Where it is used
Each entry identifies the volume, edition, chapter or appendix in which the work appears.
The evidence behind the series
Browse the scientific research and professional literature used across all seven volumes of The Science of Learning for Educators.
Search the collection
Choose an author range, search a title or concept, or narrow the collection by source type. Open any result to see where it appears in the books, read its evidence summary and follow available source links.
117 references
This meta-analysis examines whether low-stakes practice tests improve later learning compared with non-testing activities such as restudying.
Across 118 experiments and 272 independent effects, practice testing produced a robust overall benefit. The size of the benefit varied with features such as test format, the relationship between practice and final tests, and the number of retrieval opportunities. Benefits appeared with and without feedback, although feedback remains educationally important because it corrects errors. The authors distinguish retrieval practice from high-stakes accountability testing: the learning mechanism is the act of bringing information to mind.
For educators, the review supports frequent, low-stakes opportunities to retrieve knowledge and suggests varying formats while keeping retrieval aligned with later learning demands.
This practitioner-focused book translates cognitive science into four classroom “power tools”: retrieval practice, spacing, interleaving, and feedback-driven metacognition.
Agarwal and Bain combine research evidence with examples drawn from a long scientist–teacher collaboration, showing how the strategies can be embedded in ordinary lessons without extensive preparation or additional grading. The book also addresses classroom culture, anxiety, communication with students and families, and professional learning for educators. Its central message is that durable learning improves when students repeatedly bring knowledge to mind, revisit it over time, discriminate among related ideas, and receive information that helps them judge what they know.
The guidance is intended as adaptable practice rather than a scripted programme, allowing teachers to fit the principles to subject, age group, and context.
View or buy the cited work >>Read a similar article free online >>
This article reports lessons from a five-year partnership involving a cognitive scientist, a middle-school teacher, a principal, and more than 1,400 students.
The programme tested retrieval practice in authentic classrooms rather than relying only on laboratory findings. Across several studies, brief quizzes and other opportunities to retrieve course content improved long-term learning, including performance on later classroom assessments. The authors emphasize that productive applied research depends on sustained collaboration: teachers contribute knowledge of learners and curriculum, school leaders create workable conditions, and researchers provide experimental design and analysis. For educators, the work supports integrating low-stakes retrieval into normal instruction and using feedback to strengthen learning.
It also shows why school-based research must balance scientific control with the practical demands and variability of real classrooms.
Across two experiments, Agarwal and colleagues compared restudying with open-book and closed-book tests after participants studied prose passages.
When tests included feedback, both formats improved delayed retention relative to restudying or testing without feedback. Open-book testing produced higher performance during the initial test because learners could consult the material, but that immediate advantage disappeared: open- and closed-book groups retained equivalent amounts later. Participants also predicted that repeated study would produce better memory, even though retrieval practice was more effective. The findings separate access conditions from the learning benefit of retrieval and show that feedback matters when answers may be incomplete or wrong.
For educators, an open-book assessment can still function as retrieval practice when questions require active recall and feedback is supplied; simply allowing resources does not eliminate the testing effect.
This systematic review asks whether retrieval practice works under normal school and classroom conditions. The authors screened nearly 2,000 abstracts and applied narrow inclusion criteria to 50 experiments involving 5,374 learners.
Of 49 reported effect sizes, 57 percent showed medium or large benefits. Positive effects appeared across educational levels, subject areas, test delays, retrieval formats, final-test formats, and different timings of practice and feedback. The review avoids collapsing highly varied classroom designs into one pooled estimate and instead reports individual effects and confidence intervals. A key limitation is geographic representation: only six percent of experiments came from non-WEIRD countries.
For educators, the findings support low-stakes retrieval as a broadly useful strategy while underscoring the need to adapt implementation to context and expand research with more diverse learners.
Alexander presents educational psychology through problems that matter directly to teaching rather than as a catalogue of separate psychological theories.
The book compares cognitive, developmental, motivational, social, and sociocultural perspectives and applies them to knowledge acquisition, transfer, strategic processing, assessment, and classroom learning. A recurring concern is how learners' prior knowledge, beliefs, goals, interests, and strategies interact with instructional conditions. The text helps readers see that no single theory explains every educational situation and that sound instruction requires selecting concepts that fit the learner, content, and task. For educators, its practical value lies in connecting psychological research to decisions about explanation, practice, motivation, evaluation, and support for increasingly independent learning.
It is an integrative textbook rather than a report of one study.
View or buy the cited work >>Read a similar article free online >>
This handbook explains dialogic teaching as a principled approach to classroom interaction, not simply as increasing the amount of student talk.
Alexander connects classroom dialogue to views of knowledge, learning, culture, social relationships, and civic participation. The book sets out practical repertoires and indicators that teachers can adapt to circumstances, while preserving teacher agency rather than prescribing a single routine. It reviews international research and incorporates evidence from a large English trial in which the approach was paired with professional development, mentoring, planning, and video-supported reflection. It also addresses oracy, argumentation, literacy, student voice, and implementation.
For educators, the central implication is that productive dialogue requires purposeful questions, extended student contributions, cumulative development of ideas, and a classroom climate where reasoning can be examined collectively.
View or buy the cited work >>Read a similar article free online >>
This systematic review examines how human-centred principles are represented in learning analytics and artificial-intelligence systems for education.
The authors reviewed 108 papers, focusing on stakeholder participation, the balance between human control and automation, and attention to safety, reliability, and trustworthiness. Although many systems acknowledged a need for human control, students and other target users were often only lightly involved in actual design and development. Safety and trustworthiness also received less attention than usability or technical performance. The review recommends involving educators and learners throughout design and deployment, making decisions about automation explicit, and treating reliability, agency, privacy, and trust as core design requirements.
For schools, the study cautions against adopting data-driven tools solely for efficiency: educational stakeholders should help define the purposes, limits, and acceptable consequences of the technology.
Anderson develops a unified computational account of cognition centred on the ACT architecture.
The book proposes that thinking emerges from interactions between declarative knowledge, represented as facts and structured information, and procedural knowledge, represented as productions that specify actions under particular conditions. Learning changes both the availability of knowledge and the efficiency of procedures, allowing performance to move from deliberate problem solving toward skilled execution. The architecture is intended to explain memory, reasoning, language, problem solving, and the acquisition of expertise within one framework. Its educational importance lies in treating learning as changes in organized knowledge and production rules rather than as undifferentiated practice.
Later ACT-R work revised many details, so the book is best read as a foundational model whose central questions continued to guide cognitive modelling and intelligent tutoring research.
View or buy the cited work >>Read a similar article free online >>
Andrade argues that instructional rubrics can support teaching and learning, not merely score finished work. A useful rubric identifies important criteria and describes gradations of quality in language students can understand.
When learners examine examples, help articulate criteria, and apply a rubric to their own drafts, expectations become visible and feedback becomes more actionable. The article illustrates a self-assessment process in which students mark evidence in their writing against each criterion and use what they notice to revise. Rubrics therefore promote thinking when they direct attention to quality, support dialogue about standards, and become tools students use during production. Andrade cautions implicitly against treating a rubric as a checklist or grading device alone.
For teachers, the practical sequence is to co-construct or explain criteria, study models, practise applying the rubric, and give students time to revise.
This qualitative study investigates how undergraduates experienced criteria-referenced self-assessment after sustained classroom use. Fourteen students participated in gender-separated focus groups.
Students generally reported positive attitudes once they understood the teacher's expectations and had practised applying explicit criteria. They described using self-assessment to check work, guide revision, improve quality and grades, strengthen motivation, and support learning. Some also noticed tension between their own standards and those communicated by teachers. The research found no clear gender differences in the responses.
Andrade and Du interpret self-assessment as more than assigning oneself a mark: it involves internalising criteria, monitoring work, and regulating subsequent action. For educators, the study suggests that self-assessment becomes credible when criteria are transparent, learners receive repeated guided practice, and the process leads directly to opportunities for revision rather than functioning as an isolated scoring exercise.
Argyris examines why highly successful professionals can struggle to learn from failure. Their education and career histories often reward solving external problems without requiring reflection on how their own assumptions and actions contribute to difficulty.
When challenged, they may use defensive reasoning: shifting blame, avoiding embarrassment, and keeping their governing assumptions beyond discussion. This enables single-loop learning, in which people adjust actions while leaving goals and underlying beliefs unchanged, but blocks deeper double-loop learning. Argyris argues that organisations should make reasoning visible by examining the gap between what people claim to do and the theories actually evident in their behaviour. Productive reflection tests assumptions openly and links advocacy with inquiry.
For schools and professional learning communities, the implication is that expertise alone does not ensure learning; teams need psychologically safe routines for examining their own contributions to problems and revising shared norms.
Argyris and Schön develop a theory of how organisations detect and correct error. They distinguish espoused theories—the explanations people give for their actions—from theories-in-use, the assumptions and rules that actually govern behaviour.
In single-loop learning, an organisation changes actions while preserving existing goals, values, and governing variables. Double-loop learning questions and may revise those underlying arrangements. The authors also consider “deutero-learning,” or learning how an organisation learns, and show how defensive routines, undiscussable issues, and mismatches between words and actions inhibit inquiry. Organisational learning is therefore not simply the sum of individual learning; it depends on shared maps, norms, records, and practices that retain and shape knowledge.
For educational organisations, the framework supports examining policies and assumptions behind recurring problems instead of repeatedly adjusting surface-level procedures.
View or buy the cited work >>Read a similar article free online >>
This sequel revisits and extends Argyris and Schön's theory of organisational learning in light of subsequent research and practice.
It asks what makes an organisation capable of learning, what forms of learning are desirable, and how productive inquiry can be strengthened. The authors retain the distinction between espoused theories and theories-in-use and examine how defensive routines protect people from embarrassment while preventing the testing of assumptions. They develop single-loop and double-loop learning through cases, action research, and discussion of the relationship between researchers and practitioners. Organisational knowledge is treated as embedded in routines, conversations, records, and shared interpretations, not merely stored in individual minds.
For schools, the book supports disciplined inquiry into recurring problems: teams should make reasoning visible, test causal claims, and reconsider governing values rather than changing procedures while leaving the assumptions that produced the problem untouched.
View or buy the cited work >>Read a similar article free online >>
Atkinson and Shiffrin propose the influential “modal model” of human memory. The framework separates relatively permanent structural features from flexible control processes and distinguishes three stores: sensory registers, a limited short-term store, and a long-term store.
Information briefly enters sensory memory; selected material moves into short-term memory, where attention and rehearsal help maintain it; and some information is transferred to long-term memory for later retrieval. Control processes such as rehearsal, coding, search, and decision strategies vary with goals, instructions, tasks, and experience rather than operating as fixed structures. The chapter combines the general architecture with experimental models of particular memory tasks. Later research revised the idea of simple separate stores, but the model established enduring questions about capacity, control, encoding, and retrieval.
For educators, it highlights the need to manage attention and limited immediate processing while supporting organised, retrievable long-term knowledge.
Ausubel presents a cognitive theory of classroom learning built around the learner's existing knowledge structure.
New information is learned meaningfully when it can be related to relevant concepts already understood; rote learning occurs when material is stored with few substantive connections. The teacher's task is therefore to diagnose prior knowledge, organise subject matter from broad inclusive ideas toward greater detail, and help learners reconcile similarities, differences, and apparent contradictions. Advance organisers provide a conceptual bridge before instruction, preparing relevant ideas to which unfamiliar material can attach. Ausubel distinguishes meaningful reception learning from passive memorisation: well-structured explanation can demand active cognitive integration even when information is presented directly.
For educators, the book's enduring implication is to begin from what students know, make conceptual relationships explicit, and design sequences that progressively differentiate and integrate knowledge rather than presenting isolated facts.
View or buy the cited work >>Read a similar article free online >>
Ayres investigates whether learners' own ratings of mental effort can detect changes in intrinsic cognitive load within a single task. In two experiments, students solved algebra problems and rated cognitive load after each computation.
Because the problems were designed to hold extraneous and germane load relatively constant, variation in ratings was interpreted as variation in element interactivity: the number and relation of information elements that must be processed together. The ratings were highly reliable, changed significantly across steps within problems, and were strongly related to errors. They also reflected learner expertise and procedural mistakes. The study therefore extends subjective rating scales beyond comparisons between whole tasks or instructional conditions.
For educators and researchers, it shows that a brief rating can reveal where a multistep problem becomes cognitively demanding, although ratings must be interpreted alongside performance and prior knowledge.
Azevedo reviews research on learning complex science topics with open-ended hypermedia and argues that such environments become useful metacognitive tools only when they support self-regulated learning.
Learners must coordinate prior knowledge, planning, monitoring, strategy use, motivation, and adaptation while navigating nonlinear information. Studies from the author's laboratory and classrooms show that learners of different ages often struggle to deploy these processes spontaneously. The article uses self-regulated learning as a framework for analysing both learning outcomes and the moment-by-moment processes that produce them. It also considers how prompts, human tutors, and adaptive computer scaffolds can model or support productive regulation.
The main implication is that access to rich digital resources does not by itself ensure deep learning: effective environments must help students set goals, activate relevant knowledge, monitor understanding, choose strategies, and revise their approach when comprehension breaks down.
View or buy the cited work >>Read a similar article free online >>
Azevedo and Cromley tested whether explicit self-regulated learning training improves learning from hypermedia. They randomly assigned 131 undergraduates to a training or control condition before students used a hypermedia environment to learn about the circulatory system.
The intervention provided 30 minutes of instruction in specific, research-based regulatory processes; the control group received no such training. Pretests, posttests, and verbal protocols captured both conceptual change and learners' activity during study. Students who received training showed significantly greater shifts toward more sophisticated mental models. Their verbal protocols connected this advantage to use of the strategies taught in training.
The study demonstrates that learners do not necessarily regulate effectively merely because a digital environment permits flexible navigation. Brief, targeted instruction can change how students plan, monitor, and process information, with measurable effects on conceptual understanding.
View or buy the cited work >>Read a similar article free online >>
Azevedo and Gašević examine the opportunities and difficulties involved in using multimodal, multichannel data to study self-regulated learning with advanced learning technologies.
Digital environments can capture fine-grained traces such as navigation, eye movements, dialogue, physiological signals, and performance, potentially revealing how regulation unfolds over time. Yet more data do not automatically produce valid inferences. Researchers must connect observable traces to a clear theory of regulation, align channels in time, distinguish meaningful processes from noise, and account for differences among learners and tasks. The article calls for collaboration across learning science, measurement, data science, and system design, together with transparent analytic decisions.
Its practical message is that learning analytics should be designed around educational questions rather than available sensors: evidence from multiple channels is valuable when it helps explain learners' goals, strategies, monitoring, and adaptation.
View or buy the cited work >>Read a similar article free online >>
A transparent research trail
Each entry identifies the volume, edition, chapter or appendix in which the work appears.
Evidence summaries explain the central idea and why the source matters to educators.
Entries link to lawful full text where available, or to a publisher, library or purchase page.
References are checked against publisher records, scholarly indexes, DOI or ISBN metadata, repositories and author records where available. An entry marked [Citation not verified] preserves the wording found in the relevant volume without attributing findings to an unconfirmed work.
From sources to a coherent pathway