ArticlePDF Available

Chunking mechanisms and learning



Content may be subject to copyright.
Fernand Gobet
Department of Psychology, Brunel University
United Kingdom
Peter C. R. Lane
School of Computer Science, University of Hertfordshire
United Kingdom
A chunk is meaningful unit of information built from smaller pieces of information, and chunking is the
process of creating a new chunk. Thus, a chunk can be seen as a collection of elements that have strong
associations with one another, but weak associations with elements belonging to other chunks. Chunks, which
can be of different sizes, are used by memory systems and more generally by the cognitive system. Within
this broad definition, two further meanings can be differentiated. First, chunking can be seen as a deliberate,
conscious process. Here, we talk about goal-oriented chunking. Second, chunking can be seen as a more
automatic and continuous process that occurs during perception. Here, we talk about perceptual chunking.
Theoretical Background
Chunking as a mechanism was initially proposed by De Groot (1946/1978) in his study of chess experts’ <<
link to Development of expertise>> perception, memory and problem solving, to explain their ability to
recall briefly presented positions with a high level of precision. It was also a central ingredient of Miller’s
(1956) classical article about the limits on human information-processing capacity. Miller proposed that chunk
are the correct measure for the information in the human cognitive system, and that 7 ± 2 chunks can be
Gobet, F., & Lane, P. C. R. (2012). Chunking mechanisms and learning.
In N. M. Seel (Ed.), Encyclopedia of the sciences of learning.
New York, NY: Springer.
held in short-term memory. Chase and Simon (1973) proposed a general theory of processes underpinning
chunking. It is interesting to note that the approaches of De Groot as well as Chase and Simon emphasize
the implicit nature of chunks, that are seen as the product of automatic learning processing sometimes called
perceptual chunking. Miller’s view emphasizes a type of strategic, goal-oriented chunking, where chunking is
essentially re-coding of the information in a more efficient way. For example, the 9-digit binary number
101000111 can be re-coded as the 3-digit decimal number 327, making it easier to process and memorize for
humans. The presence of chunks explains how humans, in spite of strict cognitive limitations in memory
capacity, attention, and learning rate, can cope efficiently with the demands of the environment. Chunking
has been established as one of the key mechanisms of human cognition, and plays an important role in
showing how internal cognitive processes are linked to the external environment.
There is considerable empirical evidence supporting the notion of a chunk, for example in our ability to
perceive words, sentences or even paragraphs as single units, bypassing their representation as collections of
letters or phonemes; this explains, for example, how skilled readers may be insensitive to word repetition or
deletion. Particularly strong evidence is found in those studies which use information about the timing of
responses to infer the presence of chunks. The use of response times assumes that the output of elements
within a chunk will be faster than the output of elements across different chunks. This is because the
elements within a chunk belong to the same structure, as well as sharing a number of relations. There is good
empirical evidence confirming that subjects’ pauses are shorter within chunks than between chunks. For
example, timing information shows that when the alphabet is recited back, letters are grouped in clusters, and
clusters grouped in super-clusters. When trained to learn alphabets using scrambled letter orders, subjects also
recall letters in a burst of activity followed by a pause, and therefore show evidence for clusters.
The strongest empirical evidence for chunks is based on their inference from several converging methods.
For example, studies on chess have shown that chunks identified by latencies in recall or placement of chess
pieces correlate highly with chunks identified by the number of relations shared between successively placed
pieces. By analyzing the patterns picked out by chess players within a position for various natural relations
(including proximity, color, and relations of attack or defense), it is evident that within-chunk relations are
much stronger than between-chunk relations. This pattern was found whether the subjects were asked to
place pieces on the board from memory (using timings to separate the groups), or to copy a board (using the
presence of glances between the two boards to separate the groups). Further empirical evidence for chunking
has been uncovered in a number of other areas including artificial-grammar learning, problem solving, and
animal research.
The chunking theory, developed by Chase and Simon (1973) was an important attempt to formalize the
mechanisms linked to chunking. It postulated that attention is serial and short-term memory limited to about
seven items (Miller’s magical number). When individuals acquire information about a domain with practice
and study, they acquire an increasingly larger number of chunks, which themselves tend to become larger, up
to a limit of four or five items. While learning is assumed to be slow (10 seconds per chunk), recognition of
the information stored in a chunk occurs in a matter of hundreds of milliseconds. Another important
assumption is that chunks are linked to possible information. For example, in chess, the domain in which the
theory was first applied, a chunk could provide information about potentially useful moves (see Figure 1).
Chunks help in a recall task, because groups of pieces rather than individual pieces can be stored in short-
term memory. They also help in a problem-solving task, because some of the chunks, being linked to
potentially useful information, provide clues about what kind of action should be taken.
Figure 1. Top panel: examples of chunks in a chess position. Bottom panel: one of the chunks elicits a
possible move (retreating the white bishop).
There is also evidence that people, in particular experts in a domain, use higher-level representations than
chunks. For example, data from chess research indicate that sometimes the entire position, up to 32 pieces, is
handled as a single unit by grandmasters. In addition, evidence from expertise research indicates that
information can sometimes be encoded in long-term memory faster than the 10 seconds proposed by
chunking theory. Together, these results led to a revision of the chunking theory with the template theory
(Gobet & Simon, 1996). The template theory proposes that frequently used chunks become “templates”, a
type of schema. A template consists of a core, which contains constant information, and slots, where variable
information can be stored. The presence of templates considerably expands experts’ memory capability.
The image cannot be displayed. Your compu ter may not have enou gh memory to open the image, or the image may have b een corrup ted. Restart your compu ter, and then open the file again. If the red x still appears, you may have to delete the image and then insert it again.
A methodological difficulty with research on chunking has been to precisely identify the boundaries between
chunks. For example, the most direct explanation for observing a set of actions as a chunk is for the actions
to be represented internally as a single unit, i.e. a chunk, and so retrieved and output together. However, it is
also possible for a subject to plan output actions ahead, and so either break long sequences into sub-parts
(e.g., to take a breath when reciting the alphabet) or else compose short sequences into what appear as longer
ones (e.g., where a second chunk begins naturally from where the first one finished). Distinguishing between
these types is only possible with the aid of a computational model, where the precise items of information
known by the subject at a given point in time can be ascertained (Gobet et al., 2001). The advantage of using
computer models is discussed in more detail in the entry on Learning in the CHREST cognitive architecture,
a model based on the template theory.
Chunk-based theories, such as the chunking and template theories, not only provide a powerful explanation
of learning and expert behavior, but also offer useful information as to how learning occurs in the classroom
and how it could be improved (Gobet, 2005). We briefly discuss some of the implications for education
(further principles are listed in Table 1).
Table 1. Educational principles derived from chunk-based theories (after Gobet, 2005).
Teach from the simple to the complex
Teach from the known to the unknown
The elements to be learned should be clearly identified
Use an ‘improving spiral,’ where you come back to the same concepts and ideas and add
increasingly more complex new information
Focus on a limited number of types of standard problem situations, and teach the various methods
in these situations thoroughly
Repetition is necessary. Go over the same material several times, using varying points of view and
a wide range of examples
At the beginning, don’t encourage students to carry out their own analysis of well-known problem
situations, as they do not possess the key concepts yet
Encourage students to find a balance between rote learning and understanding
A first implication of chunk-based theories is that acquiring a new chunk has a time cost, and therefore time
at the task is essential, be it in mathematics or dancing. As documented by research into deliberate practice
<<link to Deliberate practice>>, practice must be tailored to the goal of improving performance. Chunk-
based theories give attention a central role – see for example the CHREST model – and such theories are
therefore suitable models of deliberate practice. In particular, conceptual knowledge is built on perceptual
skills, which in turn must be anchored on concrete examples. Thus, curricula should provide means to acquire
perceptual chunks in a given domain.
There are different useful ways to direct attention and to encourage the acquisition of perceptual chunks: to
segment the curriculum into natural components, of the right size and difficulty; to present these
components with an optimal ordering and suitable feedback; and to highlight the important features of a
If perceptual chunking is an important way of storing knowledge, then a clear consequence is that transfer
will be difficult. Unfortunately for learners, this prediction is correct, both for school knowledge and more
specific skills such as sports and arts. More than 100 years of research have established that transfer is
possible from one domain to another only when the components of the skills required in each domain
overlap. Thus, it might be helpful to augment the teaching of specific knowledge with the teaching of
metaheuristics – including strategies about how to learn, how to direct one’s attention, and how to monitor
and regulate one’s limited cognitive resources.
As noted above, an important idea in Chase and Simon’s (1973) theory is that perceptual chunks can be used
as conditions to actions, thus leading to the acquisition of productions. Then, an important aspect of
education is to balance the acquisition of the condition and action parts of productions. Another important
aspect of education is to favor the acquisition of templates (schemata). Templates are created when the
context offers both constant and variable information. As a consequence, and as is well established in the
educational literature, it is essential to have variability during learning if templates are to be created.
Finally, chunk-based theories are fairly open to the possibility of large individual differences in people’s
cognitive abilities. In particular, while they postulate fixed parameters for short-term memory capacity and
learning rates, it is plausible that these parameters vary between individuals. In addition, differences in
knowledge will lead to individual differences in performance. A clear prediction of chunk-based theories is
that individual differences play a large role in the early stages of learning, as is typical of classroom
instruction, but tend to be less important after large amounts of knowledge have been acquired through
practice and study.
Important Scientific Research and Open Questions
Chunk-based theories have spurred vigorous research is several aspects of learning and expertise. A first
aspect is the acquisition of language, where recent research has shown that chunking plays an important role
in the development of vocabulary and syntactic structures. A second aspect is related to the neurobiological
basis of chunking. Recent results indicate that perceptual chunks are stored in the temporal lobe, and in
particular the parahippocampal gyrus and fusiform gyrus.
Other issues being currently researched include the effect of order in learning, and in particular how curricula
can be designed so that that they optimize the transmission of knowledge. A possible avenue for future
research is the design of computer tutors that use chunking principles for teaching various materials,
optimizing instruction for the abilities and level of each student by providing personalized curricula,
providing judicious feedback, and teaching strategies.
→ Bounded rationality and learning
→ Decision making and learning
→ Deliberate practice
→ Development of expertise
→ Learning in the CHREST cognitive architecture
→ Schema
1. Gobet, F., & Simon, H. A. (1996). Templates in chess memory: A mechanism for recalling several
boards. Cognitive Psychology, 31, 1-40.
2. Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4, 55-81.
3. De Groot, A. D. (1978). Thought and choice in chess (first Dutch edition in 1946). The Hague: Mouton
4. Gobet, F. (2005). Chunking models of expertise: Implications for education. Applied Cognitive
Psychology, 19, 183-204.
5. Gobet, F., Lane, P. C. R., Croker, S., Cheng, P. C-H., Jones, G., Oliver, I. & Pine, J. M. (2001).
Chunking mechanisms in human learning. TRENDS in Cognitive Sciences, 5, 236-243.
6. Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on our capacity for
processing information. Psychological Review, 63, 81-97.
Chunk: A meaningful unit of information built from smaller pieces of information
Chunking: The creation of new chunks in long-term memory.
Short-term memory span: The largest amount of information that can be held in short-term memory at a
given time.
Chunking theory: Theory developed by Chase and Simon in 1973, explaining how experts circumvent the
limitations of cognitive processes through the acquisition of domain-specific knowledge, in particular
small meaningful units of inter-connected elements (chunks).
Template theory: Theory of expertise, developed in 1996 by Gobet and Simon, building on the chunking
theory and proposing that well-elaborated chunks lead to larger meaningful units (templates).
... Chunking adalah salah satu model penyimpanan memori di otak yang diperkenalkan seorang ahli bernama De Groot. Dalam psikologi model ini diimplementasikan dengan memberikan potongan informasi yang saling terkait maknanya secara berulang [10]. Model penyampaian informasi ini dapat mempercepat proses mengingat dan meningkatkan kualitas belajar. ...
Full-text available
Kemampuan belajar anak-anak di sekolah mempengaruhi prestasi belajarnya. Guru berperan penting dalam proses belajar anak-anak. Ketidakmampuan guru memahami peran otak dan bagaimana fungsi bagian-bagian otak secara keseluruhan untuk mendukung proses belajar ini menjadi salah satu hambatan mengembangkan metode belajar yang optimal. Tujuan kegiatan workshop adalah untuk meningkatan pengetahuan dan keterampilan guru tentang perkembangan dan fungsi otak. Para guru mendapatkan materi, diskusi dan simulasi tentang anatomi dan fisiologi otak dalam proses belajar. Evaluasi kegiatan dalam bentuk pre dan post tes dan laporan hasil praktek metode belajar yang sesuai dengan ritme kerja otak anak (brain base learning). Hasil kegiatan ini menguatkan kemampuan guru dalam proses belajar dan mengajar yang memacu perkembangan otak, meningkatkan kemampuan belajar serta mencegah terjadinya perlambatan proses belajar anak di sekolah.
... -The feeling of consciousness arises in CTM because all its processors, including especially those that are particularly responsible for consciousnessthe Inner Speech, Inner Vision, Inner Sensation and Model-of-the-World processorsare privy to the same (conscious) content of STM; the gists of outer speech (what we say and hear in the world), outer vision (what we see in the world), and so on, are nearly indistinguishable from the gists of inner speech (what we say to ourselves) and inner vision (what we see in dreams); and the multi-modal gists, most importantly, are expressed in Brainish, an enormously expressive language capable of generating the illusion of sensations, actions, and feelings. bb See, e.g., \Chunking Mechanisms and Learning" [Gobet and Lane, 2012]. (7) It gives some understanding of how a pain or pleasure experience, not just its simulation, is produced (Sec. ...
Full-text available
The quest to understand consciousness, once the purview of philosophers and theologians, is now actively pursued by scientists of many stripes. This paper studies consciousness from the perspective of theoretical computer science. It formalizes the Global Workspace Theory (GWT) originated by the cognitive neuroscientist Bernard Baars and further developed by him, Stanislas Dehaene, and others. Our major contribution lies in the precise formal definition of a Conscious Turing Machine (CTM), also called a Conscious AI. We define the CTM in the spirit of Alan Turing’s simple yet powerful definition of a computer, the Turing Machine (TM). We are not looking for a complex model of the brain nor of cognition but for a simple model of (the admittedly complex concept of) consciousness. After formally defining CTM, we give a formal definition of consciousness in CTM. We later suggest why the CTM has the feeling of consciousness. The reasonableness of the definitions and explanations can be judged by how well they agree with commonly accepted intuitive concepts of human consciousness, the range of related concepts that the model explains easily and naturally, and the extent of its agreement with scientific evidence.
... 55 See e.g., "Chunking Mechanisms and Learning" (Gobet, 2012). 56 Inconsistencies are detected by the Model-of-the-World processors (Chapter 3 (2)), among others. ...
Full-text available
The quest to understand consciousness, once the purview of philosophers and theologians, is now actively pursued by scientists of many stripes. This paper studies consciousness from the perspective of theoretical computer science. It formalizes the Global Workspace Theory (GWT) originated by cognitive neuroscientist Bernard Baars and further developed by him, Stanislas Dehaene, and others. Our major contribution lies in the precise formal definition of a Conscious Turing Machine (CTM), also called a Conscious AI. We define the CTM in the spirit of Alan Turing's simple yet powerful definition of a computer, the Turing Machine (TM). We are not looking for a complex model of the brain nor of cognition but for a simple model of (the admittedly complex concept of) consciousness. After formally defining CTM, we give a formal definition of consciousness in CTM. We then suggest why the CTM has the feeling of consciousness. The reasonableness of the definitions and explanations can be judged by how well they agree with commonly accepted intuitive concepts of human consciousness, the breadth of related concepts that the model explains easily and naturally, and the extent of its agreement with scientific evidence.
... Psychological and neurological theory has answered these questions by proposing two parallel mechanisms: chunking, for sensory, and conceptual or declarative knowledge, and automaticity, for behavior or procedural knowledge (Gobet & Lane, 2012;Moors, 2016). Both processes involve data compression. ...
Full-text available
The theme of IEEE ICCI*CC'16 on Cognitive Informatics (CI) and Cognitive Computing (CC) was on cognitive computers, big data cognition, and machine learning. CI and CC are a contemporary field not only for basic studies on the brain, computational intelligence theories, and denotational mathematics, but also for engineering applications in cognitive systems towards deep learning, deep thinking, and deep reasoning. This paper reports a set of position statements presented in the plenary panel (Part I) in IEEE ICCI*CC'16 at Stanford University. The summary is contributed by invited panelists who are part of the world's renowned scholars in the transdisciplinary field of CI and CC.
... Chunk decomposition is a reciprocal process to "chunking". The term "chunking" refers to grouping strongly or weakly associated information components into a meaningful pattern (Chase & Simon, 1973;De Groot, 1978;Gobet & Lane, 2012;Gobet et al., 2001;Miller, 1956). Chunking has wide application in human cognition (Gobet et al., 2001), affecting a diverse set of mental tasks including the learning of action repertoires (Graybiel, 1998) speech (Gilbert, Boucher, & Jemel, 2015), memory (Miller, 1956), and problem solving (Chase & Simon, 1973;De Groot, 1978). ...
A wealth of studies have investigated how to overcome experience-based constraints in creative problem solving. One such experience-based constraint is the tendency for people to view tightly organized visual stimuli as single, unified percepts, even when decomposition of those stimuli into component parts (termed chunk decomposition) would facilitate problem solving. The current study investigates the neural underpinnings of chunk decomposition in creative problem solving by analyzing event-related potentials. In two experiments, participants decomposed Chinese characters into the character's component elements and then used the base elements to form a new valid character. The action could require decomposing a “tight” chunk, meaning that the component elements intersected spatially, or a “loose” chunk, in which the component elements did not overlap in space. Behaviorally, individuals made more errors and responded slower to trials involving tight chunks relative to loose chunks. Analysis of the ERPs revealed that relative to loose chunks, the electrophysiological response to tight chunks contained an increased N2, an increased N400, and a decreased late positive complex. Taken together, these results suggest that chunk tightness is a principle determinant of the difficulty of chunk decomposition, and that chunk tightness provokes neural conflict and semantic violations, factors known to influence the N2 and N400 ERP components.
Conference Paper
Full-text available
This study aimed to investigate the proposed standards of designing microcontent in a digital learning environment. The research used the analytical descriptive research procedures based on deliberative inquiry approach in the presentation, study and analysis of the researches to obtain the standards. A final list of 4 standards and 62 indicators can be used to design microcontent in digital learning environments for application in digital education and training fields.
Full-text available
Teaching reading skill to the students needs applying the appropriate strategy to stimulate students’ understanding to meet the main purpose in the teaching and learning process. It was expected that the strategy itself should be able to be used by the teacher to design the better ways for the students in order to acquire important information. It is realized that many students now need as much information as possible in comprehending all the texts they read. To overcome this problem, the researcher used chunking strategy to improve students’ reading skill. The researcher used true experiment to design this research, then the pre-test was used to find out whether this group in the same level and the post-test was used to detect the progress of the students. The total numbers of the students who involved in the pre-test and post-test were 20. The data showed that the information in the post-test that; mean score was 70.50, and the pre-test got score was 61.85 then the standard deviation in the post-test was 10.32 and the standard deviation in pre-test is 8.27. It could be concluded that the students get significant progress in their understanding of reading skill by using Chunking strategy in English Department.
Full-text available
Automation, as we know is not a new phenomenon, automating the automation process that can learn, evolve, decide, and act as per context, just like a human is the new paradigm shift we all are experiencing of late with the advent of new technological enablers. The recent advances in computational capabilities, availability of historic data, low cost of sensors, and high-speed data transfer mechanisms are creating an opportunity to mimic a human brain and automate the process which matures along with the organization. Enterprises are now relying on software robotic systems which mimics the ways a repetitive task is handled by humans and preforms it in an analogous way, thereby eliminating the need of human intervention. But, these systems can neither take judgemental decisions nor can learn from previous actions and its results, thus restricting an automation process to simple repetitive process automations. This problem can be addressed by integrating cognitive features into software robotic systems, which will transform an enterprise into truly digital. In this paper we explored the evolution of cognitive systems, architecture, applications (case study) and challenges ahead in using cognitive models for industrial automations.
Learning, memorizing and recalling knowledge are the basic functions of cognitive models. These models must prioritize which stimulants to respond to as well as package acquired knowledge in an easy to retrieve manner. The human brain is a cognitive model that derives information from sensor data such as vision, associates different patterns to create knowledge, and uses chunking mechanisms to package the acquired knowledge in manageable entities. The use of chunking mechanisms by the brain aids it to overcome its short-term memory (STM) capacity limitation. Through chunking, each entity held in the STM is a chunk containing more associations (knowledge) in it. By mimicking the human brain, this study proposes an associative memory and recall (AMR) model that stores associative knowledge from sensor data. Using chunking mechanisms, AMR can organize human activity knowledge in the manner that is efficient and effective to store and recall. The knowledge–information–data (KID) model is used for learning associative knowledge while the AMR continuously looks for associations among knowledge units and merges related units using merging mechanisms. The chunking mechanisms used in this study are inspired by the chunking mechanisms of the brain i.e. goal oriented chunking and automatic chunking.
Full-text available
Chunking models offer a parsimonious explanation of how people acquire knowledge and have been validated in domains such as expert behaviour and the acquisition of language. In this paper, we review two computational theories based on chunking mechanisms (the chunking theory and the template theory) and show what insight they offer for instruction and training. The suggested implications include the importance of perception in learning, the cost of acquiring knowledge, the significance of segmenting and ordering instruction material, the role of the variability of the instructional material in acquiring schemata, and the importance of taking individual differences into account.
Full-text available
This paper addresses empirically and theoretically a question derived from the chunking theory of memory (Chase & Simon, 1973a, 1973b): To what extent is skilled chess memory limited by the size of short-term memory (about seven chunks)? This question is addressed first with an experiment where subjects, ranking from class A players to grandmasters, are asked to recall up to five positions presented during 5 s each. Results show a decline of percentage of recall with additional boards, but also show that expert players recall more pieces than is predicted by the chunking theory in its original form. A second experiment shows that longer latencies between the presentation of boards facilitate recall. In a third experiment, a Chessmaster gradually increases the number of boards he can reproduce with higher than 70% average accuracy to nine, replacing as many as 160 pieces correctly. To account for the results of these experiments, a revision of the Chase-Simon theory is proposed. It is suggested that chess players, like experts in other recall tasks, use long-term memory retrieval structures (Chase & Ericsson, 1982) or templates in addition to chunks in short-term memory to store information rapidly.
This paper develops a technique for isolating and studying the per- ceptual structures that chess players perceive. Three chess players of varying strength - from master to novice - were confronted with two tasks: ( 1) A perception task, where the player reproduces a chess position in plain view, and (2) de Groot's ( 1965) short-term recall task, where the player reproduces a chess position after viewing it for 5 sec. The successive glances at the position in the perceptual task and long pauses in tbe memory task were used to segment the structures in the reconstruction protocol. The size and nature of these structures were then analyzed as a function of chess skill. What does an experienced chess player "see" when he looks at a chess position? By analyzing an expert player's eye movements, it has been shown that, among other things, he is looking at how pieces attack and defend each other (Simon & Barenfeld, 1969). But we know from other considerations that he is seeing much more. Our work is concerned with just what ahe expert chess pIayer perceives.
What does a chessmaster think when he prepartes his next move? How are his thoughts organized? Which methods and strategies does he use by solving his problem of choice? To answer these questions, the author did an experimental study in 1938, to which famous chessmasters participated (Alekhine, Max Euwe and Flohr). This book is still usefull for everybody who studies cognition and artificial intelligence.
  • F Gobet
  • P C R Lane
  • S Croker
  • P. C-H Cheng
  • G Jones
  • I Oliver
  • Pine
Gobet, F., Lane, P. C. R., Croker, S., Cheng, P. C-H., Jones, G., Oliver, I. & Pine, J. M. (2001).