Memory and Learning

What this is. The principles of cognitive science on how people consolidate, retain, retrieve and transfer what they learn, and how Smoother applies them. It is the evidential base on which the Learning Tasks (Tareas de Aprendizaje — the cognitive work the person who is learning does), the Activities, the Materials and the pedagogical instruments are designed. Where the Reference Framework (Marco de Referencia) walks through the history of pedagogical thought, this document deals with the empirical findings of the last five decades and with how to apply them systematically.

Thirteen principles

1. Retrieval practice. Reconstructing information from memory — not recognizing it, not rereading it — strengthens long-term storage more than any other study technique. The difficulty of retrieving is precisely what produces the consolidation. Strong evidence: a meta-analysis over 159 studies with a mean effect of d = 0.50, robust across ages, materials and time frames.

2. Spaced practice. Distributing study across separate sessions retains better than massing it, with the same total time. Partial forgetting between sessions forces a more demanding retrieval effort. Strong evidence, over a meta-analysis of 184 studies.

3. Interleaving. Mixing problem types within a session improves discrimination and transfer compared with practicing a single type in blocks. It is counterintuitive: during practice performance looks worse, and later retention is higher. It works best when the concepts are similar and discriminable.

4. Cognitive load. Working memory is severely limited. Intrinsic load comes from the material and from prior knowledge; extraneous load is generated by the design and must be minimized; germane load is the effort of building schemas and must be maximized. Strong evidence, and the basis of the 4C/ID model.

5. Dual coding. Verbal plus visual encodes and retrieves better than either one alone. It is not the same thing as “learning styles”: dual coding offers both modalities to everyone; matching by individual preference has no evidence.

6. Elaborative interrogation. Asking “why is this so?” about what is being learned improves comprehension and retention. It works best with factual material about which there is already some familiarity.

7. Self-explanation. Explaining to yourself what you read, see or do forces you to fill gaps and integrate fragments. Strong evidence, with meta-analysis.

8. Worked examples and fading. For someone who is a novice, studying worked examples before attempting problems produces better learning than solving directly; the transition must be progressive. With expertise the effect attenuates and even reverses: for experienced learners, direct problems can yield more.

9. Productive failure. Struggling with a challenging problem before formal instruction deepens later conceptual understanding, even though errors are made during the struggle. It works with well-designed problems, not with just any difficult task. The apparent contradiction with principle 8 is resolved by purpose: worked examples for fast procedural acquisition, productive failure for deep conceptual understanding.

10. Feedback quality. The highest-impact feedback is specific, actionable, timely, about the process and not only about the result, and it enables self-regulation rather than dependence.

11. Transfer. It does not happen automatically: it is designed. Variation of contexts during practice, explicit identification of the underlying principles, and application to problems that are structurally similar but superficially different. Spontaneous transfer between distant domains is rare. Moderate-to-strong evidence.

12. Purposeful and deliberate practice. Expertise is not built by repetition but by structured practice: specific sub-skill goals, work at the edge of capability, immediate feedback, refinement, and mental representations that make self-monitoring possible. Two modes: purposeful practice is the floor — goal, focus, feedback, stepping out of the comfort zone — and applies in any domain; deliberate practice is the ceiling and adds expert decomposition and coaching, in domains with objective measures of performance. Honest qualification, non-negotiable: a meta-analysis shows that deliberate practice explains a substantial but partial and domain-dependent portion of expertise — smaller in education and the professions than in music, chess or sports — and the popularization of the “10,000 hours” is a simplification. It is necessary and not sufficient.

13. Social learning and the social construction of knowledge. A good part of learning is built with others. The central mechanism is the Zone of Proximal Development: the distance between what one can do alone and what one can do with the support of someone more capable, who scaffolds and then withdraws the scaffolding. Honest qualification: the evidence refutes discovery with minimal guidance. Smoother adopts social co-construction and scaffolding, but guided — the scaffolding exists and then fades; it is never absent.

What does not work, despite its popularity

For consistency, Smoother also names what the evidence debunks: matching by learning style; the learning pyramid with its percentages, which are fabricated figures; left brain / right brain as a framework for instruction; and rereading and highlighting as a primary strategy, both classified as low utility next to retrieval practice.

Apparent contradictions, resolved by context

Worked examples versus productive failure: both valid, different purposes. Worked examples versus the expertise reversal: both valid, different levels of the person who is learning. Social construction versus explicit guidance: compatible — it is guided constructivism. Spaced practice versus narrative cohesion: spacing retains better, but breaking the thread damages motivation; it is resolved with Learning Paths and Learning Blocks that give cohesion while spacing the retrieval.

Evidence

Level: Research-backed. That is the label this document holds in Smoother’s evidence index. It is a synthesis of consolidated cognitive science, with its sources cited and its honest qualifications included — notably the one on deliberate practice (§12) and the one on guided constructivism (§13). The evidence index is not a hierarchy of quality: it is a statement of provenance.

Acknowledgement. The selection and curation of these principles draw on Gareth Manning’s open-source education-agent-skills library (CC BY-SA 4.0), in particular its memory-science domain and its explicit list of exclusions. The intellectual debt is explicit and permanent.

Learning Activities · Learning Tasks · Artificial Intelligence in Learning · Learning Units · Personal Preferences · Evaluation · Study and Practice Techniques

Source

protocol/smoother-memoria-y-aprendizaje.md — version 0.35.0 (2026-07-09).