TO APPEAR IN OXFORD HANDBOOK OF HUMAN MEMORY
Expertise and Memory
Guillermo Campitelli1, David Z. Hambrick2, and Alessandro Guida3
2Michigan State University
3Université Rennes 2
Address correspondence to:
90 South Street, Murdoch
Western Australia 6150,
The study of memory is central to scientific understanding of expertise. Memory processes
underpin skilled performance in complex tasks, whether choosing a move in a chess game,
playing a musical instrument, or diagnosing a medical patient. Moreover, acquiring expertise
changes memory structures. We review major perspectives on expertise and use the umbrella
term knowledge structure to refer to all the types of memory structures proposed in theories
of memory based on expertise research, including chunks, templates, retrieval structures, and
semantic knowledge. Those theories postulate that knowledge structures reside in the long-
term memory store and accept the traditional dual models of the macro structure of memory
(i.e., models that postulate the existence of a short-term (or working) memory store and a
long-term memory store). In this chapter, we also present a recent proposal in the field of
expertise research that suggests knowledge structures constitute the macrostructure of
memory, and we review brain imaging studies investigating the interrelation between
memory and expertise. Finally, the effect of individual differences in traditional measures of
working memory capacity on expertise is discussed. We conclude with thoughts on
productive directions for future research.
There is no denying that some people are vastly more skilled than other people in
certain tasks. For example, only about 0.3% of all rated chess players are grandmasters. Not
surprisingly, there are massive individual differences in memory tasks, as well. Consider that
in the 2015 World Memory Championship the memory “athlete” Lance Tschirhart
remembered a bewildering 456 spoken digits. This is approximately 65 times more than the
average for the general population for this type of short-term memory task.
What explains this striking variability in performance? Research has left little doubt
that one factor that contributes substantially to individual differences in skilled performance
is the number of hours of intense practice in the corresponding field of expertise. For
example, in a reanalysis of previous studies, Hambrick et al. (2014) found that the
accumulated amount of practice explained roughly one-third of the individual differences in
chess expertise and music expertise. In short, expert performers have typically accumulated a
great deal more practice than novice performers.
Chase and Simon (1973; see also Simon & Chase, 1973) claimed that it takes around
a decade to reach an expert level of skill in chess. Ericsson, Krampe, and Tesch-Römer
(1993) tested this hypothesis in studies of pianists and violinists. In that pivotal article,
Ericsson et al. described the concept of deliberate practice, and in doing so, proposed a
number of criteria to determine whether practice is deliberate or not. They defined deliberate
practice as training in which the goal is to improve one’s current level of performance in a
domain. They further stipulated that deliberate practice is not inherently enjoyable and highly
effortful, and that the amount of deliberate practice that a person can engage in is limited to a
few hours a day (4 hours on average) without risking burn-out. Chase and Simon (1973)
argued that practice leads to the formation and storage of domain-specific knowledge in long-
term memory, and that differences in domain-specific knowledge mostly mediate the effect
of practice on performance (Figure 1).
Figure 1. The factor that mediates the causal relationship between practice and performance
is the accumulated domain-specific knowledge.
One of the differences between these views is that Chase and Simon (1973) indicated
that practice is the most important of a number of variables that cause individual differences
in performance, whereas Ericsson et al. (1993) went further, stating:
We agree that expert performance is qualitatively different from normal performance
and even that expert performers have characteristics and abilities that are qualitatively
different from or at least outside the range of those of normal adults. However, we
deny that these differences are immutable, that is, due to innate talent. Only a few
exceptions, most notably height, are genetically prescribed. Instead, we argue that the
differences between expert performers and normal adults reflect a life-long period of
deliberate effort to improve performance in a specific domain. (p. 400)
Thus, Ericsson et al.’s framework rejected any direct causal role of innate abilities,
with the obvious exception of height in sports such as basketball. They argued that effects of
motivation on the acquisition of expert performance are mediated through practice; that is,
motivation influences the acquisition of expert performance through its influence on
deliberate practice. Moreover, early exposure to a field was considered as a possible factor
affecting performance, but only insofar as starting early leads to more deliberate practice; that
is, the effect is fully mediated by deliberate practice. The role of deliberate practice in
explaining individual differences in expertise is still vigorously debated. In recent years,
some researchers have argued that although deliberate practice may be an important factor,
other factors make an important contribution as well (i.e., age of starting practice, general
cognitive abilities [e.g., Campitelli & Gobet, 2011; Hambrick et al., 2014]). By contrast,
other researchers question the validity of these new data and claims (e.g., Ericsson, 2014).
This chapter focuses on core questions about knowledge building and its relation to
performance: What are the characteristics of domain-specific knowledge? Is domain-specific
knowledge the only factor affecting expert performance, providing a full and parsimonious
account of individual differences in skilled performance? In the course of providing (partial)
answers to these questions, we argue that progress is made not only in understanding
expertise, but also in specifying the structure of memory in the general population.
2. Memory of experts
Nearly fifty years ago, Simon and Chase (1973) developed the chunking theory of
expertise, in which they proposed that experts spend more than a decade of exposure to their
fields of expertise, which leads to the formation and storage of domain-specific
configurations in long-term memory. They termed those structures chunks and, using chess as
a model of expertise, they estimated that experts possess 50,000 to 100,000 domain-specific
chunks in long-term memory. They defined chunks as visuo-spatial configurations that
appear frequently in a field of expertise. In the case of chess, chunks referred to
configurations of three or four pieces that appear in a specific location o n the chess board.
Moreover, these configurations are linked to typical actions (i.e., chess moves).
Research into cognitive processes involved in chess expertise typically uses chess
positions as stimuli. A chess position is constituted by a chess board and chess pieces
distributed throughout the board. That distribution of chess pieces arises as a consequence of
the moves played in a game until a specific point in time. Researchers present chess positions
to chess players of different degrees of expertise, and they request the players to perform
some sort of task. The most popular tasks have been the reconstruction task and the choose-a-
move task (DeGroot, 1946/1978 popularized these tasks in his seminal study). In the former,
participants are shown a chess position for a few seconds, and then they are presented with an
empty chess board and requested to reconstruct the previously seen position. In the latter,
participants are asked to decide what move to make.
Individual differences in level of expertise are mainly explained in terms of the
numbers of chunks stored in long-term memory. Performance is mainly explained by pattern-
recognition: in the case of chess, when a familiar pattern appears in the visual field, the
corresponding chunk is activated in long-term memory and a typical move associated to that
chunk is retrieved. Gobet and Simon (1998) proposed that experts also use pattern
recognition as part of their thinking process when deciding which action to execute. When
playing a chess game, players perform pattern recognition over the current chess position of
the game; this may lead to more than one potential move. In order to decide which move
to play, chess players may simulate that one of the potential moves is played, updating
their representation of the chessboard in their mind’s eye, accordingly. Now pattern
recognition occurs over this updated representation of the chess game, leading to one or more
potential responses of the opponent. One of the responses is chosen, and a simulation of that
response leads to a new change in the representation of the game. And this is followed by
pattern recognition on that updated representation. This recursive pattern recognition
continues until the player gathers sufficient information to decide which move to play.
Importantly, Chase and Simon’s explanation of the role of memory on expert
performance does not argue that expertise affects the fundamentals of memory’s
macrostructure –– that is, the limited capacity short-term (or working) memory store and the
large-capacity long-term memory store are not affected by expertise. The only change
produced by expertise is the content stored in long-term memory. The typical evidence shown
to support their view comes from the reconstruction task. The results show that the number of
pieces correctly reconstructed is a function of expertise (Chase & Simon, 1973; see also
DeGroot, 1946/1978 for the first time this task was reported). However, when instead of
presenting typical chess game positions, an artificial chess position is generated by randomly
allocating pieces on the chess board, the difference in accuracy among levels of expertise
dramatically decreases, with experts and non-experts both typically being able to
reconstruct the locations of about 7 chess pieces (chess experts still outperform novices
slightly). This suggests that experts, on average, do not have a higher working memory
capacity. When familiar configurations of a few chess pieces appear on the board, experts
activate chunks in long-term memory, and pointers (i.e., a conceptual or verbal label) to these
chunks are maintained in working memory. Given that stronger players (i.e., those with
greater expertise) can recognize more configurations, their performance is better. But in the
memory task with random chess positions, the typical configurations almost disappear, and
with them the accuracy differences go as well.
Gobet and Simon (1996) updated the chunking theory to deal with anomalous
findings, presenting their template theory of expertise. In their and previous studies a
modification to the DeGroot experimental paradigm was introduced: rather than viewing a
single board, participants were presented with a sequence of chess boards, each containing a
different chess position. While participants’ accuracy in reconstructing the chess positions
across multiple boards, measured in percentage of pieces correctly reconstructed, was lower
than that with one chess board, importantly, the total number of pieces correctly
reconstructed was higher than that with one chess board. This result could not be accounted
for by chunking theory because it postulated that each chunk has a maximum of 4 pieces and
that working memory can maintain only 7 chunks. This leads to a maximum capacity of 4 ×
7 = 28 chess pieces, which is far less than the up to 80 pieces that experts were able to
reconstruct correctly when presented with multiple chess boards. This dramatic increase in
the number of items remembered in a task that is typically considered to measure working
memory capacity would cast doubt on traditional models that posit that working memory
capacity is fixed and limited. However, rather than arguing the working memory capacity
limits are flexible, Gobet and Simon maintained the traditional model by adding two
important theoretical innovations to chunking theory –– one explaining the anomalous results,
and the other increasing the explanatory power of the chunking theory.
First, in their template theory experts are able to evolve chunks into larger chunks
called templates, which, not unlike the smaller chunks, are stored in long-term memory. This
is not a new proposal; the seminal paper on working memory positing the magical number 7
by Miller (1956) already argued that people are able to increase performance in memory tasks
by using their previous knowledge to recode presented information into meaningful chunks.
For example, typical configurations of letters form familiar words; typical configuration of
words form familiar phrases and sentences. Thus, laypersons increase the number of letters
remembered when those letters are presented as part of words. The second assertion is that
templates are not closed structures (like chunks) only linked to typical moves. Instead,
templates are slotted structures or schemas and the slots can be filled in with smaller
configurations of pieces (i.e., chunks) or conceptual knowledge (e.g., typical names given to
positions, general strategies, names of players who typically arrive at those positions). The
idea of slotted schemas also is not new; it can be tracked to Bartlett’s schema theory of
memory (1932) and Piaget’s (1936) cognitive schema theory of cognitive development (see
also Chapter 6.6).
Figure 2 shows an example of a template. The pieces on the board indicate the core ––
a set of pieces that are expected to be in specific locations in a typical master-level chess
game –– while the circles indicate the squares where chess pieces can be added without
changing the identity of the template (i.e., the slots).
Figure 2. Example of a chess template, as proposed by Gobet and Simon (1996). The
template contains a core and slots (black circles), which are empty squares that are typically
the location for pieces that are not in the core.
Gobet and Simon’s theoretical innovation, albeit built upon previous theories, was
profound. It put together two theoretical traditions: the computational tradition interested in
the macro structure of memory, represented by the (short and long) storage models of
memory (e.g., Atkinson & Shiffrin, 1968), and the Piagetian content-rich theory of cognitive
development. On the one hand, Gobet and Simon maintained the prevalent structural model
of short and long-term memory. On the other hand, instead of using the abstract concept of
information being transferred from one store to another, they used the content-rich concept of
template. The mechanism by which chunks are formed, network of chunks are developed and
chunks evolve into templates was implemented in a computational program named CHREST
[Chunk Hierarchy and Retrieval STructures] by Gobet and colleagues (see, Gobet et al., 2001
for more details). This program is fed with thousands of chess boards containing positions
belonging to master-level games, and it develops networks of chunks and templates. Models
with more or less exposure (e.g., exposed to 1,000 chess positions vs. exposed to 10,000
chess positions) are then tested in the memory task presented above and their performance in
those tasks is compared to human performance (see Gobet & Simon, 2000 for more details on
3. Memory expertise
In the previous section we referred to characteristics of memory that help explain
expert performance in fields in which memory processes are a subset of the many cognitive
processes involved, with others including problem solving, decision making and action.
There is an entire field of expertise in which the specific expertise is memorizing. That is,
experts in this field excel at learning and remembering vast amounts of information acquired
in a short period of time. Memory experts used to demonstrate their skills in exhibitions, but
since 1991 they have an annual international competition: The World Memory
Championships. This event involves ten memory activities –– e.g., memorizing numbers,
binary digits, historic dates, names with faces, and abstract images; learning and
remembering a shuffled 52-card deck as fast as possible is the last activity of the competition.
Remarkable achievements in this competition include memorizing the 52-card deck in 13.96
seconds, remembering 1924 cards presented in a one-hour period, remembering 302 random
words presented for 15 minutes, remembering 456 digits with a one second presentation of
each digit, and remembering 132 dates presented for 5 minutes.
A decade before the World Memory Championships were launched, William Chase
and Anders Ericsson set out to capture expert memory performance in the lab. Chase and
Ericsson (1981, 1982; Ericsson, Chase, & Faloon, 1980) presented university students (Steve
Faloon [SF] and Dario Donatelli [DD]) with a series of digits at a speed of one per second
and requested them to remember the series in the exact order. SF and DD increased their
performance from a standard score of approximately 7 digits to 82 digits following 264 hours
of training (SF) and to 106 digits following 800 hours of training (DD). Based on these
studies, Chase and Ericsson proposed skilled memory theory and subsequently expanded it
into long-term working memory theory.
In skilled memory theory, Chase and Ericsson (1981, 1982) presented three principles
to explain the acquisition of exceptional memory performance in individuals with no apparent
natural memory advantage: meaningful coding, building retrieval structures, and extensive
practice. Meaningful coding refers to the use of previous semantic knowledge to transform
incoming information (e.g., digits) into meaningful pieces of information (i.e., chunking).
Retrieval structures are memory structures that are built to improve performance in a specific
task. For example, in the digit task a retrieval structure may consist of a hierarchical tree of
nodes as shown in Figure 3.
Figure 3. Example of retrieval structure. The nodes in the bottom row are filled in
with the digits that are presented to the participants. The nodes in the middle row may consist
of labels that help maintain the digits together, such as “first”, “middle” and “last”, and the
node in the top row may be used for yet another label such as “section 1”. Memory experts
are able to construct multiple-level retrieval structures.
DD and SF combined the generation of retrieval structures, which as shown in Figure
3 only provide a structure to organize information, with semantic knowledge. Because they
were runners, they used their knowledge of running times to associate meaningful labels to
the retrieval structure and the digits presented to them. For example, the sequence of digits
“9”, “9”, “5” can be represented by the label “Current 100 m Male World Record”, which
was 9.95 seconds in the 1980s. As previously argued by Chase and Simon, semantic
knowledge enables the chunking of information, as separate elements (“9”, “9”, “5”) are
processed as one element “9.95”. The principle of extensive practice was further elaborated
into the deliberate practice framework, which was presented above.
Whereas skilled memory theory aimed at explaining performance on memory tasks,
Ericsson and Kintsch (1995) expanded the ideas in developing the long-term working
memory (LTWM) theory. The novelty of LTWM theory is that it not only aims to explain
performance in memory tasks, but also in all areas of expertise, including expertise in text
comprehension, medicine, chess, mental calculation, and waiters and waitresses’ memory for
dining orders. In other words, they developed a general theory of memory inspired by studies
on memory in experts and experts in memory. LTWM theory has the same components as
skilled memory theory, but it expands the types of retrieval structures, which differ across
fields of expertise. For example, the canonical retrieval structure for chess players is not a
tree-like structure but an empty chess board in which experts temporarily store chess pieces.
For text comprehension, Ericsson and Kintsch build upon the work of van Dijk and Kintsch
(1983) on situational models and propose those models as part of the canonical retrieval
structures for text comprehension. A situational model is a structure that combines aspects of
retrieval structures (temporal cues and cues about the topic of the text) with semantic
knowledge (e.g., knowledge of words’ meanings). The canonical retrieval structure for
waiters and waitresses would be a representation of the locations of tables in a restaurant.
This representation facilitates the recollection of orders because the food items have a spatial
localization in the retrieval structure.
Another theory developed to explain the performance of memory experts is the
cognitive architecture EPAM IV (Elementary Perceiver And Memorizer; Richman,
Staszewski, & Simon, 1995). A cognitive architecture is a computational implementation of
the mind, including memory structures and learning processes. EPAM IV comprises a short-
term memory, which can hold a limited number of chunks, and a semantic long-term
memory, which is accessed from a discrimination net. The discrimination net is a
computational implementation of the content of long-term memory. It is a tree-like structure
which contains nodes and connections among them. Some nodes are called terminal nodes,
which constitute representations of perceptual information (e.g., in the case of chess, typical
configuration of chess pieces); the non-terminal nodes contain components of the terminal
nodes (e.g., individual chess pieces). Both the discrimination net and the semantic long-term
memory increase in size by being exposed to stimuli in the environment and the internal
functioning of two types of learning processes in long-term memory: discrimination learning,
which associates individual items to create larger items or chunks; and association learning,
which associates representations of perceptual information to representations of actions (e.g.,
4. Knowledge structures
Campitelli (2015) outlined a conceptualization of memory inspired by the conception
of memory as behavior (Delaney & Austin, 1998), by models of memory that exclude the
short-term (working) memory store (e.g., Conway et al., 2005; Cowan, 1999 ; Fuster,
1997; McClelland et al., 2010; Nairne, 1992; Neath, 1998; Oberauer, 2002), and the concept
of knowledge structures that comes from the field of expertise research (e.g., Ericsson &
Kintsch, 1995; Gobet & Simon, 1996). Discussing Campitelli’s (2015) approach is beyond
the scope of this chapter ; this section focuses only on the third component of that
approach: the concept of knowledge structures introduced in that article and expanded by
Guida and Campitelli (2019). Guida and Campitelli defined knowledge structures as
meaningful blocks of knowledge that one builds while acquiring expertise. And, given their
conceptualization that we all have a degree of expertise in different fields, including everyday
life, they apply the concept of knowledge structure to explain general memory, not just expert
memory. Note the modification of the term retrieval structure to the term knowledge
structure. Campitelli (2015) provided two reasons for this change: first, knowledge structure
is a more generic term that includes not only retrieval structures but also chunks, templates
and semantic knowledge; second, these structures are not only useful to retrieve information,
but they are also used by the cognitive system to engage in many types of cognitive processes
(e.g., thinking, imagining).
Guida and Campitelli classified knowledge structures into two categories: non-slotted
schemas (e.g., semantic knowledge, chunks) and slotted schemas (e.g., retrieval structures,
templates). As indicated earlier, the term schema has a long history in psychology, it was
used by Piaget (1936) in his theory of cognitive development and by Bartlett (1932) in his
theory of memory. Definitions of this concept vary, but here we use a general
conceptualization: schemas are cognitive structures (either innate or acquired and modified
by experience) by which humans organize what they perceive and think. Slots refer to
parts of the knowledge structure to which additional information can be added very fast, and
temporarily. Schemas with slots allow temporary information to be added to them, as shown
in retrieval structures and templates in previous sections.
Chunks –– a typical configuration of items –– and semantic knowledge –– pieces of
information typically in language format –– are non-slotted schemas. Letters, words,
sentences, numbers, typical sequences of numbers, shapes, scenes, etc. are examples of non-
slotted schemas in everyday life. Typical domain-specific configurations (such as groups of
pieces in chess, musical notes in music, lines of code in programmers) are non-slotted
schemas in specific areas of expertise. Slotted schemas contain two parts: the core of the
schema and the slots (the non-slotted schemas only possess a core). Retrieval structures and
templates are examples of slotted schemas. The difference between retrieval structures and
templates resides in their core. The core in retrieval structures does not possess content, it is
only a spatial arrangement to which additional information can be added. The tree-like
structure in Figure 3 is a retrieval structure and so is the chain structure shown in Figure 4.
Contrarily, templates contain a content-rich core. An example of a template in everyday life
is a mental model of one’s home, with the more permanent components of the mental model
(e.g., the distribution of rooms and furniture) being the core, and parts of the core in which
less permanent components can be added being the slots (e.g., my car keys can be in my
bedroom now, in the living room later, and in the kitchen at another time).
Campitelli (2015) and Guida and Campitelli (2019) proposed that remembering things
presented a few seconds ago, rather than requiring a generic working memory, is
accomplished by the activation of knowledge structures. Let’s assume an expert chess player
is presented with stimuli containing chess boards with chess positions, and then asked to
immediately reconstruct the position (i.e., the reconstruction task explained earlier). She can
perform this task, which is typically considered a working memory task, with her knowledge
structures. That is, domain-specific chess templates and chunks are activated, allowing her to
perform the reconstruction task with high accuracy. A non-chess player, who has not
developed chess-related knowledge structures, cannot perform the task using the same
cognitive process. Instead, he can only activate a generic knowledge structure, like the chain
presented in Figure 4, to associate chess pieces presented in the reconstruction task. That
generic retrieval structure fulfils a similar role as traditional short-term memory models (e.g.,
Luck & Vogel, 1997).
Figure 4. A very simple knowledge structure: a generic retrieval structure, which is a
slotted schema used to associate it with generic stimuli; it is “filled in” from left to right.
5. A retrospective test of the alleged role of knowledge structures on general
Guida and Campitelli (2019) provided a retrospective test of the importance of
knowledge structures on memory, using research that uncovered the SNARC (Spatial
Numerical Association Response Codes, Dehaene et al., 1993) and SPoARC (Spatial
Positional Association Response Codes, van Dijk & Fias, 2011) effects in attention and
memory research. The SNARC and SPoARC are both effects where items in memory seem
to be organized along a vertical or horizontal line. Following previous work (Guida, Carnet,
Normandon, & Lavielle-Guida, 2016), Guida and Campitelli referred to this phenomenon as
“spatialization,” which consists of adding spatial organization to an input that did not
originally have such spatial structure.
Dehaene et al. (1993) investigated the SNARC effect in odd-even judgement
experiments. Participants had to indicate, by pressing a key, if a number presented in the
center of the screen was odd or even. In half of the trials, they made the choice using their left
hand for odd numbers and their right hand for even numbers; for the other half of the trials,
the reverse was true. The main result was that small numbers (e.g., 1, 2) were responded to
faster with the left hand and large numbers (e.g., 8, 9) were responded to faster with the right
hand. To explain the SNARC effect, Dehaene et al. (1993) proposed that it was due to the
internal spatial organization numbers have in long-term memory, which takes the form of a
left-to-right mental line.
The SNARC effect was replicated numerous times and in different circumstances. For
example, it is observed in right-handed and left-handed participants (Dehaene et al., 1993),
and in different tasks, formats, and materials: such as (a) the number comparison task
(participants indicate if a number on the screen is bigger or smaller that a reference number),
(b) with one or two-digit numbers and with words referring to numbers, and (c) with other
material that lends itself to spatialization, such as days (Gevers, Reynvoet, & Fias, 2004),
months, and letters of the alphabet (Gevers, Reynvoet, & Fias, 2003). Several articles have
provided reviews of the SNARC effect in different situations (Abrahamse, van Dijck, & Fias,
2016; Gevers & Lammertyn, 2005; Wood, Willmes, Nuerk, & Fischer, 2008). The direction
of reading of the participants affects the SNARC effect, with participants who read from right
to left, such as Iranians (Dehaene et al., 1993), Palestinians (Shaki, Fischer, & Petrusic,
2009), and Lebanese (Zebian, 2005), showing the opposite effect. Also a vertical SNARC
effect was found in Chinese (Hung, Hung, Tzeng, & Wu, 2008) and Japanese (Ito & Hatta,
2004) participants, who read top-down.
The SPoARC effect was discovered by van Dijk and Fias (2011) in a task that
required maintaining and processing information. Participants were presented with sequences
of five numbers (ranging from 1 to 10), which were displayed successively on the center of
the screen. Then, the participants had to make an odd-even judgment for those numbers.
After that, four sequences (one old and three new) were presented and the participants had to
choose the sequence they believed was the one encountered in the presentation phase. Instead
of showing the SNARC effect, the left-hand responses were faster for the numbers presented
at the beginning of the sequence, regardless of whether those numbers were smaller or larger.
The SPoARC effect has been found in tasks that only require maintenance of information
(e.g., Guida, Leroux, Lavielle-Guida, & Noël, 2016) and tasks requiring maintenance and
processing of information (van Dijck & Fias, 2011), with auditory stimuli (Guida et al., 2016)
and with visual stimuli (e.g., van Dijck & Fias, 2011; see Guida & Campitelli, 2019 for a
review of the SPoARC effect). Importantly, unlike SNARC, SPoARC occurs only with
sequences of up to five or six digits.
Knowledge structures account of SNARC and SPoARC
Guida and Campitelli (2019) proposed that the SNARC effect can be accounted for by
the activation of non-slotted schemas, and the SPoARC can be explained by the activation of
Non-slotted schemas explanation of the SNARC effect. As proposed by chunking
theory and template theory, chess players’ exposure to chess positions leads to the production
of chunks (non-slotted schemas) and templates (slotted schemas) containing typical
configurations of chess positions. The same applies to expert waiters (i.e., typical
configurations of restaurants), expert radiologists (i.e., typical configurations in X-rays),
sportspersons (typical configurations in their environments), and programmers (e.g., typical
commands). And this also applies to areas in which most people are experts, such as one’s
native language and basic knowledge of numbers, because we are all repeatedly exposed to
these stimuli and need to process them. As with having stored chunks of typical
configurations of letters forming words and chunks of typical configurations of words
forming sentences, we also possess chunks of typical sequences of numbers, such as “1-2-3-
4-5-6-7-8-9”. Importantly, this sequence occurs in our lives much more frequently than
others, such as “4-8-1-7-3-6-9-5-2”; thus, we possess a chunk of this frequent sequence, and
not of the latter. Thus, when performing a task that involves numbers, we activate this chunk,
which allows us to anticipate possible future stimuli (e.g., when one hears “1, 2”, one will
anticipate “3” because the chunk “1-2-3-4-5-6-7-8-9” gets activated and thus the number 3 is
the expected next number). The length of this canonical chunk varies as a function of
expertise with numbers, which at young ages will tend to be correlated with age. Thus, for a
young adult, but not yet for a 4-year old, the 1 to 9 sequence would be available as one chunk
(instead of 9 separate elements, also see Abrahamse et al., 2016), which is a non-slotted
schema. Other everyday chunks include week days (Monday-Tuesday-Wednesday-…-
Sunday), month sequence (January-February-March-April-…-December), and the alphabet
(A-B-C-D-…-Z). When participants carry out an even-odd judgement task they automatically
activate the corresponding non-slotted schema, and that activation influences the participant’s
behavior in the task. The number 9 presented in the center of the screen will be associated
with the right side of the canonical chunk of the 1 to 9 number sequence, and that spatial
priming will prompt a quicker action with the right hand than with the left. Equivalently, the
presentation of number 2 will prompt a quicker action with the left hand.
Slotted schema explanation of the SPoARC effect. The SPoARC effect cannot
possibly occur due to the activation of the canonical number sequence because it is related to
the specific order of presented numbers in a novel sequence, not the order of presented
numbers in a well-learned sequence. Guida and Campitelli (2019) proposed that the effect
could be caused by the activation of well-learned slotted schemas. As explained earlier,
templates are domain-specific slotted schemas with a content-rich core and slots. Another
type of slotted schemas is retrieval structures, which contain a content-less core plus slots.
The simplest version of this structure is presented in Figure 4, and it can be used to remember
any type of material presented. For example, if we are presented with the sequence 8-4-9-2
(remember that the presentation is sequential and that all the numbers appear at the same
spatial location, the center of the screen) in a working memory task, the slotted schema is
activated to maintain the following piece of information: 8 → 4 → 9 → 2
The slotted schema allows the cognitive system to maintain this piece of information, and this
will affect the behavior of the participant in the same way that the well learned canonical
number line does in the SNARC effect. Numbers temporarily associated with the left side of
the slotted schema (8 and 4 in this case) will prompt quicker actions with the left hand, and
those associated with the right side of the slotted schema (9 and 2) will target a quicker action
with the right hand. The direction of the arrows in the simplest slotted schema is acquired
through our expertise in reading; in other words, the mapping of sequential positions to
spatial locations is influenced by expertise in reading. Arabic readers would temporarily
maintain the digits in the opposite order than Roman language readers because they are used
to reading from right to left. Guida, Megreya, et al. (2018) tested Arabic literates and showed
that their SPoARC is reversed compared to Westerners. In our example, slotted schema with
the digits would look like this: 2 ← 9 ← 4 ← 8
The account put forward by Guida and Campitelli (2019) includes a hierarchy of
priorities for the cognitive system conducting a memory task. First, if the relevant stimuli for
the task contains spatial information (e.g., if the digits of the sequence 8-4-9-2 are presented
from right to left [i.e., 8 is presented first on the right-most part of the screen, 4 is presented
second in the middle-right part of the screen, 9 is presented third in the middle-left of the
screen, and 2 is presented fourth on the left-most part of the screen], participants will not
recode spatially the memoranda in a left-to-right way). Second, if the stimuli do not have
spatial information (i.e., all the items are presented in the same location), but the sequence
coincides with a well-established sequence already stored in long-term memory (e.g., the
sequence 1-2-3-4), the well-established sequence (non-slotted schema) will affect behavior
the most. Finally, if none of the above are a possibility, people will use their generic slotted
schema and the temporal order of the sequence will be encoded spatially, and this spatial
information will influence behavior.
This knowledge structures account of the SNARC and SPoARC effects is an expertise
account. For example, Guida, Megreya, et al. (2018) showed that SPoARC does not occur in
illiterates. This is because illiterates do not possess expertise in reading and writing in the
direction dictated by their native languages. On the other hand, it seems that SNARC is less
affected by expertise in reading/writing, as similar effects can be detected in 4-year-old
children (McCrink, Shaki, & Berkowitz, 2014; Opfer, Thompson, & Furlong, 2010) before
formal reading/writing acquisition.
6. Knowledge structures in the brain
Since the take-off of neuroimaging in the 1990s, researchers have used various neuro
imaging techniques (Guida, Noël, & Jonin, 2017) to examine the brain structures that are
crucial for expertise. One of the first studies (Elbert, Pantev, Wienbruch, Rockstroh, & Taub,
1995), using magnetoencephalography, showed that musician string players (when compared
with novices) had a larger cortical finger representation of the left hand in the postcentral
gyrus, which correlated with musical expertise as measured by starting age.
Subsequently, more theory-driven studies have emerged which enabled to pinpoint the
cerebral regions that were important for cognitive structures, such as chunks (non-slotted
schemas) or retrieval structures (slotted schemas). These studies can be separated in two
categories: longitudinal studies, where one measures the same individuals while they are
gaining expertise, and cross-sectional studies, where experts are compared to novices. Guida,
Gobet, Tardieu and Nicolas (2012, see also Guida, Gobet, & Nicolas, 2013), reviewing these
studies, extracted a pattern that they used to build a theoretical framework. They observed
that longitudinal studies typically reveal a decrease of brain activation in frontal and parietal
regions from before practice to after practice. This decrease could be detected even with
short-length training. For example, Landau, Schumacher, Garavan, Druzgal, and D’Esposito
(2004) used event-related functional magnetic resonance imaging (fMRI) and a delayed-
match-to-sample task, in which a stimulus set composed of four intact or scrambled faces was
shown to participants, who were asked to remember all the intact faces such that they can
make a match-to-sample decision after an 8-s delay period. During the scanning session, two
periods –– one early in task performance and the other following an hour of experience ––
were compared to contrast “before” and “after” the training. The results showed a decrease in
activation in frontal and parietal areas following an hour of training (for a more in-depth
analysis, see Guida et al., 2012). Similar results were observed in several other studies (e.g.,
Garavan, Kelley, Rosen, Rao, & Stein, 2000; Landau, Garavan, Schumacher, & D’Esposito,
2007; Sayala, Sala, & Courtney, 2006). Landau et al. (2004) explained the decrease of
activations in frontal and parietal areas in terms of chunking. They proposed that participants
used chunking as an encoding strategy. Since greater knowledge in a domain allows more
chunking of information in that domain (Gobet et al., 2001; Miller, 1956), they proposed that
their participants could be considered experts concerning faces, which enabled them to use
their face expertise to organize individual facial features into chunks more efficiently as the
task became familiar with practice.
In their review, Guida et al. (2012) concurred with this explanation to account for the
general pattern of brain decrease in longitudinal studies and contrasted it with the pattern
found in cross-sectional studies, that is, functional brain reorganization. In terms of univariate
activity1, when one compares two moments on a continuum of expertise, two changes in the
1 The univariate activity is the magnitude of neural activity averaged across voxels in a region, whereas a
multivariate pattern constitutes a spatial pattern of neural activity among voxels within a region (Formisano &
Kriegeskorte, 2012; Yang et al., 2016)
brain can occur when considering one location of the brain: increase or decrease. However,
when one takes into account multiple locations in the brain, then one must add
reorganization, which can be defined as a combined pattern of increases and decreases across
brain areas (Kelly & Garavan, 2005). This is what seems to happen in cross-sectional studies
(Guida et al., 2012, 2013). In one of the first studies of this kind, Pesenti et al. (2001) used
positron emission tomography (PET) to compare a calculating prodigy with laypeople, asking
them to carry out both simple and complex calculations. The result of a simple calculation is
posited to be retrieved directly from memory (e.g., √45 for the expert and 5x7 for the
novices), whereas complex calculation is posited to necessitate a real calculation (e.g., 48x86
for the expert and 15x12 for the novices). When brain areas activated during complex
calculation were compared to those activated in simple calculation, Pesenti et al.’s results
showed that the non-expert and the expert groups used some of the same brain areas, such as
frontal and parietal areas (for a more in-depth analysis, see Guida et al., 2012). By contrast,
the authors argued that the areas activated only by the experts, such as the medial temporal
lobe (a crucial area for episodic long-term memory), constitute a long-term working memory
(Ericsson & Kintsch, 1995), which operates through the activation of retrieval structures
(slotted schemas). This pattern is compatible with the definition of functional brain
reorganization of the brain given by Kelly and Garavan (2015). Pesenti et al. (2001) argued
that experts were not individuals for whom the processes of calculation had simply been
accelerated and improved, but individuals who used completely different brain areas which
corresponded to different cognitive structures (slotted schemas). A similar pattern of results
was observed in several other studies (e.g., Campitelli, Gobet, Head, Buckley, & Parker,
2007; Maguire, Valentine, Wilding, & Kapur, 2003; Saariluoma, Karlsson, Lyytinen, Teräs,
& Geisler, 2004).
Overall, Guida et al. (2012) argued that 1) longitudinal studies using novices and a
training program in working memory-related tasks were mainly compatible with a decrease
of cerebral activity in frontal and parietal areas, whereas 2) cross-sectional studies using
trained experts in working memory-related tasks tended to show results compatible with a
cerebral functional reorganization. Guida et al. (2012) combined these two patterns, which
did not seem compatible at first sight and were from separate domains, to describe the
evolution of expertise acquisition in a two stage-framework. The first stage—decrease of
cerebral activity—is linked to non-slotted schemas: chunks (Chase & Simon, 1973; Cowan,
2001; Gobet et al., 2001). When practice commences, individuals start binding various
elements together, which ultimately will result in a compression (Mathy & Feldman, 2012) of
the elements into one structure, a chunk. Once chunks are built, separate constituents can be
processed as one element, which means that less cognitive resources are needed. This leads to
the decrease of activation in frontal and parietal areas, which subserve a number of cognitive
processes that are typically engaged during working memory tasks. With extended practice
and expertise, chunks grow in size (e.g., Cowan, Chen, & Rouder, 2004; Chen & Cowan,
2005) and eventually become hierarchical chunks (slotted schemas), such as templates (Gobet
& Simon, 1996). These well-learned cognitive structures allow experts to store information in
a fast and reliable fashion even in tasks with fast presentation times of multiple elements and
performance examined after a brief delay, which is not possible for novices using similar cell
assemblies from the medial temporal lobe. While novices will use these areas only during
long-term memory tasks2, experts will be able to use these areas also during the working
2 The sentence “novices will use episodic long-term memory areas only during long-term memory tasks” could
be proven wrong with studies showing that the medial temporal lobe is also engaged in STM tasks (e.g.,
Nichols, Kao, Verfaellie, & Gabrieli, 2006; Ranganath & Blumenfeld, 2005; and Ranganath & D’Esposito,
2005). However, when looking at meta-analyses, this link fails to emerge (Emch, von Bastian, & Koch, 2019;
memory tasks. This is explained by the activation of well learned knowledge structures (in
this case, slotted schemas such as templates or retrieval structures). The possibility of using
slotted schemas to encode in a fast and reliable way via the medial temporal lobe would
explain the functional reorganization observed in experts, which is believed to be the
signature of the second stage of expertise acquisition (Guida et al., 2013).
More recently Guida, Campitelli and Gobet (2016) have linked the capacity of experts
to use different brain structures compared to novices to Anderson’s (2014) neural re-use
theory. Anderson (2014, p. 308) defines neural re-use as “the reuse of regions of the brain in
multiple-task contexts [which] results in the inheritance of function from one task to the
other”. As hinted above, experts via functional brain reorganization have developed the
capacity to re-use the medial temporal lobe differently compared to novices. Whereas novices
typically use the medial temporal lobe for performing long-term memory tasks, experts are
able to (re)use these areas also for performing working memory tasks.
The relation between working memory and episodic long-term memory also relates to
another interesting feature put forward by Anderson (2014): unmasking. The basic idea is that
regions are specialized through the dominance of one input. However, under some conditions
that allow the source of dominant input to be disrupted (e.g., injury, sensory deprivation),
new input can be processed, unmasking new processing capacities. Expertise through the
two-stage framework offers a good example of this. It is indeed possible that the decrease of
activity in the first stage may help the unmasking and thus the re-use of the medial temporal
lobe for working memory tasks. This would enable the occurrence of the second stage,
Kim, 2019; Owen, McMillan, Laird, & Bullmore, 2005; Rottschy, et al., 2012; Yaple, Stevens, & Arsalidou,
through the use of slotted schemas. The efficiency of these cognitive structures is a necessary
condition. The biological reasons that undergird such processes are unknown.
7. Effect of working memory on performance
In previous sections we discussed the effect of expert performance on memory
structure. In this last section, we invert the relationship and we focus on research on the effect
of individual differences in traditional measures of working memory capacity (WMC) on
complex task performance. WMC is the ability to maintain task-relevant information in a
highly active state (see also Chapters 2.9 and 2.10). One way of measuring it is with complex
span tasks, such as operation span (Engle, 2002), in which participants solve equations while
remembering words. Scores on such tasks correlate with success in complex cognitive tasks
such as reasoning and language comprehension, and with each other (Kane & Engle, 2003),
which suggests that WMC is highly general.
As we have already mentioned, domain-specific knowledge is an important
determinant of expertise. In fact, many tasks cannot be performed at all without some level of
domain-specific knowledge. A person must have learned the rules of chess to play a game of
chess, and how to position their fingers on a piano’s keyboard to play a song on the piano.
Not surprisingly, then, domain-specific knowledge is an important predictor of individual
differences in performance on complex tasks. To put it another way, a major factor in
explaining why some people perform more highly on complex tasks than others is simply
knowledge. One of us found that knowledge of esoteric terms, such as aril (a seed covering)
and etui (a needle case), frequently referenced in crossword puzzles (but seldom elsewhere)
accounted for a large amount of the variance in people’s success in solving New York Times
crossword puzzles (e.g., Hambrick, Salthouse, & Meinz, 1999).
That research has focused on characterizing the interplay between working memory
capacity and domain-specific knowledge in complex cognitive tasks. The first is a differential
approach, which asks how working memory and domain-specific knowledge interact with
each other in predicting performance in some task. Hambrick and Meinz (2011) have focused
specifically on testing what we have called the circumvention-of-limits hypothesis. This
hypothesis holds that as performers acquire specialized knowledge in a domain through
training, it becomes possible to circumvent or “bypass” general constraints on performance
such as the limited capacity of working memory. This hypothesis predicts an interaction
between working memory and domain knowledge, such that the effect of working memory
on performance is smaller at high levels of domain knowledge than at lower levels.
Alternatively, if domain knowledge is treated as a group variable (i.e., low knowledge vs.
high knowledge), then the prediction is a larger correlation between working memory and
performance in the low knowledge group than in a high knowledge group.
The circumvention-of-limits hypothesis is appealing because it implies that
performance limitations stemming from working memory can be overcome with training.
That said, there is little evidence to support this hypothesis. Hambrick and Engle (2002) had
participants representing a wide range of knowledge of the game of baseball listen to and
attempt to remember information from fictitious (but realistic sounding) radio broadcasts of
baseball games, including the sequence of events in each half-inning (i.e., which bases were
occupied after each at-bat), as well as game-relevant details (e.g., the batting averages of the
players) and non-game-relevant details (e.g., the size of the crowd). Predicting memory for
game sequences, effects of working memory and baseball knowledge were additive, while
predicting memory for game-relevant details, they were over-additive. Thus, there was no
evidence that a high level of domain-specific knowledge attenuated, much less eliminated,
the effect of working memory on performance.
More recently, Hambrick, Burgoyne, and Oswald (2018) carried out a review of
evidence relevant to the circumvention-of-limits hypothesis, conducting systematic searches
for relevant articles in the literature on expertise in six domains (games, music, science,
sports, surgery/medicine, and aviation). Altogether, Hambrick et al. searched approximately
1,300 documents. On balance, evidence from the expertise literature does not support the
circumvention-of-limits hypothesis: only three of fifteen studies provide support for this
hypothesis, either in the form of significantly different ability-performance correlations
across skill groups or significant ability × skill interactions on performance. In short, though
the circumvention-of-limits hypothesis is appealing, it does not receive compelling support
from the available evidence.
The study of memory in experts and experts in memory has provided very rich
information regarding not only acquisition of expertise but also the general structure of the
memory system in all humans. The first approach in expertise research integrated extant
traditional models of memory with new findings from the science of expertise. In this chapter
we presented a recent approach by which expertise is not a rare phenomenon that needs an
ad-hoc explanation within extant models; rather, it is a common phenomenon that needs to be
taken into account to develop a general theory of memory. This approach emphasizes the role
of knowledge structures not only as the carriers of content of memory but also as constituting
the macrostructure of memory. We also presented a retrospective test of this approach, in
which we use it to explain two effects in attention and memory research: the SNARC effect
and the SPoARC effect. Furthermore, we presented studies that aim at identifying the
localization of knowledge structures in the brain, and how this localization changes as a
function of expertise. The fact that expertise modifies the general structure of memory via
creating knowledge structures does not mean that individual differences in working memory
capacity disappear at high levels of expertise, as shown by the lack of evidence for the
circumvention-of-limits hypothesis. This finding suggests that two experts who possess the
appropriate knowledge structures for the task at hand might still differ on how efficiently they
Since its inception, expertise research aimed at providing general accounts of
cognitive functioning, especially about memory functioning: EPAM IV and long-term
working memory are theories of general memory, not just memory of experts or expertise in
memorizing. Those theories proposed the existence of knowledge structures (chunks,
templates, retrieval structures, semantic knowledge) as residing within the two stores (short-
term or working memory store and long-term store), which are prevalent in traditional models
of the structure of memory. In this chapter, we presented an approach by which the
knowledge structures are not contained within a macrostructure, rather they themselves
constitute the macro structure of memory. This conceptualization of memory macro
structure is compatible with theories of neural re-use and the concept of neural reorganization
which suggest that the macro structure of memory goes through processes of
reorganization, rather than being fixed. The retrospective test of this approach using the
SNARC and SPOARC effects is promising; however, research testing predictions of the
approach must be conducted for this approach to be considered a serious alternative to more
List of references
Abrahamse, E., van Dijck, J.-P., & Fias, W. (2016). How does working memory enable
number-induced spatial biases? Frontiers in Psychology, 7.
Anderson, M. L. (2014). After phrenology: Neural reuse and the interactive brain. MIT
Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its
control processes. In Psychology of learning and motivation (Vol. 2, pp. 89-195).
Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. New
York, NY, US: Cambridge University Press.
Campitelli, G. (2015). Memory behavior requires knowledge structures, not memory stores.
Frontiers in Psychology, 6.
Campitelli, G., & Gobet, F. (2011). Deliberate practice: Necessary but not sufficient. Current
Directions in Psychological Science, 20(5), 280-285.
Campitelli, G., Gobet, F., Head, K., Buckley, M., & Parker, A. (2007). Brain localization of
memory chunks in chessplayers. International Journal of Neuroscience, 117(12), 1641-
Chase, W. G., & Ericsson, K. A. (1981). Skilled memory. Cognitive Skills and Their
Chase, W. G., & Ericsson, K. A. (1982). Skill and working memory. Psychology of Learning
and Motivation, 16, 1–58.
Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4(1), 55–
Chen, Z., & Cowan, N. (2005). Chunk limits and length limits in immediate recall: a
reconciliation. Journal of Experimental Psychology: Learning, Memory, and Cognition,
Conway, A. R., Kane, M. J., Bunting, M. F., Hambrick, D. Z., Wilhelm, O., and Engle, R. W.
(2005). Working memory span tasks: a methodological review and user's
guide. Psychonomic Bulletin and Review. 12, 769–786.
Cowan, N. (1999). “An embedded-processes model of working memory,” in Models of
Working Memory: Mechanisms of Active Maintenance and Executive Control, eds A.
Miyake and P. Shah (Cambridge, UK: Cambridge University Press), 62–101.
Cowan, N. (2001). The magical number 4 in short-term memory: a reconsideration of mental
storage capacity. The Behavioral and Brain Sciences, 24(1), 87–114; discussion 114–
Cowan, N., Chen, Z., & Rouder, J. N. (2004). Constant capacity in an immediate serial-recall
task: A logical sequel to Miller (1956). Psychological science, 15(9), 634-640.
De Groot, A. D. (1946/1978). Thought and choice in chess. The Hague: Mouton. (2nd
English ed.; first Dutch edition published in 1946).
Dehaene, S., Bossini, S., & Giraux, P. (1993). The mental representation of parity and
number magnitude. Journal of Experimental Psychology: General, 122(3), 371.
Delaney, P. F., and Austin, J. (1998). Memory as behavior: the importance of acquisition and
remembering strategies. Analysis of Verbal Behavior, 15, 75–91.
Elbert, T., Pantev, C., Wienbruch, C., Rockstroh, B., & Taub, E. (1995). Increased cortical
representation of the fingers of the left hand in string players. Science, 270(5234), 305-
Emch, M., von Bastian, C. C., & Koch, K. (2019). Neural correlates of verbal working
memory: an fMRI meta-analysis. Frontiers in human neuroscience, 13, 180.
Engle, R. W. (2002). Working memory capacity as executive attention. Current Directions in
Psychological Science, 11(1), 19–23.
Ericsson, K. A. (2014). Why expert performance is special and cannot be extrapolated from
studies of performance in the general population: A response to criticisms. Intelligence,
Ericcson, K. A., Chase, W. G., & Faloon, S. (1980). Acquisition of a memory skill. Science,
Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in
the acquisition of expert performance. Psychological Review, 100(3), 363-406.
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review,
Formisano, E., & Kriegeskorte, N. (2012). Seeing patterns through the hemodynamic veil—
the future of pattern-information fMRI. Neuroimage, 62(2), 1249-1256.
Fuster, J. M. (1997). Network memory. Trends in Neurosciences. 20, 451–459. doi:
Garavan, H., Kelley, D., Rosen, A., Rao, S. M., & Stein, E. A. (2000). Practice‐related
functional activation changes in a working memory task. Microscopy research and
technique, 51(1), 54-63.
Gevers, W., & Lammertyn, J. (2005). The hunt for SNARC. Psychology Science, 47(1), 10–
Gevers, W., Reynvoet, B., & Fias, W. (2003). The mental representation of ordinal sequences
is spatially organized. Cognition, 87(3), B87–B95.
Gobet, F., Lane, P. C., Croker, S., Cheng, P. C., Jones, G., Oliver, I., & Pine, J. M. (2001).
Chunking mechanisms in human learning. Trends in Cognitive Sciences, 5(6), 236–243.
Gobet, F, & Simon, H. A. (1996). Templates in chess memory: a mechanism for recalling
several boards. Cognitive Psychology, 31(1), 1–40.
Gobet, F., & Simon, H. A. (1998). Pattern recognition makes search possible: Comments on
Holding (1992). Psychological Research, 61, 204-208.
Gobet, F., & Simon, H. A. (2000). Five seconds or sixty? Presentation time in expert
memory. Cognitive Science, 24(4), 651-682.
Guida, A., & Campitelli, G. (2019). Explaining the SPoARC and SNARC effects with
knowledge structures: An expertise account. Psychonomic Bulletin & Review, 26, 434-
Guida, A., Campitelli, G., & Gobet, F. (2016). Becoming an expert: Ontogeny of expertise as
an example of neural reuse. Behavioral and Brain Sciences, 39.
Guida, A., Carnet, S., Normandon, M., & Lavielle-Guida, M. (2018). Can spatialisation be
extended to episodic memory and open sets? Memory, 1–14.
Guida, A., Gobet, F., & Nicolas, S. (2013). Functional cerebral reorganization: a signature of
expertise? Reexamining Guida, Gobet, Tardieu, and Nicolas’(2012) two-stage
framework. Frontiers in Human Neuroscience, 7.
Guida, A., Gobet, F., Tardieu, H., & Nicolas, S. (2012). How chunks, long-term working
memory and templates offer a cognitive explanation for neuroimaging data on expertise
acquisition: a two-stage framework. Brain and Cognition, 79(3), 221–244.
Guida, A., Noël, A., & Jonin, P. Y. (2017). Primer: Neural Approaches in Research on
Expertise. In The Science of Expertise (pp. 171-181). Routledge.
Guida, A., Leroux, A., Lavielle-Guida, M., & Noël, Y. (2016). A SPoARC in the dark:
Spatialization in verbal immediate memory. Cognitive Science, 40, 2108–2121.
Guida, A., Megreya, A. M., Lavielle-Guida, M., Noël, Y., Mathy, F., van Dijck, J.-P., &
Abrahamse, E. (2018). Spatialization in working memory is related to literacy and
reading direction: Culture “literarily” directs our thoughts. Cognition, 175, 96–100.
Hambrick, D. Z., Oswald, F. L., Altmann, E. M., Meinz, E. J., Gobet, F., & Campitelli, G.
(2014). Deliberate practice: Is that all it takes to become an expert? Intelligence, 45, 34-
Hambrick, D. Z., Burgoyne, A. P., Oswald, F. L. (2018). Domain-general models of
expertise: The role of cognitive ability. In Ward, P., Schraagen, J. M., Gore, J., & Roth,
E., (Eds.), Oxford handbook of expertise: Research and application. Oxford: Oxford
Hambrick, D. Z., & Engle, R. W. (2002). Effects of domain knowledge, working memory
capacity, and age on cognitive performance: An investigation of the knowledge-is-
power hypothesis. Cognitive psychology, 44(4), 339-387.
Hambrick, D. Z., & Meinz, E. J. (2011). Limits on the predictive power of domain-specific
experience and knowledge in skilled performance. Current Directions in Psychological
Science, 20, 275-279.
Hambrick, D. Z., Salthouse, T. A., & Meinz, E. J. (1999). Predictors of crossword puzzle
proficiency and moderators of age–cognition relations. Journal of Experimental
Psychology: General, 128(2), 131-164.
Hung, Y., Hung, D. L., Tzeng, O. J.-L., & Wu, D. H. (2008). Flexible spatial mapping of
different notations of numbers in Chinese readers. Cognition, 106(3), 1441–1450.
Ito, Y., & Hatta, T. (2004). Spatial structure of quantitative representation of numbers:
Evidence from the SNARC effect. Memory & Cognition, 32(4), 662–673.
Kane, M. J., & Engle, R. W. (2003). Working-memory capacity and the control of attention:
the contributions of goal neglect, response competition, and task set to Stroop
interference. Journal of Experimental Psychology: General, 132(1), 47.
Kelly, A. M. C., & Garavan, H. (2005). Human functional neuroimaging of brain changes
associated with practice. Cerebral Cortex, 15, 1089–1102.
Landau, S. M., Garavan, H., Schumacher, E. H., & D’Esposito, M. (2007). Regional
specificity and practice: Dynamic changes in object and spatial working memory. Brain
Research, 1180, 78–89.
Landau, S. M., Schumacher, E. H., Garavan, H., Druzgal, T. J., & D’Esposito, M. (2004). A
functional MRI study of the influence of practice on component processes of working
memory. Neuroimage, 22, 211–221.
Luck, S. J., & Vogel, E. K. (1997). The capacity of visual working memory for features and
conjunctions. Nature, 390(6657), 279.
Maguire, E. A., Valentine, E. R., Wilding, J. M., & Kapur, N. (2003). Routes to
remembering: The brains behind superior memory. Nature Neuroscience, 6, 90–95.
Mathy, F., & Feldman, J. (2012). What’s magic about magic numbers? Chunking and data
compression in short-term memory. Cognition, 122(3), 346–362.
McCrink, K., Shaki, S., & Berkowitz, T. (2014). Culturally driven biases in preschoolers’
spatial search strategies for ordinal and non-ordinal dimensions. Cognitive
Development, 30, 1–14.
McClelland, J. L., Botvinick, M. M., Noelle, D. C., Plaut, D. C., Rogers, T. T., Seidenberg,
M. S., et al. (2010). Letting structure emerge: connectionist and dynamical systems
approaches to cognition. Trends Cogn. Sci. 14, 348–356. doi:
Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits of our
capacity for processing information. Psychological Review, 63, 81–97.
Nairne, J. S. (1992). The loss of positional certainty in long-term memory. Psychol. Sci. 3,
199–202. doi: 10.1111/j.1467-9280.1992.tb00027.x
Neath, I. (1998). Human Memory: An Introduction to Research, Data, and Theory. Belmont,
CA: Thomson Brooks/Cole Publishing Co.
Nichols, E. A., Kao, Y.-C., Verfaellie, M., & Gabrieli, J. D. E. (2006). Working memory and
long-term memory for faces: Evidence from fMRI and global amnesia for involvement
of the medial temporal lobes. Hippocampus, 16, 604–616.
Oberauer, K. (2002). Access to information in working memory: exploring the focus of
attention. J. Exp. Psychol. Learn. Mem. Cogn. 28, 411–421. doi: 10.1037/0278-
Opfer, J. E., Thompson, C. A., & Furlong, E. E. (2010). Early development of spatial-
numeric associations: evidence from spatial and quantitative performance of
preschoolers: Spatial-numeric associations. Developmental Science, 13(5), 761–771.
Owen, A. M., McMillan, K. M., Laird, A. R., & Bullmore, E. (2005). N‐back working
memory paradigm: A meta‐analysis of normative functional neuroimaging studies.
Human Brain Mapping, 25(1), 46-59.
Pesenti, M., Zago, L., Crivello, F., Mellet, E., Samson, D., Duroux, B., et al. (2001). Mental
calculation in a prodigy is sustained by right prefrontal and medial temporal areas.
Nature Neuroscience, 4, 103–107.
Piaget, J. (1936). Origins of intelligence in the child. London: Routledge & Kegan Paul.
Ranganath, C., & Blumenfeld, R. S. (2005). Doubts about double dissociations between
short- and long-term memory. Trends in Cognitive Science, 9, 374–380.
Ranganath, C., & D’Esposito, M. (2005). Directing the mind’s eye: Prefrontal, inferior and
medial temporal mechanisms for visual working memory. Current Opinion in
Neurobiology, 15, 175–182.
Richman, H. B., Staszewski, J. J., & Simon, H. A. (1995). Simulation of expert memory
using EPAM IV. Psychological Review, 102(2), 305–30.
Rottschy, C., Langner, R., Dogan, I., Reetz, K., Laird, A. R., Schulz, J. B., ... & Eickhoff, S.
B. (2012). Modelling neural correlates of working memory: a coordinate-based meta-
analysis. Neuroimage, 60(1), 830-846.
Saariluoma, P., Karlsson, H., Lyytinen, H., Teräs, M., & Geisler, F. (2004). Visuospatial
representations used by chess experts: A preliminary study. European Journal of
Cognitive Psychology, 16, 753–766.
Sayala, S., Sala, J. B., & Courtney, S. M. (2006). Increased neural efficiency with repeated
performance of a working memory task is information-type dependent. Cerebral
Cortex, 16, 609–617.
Shaki, S., Fischer, M. H., & Petrusic, W. M. (2009). Reading habits for both words and
numbers contribute to the SNARC effect. Psychonomic Bulletin & Review, 16(2), 328–
Simon, H. A., & Chase, W. G. (1973). Skill in chess. American Scientist, 61, 394-403.
van Dijck, J.-P., Abrahamse, E. L., Majerus, S., & Fias, W. (2013). Spatial attention interacts
with serial-order retrieval from verbal working memory. Psychological Science, 24(9),
van Dijck, J.-P., & Fias, W. (2011). A working memory account for spatial–numerical
associations. Cognition, 119(1), 114–119.
van Dijk, T. A., & Kintsch, W. (1983). Strategies of discourse comprehension. Academic
Press New York.
Wood, G., Willmes, K., Nuerk, H.-C., & Fischer, M. H. (2008). On the cognitive link
between space and number: A meta-analysis of the SNARC effect. Psychology Science
Quarterly, 50(4), 489.
Yang, Z., Zhen, Z., Huang, L., Kong, X. Z., Wang, X., Song, Y., & Liu, J. (2016). Neural
univariate activity and multivariate pattern in the posterior superior temporal sulcus
differentially encode facial expression and identity. Scientific reports, 6, 23427.
Yaple, Z. A., Stevens, W. D., & Arsalidou, M. (2019). Meta-analyses of the n-back working
memory task: fMRI evidence of age-related changes in prefrontal cortex involvement
across the adult lifespan. NeuroImage, 196, 16-31.
Zebian, S. (2005). Linkages between number concepts, spatial thinking, and directionality of
writing: The SNARC effect and the reverse SNARC effect in English and Arabic
monoliterates, biliterates, and illiterate Arabic speakers. Journal of Cognition and
Culture, 5(1–2), 1–2.