access.
The claim is scoped to declarative memory. Non-declarative memory (procedural skill, conditioning, priming) is a dissociable system supported by different neural structures, with fast, automatic, reliable access rather than the slow, flexible, fallible access of declarative recall; the framework does not claim it is reconstructive and does not adjudicate the boundary between the two systems. Nor does the framework claim a mechanism for reconstruction — how or why the brain reconstructs rather than stores-and-replays is not answered and is not load-bearing. What is load-bearing is narrow: wherever declarative recall enters learning, it enters as reconstruction, not playback.
Procedures derived from theory. Each procedure references the theory claim(s) it operationalizes. A procedure without theory backing is either a finding (theory needs extension) or a flag (folk practice that snuck in unjustified). A theory claim without a procedure is either non-actionable (acceptable but limits usefulness) or an open practical question (note for later).
The practice section uses three related but distinct words deliberately:
Substrate is the framework's theory-layer term, used in T-claims (especially T10). Substrate refers to the cognitive precondition for new learning: prerequisite vocabulary, conceptual repertoire, accumulated schemas. It is what is in the learner's head as a precondition for new acquisition. Substrate is not directly observable; the teacher never has substrate certainty.
Capability is the practice-layer observable. Capability is what the learner can demonstrate — observable performance under specific conditions. The teacher sees capability; capability is the evidence the teacher has access to.
Readiness is the teacher's judgment that the learner has the capability to attempt something new. Readiness is necessarily uncertain because it extrapolates from observed capability (what the teacher has seen) to potential capability (whether observed capability extends to something not yet tried). Readiness judgments are where most teacher decisions about activities happen.
The three terms form a stack: substrate is the cognitive reality; capability is what the teacher can observe of substrate's effects; readiness is the teacher's inferred judgment from observed capability about whether to attempt something new. P-claims operate primarily at the capability and readiness layers because that's what the teacher works with. Where P-claims reference T-layer cognitive claims explicitly, the word "substrate" still appears because that is the theoretical concept being invoked.
This is a teaching framework, not an assessment framework. The questions the practice claims answer are about how to support a specific learner whose history is observed across contexts: what to do next, how to read state, when to push, when to pull back. Questions about verifying claimed capabilities under controlled conditions — the standard assessment-framework problem — are not in scope.
Where formal instruments appear in the framework, they are used as instruments rather than authorities (T2). Standardized tests, curricular checklists (Florida BEST, Common Core, similar standards), and adaptive learning platforms are useful as coverage prompts and as one input among many; their results are weighted against contextual observation rather than overriding it. A periodic check against a curricular checklist to verify breadth is consistent with this framing — the checklist tells the teacher what coverage the standards expect; the teacher's accumulated observation tells them what the learner actually knows.
The framework's practice procedures are designed to work with measurement bounds rather than against them. Short-pace operations across multiple P-claims (warmup, surface-form perturbation, stage binning) function as small repeated observations: each carries low information per instance but low disturbance, accumulating longitudinally toward state estimates useful for action without claiming exactness. Long-pace assessment is approached through accumulation of contextual observations across activities rather than through formal diagnostics, with diversity of contexts substituting for the precision a single instrument cannot provide. The classroom-assessment regime contains the psychometric regime as a tool, not the other way around.
Essence: rigid execution of a fixed plan regardless of what occurs is strictly inferior to flexible teaching that calibrates help-amount and recognizes when to stop.
Within any teaching activity, the teacher's options as the learner responds are: give more help, give less help, or abandon the activity. Rigid execution of a fixed plan regardless of what occurs is strictly inferior to flexible teaching that calibrates help-amount and recognizes when to stop.
Switching teaching medium mid-activity is generally not available. Changing the medium — paper practice to physical manipulatives, worksheets to cooking, whiteboard to block-building — requires planning, materials, setup, and attention transition. Procedure switching within a medium has its own constraints: once committed to a specific procedure, the procedure's cognitive structure determines what makes sense. Both medium and procedure selection are planning-time work; in-the-moment flexibility is in calibrating help-amount within whatever was chosen, plus the abandonment option when the teacher's readiness judgment warrants stopping.
Adaptations available in real time:
- Add help when struggle is unproductive (heading toward T13's failure-driven selection bias)
- Reduce help when the learner is taking over (the handover knob from full-execution scaffolding, generalized)
- Reduce difficulty within the procedure if frustration is mounting toward disengagement
- Increase difficulty within the procedure if fluency is appearing earlier than expected
- Abandon the activity when the teacher's readiness judgment warrants it; plan a different medium or procedure for next time
P1a. Ignoring learner frustration prevents learning → operationalizes T9, T9b [type: execution, motivation]
Essence: a fixed plan that doesn't respond to rising frustration causes the session to fail.
When a learner becomes frustrated and the teacher continues executing the planned activity without responding to the emotional state, productive engagement collapses. Frustration occupies working memory, displaces engagement with the material, and feeds forward into reduced motivation for the next encounter (T9b). Camacho-Morles et al. 2021 [77] meta-analysis documents the directional pattern: anger correlates negatively with academic performance (ρ = -0.35), boredom correlates negatively (ρ = -0.25). The teacher's response options range from light (interpretive coaching) to heavy (abandon the plan), but ignoring is not on the menu.
Essence: the first couple of problems of an activity or activity sequence should be easier than the targeted activity.
Begin every activity with the simplest form of the planned task. The first minute is the assessment.
- Breezes through → push harder; good day
- Lost on easy version → drop down or substitute; not today
- Partial recognition → meet there, recalibrate
This is a small weak observation under T2's measurement bounds. Prior session performance is not carried in; today's state is read fresh from today's behavior (T1). It is structurally distinct from formal placement assessment; it assesses readiness for the immediate activity, not curricular position.
Essence: one can fail a test due to not having skill in test-taking even though they have the skill being tested.
Test-taking conditions are sufficiently different from practical experience that treating it as a skill unto itself is warranted. Tests are typically timed, scaffolding including self-scaffolding is purposefully withheld, information is generally distilled to ensure the principle under test is understood, and feedback occurs at the test giver's preference not at the test taker's preference. Each item listed can be considered a sub-skill to manage.
Direct Instruction (Engelmann's methodology) is a tightly-scripted instructional approach with a substantial empirical record on standardized achievement measures — Project Follow Through used the Metropolitan Achievement Test, Raven's Coloured Progressive Matrices, and Coopersmith Self-Esteem Inventory; Stockard et al. 2018 meta-analyzed 328 studies. Their methodology seems primed to address any test taking shortcomings a person might have.
Essence: when the learner's outputs appear correct in a familiar form, the next move is to vary the surface form to test whether the underlying schema generalizes.
When outputs appear correct, vary the surface form of the task in ways the training did not. Useful perturbations:
- Explanation to a third party (real or imagined: stuffed animal, sibling)
- Translation between representations (verbal → visual → symbolic)
- Application to adjacent problems the training did not cover
- Significant temporal delay combined with format change
If outputs survive, reconstruction has produced an abstraction robust to the variation tested (T5). If outputs collapse, additional practice in the original format is contraindicated; the response is to provide additional instances with surface variation rather than additional repetitions in the original form. Repetition in original form deepens format-bound brittleness (T3b) without producing the multiple-instance comparison from which abstraction emerges.
P4a. Use three different high-element-interactivity contexts to verify generalization → operationalizes T5 [type: judgment]
Essence: to verify a schema has truly generalized rather than remaining context-bound, test it across three different high-element-interactivity contexts.
When testing whether a schema has truly generalized, use at least three contexts that each involve high element interactivity but differ in surface form. Three is a verification heuristic — below this floor, surface differences may not be sufficient to distinguish genuine schema reconstruction from instance-bound performance.
The triangulation logic: a single instance contains contingent details that look structural from inside it. Two instances differ in ways that can be either contingent or structural; the difference between them cannot itself determine which. Three instances — with two prior reference points to compare against — let the contingent and structural separate, since contingent details vary across the three while structural ones persist. Brooks (1975) [27] articulated this in software architecture: "his prior experiences will confirm each other as to the general characteristics of such systems, and their differences will identify those parts of his experience that are particular and not generalizable." The same triangulation logic applies to schema verification — three high-element-interactivity contexts let the verifier separate "the schema generalizes" from "it works in these specific surface conditions."
Each context must involve the learner's own active construction (their own predictions, their own choices under uncertainty, their own consequences). Passive observation of someone else's contexts does not satisfy the heuristic; it produces recognition, vocabulary, and partial models but not verification of genuine generalization.
This is an inversion of the engineering rule of three: where engineering plans for three iterations to build something complex, learning uses three to verify the schema reconstructs across genuinely different instances.
P4 is supported in the desirable-difficulties literature: variation in surface form during learning produces better long-term retention than repeated practice in the same form (Bjork [10]–[13]).
Essence: when uncertain about learner readiness, default to more assistance rather than less.
When the teacher's judgment about learner readiness for an activity is uncertain, default to providing more assistance rather than less. The expertise reversal effect's asymmetry (Tetzlaff et al. 2025 [50]): under-providing assistance to learners who actually needed it is more harmful than over-providing assistance to learners who didn't. The Tetzlaff meta-analysis (60 studies, 5,924 participants) found d = 0.505 for low prior knowledge learners benefiting from high-assistance instruction and d = -0.428 for high prior knowledge learners benefiting from low-assistance instruction, but the asymmetric harm structure means erring toward assistance is the safer default when readiness is unclear.
The asymmetric cost has multiple supporting mechanisms. Under-helping creates failure that erodes the learner's substrate (T10) AND erodes the learner's belief about their capability; over-helping has smaller costs on both dimensions and primarily costs efficiency. Whether the load-bearing mechanism is cognitive substrate, motivational belief, or both, the asymmetric default points the same direction.
This principle applies at both planning time and execution time. At planning time, when uncertainty exists about whether the learner is ready for an intended activity, build the activity with more scaffolding rather than less; the cost of unnecessary scaffolding is small compared to the cost of insufficient scaffolding for a learner who needed it. At execution time, when observed performance is ambiguous and could indicate either struggle that warrants more help or productive difficulty that warrants less, the default direction is toward more help.
P5a. Route help through external attribution when direct help threatens self-construct → operationalizes T9b, T10 [type: judgment, motivation]
Essence: receiving help has motivational costs (per T9b); when direct teacher-to-learner help would threaten the learner's self-construct, providing the same content through an external source (book, video, third party) preserves the help while reducing competence-cost.
Sierksma and Brummelman (2025) [100] found in three preregistered experiments (N = 619, ages 7-9) that both direct help (providing the answer) and indirect help (hints) reduce children's competence feelings, task enjoyment, and challenge-seeking. However, children disliked and misreported receiving direct help more than indirect help, suggesting direct help carries additional competence-cost beyond the cost of receiving any help.
When the teacher's observation suggests direct teacher-to-learner help would threaten the learner's self-construct — refusal of direct help, signs of fragile competence belief, learner seeking to demonstrate independence — routing the same content through an external source (a book, video, third party) preserves the help while reducing the competence-cost. The teacher's role becomes pointing the learner's attention toward what the external source already said, rather than positioning themselves as the source of the answer.
This is not a categorical preference for indirect over direct help: both have motivational costs, and direct help is sometimes the right choice (when the learner is requesting it, when no plausible external source exists, when the relational cost of indirection outweighs the competence-cost of directness).
P5b. Develop the learner's help-seeking metacognition → operationalizes T9b, T15 [type: judgment, motivation]
Essence: knowing when to ask for help versus persist independently is itself a learnable skill; the teacher's role includes developing this metacognition.
The motivational cost of receiving help is reduced when the learner has initiated the help-seeking. This makes developing the learner's help-seeking metacognition — knowing when they're stuck enough to warrant external input versus when they can productively persist — a goal distinct from when to offer help.
Help-seeking is a recognized developmental skill (Nelson-Le Gall 1981 [101]; Newman 1990 [102]; Karabenick 2003 [103]). It involves multiple components: recognizing one needs help, deciding to seek help, identifying who or what to ask, asking appropriately, and processing the help received. Like other metacognitive skills, it develops through practice with feedback over years.
The teacher's role in developing help-seeking includes: respecting refusal of help while signaling availability; making asking for help low-cost (no shame, no implication of incompetence); modeling asking for help in own context (the teacher asking others for help demonstrates help-seeking as normal); and gradually increasing the threshold for offering unsolicited help as the learner develops their own help-seeking patterns.
This goal constrains teacher intervention frequency. A teacher who always provides help prevents the learner from developing help-seeking skill; a teacher who never provides help leaves the learner without needed support. The development trajectory is one where teacher-initiated intervention decreases over time as learner-initiated help-seeking develops.
Essence: embedding skill encounters into the learner's actual life is the most powerful teaching aid available.
Full immersion in a domain provides dense varied inputs — both deliberate mentorship or even ambient exposure across many contexts — makes learning easier. By embedding oneself, one cannot help but pick up language by being around those doing (T7), providing a vocabulary they can use to accelerate learning (T5), and opportunities for natural use reducing the probability of forgetting (T6).
Embedded learning is not unguided learning. The teacher is only teaching if actively monitoring ambient encounters and raising them to active engagement when aligned with teaching goals.
Embedded learning is superior for both teaching and assessment. Isolated evaluation samples a narrow slice of the schema space ensuring the concept is understood in isolation; immersive evaluation samples the full skill in context showing they can utilize it where it matters. A learner with strong isolated performance and no immersion produces brittle representations (T3) — book smart without field competence. Educational systems default to isolated assessment because it scales; for teacher-mediated contexts, this scaling constraint does not bind.
The implication for effort allocation: skills life supplies through ambient encounter may get away with less deliberate practice. However, a person who looks brilliant under the first context may suddenly suffer if they transition to learning skills with no full life embedding due to skill decay becoming far more dominant (T6).
Essence: blocked practice (same form repeatedly) builds initial fluency; only after fluency is established should variation be introduced.
For new skills, establish initial fluency before introducing variation. Premature interleaving produces noise. (See Stage binning: stage 2 calls for blocked practice; stage 3 calls for interleaved.)
The transition from blocked to interleaved practice is itself a desirable-difficulty manipulation; expect performance during practice to degrade when the transition is made (T1), with long-term retention and the moving of nearby skills out of stage 1 into stage 2 (T5). Empirical support for the variation-phase effect comes from multiple research lines on within-subject interleaving (alternating between different items or categories within the same subject area). Paas and van Merriënboer [18] showed that high-variability worked examples produce better transfer than low-variability examples in geometry problem-solving. Brunmair and Richter's meta-analysis [48] of category interleaving (a related but methodologically distinct intervention) found g = 0.42 across 59 studies, with effect size moderated by domain similarity — interleaving works best when items are similar enough to support cross-instance comparison, consistent with T5's reconstruction account and with Kornell and Bjork's discriminative-contrast hypothesis [78]. For mathematics specifically, Rohrer, Dedrick, and Stershic [49] provide direct empirical support for interleaved over blocked practice in school settings.
Activity integration is a continuous spectrum, not a binary. At one end is manufactured practice (single skill, narrow context, decontextualized) where schema forms easily but brittlely. In the middle are composite activities — where multiple skills genuinely combine into a unified task, like a science experiment requiring reading, math, and domain knowledge to accomplish something. The composite is a different cognitive object than the sum of its components: a new integration schema forms, and the component skills are exercised in a context that connects them. T5 (reconstruction across instances) and T7 (active ambient encounter) operate strongly here — the components experience reconstruction in a genuinely shared context, and the integration itself becomes a robust schema. At the far end is incoherent heterogeneity (entirely unrelated activities sequenced together) where there is no shared underlying pattern for reconstruction to operate on, and the benefit collapses.
The framework's expectation is that productive composite activities sit in the middle of this spectrum. The within-subject interleaving literature (Brunmair & Richter [48], Rohrer et al. [49]) directly studies one specific kind of integration: alternating between categories that share an underlying domain. Composite activities of the science-experiment kind extend this principle: the reconstruction mechanism (T5) operates whenever instances of an underlying pattern appear together, regardless of whether the integration is within or across nominal subject boundaries. The empirical support is direct for the within-subject case and follows by extension for the composite-activity case.
P7 is supported by both major EdPsych traditions: blocked-then-interleaved sequencing aligns cognitive load tradition prescriptions for novice acquisition with desirable-difficulties tradition prescriptions for post-acquisition variation (Sweller [1]; Bjork [10]–[13]; Rohrer et al. [49]; Pyke et al. [70]).
Essence: the learner should understand why an activity matters, expressed in vocabulary they can understand.
When introducing a skill, provide an honest answer to "why are we doing this." Acceptable honest answers include practical use, prerequisite for later work that has practical use, and credentialing requirement.
The answer must be SMART from the learner's current language. Goals require substrate to parse (T10), the same way skills do. A goal articulated in vocabulary the learner does not yet have — "learn multiplication" to a kid who has heard the word but has no atomics that map onto multiplication — does not function as a goal regardless of how clearly stated. For learners who lack the verbal substrate for an articulated goal, demonstration substitutes (see full-execution scaffolding): the goal is conveyed in a representation the learner can see without first needing to learn the language for it. As substrate accumulates, the verbal channel becomes available and articulated goals begin to function. Goal-setting research (Locke and Latham [69]) confirms that visible distal goals combined with calibrated proximal engagement accelerate learning, while goals too distant from current capability produce avoidance and learned helplessness — exactly what verbal goal-setting alone produces when the learner lacks substrate to parse the verbal goal.
This translation work is the teacher's responsibility. The teacher knows the learner's current language; an automated layer does not and cannot. The framework provides information about activity options and the learner's accumulated state; the teacher uses this to choose how to introduce the activity in vocabulary the learner can parse. This is one reason the framework's architecture is teacher-mediated rather than direct-to-learner.
Essence: if you must manufacture a why for the student because they won't engage, know it has a cost.
Teachers sometimes need to install a why the learner doesn't have. The learner won't engage with the activity, the learning matters, and an installed why is the only available tool to produce engagement. Common installed whys include threats ("if you don't do this, X happens"), bribes ("if you do this, Y happens"), comparisons ("your sibling could do this at your age"), and manufactured significance ("this is going to be really fun"). While some learning can be better than no learning, using this technique when the learner is already engaged is worse than having no why at all. Effect sizes are modest (correlations around r = 0.10 to 0.30) but reliable across hundreds of studies (Linnenbrink-Garcia et al. 2008 [72]; Richardson et al. 2012 [73]; Van Yperen et al. 2014 [74]).
P8b. Prompt the learner to affirm their own why when it's under pressure → operationalizes T9 [type: motivation]
Essence: prompting the learner to articulate their own commitment is a high-leverage motivation intervention; learner self-affirmation produces substantially larger effects than teacher-imposed affirmation.
Self-affirmation interventions — having the learner articulate what they value and why — produce d = 0.41 (95% CI: 0.36-0.45) across 144 experimental studies in educational settings (Escobar-Soler et al. 2024 [79]). This is substantially larger than the belief-installation interventions covered in T9c (growth mindset d = 0.03 to 0.10; grit r = 0.18; self-efficacy intervention β = 0.071). The structural distinction matters: belief installation tries to update the learner's beliefs from outside; self-affirmation has the learner generate their own articulation, which updates their own commitment more effectively than external speech can.
Effects are moderated by identity threat, participant age, and intervention procedure: self-affirmation is more effective when identity threat is present. When teachers participated directly in the affirmation, effectiveness decreased — teacher-manufactured affirmation is less effective than learner-initiated affirmation. The teacher's role is creating the prompt, not providing the content of the affirmation.
When the teacher detects pressure on the learner's commitment — whether from academic difficulty, social context, or internal doubt — the teacher can prompt the learner to confirm their own why. "Are you sure you want to learn this?" or other prompts that elicit self-articulation from the learner produce the affirmation as the learner's own statement. The answer either confirms commitment or lets the teacher set expectations accordingly. Sometimes the right answer is to not teach until the learner has a why. Sometimes there is no choice but to teach. Sometimes the teaching might be focused on exploration so the learner can find a why.
Essence: do not be afraid to use bin judgments to get a quick snapshot of where a learner is, even though learning itself is continuous and bin judgments cannot be precisely accurate.
Cognition is continuous (T1's non-monotonic trajectories), and the cognitive science consensus correctly rejects strict staircase models of learning [25]. Bin judgments — placing a learner in stage 1, stage 2, stage 3, etc. — cannot be precisely accurate descriptions of cognitive state. The framework agrees with the cognitive science position at the theoretical layer.
At the practice layer, binning is a useful tool that survived in the folk tradition for a reason. Teaching requires action; action requires categorical decisions; categorical decisions require categorical descriptions of state. Bin judgments give a teacher a quick read that they can act on. The alternative — refusing to bin because cognition is continuous — produces analysis paralysis or implicit binning that the teacher hasn't articulated and therefore can't audit.
The framework's stance is: bin freely, but know what you are doing. A bin judgment is a decision-support overlay on continuous reality, not a claim about underlying cognition. The teacher who bins a learner as "stage 3 on long division" is saying "I can act as if this learner is somewhere around the conscious-competence region for this skill," not "this learner has completed stage 3." The first is useful and defensible; the second is a category error.
Stage theories, expertise level frameworks, and classification traditions typically claim their categories describe real psychological states, developmental milestones, or natural kinds. The framework accepts that there is some level of truth to these assertions but stops short of treating these binning frameworks as anything more than engineering shortcuts designed to take impossibly complicated theories and make them tractable for quick decisions.
The framework leverages four stages out of convenience, modifying them from Broadwell's standard descriptions to fit the framework's T-claims:
Stage 1: not ready to learn. The learner is not in a cognitive state where teaching moves apply. Possible structural causes include: substrate absence (no schema for imagining the skill, no relevant vocabulary, no related-skill scaffolding to attach to), corrupted self-model (Dunning-Kruger position [40] where confident imagination is decoupled from actual demands of the skill, evidence-resistant), social-cultural defense (the learner cannot accept being taught for reasons of identity, authority, or relationship), motivational void (the learner has no genuine reason to engage), or developmental constraint outside the framework's typical-development scope (clinical conditions requiring professional assessment). The teacher's first job is diagnostic — figure out which structural problem is operating before any teaching move can land.
Stage 2: gap-aware. The learner has enough metacognitive substrate to recognize gaps in their knowledge. Whether arrived at through demonstration, attempted execution, or other gap-evidence, the learner can now use evidence about their performance to update their self-model. The instructor's job is to provide enough help that they can fill the gaps without becoming frustrated and giving up. Blocked practice is preferred over interleaved here (T5): the learner needs sustained engagement with one type of problem to construct the schema before variation becomes productive.
Stage 3: can perform with conscious thought. The learner can roughly perform the task on their own but performing requires conscious attention. Too much help here hinders further learning by preventing the learner from doing the cognitive work. Interleaved practice is preferred over blocked here (T5): the learner has constructed enough schema that variation across instances now produces the abstraction-via-comparison T5 describes.
Stage 4: forgot how they figured it out. The conscious work that built the schema is no longer accessible to the learner. Execution is automatic. Stage 4 operationalizes T15 (procedural integration through practice) as a binning category. Reachable through sufficient practice; most practitioners reach stage 4 in their domain by accumulating years of varied work. Stage 4 alone often makes for poor teaching of the same material — the integration that produces fluent execution leaves the practitioner without articulation of why each step matters.
Stage 5 (framework's addition): deliberately reconstructed what stage 4 forgot. The practitioner has chosen to take their integrated procedural knowledge apart for transmission purposes. Stage 5 operationalizes T14 (teaching is a different knowledge type than the subject itself). Specifically the cognitive work of taking an engrained schema and reconstructing it as explicit strategies. Includes stage 4 as substrate — the decomposition is added on top of fluent execution rather than replacing it. This is what makes stage 5 instruction superior to stage 3 instruction: the stage-5 instructor has both fluent execution (for smooth demonstration) and decomposition (for articulation), while stage 3 has only the decomposition.
Stage 4 and stage 5 specifically have substantive theoretical grounding (T15 for stage 4's automaticity, T14 for stage 5's decomposition-for-transmission). Stages 1-3 are decision-support binning useful for casual reasoning; the 2-vs-3 boundary in particular is judgment-based, distinguishing "needs help to fill gaps" from "needs space to consolidate."
Stage 5 is local to procedure-novice combinations. Decomposition done for one procedure does not transfer to another, and decomposition for one novice level does not yield decomposition for a different novice level.
We came up with our 5-stage system to help us work through what cognitive science said vs what our intuition said. Practitioners may find this definition unworkable; in such case adopt whichever binning framework better supports your decision making.
Several alternative systems could serve equally well with different trade-off profiles:
- Multiple Intelligences (Gardner). Categorical, trait-based — eight intelligence types used as instantaneous priors for activity-type variety. Structurally different from stage-based systems.
- Vygotsky's Zone of Proximal Development (ZPD). Zone-based, binary or simply-graded. Coarser than stage-based systems; trades expressiveness for ease of application.
- Mastery learning binary. Two states: not-yet-mastered vs mastered. Coarsest of the alternatives but operates well in contexts with limited observation bandwidth.
Substantive treatment of these systems and their epistemic statuses lives in the binning paper.
Essence: trait measurements capture both the learning and non-learning environment into a snapshot that can predict success.
Big Five or similar trait measurements provide compressed initial summaries of biological substrate plus accumulated experience plus current environmental supports for a learner. Use as starting prior for activity recommendations when no direct longitudinal data exists, with explicit recognition that the input compresses three factors the framework cannot disentangle.
For elementary learners, the framework predicts that Big Five measurements have measurable signal even though rank-order stability is lower in childhood than in adulthood. Roberts & DelVecchio (2000) [80] meta-analyzed 152 longitudinal studies (3,217 test-retest correlations) finding stability increases from r = 0.31 in childhood to r = 0.54 in college years to 0.64 by age 30, plateauing at 0.74 by ages 50-70. Big Five effect sizes for academic performance are generally larger at the elementary/middle school level than at later education levels (Mammadov 2022 [81]).
Higher conscientiousness predicts willingness to engage with structured tasks (use targeted practice, longer activities); this is the strongest and most robust Big Five → academic performance association in the literature, accounting for approximately 28% of the variance explained by personality (Mammadov 2022 [81]; Gul-E-Zahra et al. 2026 [82]: β = 0.199). Higher openness predicts willingness to engage with novelty (use varied stages 1-2 transitions); this prediction has particular support at the elementary/middle school level (Mammadov 2022 [81]). Current meta-analyses show small, mixed, or null associations between extraversion, agreeableness, neuroticism and academic performance (Gul-E-Zahra et al. 2026 [82]: extraversion β = -0.062, agreeableness β = 0.034, neuroticism not significant); the framework therefore makes no specific predictions for these traits as starting priors.
As longitudinal observation accumulates, direct evidence supplants the trait input as primary signal. The trait remains as one input among many but no longer dominates routing decisions. By the framework's account, this is structurally honest: trait measurements compress what longitudinal observation reveals more precisely, so when longitudinal data exists, it should dominate.
P11. Composite activities produce integration schemas distinct from component schemas → operationalizes T5, T7, T10 [type: planning]
Essence: a composite activity exercises multiple skills together and produces an integrated capability that none of the component skills produce alone.
Composite activities — cooking, science experiments, projects, community-of-practice work, multi-skill creative tasks — exercise multiple skills together. The integrated capability differs qualitatively from any of the component skills practiced separately. This aligns with Cognitive Flexibility Theory (Spiro et al. 1988 [107]): advanced knowledge in complex domains requires multiple alternative linkages among knowledge elements and schema assembly rather than retrieval of prepackaged schemas. A learner who cooks regularly is building component schemas (math, reading, procedural) AND the linkages that connect them in the cooking context.
T5 (reconstruction across instances) and T7 (active ambient encounter) jointly produce these linkages when activities genuinely combine components rather than sequencing them. A single composite activity does qualitatively different work than the same components practiced separately.
Composite activities are high-element-interactivity tasks (Pyke et al. [70]); T10's substrate-gating handles the cognitive load consideration. P5's prefer-helping default applies when composite activities approach substrate limits.
P12. Full-execution scaffolding for procedure-level substrate gaps → operationalizes T7a, T10, T1, T13, T14 [type: execution] [scoped to one-to-one mentorship requiring both fluent execution (T15) and articulated decomposition (T14) of the specific procedure]
Essence: a mentor can always walk the activity by hand, letting the student engage in only the parts they can handle.
When a procedure exceeds independent attempt but the learner has substrate for the procedure's atomic operations, full-execution-with-atomic-engagement can build procedure-level substrate while maintaining substrate-appropriate engagement. The expert executes the full procedure while the learner participates only at the level of atomic operations they can already perform. T7a (passive observation produces partial cognitive effects) builds the procedure schema components through observation; T10 (substrate as precondition) is satisfied at the atomic level by the learner's existing capability; T1's non-monotonic trajectories are respected because the learner is operating at their own substrate level on the atomic pieces while the procedure structure forms in the background.
The demonstration phase serves multiple functions simultaneously: it builds procedure-level substrate through observation (T7a), it maintains learner engagement through atomic-level participation, and it sets a visible distal goal in a form the learner can apprehend without requiring meta-vocabulary substrate. This last function is important for learners who lack the verbal substrate to parse articulated goals (T10): "you need to learn double-digit addition" requires substrate for "double-digit" and "addition" beyond what the novice can already do. The demonstration substitutes for verbal goal-setting by carrying the goal in a representation the learner can see without first having to learn the language for describing it.
The empirical grounding has two components. First, the worked examples effect from cognitive load theory: studying worked examples produces better learning outcomes than unguided problem-solving for novices, with a meta-analytic mean effect around d = 0.57 (Crissman 2006 meta-analysis [104]; foundational: Sweller & Cooper 1985 [105]). The expertise reversal effect (T10) explains the boundary: worked examples are most beneficial for novices and become redundant or counterproductive for advanced learners. Second, Rogoff's guided participation (1990 [106]) describes the collaborative process where a more skilled partner and learner engage in shared activity, with the expert handling parts beyond the learner's current capability while the learner participates at parts they can perform — a process documented across cultural contexts including caregiving, traditional crafts (Mayan mothers weaving with children), and motor skill acquisition. P12 operationalizes guided participation for the specific case where procedure-level substrate is the gap: the expert executes the full procedure while the learner participates at atomic operations.
The transition from full-execution to learner-attempt is the operative adjustment knob: handover happens when the learner shows recognition of procedure structure (predicting next moves, wanting to take over an atomic piece, asking about the structure rather than the atomics). This corresponds to the fading guidance technique in worked examples research: gradually removing scaffolding as the learner develops independent capability. Premature handover puts the learner into attempt mode before procedure-level substrate is sufficient, risking failure-driven selection bias against the activity (T13). Delayed handover wastes practice opportunity once substrate is sufficient.
Scope: this procedure requires the expert to have both fluent integrated execution (T15) and articulated decomposition for teaching (T14) of the specific procedure. Fluent execution is required for monitoring — recognizing correct atomic outputs from the novice without conscious effort, so that monitoring doesn't consume the cognitive bandwidth that execution requires. Articulated decomposition is required for surfacing procedure structure as the execution unfolds, narrating in ways that make the schema visible to the novice. This is the pedagogical content knowledge Shulman 1986 [63] describes — the subject- and topic-specific teaching knowledge that experienced teachers develop, distinct from subject matter expertise alone. The medical training literature on observational learning of surgical procedures (Custers et al. 1999 [58]) provides one specific application context for the demonstration-then-attempt pattern; the cognitive apprenticeship framework (Collins, Brown & Newman 1989 [62]) describes a related model.
P13. Spaced review with active recall for non-ambient skills → operationalizes T6, T7 [type: planning]
Essence: when life isn't supplying the spacing through ambient encounter, the teacher must supply it through scheduled review with active recall.
Spacing is a universal requirement for skill maintenance. Skills decay through disuse (T6) — Arthur, Bennett, Stanush, and McNelly's meta-analysis [28] (189 data points from 53 studies) found effect sizes for skill loss ranging from approximately zero immediately post-training to d = -1.4 after more than 365 days of nonuse. Cepeda et al.'s meta-analysis [9] of 254 effect sizes across 14-day to 350-day retention intervals confirms that distributed (spaced) practice produces robust retention gains over massed practice across multiple delay intervals. Active production substantially outperforms passive exposure for retention (T7): the Rowland meta-analysis [46] found g = 0.50 across 61 studies; the Adesope meta-analysis [47] found g = 0.61. These claims jointly establish that any skill the framework wants to maintain requires periodic spaced encounters with active recall, not occasional passive review. Effect sizes vary by domain and setting; recent meta-analyses in mathematics [92] and online learning contexts [93] suggest smaller effects than the original lab tradition reported.
Spacing can be satisfied two ways:
-
Ambient encounter (T7's ambient classification). Life supplies spaced encounters automatically. A learner who counts daily through real activities is getting spaced practice on counting whether or not anyone scheduled it. When ambient encounter is dense enough to satisfy the spacing requirement, scheduled review is unnecessary — life is doing the work.
-
Scheduled review. When ambient encounter is absent or sparse for a skill, the teacher must supply the spacing through deliberate review sessions. Active recall (retrieval practice) is preferred over restudy (passive re-exposure); the active-recall advantage is among the largest and most replicated effects in cognitive psychology, with Roediger and Karpicke's foundational study [45] finding that after a one-week delay students recalled more of the material they had been tested on than material they had restudied — 56% versus 42% after a single test, and 61% versus 40% with repeated retrieval.
The classification step (T7's ambient vs explicit distinction) determines which path applies. For ambient skills, the framework's prescription is to leave well enough alone — life is providing the spacing automatically. For explicit skills (curricular-only skills receiving no ambient encounter), scheduled review with active recall is the framework's prescription, consistent with the broader spaced-repetition tradition that has accumulated since Ebbinghaus (1885) [8].
The framework's stance is that spacing is universal but the mechanism for achieving it is context-dependent. Tools that schedule review based on spacing intervals (whether classroom flashcard practice, modern spaced-repetition software, or AIED scheduling systems) are doing the right thing for explicit skills. They risk over-application when used for skills the learner is also encountering ambiently — partial duplication of maintenance work, which is wasteful but not harmful. They risk under-application when skills classified as ambient turn out not to be (the brilliant kindergartener who hits a wall when life specializes and ambient encounters dry up; T7 covers this transition).
Specific scheduling intervals are not prescribed. The forgetting-curve literature suggests expanding intervals (short initial review, longer subsequent reviews as storage strength builds) but the optimal schedule depends on the specific learner, skill, and context. Modern algorithms use response feedback to adapt intervals; the framework treats these as legitimate engineering tools without endorsing any specific implementation.
Essence: practicing past mastery has sharp diminishing returns; the practical stop-condition for most skills is mastery, not automation.
For competent learners on low-element-interactivity tasks, the literature documents a ceiling effect: increased difficulty no longer produces measurable learning gains (Pyke et al. 2024 [70]). Practice continues to produce automation, execution requires no conscious effort, as a different outcome (T15), but the gain per unit of practice drops sharply. For most skills, mastery (functional usability) is the appropriate stop-condition; further focused practice doesn't pay for itself and time can be better positioned on downstream skills or encouraging independent study. The exception: skills that need to become substrate for further learning (basic arithmetic for problem solving, letter recognition for reading, scales for music) warrant the additional practice because automation frees cognitive resources for higher-level work.
Essence: mastery experience — actual performance success — is the strongest single source of self-efficacy in the literature; build motivation through enabling success rather than through belief installation.
Belief-based motivation interventions (growth mindset, grit training, self-efficacy work) produce real but modest effects (per T9c). Mastery experience — actual performance success — is the strongest single source of self-efficacy (per T9b). Build motivation through enabling performance success rather than through belief installation.
P17. Recognize burnout signs and respond with workload reduction plus recovery support → operationalizes T16 [type: motivation]
Essence: burnout requires active intervention, not just workload removal; recognize early signs and respond with reduced load plus recovery support.
Burnout protective factors identified in the literature (Sukamto et al. 2026 [89], systematic review of 50 studies) include sustainable workload, adaptive learning and coping strategies, autonomous motivation (the learner's own why for the work), social support, and resilience-supportive environment — these are the conditions to maintain to prevent burnout. The type of social support matters: Messina et al. (2025) [99] found that venting through interpersonal emotion regulation paradoxically exacerbates burnout by reducing social support, while reassurance-seeking helps. When early signs appear (declining engagement despite continued effort, cynicism toward previously-engaging activities, sense of inability to accomplish despite effort), respond with reduced workload plus recovery support structured around reassurance about capability and the temporary nature of the challenge rather than open-ended venting about how difficult or unpleasant the work is (mindfulness, CBT, or REBT interventions are evidence-backed per T16), not just curriculum changes.
- Composite activities as the typical case: T5 (reconstruction) and T7 (ambient encounter) together establish that composite activities are not a special design challenge but the structural default for skills with dense ambient exposure. T3 adds that deliberately narrowing toward isolation is counterproductive even when achievable. A practical procedure for designing composite activities deliberately, with attention to which abstractions the variation across components is meant to produce, is open work, but not blocked on missing theory.
Note (audit in progress): Reference annotations may contain stale P-number cross-references. The framework is currently being decoupled so that T-claims, the appendix, and other stable content do not reference specific P-numbers (which are subject to restructuring). Reference annotations were left for a single cleanup pass after P-claim restructuring is complete. Where an annotation says "Cited for Pn..." this should be read as descriptive of which claim the reference supports, with the P-number expected to be re-checked.
[1] J. Sweller, "Cognitive load during problem solving: Effects on learning," Cognitive Science, vol. 12, no. 2, pp. 257–285, 1988.
[2] J. J. G. van Merriënboer and P. A. Kirschner, Ten Steps to Complex Learning, 3rd ed. Routledge, 2018.
[3] J. R. Anderson, "Acquisition of cognitive skill," Psychological Review, vol. 89, no. 4, pp. 369–406, 1982.
[4] M. Broadwell, "Teaching for learning (XVI)," The Gospel Guardian, 1969. Cited as structural source; folk-pedagogical, not directly empirically validated, but consistent with [3].
[5] (intentional gap; reference removed — the framework is silent on the Vygotskian sociocultural/ZPD account, which operates at the learner–teacher dyad and would need its own framework to treat faithfully; the practice layer's scaffolding content is grounded in worked-examples and cognitive-apprenticeship literature instead.)
[6] (intentional gap; reference removed alongside [5] when the sociocultural-foundations note was dropped — Chaiklin's ZPD-interpretation critique is no longer cited.)
[7] (intentional gap; reference removed during citation audit — Smagorinsky 2011 concerns Vygotsky's psychology of art and perezhivanie, not the zone of proximal development; the ZPD-interpretation point it was cited for is fully supported by Chaiklin [6].)
[8] H. Ebbinghaus, Memory: A Contribution to Experimental Psychology. Teachers College, Columbia University, 1913 (orig. 1885).
[9] N. J. Cepeda, H. Pashler, E. Vul, J. T. Wixted, and D. Rohrer, "Distributed practice in verbal recall tasks: A review and quantitative synthesis," Psychological Bulletin, vol. 132, no. 3, pp. 354–380, 2006.
[10] R. A. Bjork, "Memory and metamemory considerations in the training of human beings," in Metacognition: Knowing about Knowing, J. Metcalfe and A. Shimamura, Eds. MIT Press, 1994, pp. 185–205.
[11] E. L. Bjork and R. A. Bjork, "Making things hard on yourself, but in a good way: Creating desirable difficulties to enhance learning," in Psychology and the Real World, 2nd ed., M. A. Gernsbacher and J. Pomerantz, Eds. Worth, 2014, pp. 59–68.
[12] R. A. Bjork and E. L. Bjork, "Desirable difficulties in theory and practice," Journal of Applied Research in Memory and Cognition, vol. 9, no. 4, pp. 475–479, 2020.
[13] N. C. Soderstrom and R. A. Bjork, "Learning versus performance: An integrative review," Perspectives on Psychological Science, vol. 10, no. 2, pp. 176–199, 2015. Canonical articulation of the learning-vs-performance distinction: performance during practice is observable but unreliable as an index of long-term learning. Certain manipulations improve performance while reducing long-term learning; others reduce performance while improving long-term learning. The framework draws on this distinction throughout but operationalizes it through immersion observation (P6) rather than through controlled-conditions testing — the framework is a teaching framework, and the controlled-conditions testing approach belongs in assessment frameworks instead.
[14] D. K. Detterman, "The case for the prosecution: Transfer as an epiphenomenon," in Transfer on Trial: Intelligence, Cognition, and Instruction, D. K. Detterman and R. J. Sternberg, Eds. Ablex, 1993, pp. 1–24.
[15] S. M. Barnett and S. J. Ceci, "When and where do we apply what we learn? A taxonomy for far transfer," Psychological Bulletin, vol. 128, no. 4, pp. 612–637, 2002.
[16] G. Sala and F. Gobet, "Does far transfer exist? Negative evidence from chess, music, and working memory training," Current Directions in Psychological Science, vol. 26, no. 6, pp. 515–520, 2017.
[17] G. Sala et al., "Near and far transfer in cognitive training: A second-order meta-analysis," Collabra: Psychology, vol. 5, no. 1, art. 18, 2019.
[18] F. Paas and J. J. G. van Merriënboer, "Variability of worked examples and transfer of geometrical problem-solving skills: A cognitive-load approach," Journal of Educational Psychology, vol. 86, no. 1, pp. 122–133, 1994. Empirical study on worked-example variability in geometry; cited for P6 as one line of evidence for variation-supports-transfer. Methodologically distinct from category interleaving studies aggregated in [48], so represents independent supporting evidence rather than the same evidence cited twice.
[19] R. Pekrun, "The control-value theory of achievement emotions: Assumptions, corollaries, and implications for educational research and practice," Educational Psychology Review, vol. 18, no. 4, pp. 315–341, 2006.
[20] (intentional gap; reference removed during verification — Craik & Lockhart 1972 levels-of-processing not load-bearing for current framework, which uses Bjork's storage/retrieval strength account [29] instead.)
[21] S. Kalyuga, P. Ayres, P. Chandler, and J. Sweller, "The expertise reversal effect," Educational Psychologist, vol. 38, no. 1, pp. 23–31, 2003.
[22] S. Strauss, Ed., U-Shaped Behavioral Growth. Academic Press, 1982.
[23] F. Pauls, T. Macha, and F. Petermann, "U-shaped development: An old but unsolved problem," Frontiers in Psychology, vol. 4, art. 301, 2013.
[24] L. B. Smith and E. Thelen, "Development as a dynamic system," Trends in Cognitive Sciences, vol. 7, no. 8, pp. 343–348, 2003.
[25] R. S. Siegler, Emerging Minds: The Process of Change in Children's Thinking. Oxford University Press, 1996.
[26] S. Phillips, "A reconstruction theory of relational schema induction," PLOS Computational Biology, vol. 17, no. 1, art. e1008641, 2021.
[27] F. P. Brooks Jr., The Mythical Man-Month: Essays on Software Engineering, Anniversary Ed. Addison-Wesley, 1995 (orig. 1975). The second-system effect and third-system convergence are in Ch. 5; "plan to throw one away" is in Ch. 11; its retraction as too simplistic appears in the 1995 anniversary additions (Ch. 19).
[28] W. Arthur Jr., W. Bennett Jr., P. L. Stanush, and T. L. McNelly, "Factors that influence skill decay and retention: A quantitative review and analysis," Human Performance, vol. 11, no. 1, pp. 57–101, 1998.
[29] R. A. Bjork and E. L. Bjork, "A new theory of disuse and an old theory of stimulus fluctuation," in From Learning Processes to Cognitive Processes: Essays in Honor of William K. Estes, vol. 2, A. F. Healy, S. M. Kosslyn, and R. M. Shiffrin, Eds. Erlbaum, 1992, pp. 35–67.
[30] C. Raffel, N. Shazeer, A. Roberts, K. Lee, S. Narang, M. Matena, Y. Zhou, W. Li, and P. J. Liu, "Exploring the limits of transfer learning with a unified text-to-text transformer," Journal of Machine Learning Research, vol. 21, no. 140, pp. 1–67, 2020. Cited as analogical reference for the unified-representation argument; not invoked as direct evidence for cognitive claims.
[31] J. S. Brown, A. Collins, and P. Duguid, "Situated cognition and the culture of learning," Educational Researcher, vol. 18, no. 1, pp. 32–42, 1989.
[32] J. Lave, Cognition in Practice: Mind, Mathematics and Culture in Everyday Life. Cambridge University Press, 1988.
[33] K. Erdocia, A. Zawiszewski, and I. Laka, "Word order processing in a second language: From VO to OV," Journal of Psycholinguistic Research, vol. 43, no. 6, pp. 815–837, 2014. Cited as representative of the broader ERP literature on L1 transfer effects on L2 processing at high proficiency.
[34] J. Rothman, J. González Alonso, and E. Puig-Mayenco, Third Language Acquisition and Linguistic Transfer. Cambridge University Press, 2019. Linguistic Proximity Model and other formal accounts of how prior languages shape new-language acquisition.
[35] (intentional gap; reference removed during T12 dissolution.)
[36] (intentional gap; reference removed during T12 dissolution.)
[37] M. K. Singley and J. R. Anderson, The Transfer of Cognitive Skill. Harvard University Press, 1989. Cited as evidence that isolation is constructable under laboratory conditions, supporting T3's claim that engineered isolation is achievable. The framework treats this work as compatible with T3 rather than as opposition: it shows isolation can be constructed, while T3 adds that the resulting representations are brittle compared to embedded learning.
[38] F. Gobet and G. Sala, "Cognitive training: A field in search of a phenomenon," Perspectives on Psychological Science, vol. 18, no. 1, pp. 125–141, 2023. Most current statement of the position the framework sides with at RT5; explicit response to Green et al. [39] and to the field's continued investment in transfer-as-portage despite the negative-evidence wave.
[39] C. S. Green, D. Bavelier, A. F. Kramer, et al., "Improving methodological standards in behavioral interventions for cognitive enhancement," Journal of Cognitive Enhancement, vol. 3, no. 1, pp. 2–29, 2019. Consensus article by 48 researchers representing the position RT5 rejects. Cited so that future readers (and future-me) can see what the framework has chosen to reject and verify the framework's characterization of it.
[40] J. Kruger and D. Dunning, "Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments," Journal of Personality and Social Psychology, vol. 77, no. 6, pp. 1121–1134, 1999. Foundational paper on the Dunning-Kruger effect; cited as the grounding for stage 1 in the practical stage binning. The structural claim that the skills needed to perform are the same skills needed to evaluate one's performance is what makes unconscious incompetence a stable state rather than a self-correcting one.
[41] R. M. Ryan and E. L. Deci, "Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being," American Psychologist, vol. 55, no. 1, pp. 68–78, 2000. Foundational paper on self-determination theory; cited as the grounding for T9. The claim that providing rationale for tasks supports autonomous motivation, and that autonomous motivation produces better learning outcomes than controlled motivation, is the canonical SDT finding.
[42] J. P. Hausknecht, J. A. Halpert, N. T. Di Paolo, and M. O. Moriarty Gerrard, "Retesting in selection: A meta-analysis of coaching and practice effects for tests of cognitive ability," Journal of Applied Psychology, vol. 92, no. 2, pp. 373–385, 2007. Meta-analysis establishing the magnitude and consistency of practice effects across cognitive ability tests; cited as the empirical grounding for T2's claim that repeated formal assessment produces score changes attributable to format familiarity rather than true cognitive change.
[43] A. Scharaschkin, "Educational assessment without numbers," Frontiers in Psychology, vol. 15, art. 1399317, 2024. Cited for T2's multidimensionality claim. Argues that the assumption of quantitative structure for cognitive proficiency is unlikely to be tenable, and that the projection of multidimensional competence onto a single number for measurement convenience loses information that matters.
[44] S. M. Brookhart and J. H. McMillan, Eds., Classroom Assessment and Educational Measurement. Routledge, 2019. NCME volume on the distinction between classroom assessment and large-scale psychometric measurement. Cited for T2's claim that these are different operations with different epistemic properties, and that classroom assessment relies on contextual information that psychometric measurement cannot replicate.
[45] H. L. Roediger and J. D. Karpicke, "Test-enhanced learning: Taking memory tests improves long-term retention," Psychological Science, vol. 17, no. 3, pp. 249–255, 2006. Foundational empirical study on the testing effect; cited for T7's vocabulary and the empirical claim that retrieval practice substantially outperforms restudy. At a one-week delay the study found higher recall for tested than restudied material: 56% versus 42% after a single test (Exp. 1) and 61% versus 40% with repeated retrieval (Exp. 2) — the study's reported recall rates, not a single headline figure.
[46] C. A. Rowland, "The effect of testing versus restudy on retention: A meta-analytic review of the testing effect," Psychological Bulletin, vol. 140, no. 6, pp. 1432–1463, 2014. Meta-analysis across 61 studies; cited for T7. The mean effect size of g = 0.50 establishes the testing effect's robustness.
[47] O. O. Adesope, D. A. Trevisan, and N. Sundararajan, "Rethinking the use of tests: A meta-analysis of practice testing," Review of Educational Research, vol. 87, no. 3, pp. 659–701, 2017. Larger meta-analysis comparing retrieval practice to other learning strategies; cited for T7. Mean effect size g = 0.61 confirms and extends the Rowland findings.
[48] M. Brunmair and T. Richter, "Similarity matters: A meta-analysis of interleaved learning and its moderators," Psychological Bulletin, vol. 145, no. 11, pp. 1029–1052, 2019. Comprehensive meta-analysis of the interleaving effect; cited for P6. Overall effect size g = 0.42 across 59 studies, with similarity between items moderating effect substantially. Provides modern empirical grounding for the variation-phase claim, replacing older Paas & van Merriënboer 1994 work that was load-bearing in earlier versions of the framework.
[49] D. Rohrer, R. F. Dedrick, and S. Stershic, "Interleaved practice improves mathematics learning," Journal of Educational Psychology, vol. 107, no. 3, pp. 900–908, 2015. Direct empirical study showing interleaved practice improves mathematics learning in school settings; cited for P6 as domain-specific support for the variation-phase claim.
[50] L. Tetzlaff, B. A. Simonsmeier, T. Peters, and G. Brod, "A cornerstone of adaptivity – A meta-analysis of the expertise reversal effect," Learning and Instruction, vol. 98, Article 102142, 2025. Meta-analysis of 60 studies and 5924 participants confirming the expertise reversal effect (d = 0.505 for low prior knowledge with high assistance; d = -0.428 for high prior knowledge with low assistance) and establishing its asymmetry — under-providing assistance to novices is more harmful than over-providing to experts. Effect is moderated by educational status and by content domain. Cited for T10 as modern empirical grounding and for P5 as practical guidance under uncertainty.
[51] (intentional gap; reference removed during T12 dissolution.)
[52] K. L. Jang, W. J. Livesley, and P. A. Vernon, "Heritability of the Big Five personality dimensions and their facets: A twin study," Journal of Personality, vol. 64, no. 3, pp. 577–592, 1996. Foundational twin study estimating heritability of Big Five traits at 41-61%; cited for T13's empirical basis on heritability of cognitive ability and personality. Subsequent meta-analyses have produced consistent estimates in the 40-60% range across studies.
[53] D. A. Briley and E. M. Tucker-Drob, "Explaining the increasing heritability of cognitive ability across development: A meta-analysis of longitudinal twin and adoption studies," Psychological Science, vol. 24, no. 9, pp. 1704–1713, 2013. Meta-analysis of 16 longitudinal studies with 11,500 twin and sibling pairs establishing the Wilson Effect (heritability increasing across development) and identifying amplification of early genetic influences as the primary mechanism rather than innovation of new genetic influences at later ages. Cited for T13.
[54] E. M. Tucker-Drob, D. A. Briley, and K. P. Harden, "Genetic and environmental influences on cognition across development and context," Current Directions in Psychological Science, vol. 22, no. 5, pp. 349–355, 2013. Review article articulating the transactional model of gene-environment correlation that underlies the Wilson Effect. Cited for T13 as theoretical grounding for active gene-environment correlation as the mechanism behind increasing heritability with age.
[55] B. Sauce, S. Bendrath, M. Herzfeld, D. Siegel, C. Style, S. Rab, J. Korabelnikov, and L. D. Matzel, "The impact of environmental interventions among mouse siblings on the heritability and malleability of general cognitive ability," Philosophical Transactions of the Royal Society B: Biological Sciences, vol. 373, no. 1756, art. 20170289, 2018, doi: 10.1098/rstb.2017.0289. Sibling intervention study showing heritability of cognitive ability dropping from 55% (control conditions) to non-significant 15% (environmental enrichment), same genetic differences, attenuated genetic effect under expanded environmental range. Cited for T13's claim that heritability is environment-dependent, not a fixed property of the trait.
[56] D. J. Harris, S. J. Vine, M. R. Wilson, J. S. McGrath, M.-E. LeBel, and G. Buckingham, "The effect of observing novice and expert performance on acquisition of surgical skills on a robotic platform," PLOS ONE, vol. 12, no. 11, art. e0188233, 2017. Surgical skill acquisition study showing measurable performance improvements from observational learning alone, with comparable benefits from observing expert and error-strewn performance. Cited for T7a as evidence that passive observation produces real cognitive effects on subsequent execution.
[57] R. M. Hardwick, S. Caspers, S. B. Eickhoff, and S. P. Swinnen, "Neural correlates of action: Comparing meta-analyses of imagery, observation, and execution," Neuroscience & Biobehavioral Reviews, vol. 94, pp. 31–44, 2018, doi: 10.1016/j.neubiorev.2018.08.003. Reference for the claim that observational learning is subserved by partial activation of execution-related neural circuits, distinct from but overlapping with active practice. Cited for T7a's mechanism account.
[58] E. J. F. M. Custers, G. Regehr, W. McCulloch, C. Peniston, and R. Reznick, "The effects of modeling on learning a simple surgical procedure: 'See one, do one' or 'see many, do one'?," Advances in Health Sciences Education, 1999. Foundational study on observational learning in surgical skill acquisition establishing that combined observation-and-practice outperforms practice alone. Cited for T7a's complementarity claim.
[59] D. L. Feltz and D. M. Landers, "The effects of mental practice on motor skill learning and performance: A meta-analysis," Journal of Sport Psychology, vol. 5, pp. 25-57, 1983. Foundational meta-analysis of 60 studies establishing mental practice produces measurable performance effects (g = 0.48). Cited for T7a's mental practice claim. Subsequent replication in 2020 confirmed effect holds.
[60] J. E. Driskell, C. Copper, and A. Moran, "Does mental practice enhance performance?," Journal of Applied Psychology, vol. 79, no. 4, pp. 481-492, 1994. Rigorous meta-analytic follow-up to Feltz and Landers reporting d = 0.527 effect size. Cited for T7a as confirmation that mental practice effects are robust under tighter methodological controls.
[61] R. S. Lindsay, P. Larkin, A. Kittel, and M. Spittle, "Mental imagery training programs for developing sport-specific motor skills: a systematic review and meta-analysis," Physical Education and Sport Pedagogy, vol. 28, no. 4, pp. 444–465, 2023, doi: 10.1080/17408989.2021.1991297. Recent meta-analysis of 58 effect sizes finding g = 0.476 overall, with combined mental-plus-physical practice (g = 0.579) outperforming mental practice alone (g = 0.298). Cited for T7a's claim that mental practice is complementary with physical practice and effects vary by task type (cognitive content moderates effect size).
[62] A. Collins, J. S. Brown, and S. E. Newman, "Cognitive apprenticeship: Teaching the crafts of reading, writing, and mathematics," in Knowing, Learning, and Instruction: Essays in Honor of Robert Glaser, L. B. Resnick, Ed. Hillsdale, NJ: Erlbaum, 1989, pp. 453-494. Foundational paper articulating the cognitive apprenticeship framework — modeling, coaching, scaffolding, articulation, reflection, exploration. Cited for P11 as adjacent literature; the framework's full-execution scaffolding overlaps with cognitive apprenticeship's modeling phase but operates with somewhat different assumptions about learner role and explicit attention to expert stage 4 + stage 5 requirements.
[63] L. S. Shulman, "Those who understand: Knowledge growth in teaching," Educational Researcher, vol. 15, no. 2, pp. 4-14, 1986; and L. S. Shulman, "Knowledge and teaching: Foundations of the new reform," Harvard Educational Review, vol. 57, no. 1, pp. 1-22, 1987. Foundational papers introducing pedagogical content knowledge (PCK) as a specialized form of professional knowledge distinguishing experienced teachers from subject matter experts. Cited for T14 (teaching is a different knowledge type) and stage 5 binning. PCK research observes related phenomena (subject-and-topic-specific teaching knowledge that develops through teaching practice) and provides empirical grounding across multiple subjects.
[64] P. M. Fitts and M. I. Posner, Human Performance. Belmont, CA: Brooks/Cole, 1967. Foundational text introducing the three-stage skill acquisition model: cognitive (figuring out what to do), associative (working out how), autonomous (execution becomes automatic). Cited for T15 as the canonical articulation of the conscious-to-automatic transition in skill acquisition. The autonomous stage corresponds to what Broadwell calls unconscious competence and what the framework's stage 4 binning category names.
[65] J. R. Anderson, The Architecture of Cognition. Cambridge, MA: Harvard University Press, 1983; and J. R. Anderson, D. Bothell, M. D. Byrne, S. Douglass, C. Lebiere, and Y. Qin, "An integrated theory of the mind," Psychological Review, vol. 111, no. 4, pp. 1036-1060, 2004. Foundational papers articulating ACT-R (Adaptive Control of Thought–Rational) as a cognitive architecture with explicit mechanisms for skill acquisition: proceduralization (declarative-to-procedural transition) and composition (combining procedural rules into integrated execution). Cited for T15 as theoretical grounding for the cognitive mechanisms by which integration develops. ACT-R operationalizes the Fitts-Posner stages computationally.
[66] G. D. Logan, "Toward an instance theory of automatization," Psychological Review, vol. 95, no. 4, pp. 492-527, 1988. Foundational paper proposing accumulated instances as the mechanism for automaticity development. Performance shifts from algorithm-based execution to memory-retrieval-based execution as instances accumulate. Cited for T15 as alternative theoretical mechanism (compatible with ACT-R but differing in computational details) for how integration develops through practice.
[67] A. Floyer-Lea and P. M. Matthews, "Changing brain networks for visuomotor control with increased movement automaticity," Journal of Neurophysiology, vol. 92, no. 4, pp. 2405–2412, 2004, doi: 10.1152/jn.01092.2003. fMRI study of motor skill learning: as performance becomes automatic, activity in prefrontal and other cortical regions progressively decreases while activity in subcortical motor regions (putamen/basal ganglia, thalamus, cerebellum) increases. Cited for T15's claim that brain-region involvement shifts as skills automate.
[68] (intentional gap; duplicate of [13], removed during verification.)
[69] E. A. Locke and G. P. Latham, "Building a practically useful theory of goal setting and task motivation: A 35-year odyssey," American Psychologist, vol. 57, no. 9, pp. 705-717, 2002. Foundational synthesis of Locke and Latham's goal-setting research program (40+ years, hundreds of studies). Key findings: specific challenging goals produce better performance than vague or easy goals; proximal goals outperform distal-only goals during early skill acquisition; goals too distant from current capability produce avoidance and learned helplessness rather than acceleration. Cited for P8 (honest purpose, including the SMART-from-current-language constraint). The framework's commitment that goals require substrate to parse (T10) is consistent with Locke and Latham's finding that goals too distant from current capability backfire — the framework's account explains why through cognitive substrate; theirs through self-efficacy and expectancy mechanisms.
[70] W. Pyke, J. Lunau, and A.-H. Javadi, "Does difficulty moderate learning? A comparative analysis of the desirable difficulties framework and cognitive load theory," Quarterly Journal of Experimental Psychology, vol. 78, no. 10, pp. 2181-2195, 2024. Recent comparative analysis of the field's two major traditions on the role of difficulty in learning. The Desirable Difficulties Framework (DDF, Bjork tradition) holds that introducing difficulty during learning produces deeper encoding and better long-term retention. Cognitive Load Theory (CLT, Sweller tradition) holds that conditions for learning are best when extraneous load is reduced and intrinsic load optimized. The paper proposes a unified model conditioned on element interactivity (low vs high) and learner expertise (novice vs expert): low element interactivity + novice → increase difficulty; low + expert → ceiling; high + novice → decrease difficulty; high + expert → increase for refinement. Cited as evidence the field has begun synthesizing rather than choosing between DDF and CLT, supporting the framework's joint use of both traditions through P-claim conditions.
[71] F. C. Bartlett, Remembering: A Study in Experimental and Social Psychology. Cambridge University Press, 1932. Foundational work establishing memory as reconstructive rather than reproductive. Cited for T5 (reconstruction-not-transfer) and refutation log entry on convergent support for T5.
[72] L. Linnenbrink-Garcia, D. F. Tyson, and E. A. Patall, "When are achievement goal orientations beneficial for academic achievement? A closer look at main effects and moderating factors," Revue Internationale de Psychologie Sociale, vol. 21, pp. 19-70, 2008. Vote-count review of 90+ peer-reviewed articles on achievement goal orientation and academic achievement. Found ~40% of effects positive for mastery-approach and performance-approach orientations vs unspecified comparison conditions; consistently negative effects for performance-avoidance. Cited for T9 as evidence motivation affects learning with adaptive orientations producing better outcomes; cited for P8a as evidence manufactured maladaptive whys produce worse outcomes than absent whys.
[73] M. Richardson, C. Abraham, and R. Bond, "Psychological correlates of university students' academic performance: A systematic review and meta-analysis," Psychological Bulletin, vol. 138, no. 2, pp. 353-387, 2012. Comprehensive meta-analysis of motivational predictors of academic performance, with correlations ranging from -0.14 (performance-avoidance orientation) to 0.59 (performance self-efficacy). Cited for T9 as cross-construct evidence on motivation-learning effects; cited for P8a as evidence on the negative side of the motivation-orientation spectrum.
[74] N. W. Van Yperen, M. Blaga, and T. Postmes, "A meta-analysis of self-reported achievement goals and nonself-report performance across three achievement domains (work, sports, and education)," PLOS ONE, vol. 9, no. 4, art. e93594, 2014. Cross-domain meta-analysis replicating directional pattern of achievement goal effects across education, sport, and work. Cited for T9 as evidence the directional finding generalizes beyond educational contexts; cited for P8a.
[75] K. D. Multon, S. D. Brown, and R. W. Lent, "Relation of self-efficacy beliefs to academic outcomes: A meta-analytic investigation," Journal of Counseling Psychology, vol. 38, no. 1, pp. 30-38, 1991. Foundational meta-analysis of 36 studies establishing self-efficacy and academic performance correlation r = 0.38. Cited for T9 as evidence positive motivational orientations (self-efficacy specifically) increase learning outcomes.
[76] J. Honicke and J. Broadbent, "The influence of academic self-efficacy on academic performance: A systematic review," Educational Research Review, vol. 17, pp. 63-84, 2016. Replication and extension of Multon et al. with r = 0.33 across more recent studies. Cited for T9 as confirmation of self-efficacy effects on academic outcomes.
[77] J. Camacho-Morles, G. R. Slemp, R. Pekrun, K. Loderer, H. Hou, and L. G. Oades, "Activity achievement emotions and academic performance: A meta-analysis," Educational Psychology Review, vol. 33, pp. 1051-1095, 2021. Meta-analysis of 68 studies on activity achievement emotions and academic performance. Sample sizes: 57 independent samples for enjoyment (N=31,868); 25 for anger (N=11,153); 9 for frustration (N=1,418); 66 for boredom (N=28,410). Findings: enjoyment ↔ performance ρ = 0.27; anger ↔ performance ρ = -0.35; boredom ↔ performance ρ = -0.25; frustration ↔ performance ρ = -0.02 (near zero). Cited for T9b as empirical grounding for control-value theory's directional predictions about achievement emotions and outcomes.
[78] N. Kornell and R. A. Bjork, "Learning concepts and categories: Is spacing the 'enemy of induction'?" Psychological Science, vol. 19, no. 6, pp. 585-592, 2008. Foundational paper proposing the discriminative-contrast hypothesis: interleaving works because it forces the learner to discriminate between categories, producing the cross-instance comparison that supports inductive learning. Cited in P7 as theoretical mechanism explaining why interleaving outperforms blocked practice once initial fluency is established. Connects to T5's reconstruction account.
[79] C. Escobar-Soler, R. Berrios, G. Peñaloza-Díaz, C. Melis-Rivera, A. Caqueo-Urízar, F. Ponce-Correa, and J. Flores, "Effectiveness of self-affirmation interventions in educational settings: A meta-analysis," Healthcare, vol. 12, no. 1, art. 3, 2024. Meta-analysis of 144 studies (N = 36,419) finding self-affirmation interventions in educational settings produce d = 0.41 (95% CI: 0.36-0.45). Notable sub-finding: when teachers or educational assistants participated directly in self-affirmation interventions, effectiveness decreased — teacher-led affirmation undermined effectiveness of the activity. Cited in P8b for both the positive direction (learner self-affirmation works) and the inverse (teacher-manufactured affirmation backfires). [80] B. W. Roberts and W. F. DelVecchio, "The rank-order consistency of personality traits from childhood to old age: A quantitative review of longitudinal studies," Psychological Bulletin, vol. 126, no. 1, pp. 3-25, 2000. Canonical meta-analysis of rank-order stability of personality traits across the lifespan. Compiled 3,217 test-retest correlations from 152 longitudinal studies. Found stability increases monotonically from r = 0.31 in childhood to r = 0.54 in college years to r = 0.64 at age 30, plateauing at r = 0.74 between ages 50-70. Cited in P10 as the empirical grounding for the framework's claim that trait measurements have measurable signal in childhood while being less stable than in adulthood. [81] J. Mammadov, "Big Five personality traits and academic performance: A meta-analysis," Journal of Personality, vol. 90, no. 2, pp. 222-255, 2022. Most comprehensive meta-analysis to date of Big Five → academic performance associations. Combined effect of cognitive ability and personality traits explained 27.8% of academic performance variance, with cognitive ability accounting for 64% relative importance and conscientiousness accounting for 28% of the personality-explained variance. Education level moderates: openness, extraversion, and agreeableness show significantly larger effect sizes at elementary/middle school level than later. Cited in P10 as primary evidence for conscientiousness-as-strongest-trait-predictor and for the framework-relevant finding that effects are larger in younger learners. [82] Gul-E-Zahra, J. Dang, Y. Cui, J. Liu, M. Qi, and W. Wang, "Revisiting the big five–academic performance association: a one-stage meta-analytic structural equation modeling reanalysis of 84 studies," Frontiers in Psychology, vol. 17, art. 1769823, 2026, doi: 10.3389/fpsyg.2026.1769823. One-stage MASEM reanalysis of the 84-study dataset (N = 45,477 university students) originally compiled by Chen, Cheung & Zeng (2025). Standardized path coefficients controlling for trait intercorrelations: conscientiousness β = 0.199 (strongest); openness β = 0.060; agreeableness β = 0.034; extraversion β = -0.062; neuroticism not significant. Cited in P10 as evidence for the framework's decision to make predictions only for conscientiousness and openness. [83] V. F. Sisk, A. P. Burgoyne, J. Sun, J. L. Butler, and B. N. Macnamara, "To What Extent and Under Which Circumstances Are Growth Mind-Sets Important to Academic Achievement? Two Meta-Analyses," Psychological Science, vol. 29, no. 4, pp. 549–571, 2018. Two meta-analyses on growth mindset and academic achievement. First meta-analysis (k = 273, N = 365,915) examined the strength of the relationship between mindset and achievement. Second meta-analysis (k = 43, N = 57,155) examined effectiveness of mindset interventions on academic achievement. Overall effects were weak for both meta-analyses. Some results supported that students with low socioeconomic status or who are academically at risk might benefit more from interventions. Cited in T9c as primary evidence that growth mindset interventions produce small effects (d = 0.03 to 0.10). [84] D. S. Yeager, P. Hanselman, G. M. Walton, J. S. Murray, R. Crosnoe, C. Muller, et al., "A national experiment reveals where a growth mindset improves achievement," Nature, vol. 573, no. 7774, pp. 364–369, 2019. National Study of Learning Mindsets (NSLM): nationally representative sample of 65 regular public schools in the United States, n = 12,490 ninth-grade adolescents individually randomized to condition. Short (<1 hour) online growth mindset intervention improved grades among lower-achieving students and increased advanced mathematics course enrollment. Cited in T9c as confirmation of the small effect magnitude found in Sisk et al. 2018. [85] M. Credé, M. C. Tynan, and P. D. Harms, "Much ado about grit: A meta-analytic synthesis of the grit literature," Journal of Personality and Social Psychology, vol. 113, no. 3, pp. 492–511, 2017. Meta-analytic review based on 584 effect sizes from 88 independent samples representing 66,807 individuals. Found that the higher order structure of grit is not confirmed, grit is only moderately correlated with performance and retention, and grit is very strongly correlated with conscientiousness (ρ = 0.842 across 22 studies). Authors suggest interventions designed to enhance grit may only have weak effects on performance. Cited in T9c as evidence that grit shows r = 0.18 correlation with academic performance and largely overlaps with conscientiousness. [86] K. Talsma, B. Schüz, R. Schwarzer, and K. Norris, "I believe, therefore I achieve (and vice versa): A meta-analytic cross-lagged panel analysis of self-efficacy and academic performance," Learning and Individual Differences, vol. 61, pp. 136–150, 2018. First fully systematic meta-analytic cross-lagged panel analysis (k = 11, N = 2688) of the reciprocal relationship between self-efficacy and academic performance. Found performance has a net positive effect on subsequent self-efficacy (β = 0.205, p < 0.001), significantly larger than the effect of self-efficacy on subsequent performance (β = 0.071, p < 0.001). Reciprocity holds for adults but not for children; in the child subsample, performance uniquely impacts subsequent self-efficacy, but self-efficacy does not measurably impact subsequent performance. Cited in T9b and T9c as the methodologically cleanest available estimate of the bidirectional self-efficacy/performance relationship. [87] D. J. Madigan and T. Curran, "Does Burnout Affect Academic Achievement? A Meta-Analysis of over 100,000 Students," Educational Psychology Review, vol. 33, no. 2, pp. 387–405, 2021. First meta-analysis of the relationship between burnout and academic achievement. Literature search returned 29 studies (N = 109,396) and 89 effect sizes. Robust variance meta-analyses indicated that total burnout had a significant negative relationship with academic achievement (r = -0.24). Sub-dimension effects: exhaustion r = -0.15, cynicism r = -0.24, reduced efficacy r = -0.39. Cited in T16 as primary evidence that academic burnout produces measurable negative effects on achievement. [88] D. J. Madigan, L. E. Kim, and H. L. Glandorf, "Interventions to reduce burnout in students: A systematic review and meta-analysis," European Journal of Psychology of Education, 2023. First systematic review and meta-analysis of interventions aimed at reducing burnout in students. Systematic literature search returned 17 studies (10 RCT, 7 quasi-experimental), 2,462 students from secondary and tertiary education. Interventions included mindfulness, REBT, and CBT. Aggregated effect across interventions: g+ = 0.90, p = 0.02, 95% CI: [0.04, 1.75], k = 14. Cited in T16 as evidence that burnout-reduction interventions can be effective. [89] I. Sukamto, Y. Ayriza, Hiryanto, R. Kumullah, Murjainah, P. E. L. Ero, S. Purwandari, R. Lince, and D. Maulina, "Understanding academic burnout: A systematic literature review of its risk and protective factors," Social Sciences and Humanities Open, 2026. Systematic literature review following PRISMA guidelines, search through July 2025, identified 1744 citations of which 50 studies were included. Synthesis identified risk factors across academic, psychological, personal, social, and structural domains, and protective factors including strengthening individual character, strong social support, developing adaptive learning and coping strategies, autonomous motivation, and providing a healthy academic environment. Cited in P17 as primary evidence for the framework's identification of burnout protective factors. [90] E. J. Webb, D. T. Campbell, R. D. Schwartz, and L. Sechrest, Unobtrusive Measures: Nonreactive Research in the Social Sciences. Chicago: Rand McNally, 1966. Foundational text introducing the concept of unobtrusive measurement in social sciences. Argued that traditional measurement methods (interviews, questionnaires) systematically disturb the phenomena they attempt to measure, and that the field should develop measurement approaches that minimize this disturbance through nonreactive observation, archival data, and physical traces. Cited in T2 as the foundational reference for the engineering-stance principle that measurement disturbs state but can be designed to make the disturbance acceptably small for the sensitivity required. [91] J. R. Anderson, L. M. Reder, and H. A. Simon, "Situated Learning and Education," Educational Researcher, vol. 25, no. 4, pp. 5–11, 1996. Critical engagement with situated cognition arguments. Argued that the dichotomy between "situated" learning and "abstract" learning is false: skills can be acquired through deliberate practice in one context and applied in others via composition of sub-skills, with transfer dependent on the degree of shared cognitive components. Critiqued exclusive reliance on apprenticeship-style learning while acknowledging the importance of practice in realistic contexts. Cited in RT5 for the position that transfer depends on shared cognitive sub-skills accumulated through practice, supporting the framework's view that practice in instances builds repertoires that subsequent reconstruction draws on. [92] E. Murray, A. J. Horner, and S. M. Göbel, "A Meta-analytic Review of the Effectiveness of Spacing and Retrieval Practice for Mathematics Learning," Educational Psychology Review, vol. 37, art. 75, 2025. Meta-analysis investigating spaced and retrieval practice effects in mathematics learning. Found robust small-to-medium effect of spacing overall (g = 0.28, 27 studies, 53 effect sizes), with larger isolated-learning effects (g = 0.43) than course-embedded (g = 0.24). Retrieval practice (testing vs restudy) in mathematics: g = 0.18 across 7 studies, 32 effect sizes — but 95% CI crossed zero, suggesting effect is not robust in mathematics specifically. Cited in P13 as evidence that retrieval practice effects are smaller in mathematics than in other domains. [93] K. Sigayret, J.-F. Parmentier, and F. Silvestre, "Testing the testing effect on prolific: when retrieval practice fails to boost learning," Frontiers in Psychology, vol. 17, art. 1727423, 2026. Two preliminary experiments examining whether the testing effect can be detected in studies conducted on Prolific, a platform widely used for online behavioral and educational research. Failed to replicate the standard testing effect in online crowdsourced samples under certain conditions. Suggests that engagement, motivation, and task structure are non-trivial moderators of the testing effect in online settings. Cited in P13 as evidence that retrieval practice effects can fail to replicate in online learning environments. [94] S. L. Britner and F. Pajares, "Sources of science self-efficacy beliefs of middle school students," Journal of Research in Science Teaching, vol. 43, no. 5, pp. 485–499, 2006. Empirical study of self-efficacy sources in middle school science learners (N = 319). Among Bandura's four posited sources (mastery experience, vicarious experience, social persuasion, physiological arousal), mastery experience emerged as the only significant predictor of self-efficacy beliefs. Cited in T9b alongside Usher & Pajares 2008 as evidence that mastery experience is the strongest single source of self-efficacy in the empirical literature. [95] E. L. Usher and F. Pajares, "Sources of self-efficacy in school: Critical review of the literature and future directions," Review of Educational Research, vol. 78, no. 4, pp. 751–796, 2008. Critical review of self-efficacy sources research in academic contexts since Bandura's 1977 theoretical formulation. Synthesis: mastery experience is typically the most influential source of self-efficacy, though the strength and influence of the four sources differ as a function of contextual factors including gender, ethnicity, academic ability, and academic domain. Cited in T9b as the comprehensive review establishing mastery experience as the strongest source of self-efficacy. [96] J. H. Flavell, "Metacognition and cognitive monitoring: A new area of cognitive-developmental inquiry," American Psychologist, vol. 34, no. 10, pp. 906–911, 1979. Foundational paper establishing metacognition as a distinct cognitive capacity that develops through childhood and adolescence. Metacognition includes monitoring one's own cognitive processes, evaluating one's own knowledge and capability, and regulating one's own learning. Cited in T9c as the theoretical grounding for why self-efficacy intervention is age-moderated: intervention requires metacognitive reflection capacity that children have not yet developed. [97] O. Chen, S. Kalyuga, and J. Sweller, "The expertise reversal effect is a variant of the more general element interactivity effect," Educational Psychology Review, vol. 29, no. 2, pp. 393–405, 2017. Theoretical paper from the cognitive load theory tradition reframing the expertise reversal effect as a special case of the broader element interactivity principle. Argues that the same instructional content imposes different cognitive loads at different levels of prior knowledge because prior schemas chunk content into manageable element groups; what appears as a reversal between novice and expert is actually a continuous gradient driven by element interactivity. Cited in T10 to position the expertise reversal effect as evidence for a continuous substrate-instruction relationship rather than a discrete binary. [98] J. Siegrist, "Adverse health effects of high-effort/low-reward conditions," Journal of Occupational Health Psychology, vol. 1, no. 1, pp. 27–41, 1996. Foundational paper introducing the effort-reward imbalance model of work stress: chronic mismatch between effort invested and reward perceived (in compensation, recognition, security, or career opportunities) produces depleted state and adverse health outcomes. Originally developed for occupational health; subsequently imported to academic burnout contexts. Cited in T16 as the canonical reference for the effort-reward imbalance causal model. [99] I. Messina, T. Rossi, R. Maniglio, C. Loconsole, and P. Spataro, "Risk and protective factors in academic burnout: exploring the mediating role of interpersonal emotion regulation in the link with social support," Frontiers in Psychology, vol. 16, art. 1536951, 2025. Empirical study (N = 156 undergraduate students) examining the mediating role of interpersonal emotion regulation in the social-support → burnout relationship. Found that venting through interpersonal emotion regulation paradoxically exacerbates burnout by reducing perceived social support, while reassurance-seeking serves a protective function. Cited in P17 to refine the social-support protective factor: type of support matters, with venting harmful and reassurance-seeking helpful. [100] J. Sierksma and E. Brummelman, "'Here, Let Me Do It for You': Psychological Consequences of Receiving Direct and Indirect Help in Childhood," Child Development, 2025, DOI 10.1111/cdev.14259. Three preregistered experiments (N = 619, ages 7-9) on the motivational consequences of receiving help in childhood. Internal meta-analysis: children who received help (both direct and indirect) felt less competent, liked the task less, and felt more in need of help; they sought fewer challenges. Effect sizes were modest. Direct and indirect help had largely similar effects, except that children disliked and misreported receiving direct help more. Cited in P5a as empirical grounding for the prescription that when direct help would threaten the learner's self-construct, indirect routing through external attribution preserves the help while reducing competence-cost. [101] S. Nelson-Le Gall, "Help-seeking: An understudied problem-solving skill in children," Developmental Review, vol. 1, no. 3, pp. 224–246, 1981. Foundational paper establishing help-seeking as a problem-solving skill that develops through childhood, distinct from passive help-receiving. Distinguished instrumental help-seeking (asking for the means to solve oneself) from executive help-seeking (asking for the answer). Cited in P5b as the canonical reference for help-seeking as a developmental skill. [102] R. S. Newman, "Children's help-seeking in the classroom: The role of motivational factors and attitudes," Journal of Educational Psychology, vol. 82, no. 1, pp. 71–80, 1990. Empirical study establishing that children's help-seeking is shaped by motivational factors including self-perceived ability, goal orientation, and beliefs about the costs and benefits of asking for help. Cited in P5b for the motivational dimensions of help-seeking development. [103] S. A. Karabenick, "Seeking help in large college classes: A person-centered approach," Contemporary Educational Psychology, vol. 28, no. 1, pp. 37–58, 2003. Empirical study using person-centered analysis to identify distinct profiles of help-seeking among learners. Established that help-seeking varies systematically by motivational profile and that adaptive help-seeking is correlated with achievement. Cited in P5b as evidence for help-seeking as a learnable adaptive skill with motivational components. [104] J. Crissman, "The Design and Utilization of Effective Worked Examples: A Meta-Analysis," doctoral dissertation, University of Nebraska-Lincoln, 2006. Meta-analysis of the worked examples effect across educational contexts. Found a mean effect of roughly d = 0.57 for worked examples versus problem-solving, with intra-example design features moderating the benefit — fading of solution steps produced the strongest learning gains, followed by self-explanation, with conventional worked examples producing moderate effects. The effect is strongest for novices and reduces with increasing expertise (the expertise reversal effect). Cited in P12 as primary empirical grounding for the worked examples mechanism that underlies full-execution scaffolding.
[105] J. Sweller and G. A. Cooper, "The use of worked examples as a substitute for problem solving in learning algebra," Cognition and Instruction, vol. 2, no. 1, pp. 59–89, 1985. Foundational paper establishing the worked examples effect: studying worked examples produces better learning outcomes than unguided problem-solving for novices learning algebra. The seminal empirical demonstration that informed cognitive load theory's instructional design recommendations. Cited in P12 as the foundational reference for the worked examples mechanism. [106] B. Rogoff, Apprenticeship in Thinking: Cognitive Development in Social Context. New York: Oxford University Press, 1990. Foundational text introducing guided participation as the process through which children learn through active engagement in shared activities with more skilled partners. Documented across multiple cultural contexts including Mayan mothers weaving with their children, caregiving practices, and motor skill acquisition. Guided participation involves the more skilled partner handling parts of an activity beyond the learner's current capability while the learner participates at parts they can perform, with responsibilities transferring as the learner develops capability. Cited in P12 as the canonical reference for the participation pattern P12 operationalizes for procedure-level substrate gaps. [107] R. J. Spiro, R. L. Coulson, P. J. Feltovich, and D. K. Anderson, "Cognitive flexibility theory: Advanced knowledge acquisition in ill-structured domains," in Proceedings of the 10th Annual Conference of the Cognitive Science Society, 1988, pp. 375–383. Foundational paper introducing Cognitive Flexibility Theory (CFT). Argues that advanced knowledge acquisition in ill-structured domains requires multiple mental and pedagogical representations, multiple alternative systems of linkage among knowledge elements, and schema assembly rather than retrieval of prepackaged schemas. The framework cites CFT in P11 as the theoretical grounding for the claim that composite activities produce integration schemas distinct from component schemas.
[108] M. L. Gick and K. J. Holyoak, "Analogical problem solving," Cognitive Psychology, vol. 12, no. 3, pp. 306-355, 1980. Foundational study of analogical transfer using Duncker's radiation problem. Without a source analog, roughly 10% of participants produced the convergence solution; given an analogous source story but no hint to use it, roughly 30% did. Demonstrates that spontaneous recognition of an available analogy is rare. Cited in T5 for the recognition-gate claim and in RT5 context.
[109] M. L. Gick and K. J. Holyoak, "Schema induction and analogical transfer," Cognitive Psychology, vol. 15, no. 1, pp. 1-38, 1983. Establishes the "schema induction" construct: comparing two analogous source problems leads learners to induce an abstract schema that supports later transfer. (The ~30%-to-75% spontaneous-versus-cued figures cited in T5 come from the 1980 study [108]; this 1983 paper is cited for schema induction via analogical comparison.)
[110] J. D. Bransford and D. L. Schwartz, "Rethinking transfer: A simple proposal with multiple implications," Review of Research in Education, vol. 24, pp. 61-100, 1999. Proposes "preparation for future learning" (PFL) as the right measure of transfer: prior learning should be judged by how well it equips a learner to acquire new material, not by what content it lets them port directly. Cited in T5 for the claim that breadth of knowledge provides more material for new learning.
[111] E. F. Loftus and J. C. Palmer, "Reconstruction of automobile destruction: An example of the interaction between language and memory," Journal of Verbal Learning and Verbal Behavior, vol. 13, no. 5, pp. 585-589, 1974. Classic demonstration of the misinformation effect: the verb used in a leading question ("smashed" vs. "hit") altered participants' speed estimates and led some to report broken glass that was not present. Cited in T17 for the reconstructive and manipulable nature of declarative memory.
[112] E. F. Loftus and J. E. Pickrell, "The formation of false memories," Psychiatric Annals, vol. 25, no. 12, pp. 720-725, 1995. The "lost in the mall" study: a substantial fraction of participants developed confident, detailed memories of a childhood event (being lost in a shopping mall) that never occurred. Cited in T17 for the claim that entire events can be implanted as declarative memories.
[113] M. K. Singley and J. R. Anderson, The Transfer of Cognitive Skill. Cambridge, MA: Harvard University Press, 1989. Develops the identical-elements / shared-productions account of transfer within the ACT framework and reports the text-editor transfer experiments (transfer between editors sharing components). Cited in T5 for the identical-elements tradition. (Corrects an earlier body-text reference dated 1985; the text-editor transfer work is the 1989 book.)*
[114] T. M. Amabile, "Effects of external evaluation on artistic creativity," Journal of Personality and Social Psychology, vol. 37, no. 2, pp. 221-233, 1979. Experimental study (95 female undergraduates) of how expecting external evaluation affects creativity of an art activity. Expecting evaluation reduced judged creativity relative to a no-evaluation condition. Later work qualifies this: evaluation expectation paired with explicit creativity criteria can preserve or improve creativity; it is uninformed/premature evaluation during the generative phase that suppresses output. Cited in T5a for the claim that premature judgment during generation suppresses candidate-surfacing.
[115] T. M. Amabile, P. Goldfarb, and S. C. Brackfield, "Social influences on creativity: Evaluation, coaction, and surveillance," Creativity Research Journal, vol. 3, no. 1, pp. 6-21, 1990. Demonstrated that expected evaluation, the presence of others working alongside (coaction), and surveillance each reduce creative performance. Cited in T5a as converging evidence that evaluative/social pressure during the generative phase suppresses creative output.
[116] J. S. Mueller, S. Melwani, and J. A. Goncalo, "The bias against creativity: Why people desire but reject creative ideas," Psychological Science, vol. 23, no. 1, pp. 13-17, 2012. Two experiments showing that a motivation to reduce uncertainty activates an implicit bias against creativity (relative to practicality), and that this bias interferes with people's ability to recognize a creative idea even when present. Cited in T5a for the point that premature, uncertainty-driven judgment not only suppresses generation but degrades recognition of good candidates — reinforcing why judgment must be deferred during the generative phase.
[117] L. Rozenblit and F. Keil, "The misunderstood limits of folk science: An illusion of explanatory depth," Cognitive Science, vol. 26, no. 5, pp. 521-562, 2002. Across multiple studies, people rated their understanding of how things work far higher than they could justify when asked to produce a step-by-step explanation; confidence collapsed on attempting the explanation. The illusion is specific to explanatory/causal knowledge (how and why something works) and is markedly weaker for facts, procedures, and narratives. Cited in T5a for the external-validation requirement: because the feeling of understanding a self-generated causal connection systematically outruns the actual understanding — and the illusion is strongest for exactly the causal-explanatory knowledge personal innovation produces — the internal filter cannot certify itself, so surviving links must be checked against an external source.
[118] L. C. Ehri, S. R. Nunes, S. A. Stahl, and D. M. Willows, "Systematic Phonics Instruction Helps Students Learn to Read: Evidence from the National Reading Panel's Meta-Analysis," Review of Educational Research, vol. 71, no. 3, pp. 393-447, 2001. National Reading Panel meta-analysis of 66 treatment-control comparisons from 38 experiments. Systematic/explicit phonics instruction produced a moderate overall effect on reading (d = 0.41), larger when instruction began early (d = 0.55) than after first grade (d = 0.27), with effects persisting after instruction ended and benefiting decoding, word reading, comprehension, and spelling. Cited in T4 as evidence that decoding requires explicit instruction in the letter-sound code rather than being acquired through text immersion alone. Note: the contested question in the literature (e.g., Bowers 2020) concerns whether systematic phonics outperforms other phonics orderings, not whether explicit decoding instruction is required — the latter, which T4 asserts, is not the contested claim.
[119] N. M. McNeil, N. C. Jordan, A. A. Viegut, and D. Ansari, "What the Science of Learning Teaches Us About Arithmetic Fluency," Psychological Science in the Public Interest, vol. 26, no. 1, pp. 10-57, 2025, doi:10.1177/15291006241287726. Synthesis drawing on developmental cognitive science (Sfard's process-object duality, Karmiloff-Smith's implicit-explicit continuum) arguing that memorization and thinking strategies are complementary rather than opposing forces in building arithmetic fluency. Automaticity (immediate retrieval from long-term memory) frees working memory for higher-level mathematics and is built specifically by well-structured retrieval practice, with time-limited practice introduced only after accuracy is established. Cited in T4 as evidence that arithmetic-fact automaticity is built by explicit retrieval practice rather than by ambient numerical activity alone, and that conceptual understanding and retrieval practice are complementary, not substitutes.