Key ideas
Core concepts
- Memory has two components that work together: long-term memory with effectively unlimited storage, and working memory limited to about four novel items (Cowan, 2001).
- Learning is the transfer of information from working memory to long-term memory; instruction succeeds or fails on how well it manages this bottleneck through chunking, automation, and systematic sequencing.
Connected to
Cognitive Load Theory | Schema | Chunking | Fluency | Part-Whole Approach | Prior Knowledge | Learning | Retrieval Practice
Memory is a two-component cognitive system: long-term memory provides unlimited storage capacity organised as schemas, whilst working memory is a limited-capacity processor through which new learning must pass. The interaction between the two determines all learning, and the architecture creates both constraints and opportunities for instruction.
Memory is the residue of thought
People remember what they think about during an experience, not what they want to remember or what teachers intend them to remember (Willingham, 2009). Memory formation depends on where attention focuses during processing.
The penny demonstration (Nickerson and Adams, 1979) illustrates this. Most people cannot accurately recall the details of a penny despite seeing thousands throughout their lives. People think about the value of coins and how to differentiate them by size and colour, not their visual details. Memory reflects what people attend to and process, not mere exposure, and superficial processing leads to weak or absent memories.
Repetition alone does not guarantee memory formation. Students can engage with material repeatedly yet remember little if their attention focuses on irrelevant features rather than the underlying concepts. Thinking about meaning creates stronger memories than thinking about surface features (Craik & Tulving, 1975). An activity that engages students in cutting, arranging, or decorating may produce high engagement but minimal learning if students think about those surface features rather than mathematical concepts.
Long-term memory
Long-term memory is the permanent knowledge storage component. Research has identified no practical limit to how much information it can store (Landauer, 1986). Students continue learning throughout their lives without “filling up” their memory or needing to delete old information to make room for new. A teacher’s long-term memory contains millions of pieces of information (subject knowledge, pedagogical strategies, student names and characteristics, classroom procedures, personal experiences) yet continues accommodating new learning daily.
Information does not exist in long-term memory as isolated facts but as interconnected networks called schemas (Bartlett, 1932; Anderson, 1977). These organised knowledge structures represent complex information as single functional units. A mathematics teacher’s schema for “solving equations” contains thousands of connected pieces (algebraic notation, inverse operations, maintaining equality, common errors, effective explanations, typical student difficulties) yet functions as coherent, accessible knowledge.
Accessing well-established information from long-term memory imposes negligible cognitive load (Sweller, van Merriënboer, & Paas, 2019). Expert teachers simultaneously draw on vast domain knowledge, monitor student understanding, manage classroom dynamics, and adjust instruction in real time because retrieving automated knowledge does not burden the limited resources of working memory.
Information successfully encoded into long-term memory can remain accessible for decades (Bahrick, 1984). Adults recall childhood experiences, students remember foundational concepts learnt years earlier, and skills practised extensively remain available even after long periods without use.
This seemingly unlimited storage creates education’s opportunity: students can develop extensive knowledge networks that support increasingly sophisticated thinking. If storage is unlimited, the difficulty of learning must come from somewhere else, and it does: the second component.
Working memory
Working memory is where conscious thinking occurs: the mental workspace where information gets processed, manipulated, and integrated. Unlike long-term memory’s abundance, working memory operates under severe constraints.
Working memory can consciously process only about four novel elements simultaneously (Cowan, 2001). Earlier research by Miller (1956) identified approximately seven items (plus or minus two), though this included both working memory and chunking effects. Miller’s research at Bell Telephone Labs in the 1950s was commissioned to determine empirically how many random digits a person could remember, to inform telephone number design; his finding that most people can remember five to nine items directly influenced the design of seven-digit American telephone numbers. Modern research separates chunking effects from pure capacity, establishing four items as the true working memory limit (Cowan, 2001). This is not an average that improves with practice, training, or intelligence. It is a fixed architectural feature of human cognition. When encountering new information, students can juggle roughly four pieces at once before the system overloads and learning fails.
Everything processed in working memory demands attention and cognitive resources. Unlike the effortless retrieval from long-term memory, working memory operations actively consume limited capacity: thinking, analysing, integrating, and manipulating information all draw on the same resources.
Working memory is also the necessary gateway to long-term memory (Atkinson & Shiffrin, 1968). For information to transfer into permanent storage, it must first be processed in working memory. There is no shortcut; learning occurs when information successfully makes this transfer from conscious processing to permanent storage. Information held only in working memory disappears rapidly without rehearsal or transfer (Peterson & Peterson, 1959). Students might successfully process information during a lesson yet demonstrate no retention later because the transfer never occurred.
This four-item limitation creates education’s central challenge: vast amounts of knowledge must pass through a narrow bottleneck to reach long-term storage. Managing this constraint separates effective instruction from approaches that overwhelm students regardless of teacher enthusiasm or student motivation.
The multi-component model
Baddeley and Hitch (1974) developed a more detailed model of working memory, moving beyond simple capacity limits to explore how working memory functions. Their model proposed multiple components rather than a single unitary system.
The central executive controls attention and coordinates the other components. It manages which information receives processing resources, switches attention between tasks, and integrates information from different sources. It has limited capacity and becomes overloaded when managing too many competing demands.
The phonological loop processes verbal and acoustic information. It consists of a phonological store (holding speech-based information for a few seconds) and an articulatory rehearsal process (refreshing information through subvocal repetition). When students solve problems involving verbal information, the phonological loop holds intermediate steps whilst processing continues.
The visuospatial sketchpad handles visual and spatial information, storing and manipulating visual images and spatial relationships. Students use this component when visualising geometric shapes, imagining spatial arrangements, or working with diagrams and graphs.
Through ten experiments, Baddeley and Hitch examined how cognitive load affects reasoning and problem-solving. They manipulated cognitive load using verbal memory preloads (requiring participants to remember digit strings whilst completing other tasks) and examined effects of phonemic similarity and articulatory suppression. A memory load of one to three items showed minimal effects on reasoning performance, whilst a load of six items caused decrements in both verbal reasoning and comprehension. Phonemic similarity (similar-sounding items) affected verbal reasoning and comprehension but enhanced long-term recall. Articulatory suppression (preventing subvocal rehearsal) impaired reasoning and recency effects but not comprehension. These findings supported the existence of separate components with distinct functions.
The multi-component model explains why certain instructional approaches work better than others. Because visual and verbal processing are separate systems, teachers can design instruction that uses both channels without overloading either one. Presenting information simultaneously through visual diagrams and verbal explanations can reduce cognitive load compared to verbal-only presentation, provided the visual and verbal information integrate rather than compete for attention. This supports multimedia learning principles whilst cautioning against asking students to process verbal information whilst maintaining other verbal information in memory.
How the components work together
Learning occurs through a cycle of interactions between the two memory components. New information enters working memory through instruction, reading, or experience: a teacher’s explanation, a textbook passage, observations during an activity. Simultaneously, working memory retrieves related information from long-term memory to provide context; students trying to understand algebraic equations retrieve their knowledge of arithmetic operations, equality concepts, and number properties. The new information then combines with retrieved prior knowledge as students connect the new concept to existing understanding and begin constructing coherent understanding. When integration succeeds within working memory’s capacity, the new understanding moves into long-term storage, becoming part of the student’s permanent knowledge network. Over time and with extensive practice, retrieving this knowledge becomes automatic, requiring minimal working memory resources.
This cycle shows why learning depends on both components and their interaction. Strong prior knowledge in long-term memory reduces the novelty that working memory must process, and automated retrieval frees working memory capacity for new learning. When either component fails its function (missing prior knowledge in long-term memory, or cognitive overload in working memory), learning cannot occur.
Managing the bottleneck
Cognitive overload is the most common cause of learning failure. When new content contains more than about four elements that have not been chunked into existing schemas, students cannot process everything simultaneously. Some information gets dropped, integration fails, and transfer to long-term memory becomes impossible.
Missing prior knowledge compounds the problem. If students lack prerequisite knowledge, what teachers present as a few simple steps actually contains numerous novel elements. Solving seems like one or two elements to an expert but represents seven or eight to a novice missing foundational schemas: equality meaning, algebraic notation, multiplication representation, inverse operations, maintaining balance, division execution, simplification recognition. The result is predictable. Students become confused, frustrated, and disengaged. They might appear to understand during instruction but demonstrate minimal retention later; the performance was temporary processing, not permanent learning, because the transfer to long-term memory failed.
Effective instruction respects working memory limitations whilst enabling knowledge transfer. Content is presented in chunks that fit within the four-item capacity, with complex procedures broken into components, taught separately, then integrated gradually. Instruction ensures prerequisites are secure and automated, so students retrieve foundational knowledge effortlessly rather than consuming working memory processing it. Well-designed instruction also leverages information already in long-term memory (analogies to familiar concepts, connections to prior learning, building on existing schemas), reducing what working memory must process as novel. Students receive adequate time for working memory to complete its integration work before new information arrives; rushing through content prevents the processing required for transfer.
Three strategies for managing cognitive load
Since working memory capacity cannot be expanded, teaching must optimise these limited resources.
The first strategy is Chunking: grouping related information together to reduce the number of elements working memory must handle simultaneously (Miller, 1956). Instead of processing nine individual letters (C, A, T, D, O, G, B, A, T), working memory can chunk them into three familiar words: CAT, DOG, BAT. The same information requires far less capacity when organised into meaningful units. Teachers help students chunk by teaching related concepts together so students see relationships, using organising frameworks that group related ideas, making explicit the patterns that allow multiple facts to function as single units, and building schemas systematically so complex information becomes a single retrievable unit. Well-developed schemas are the ultimate chunking achievement: an expert’s schema for “solving quadratic equations” contains hundreds of individual pieces of knowledge yet functions as a single chunk when retrieved from long-term memory.
The second strategy is careful instructional design. The Part-Whole Approach and related strategies prevent overload by managing how information is presented. When content exceeds working memory capacity, components are taught separately before integration: complex elements are isolated for individual instruction, sub-skills are taught to fluency before combination, components are practised until automated before complexity is added, and integration happens gradually as individual pieces become secure. Sequencing matters too: each new lesson builds on previous learning that is now automated in long-term memory, so working memory processes only genuinely new content whilst retrieving foundations effortlessly. Scaffolding provides temporary support that manages complexity whilst students build the schemas needed for independent performance, gradually withdrawing as knowledge moves into long-term memory and retrieval becomes automated.
The third strategy is automating knowledge through practice. Fluency means developing automatic retrieval that frees working memory for new learning. Through extensive practice, knowledge that initially required conscious processing becomes retrievable automatically from long-term memory with minimal working memory demand (Ericsson & Kintsch, 1995). Basic arithmetic facts, common procedures, and fundamental concepts should all become automated through distributed practice. Every element students can retrieve automatically frees capacity for processing genuinely new content: a student with automated multiplication facts has more working memory available for learning algebra than one who must consciously calculate each product. Automation also accelerates learning. Students with automated foundations can handle more complex content because their working memory is not consumed processing prerequisites (Stanovich, 1986), so strong students learn progressively more easily whilst those lacking automated foundations struggle increasingly. Time spent building fluency in foundational knowledge is not time lost from “real” learning; it is the investment that makes subsequent learning possible by reducing future working memory demands.
Story structure and memory
Information organised as stories is easier to remember than isolated facts. Stories have a specific structure: causation links events through cause and effect, conflict creates interest and drives the narrative forward, complications add depth and maintain attention, and character provides relatable elements that anchor abstract information in concrete contexts. When content lacks this structure, memory formation suffers.
The four Cs of story structure explain why students remember television programmes and films more readily than classroom content. Television narratives use causation, conflict, complications, and character automatically. Classroom content often presents information as lists of facts without causal connections or narrative structure. Teachers can apply story structure by emphasising causal relationships between concepts, presenting content through problem-solving scenarios that create natural conflict, introducing complications that deepen understanding, and using concrete examples with relatable contexts.
Emotional connections strengthen memory formation. Content linked to emotions encodes more strongly than neutral information, but the emotion must connect to the content itself rather than peripheral elements. An engaging story about a teacher’s holiday might be memorable yet teach nothing about mathematics if students think about the story rather than the mathematical concepts. Emotional connections should direct attention towards the learning content, not away from it.
Types of long-term memory
Long-term memory stores different types of information serving distinct functions in thinking and learning (Tulving, 1972; Squire, 1987).
Semantic memory
Semantic memory contains general knowledge, concepts, and facts independent of specific personal experiences (Tulving, 1972): facts (capitals of countries, mathematical properties, scientific principles), concepts (democracy, photosynthesis, justice, multiplication), vocabulary and language rules, and general knowledge accumulated through schooling and life.
Most academic learning targets semantic memory, building students’ knowledge of subject content, concepts, and procedures they will use for thinking and problem-solving. The goal is constructing robust semantic memory networks that support increasingly sophisticated understanding.
Well-established semantic memory retrieves without conscious recall of when or how it was learnt. Students know that 7 × 8 = 56 without remembering the specific lesson where they learnt it. Semantic memory functions as a mental thesaurus containing organised knowledge about words, symbols, meanings, relations, rules, and formulas, and retrieval leaves its contents unchanged: the knowledge remains stable through repeated access.
The semantic system also allows retrieval of information not directly stored, through inductive and deductive reasoning. Students can answer questions they have never encountered by reasoning from stored knowledge. Understanding that parallel lines never intersect allows students to determine properties of parallel lines in novel configurations without having explicitly learnt each specific case.
Episodic memory
Episodic memory stores specific events and experiences: memories of things that happened at particular times and places (Tulving, 1972). It holds personal experiences (yesterday’s lesson, last summer’s holiday), specific events and their contexts, memories tied to particular times, places, and circumstances, and information about experienced occurrences of episodes or events. Remembering learning about triangles in Year 7 with a specific teacher using particular examples creates an episodic memory.
Unlike semantic memory, retrieval from episodic memory increases accessibility but also changes the memory contents. Episodic memories are susceptible to transformation and information loss over time; each retrieval can modify the memory, incorporating new details or losing original information.
Whilst academic learning primarily targets semantic memory, episodic memories can support it: remembering the lesson where a concept was introduced, recalling specific examples or demonstrations, associating content with memorable experiences or explanations, and providing contextual anchors for semantic knowledge. Often, information begins as episodic (remembering the lesson where fraction addition was taught) but transitions to semantic (knowing how to add fractions without remembering learning it). This transition represents successful long-term learning. Effective instruction should help students develop robust semantic knowledge whilst recognising that episodic memories can support initial retrieval and provide meaningful context.
Procedural memory
Procedural memory stores how to perform skills and procedures, often operating without conscious awareness (Squire, 1987). It holds physical skills (writing, typing, riding a bicycle) and automated procedures (solving familiar problem types, executing algorithms): “knowing how” rather than “knowing that”.
Procedural memory develops through extensive repetition. Initially, procedures require conscious working memory processing; with practice, they become automated, executing with minimal conscious effort. Many academic skills should ultimately reside in procedural memory (basic calculation procedures, reading processes, common problem-solving algorithms), freeing working memory for higher-order thinking.
Dual coding
Dual coding theory proposes that information is processed and stored in two separate but interconnected systems: a verbal system specialised for language and a non-verbal system specialised for imagery (Paivio, 1971).
The verbal system processes linguistic information, including written and spoken language, phonological representations, and abstract concepts expressed through words. The non-verbal system processes visual and spatial information, including images, diagrams, spatial relationships, and concrete visual representations. These systems can work independently or together, with referential connections allowing translation between verbal and visual representations.
Concrete information that can be coded both verbally and visually is better remembered than abstract information coded only verbally. When information is dual-coded, memory and learning improve because multiple retrieval routes exist. A student learning about triangles benefits from both the verbal definition (“a polygon with three sides and three angles”) and visual representations showing different triangle types; either code can trigger recall of the complete concept.
Presenting information in both visual and verbal formats can enhance learning, mainly for concrete concepts, but the benefit depends on the visual and verbal information being complementary and integrated rather than redundant or contradictory. Teachers should use diagrams, illustrations, and other visual supports alongside verbal explanations. The approach is more effective for novices than for experts, who can generate their own mental imagery (Paivio, 1971). When teaching fractions, for instance, teachers can combine verbal explanations of numerators and denominators with visual models showing fraction bars or circles divided into parts.
Not all concepts suit dual coding. Abstract concepts like justice or democracy resist visual representation without oversimplification. Visual and verbal information must be integrated to avoid split attention effects where students divide working memory between competing sources, and the quality of both the verbal explanation and the visual representation matters: poor diagrams or unclear descriptions undermine the advantage.
Forgetting
Forgetting has two main causes, and understanding them helps teachers design instruction that promotes retention rather than temporary performance.
The most common form of “forgetting” occurs when information never transfers to long-term memory in the first place. When instruction exceeds working memory capacity, information gets processed temporarily but fails to transfer; students appear to understand during the lesson but demonstrate no retention later because encoding never occurred. Even within capacity, information needs adequate processing time, and rushed instruction prevents the integration work required for encoding. Passive exposure fails for the same reason: simply hearing or seeing information does not guarantee encoding, and students need active engagement (connecting to prior knowledge, explaining, practising) to facilitate transfer. Prevention means respecting working memory limits, providing sufficient processing time, requiring active engagement, and building on automated prior knowledge.
Sometimes information successfully encoded into long-term memory becomes inaccessible (Tulving & Thomson, 1973). Information encoded weakly, without strong connections to existing knowledge or without elaboration, becomes difficult to retrieve even though it exists in long-term memory (Craik & Tulving, 1975). Knowledge that is not retrieved regularly becomes progressively harder to access as the neural pathways weaken without use. Information encoded in one context might also be inaccessible in different contexts if learning was too narrowly focused. Prevention means building strong initial encoding through elaboration and connection-making, providing distributed practice for regular retrieval, and teaching concepts in varied contexts to enable flexible access.
Working memory’s limited capacity may itself serve an adaptive function. The limitation ensures only information encountered repeatedly transfers to long-term memory, a selectivity that helps build useful schemas rather than cluttering memory with every fleeting experience. By requiring multiple exposures for encoding, the memory system naturally emphasises patterns and regularities rather than isolated events, which supports schema development. And if everything entered long-term memory equally, retrieval would become inefficient, a search through vast amounts of irrelevant information to find what is needed.
Working with memory architecture
When planning lessons, count the novel elements students must process simultaneously. If content exceeds about four new elements, break it into smaller parts or ensure some elements are already automated in students’ long-term memory. Before teaching new content, check that prerequisite knowledge is automated; students cannot process both prerequisites and new content simultaneously in working memory. Single exposure rarely creates robust encoding, so plan for distributed practice and repeated retrieval over time.
During instruction, chunk information strategically: make patterns explicit, teach organising frameworks, and help students see relationships. Manage presentation pace so working memory has time to integrate new information with prior knowledge before transfer can occur. Connect explicitly to prior knowledge, helping students retrieve relevant information from long-term memory to provide context; these connections reduce working memory burden whilst strengthening encoding. Minimise extraneous load through better design: integrate text and diagrams, provide complete materials, create quiet environments. Every bit of extraneous load wastes working memory capacity.
During practice, focus on building automation. Practice should continue until retrieval becomes automatic, not just accurate. Distribute practice over time: massed practice might produce temporary performance, but distributed practice creates stronger long-term encoding (Cepeda et al., 2006), as the Spacing Effect shows. Require active retrieval, since having students retrieve information from long-term memory strengthens encoding more effectively than passive review (Roediger & Karpicke, 2006). Retrieval Practice forces the encoding-retrieval cycle that consolidates learning.
During assessment, evaluate long-term retention rather than temporary performance. Performance during or immediately after instruction might reflect working memory processing rather than long-term encoding, so assess after a delay. Check for accessibility, not just storage: can students retrieve knowledge flexibly in varied contexts, or only in specific familiar situations? When students demonstrate poor retention, determine whether encoding failed initially (working memory overload) or retrieval is difficult (weak encoding); this diagnostic information guides the instructional response.
Misconceptions about memory
Several common beliefs about memory misread its architecture.
“Students just need to memorise better” treats memory as a willpower issue rather than an architectural constraint. When students fail to remember, the information usually never transferred to long-term memory (encoding failure due to working memory overload) or was encoded weakly (insufficient processing or poor connections). Trying harder to “memorise” does not address these causes. The remedy is instruction that respects working memory limitations, builds strong encoding through connection-making and elaboration, and provides distributed practice for consolidation.
“Working memory can be trained to expand” suggests the constraint can be overcome through practice. Extensive research attempting to expand working memory capacity through training has consistently failed (Melby-Lervåg & Hulme, 2013). The four-item limitation is a fixed feature of human cognition, so instruction must work within it through chunking, automation of prerequisites, and careful presentation design.
“Understanding means students will remember” assumes temporary working memory processing guarantees long-term encoding. Students can understand information in the moment, processing it successfully in working memory, yet demonstrate no retention later because transfer to long-term memory never occurred or was weak. Plan for extended practice and distributed retrieval rather than assuming apparent understanding during instruction means robust memory formation.
“Some students just have bad memories” attributes memory failures to inherent student characteristics rather than instructional design. Some individual differences exist, but most memory failures result from working memory overload (too many novel elements), missing prerequisites (weak long-term memory foundations), or insufficient practice (weak encoding). Analyse memory failures for their underlying causes, then adjust instruction accordingly.
Memory as the foundation for learning
Memory architecture explains how learning occurs and what effective teaching requires. The two-component system, unlimited long-term memory for storage and severely limited working memory for processing, defines both the challenge and the pathway.
The challenge: vast knowledge must pass through working memory’s narrow bottleneck to reach long-term storage. When instruction exceeds this capacity, learning fails regardless of teacher quality or student effort. The architecture imposes a hard constraint that cannot be overcome through willpower, training, or increased effort.
The pathway: instruction must work with memory architecture rather than against it. This means respecting working memory limitations through careful sequencing and chunking, building automated prior knowledge that frees working memory for new learning, eliminating extraneous cognitive load through better design, and providing sufficient practice for robust encoding.
Memory architecture also explains patterns teachers observe daily. Strong students seem to learn progressively more easily because their automated long-term memory knowledge reduces working memory demands for new learning. Struggling students fall further behind because missing prerequisites consume working memory that should process new content. Some teaching approaches consistently succeed whilst others fail because they either work with or violate memory architecture. For teachers committed to student learning, this knowledge moves practice from intuitive guessing to evidence-based design: how to sequence content, when to provide practice, how much to present simultaneously, and how to assess learning all rest on how memory functions.
References
Anderson, R. C. (1977). The notion of schemata and the educational enterprise: General discussion of the conference. In R. C. Anderson, R. J. Spiro, & W. E. Montague (Eds.), Schooling and the acquisition of knowledge (pp. 415-431). Lawrence Erlbaum Associates.
Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation (Vol. 2, pp. 89-195). Academic Press.
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (Vol. 8, pp. 47-89). Academic Press.
Bahrick, H. P. (1984). Semantic memory content in permastore: Fifty years of memory for Spanish learned in school. Journal of Experimental Psychology: General, 113(1), 1-29. https://doi.org/10.1037/0096-3445.113.1.1
Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. Cambridge University Press.
Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. https://doi.org/10.1037/0033-2909.132.3.354
Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87-114. https://doi.org/10.1017/S0140525X01003922
Craik, F. I. M., & Tulving, E. (1975). Depth of processing and the retention of words in episodic memory. Journal of Experimental Psychology: General, 104(3), 268-294. https://doi.org/10.1037/0096-3445.104.3.268
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review, 102(2), 211-245. https://doi.org/10.1037/0033-295X.102.2.211
Landauer, T. K. (1986). How much do people remember? Some estimates of the quantity of learned information in long-term memory. Cognitive Science, 10(4), 477-493. https://doi.org/10.1207/s15516709cog1004_4
Melby-Lervåg, M., & Hulme, C. (2013). Is working memory training effective? A meta-analytic review. Developmental Psychology, 49(2), 270-291. https://doi.org/10.1037/a0028228
Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81-97. https://doi.org/10.1037/h0043158
Nickerson, R. S., & Adams, M. J. (1979). Long-term memory for a common object. Cognitive psychology, 11(3), 287-307.
Paivio, A. (1971). Imagery and verbal processes. Holt, Rinehart and Winston.
Peterson, L. R., & Peterson, M. J. (1959). Short-term retention of individual verbal items. Journal of Experimental Psychology, 58(3), 193-198. https://doi.org/10.1037/h0049234
Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249-255. https://doi.org/10.1111/j.1467-9280.2006.01693.x
Squire, L. R. (1987). Memory and the brain. American Psychologist, 42(2), 170-179. https://doi.org/10.1037/0003-066X.42.2.170
Stanovich, K. E. (1986). Matthew effects in reading: Some consequences of individual differences in the acquisition of literacy. Reading Research Quarterly, 21(4), 360-407. https://doi.org/10.1598/RRQ.21.4.1
Sweller, J., van Merriënboer, J. J. G., & Paas, F. (2019). Cognitive architecture and instructional design: 20 years later. Educational Psychology Review, 31(2), 261-292. https://doi.org/10.1007/s10648-019-09465-5
Tulving, E. (1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory (pp. 381-403). Academic Press.
Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352-373. https://doi.org/10.1037/h0020071
Willingham, D. T. (2009). Why don’t students like school? A cognitive scientist answers questions about how the mind works and what it means for the classroom. Jossey-Bass.