Advanced Review

Working memory
Timothy J. Ricker, Angela M. AuBuchon and Nelson Cowan∗

The working memory system maintains the limited information that can be kept in
mind at one time. These memories are distinct from the vast amount of information
stored in long-term memory. Here we give a brief summary of findings over the past
half-century in the areas of working memory that we see as particularly important
for understanding its nature. We discuss several current controversies, including
whether there are different systems or brain modules for different kinds of working
memory, why we lose items from working memory, and how individuals and age
groups differ. We try to describe what is and is not known. Last, a discussion of
findings from neuroimaging helps to constrain working memory theory.  2010 John
Wiley & Sons, Ltd. WIREs Cogn Sci 2010 1 573–585

D iscussions of working memory often suffer

from the wide variety of usages of the term.
Some have described working memory literally, as
you get in a conversation during the drive. To succeed
you need to write down the directions, and even
after doing so working memory is needed to execute
information maintained temporarily and used to the next turn in as much, as one cannot stare at
accomplish mental work.1 Others have used the the directions constantly while driving. As another
term more specifically in reference to a proposed example, while doing arithmetic mentally, one must
multi-component system that temporarily maintains hold partial results in mind while doing calculations.
information and manipulates it to carry out cognitive A little reflection on these two examples tells us that
tasks.2 In the present article, when we discuss working the number of turns, digits, or other items that can
memory, we are referring to both the information that be held in conscious memory is quite small. The items
can be kept available in mind at any given time and the held in this limited cognitive space are said to be in
processing that occurs to maintain this information, working memory.
but not other processing. Working memory is important because it
The amount of information that can be mediates most of our conscious interactions with
maintained for quick and easy access is clearly limited. the world. On a more practical level, working
This is apparent when you try to remember directions memory is critical for a wide variety of cognitive
that have too many turns. Imagine that you are functions. Because it holds information that is being
leaving work for the day and realize that you do processed in an available state, its size and functioning
not have directions to this evening’s celebration at
affect how we are able to think about and solve
an acquaintance’s house. Because you are on the way
problems. Individual differences in working memory
out and you do not have any paper to write down
are predictive of performance on complex cognitive
the directions when a coworker tells them to you.
tasks, such as reasoning,3 and clinical deficits in
Instead, you must remember them until you drive
working memory are related to conditions, such as
home and can write them down: left on Fifth Street,
attention deficit disorder4 and schizophrenia.5
right onto Elm, left onto Bircham, left onto Lupine,
It is agreed virtually by all cognitive psycholo-
right onto Ridge, third house on the right. When you
get home and write the directions down you would gists that the processes attributed to working memory
probably find that you have a hard time remembering are essential in human cognition. One must keep
a few parts, like the name of the third- or fourth- information in mind while processing it to function
mentioned street, or whether you turned right or left intellectually and socially. There are, however, still
onto elm. You may even forget the whole thing if fundamental issues and disagreements regarding just
how working memory is limited, and how it operates.
∗ Correspondence to: CowanN@missouri.edu We have chosen to present the research in this area
University of Missouri, 210 McAlester Hall, Columbia, with the controversies firmly in mind and therefore
Missouri, USA organize the discussion around these controversies.
DOI: 10.1002/wcs.50 The questions we consider include, whether working

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 573

 Advanced Review wires.wiley.com/cogsci

memory is a single system or is composed of several (a) Atkinson & Shiffrin (1968)
independent systems, why there is a limit to the num-
ber of items that can be held in memory, whether
or not the passage of time causes us to forget, why Sensory Working Long-term
the capacity of working memory differs among indi- memory memory memory
viduals, and why it changes with development. We
concentrate on behavioral research but conclude with
a glimmer of the recent brain imaging research and
how it relates to more traditional behavioral research. (b) Baddeley (1986, 2000)

Central
IS WORKING MEMORY ONE SYSTEM executive
OR SEVERAL SYSTEMS?
Although researchers at first conceived of working
memory as a single system or process,6,7 many have
Episodic
subsequently thought that the data compel us to Phonological
buffer
Visuo-spatial
loop sketch pad
propose that there are multiple systems or processes (2000 model)
that are involved in our ability to retain a small
amount of information in a highly accessible state.8–11
In this section we consider whether there are separate Long-term memory system
visual and verbal modules for working memory; a
general, central faculty; or both general and modality-
(c) Cowan (1988)
or code-specific aspects of working memory. To
preview our conclusion, we believe that there is a
Central executive processes
general working memory and that some aspects of
working memory also are code-specific, although these
code-specific properties do not necessarily constitute
separate systems. Here we briefly summarize the Long-term Activated
memory (LTM) LTM Focus of
history of this key issue and present evidence that attention
the central memory is limited in terms of the number
of chunks. Later, we will examine other kinds of
working memory limits for code-specific information,
FIGURE 1 | Models of working memory as conceptualized by three
in particular time-based forgetting.
researchers: (a) Atkinson and Shiffrin.7 (b) Baddeley.8,10 (c) Cowan.11
Miller6 famously reviewed evidence that the
amount that can be remembered and repeated is
limited to about seven items, give or take a few such strict capacity limits occur. To try to find out,
depending on the type of items that are to be we will need to investigate further the nature of
remembered and the particular person. In the early the working memory system or systems, as the case
days of cognitive psychology, this phenomenon was may be.
illustrated within models of human information In a seminal article that has guided the field for
processing by a single box that represented the many years, Baddeley and Hitch2 reported on many
limited number of items that could be recalled (e.g., experiments suggesting that there is less interference
Ref 7; Figure 1(a)). Miller also discussed a trick that between tasks than one would expect on the basis
individuals could use to increase their memory span: of the single-system working memory box model
items could be grouped together to form larger chunks of Atkinson and Shiffrin.7 For example, Baddeley
that act like individual items. For example, one can and Hitch found that remembering a series of digits
recall nine random letters easily if they form acronyms, interfered with verbal reasoning or comprehension,
such as IBM, CIA, FBI. Another trick that was but only if the number of digits was challenging (six
discussed later was silently repeating items to one’s digits, but not three). They also demonstrated that
self or rehearsing them.8 verbal distraction interfered with verbal memory more
Cowan9 reviewed research indicating that, when than did visual distraction, and that visual distraction
one prevents chunking and rehearsal, the number of interfered with visual memory more than did verbal
independent items that can be recalled is typically distraction. On the basis of this information, Baddeley
only about 3–5. It remains unclear, however, why and Hitch theorized that working memory has a

574  2010 Jo h n Wiley & So n s, L td. Vo lu me 1, Ju ly /Au gu s t 2010

 WIREs Cognitive Science Working memory

central storage resource where information could a central, abstract kind of working memory in which
be held while it was needed for processes such as information from any source can be entered, i.e.,
reasoning and comprehension. However, they also a domain-general working memory store. Baddeley8
thought that there were other types of storage (verbal- said no; Cowan11 and Baddeley10 said yes. Cowan
phonological and visual-spatial) that were not shared further proposed that the domain-general store is
with processing and were separate from one another. the focus of attention and therefore is an especially
Small verbal memory loads did not interfere with noteworthy form of working memory.
reasoning or comprehension, the explanation went, Cocchini, Logie, Della Sala, MacPherson, and
because the information could be saved as a series of Baddeley12 conducted a well-designed study in which
speech sounds in a phonological store that is separate there were two tasks on some trials, often including a
from what is used for reasoning or comprehension. verbal memory task and a spatial pattern memory
Baddeley8 crystallized the model so that it included task. There was little interference between these
a phonological loop (so called because it involved different tasks (i.e., little cross-modal interference),
covert rehearsal of the phonological information leading Cocchini et al. to reject the idea of a domain-
in a repeating loop) and a visuo-spatial sketchpad general store. That was the case in other dual-task
operating similarly. He also removed the central store studies as well (e.g., Ref 13). Yet, other studies with
for the sake of parsimony, though he later added one slightly different designs have shown more substantial
back, called an episodic buffer10 (Figure 1(b)). This cross-modal interference between visual and verbal
left a central executive carrying out all the processing materials.14–17 Theorists accept that there is some such
and managing the storage of all the information in interference but differ dramatically in the explanation
the phonological and visuo-spatial stores. Whereas for that interference.
the central executive was attention-demanding, in According to Baddeley,8 the interference should
this model the stores themselves were attention-free, not occur. According to Baddeley10 or Cowan9,11 the
including the rehearsal process. It is not yet clear if interference could occur in a domain-general working
attention is needed for the proposed episodic buffer. memory store. Cowan goes further in predicting that if
Cowan11 was uncomfortable with the division one could get rid of most sources of activated memory,
of labor in the Baddeley8 model. Although he accepted such as memory for the way the items look or sound,
that there is more interference between like kinds of what is left over is a store limited to a fixed number of
items in working memory, it did not seem likely conceptual items. Saults and Cowan18 addressed this
that the only important distinctions are between issue in an experiment in which participants received
verbal-phonological and visual-spatial information. spoken digits and arrays of visual colored spots. In
How could we understand, for example, memory for order to make rehearsal difficult, the spoken digits
tones in various spatial locations or memory for touch were presented simultaneously from four different
information? In the absence of definitive knowledge loudspeakers in different voices. Participants were
of the taxonomy of stores, Cowan lumped together sometimes responsible for only one sensory modality
many kinds of information storage and hypothesized but at other times they were responsible for both.
that they are all instances in which information from It turned out that a fixed limit occurred provided
the vast banks of long-term memory are temporarily that sensory memory activation was eliminated with
in an activated state. Also in contrast to the Baddeley8 a mask. The mask was a combination of multicolored
model, Cowan proposed that a limited amount of spots and garbled speech, which interfered with how
information is also held in the focus of attention the memory of the stimuli looked and sounded. After
(Figure 1(c)) rather than simply held in passive stores. the mask, there was a repetition of the stimuli to
Whereas the activated memory was supposed to be be remembered, but sometimes with a change in one
limited by the amount of time that has passed and item. When only visual memory was required, people
the types of interference that have occurred, the could remember 3–4 visual items on average. When
information in the focus of attention was supposed to auditory memory was also required, people could still
be limited to just a few (3–5) separate items, a theme remember a total of 3–4 items, but now some of
that Cowan9 developed further. those items were the spoken digits and others were
All of the models after the era of Atkinson and the colored spots. This seems like good evidence for a
Shiffrin7 seem to postulate more than one working domain-general memory.
memory storage system (verbal-phonological and Perhaps the most difficult part of a domain-
visual-spatial in Figure 1(b); activated memory and general memory to prove is that the domain-generality
focus of attention in Figure 1(c)). One critical point comes in the form of storage as opposed to processing.
that distinguishes the models is whether there exists If attention is needed to carry out a refreshing process

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 575

 Advanced Review wires.wiley.com/cogsci

(similar to thinking of an item, thereby reactivating its they predicted general fluid intelligence (gF), the ability
representation) for any stimuli19,20 then the limit could to solve new problems. The best latent variable model
be in the capability of the refreshing process rather explaining variability in these tasks was composed of a
than the contents of the focus of attention. We cannot general working memory capacity with contributions
yet clearly distinguish between these hypotheses. from spatial and verbal long-term memory. Working
Other research methods do support the notion memory capacity then predicted gF. This organization
that attention is involved in storing information from describes a domain-general working memory model
multiple modalities. Just et al.14 had participants in which information from all modalities is utilized
perform a verbal sentence comprehension task and by common maintenance and processing components
a visual mental rotation task and observed dual-task while retaining its modality specific attributes.
deficits in performance relative to when the tasks If we are correct that a domain-general capacity-
were performed alone. In this experiment functional limited working memory store exists it does not mean
magnetic resonance imaging (fMRI) data were also that working memory cannot also contain information
obtained and demonstrated tradeoffs similar to those that is specific to a modality (e.g., how a word looks
found in the behavioral data. When performed vs. how it sounds). Also, certain types of special
alone, the verbal and visual tasks activated separate processing may be specific to some kinds of stimuli;
brain areas in the temporal and parietal lobes, covert verbal rehearsal, for example, can help in
respectively, indicating separate resources for memory the recall of the serial order of items only if they
representations. When both tasks were performed can be labeled easily. It would be easy to rehearse
simultaneously, the same areas were again activated, a phone number but not the shape of an irregular
but activation levels were significantly lower in both polygon. These possible modality-specific features of
of these areas in the dual-task situation compared working memory do not negate the idea that there
to when the tasks were performed individually. This is a domain-general form of storage. The discussion
suggests that a common resource, namely attention, of whether this domain-general resource exists, or if
had to be shared between the modality-specific working memory is instead several independent, yet
representations in the temporal and parietal cortex, related, systems is clearly critical to understanding
consistent with the theory that a common attention- the organization of the brain and our own limits.
related resource influences both modalities. Encouragingly, it appears that progress is being made
In a study examining event-related brain towards a consensus on some sort of general-resource
potentials during a dual-task procedure, the amplitude for all modalities. The fact that we have a limited
of the P300 component elicited by an auditory central working memory system, however, naturally
secondary task was shown to decrease as the difficulty leads to the question, why is this system limited?
of a visual primary task increased.16 This is important
because the P300 component is considered to be
a measure of working memory updating, and is
WHY CAN WE RECALL ONLY A
not affected by response selection or execution.21 LIMITED NUMBER OF CHUNKS OF
The decrease in P300 amplitude for the secondary INFORMATION?
task suggests that increasing the resources required Earlier, we stated that working memory capacity is
to perform a visual task decreases the amount of generally limited to 3–5 meaningful items or chunks.
resources available to perform a concurrent auditory A good example of this constant capacity limit is
task. Critically, Sirevaag et al.16 also found that the demonstrated by Chen and Cowan.23 They presented
visual task P300 increased as the visual task increased lists of single words or learned pairs of words
in difficulty and that the magnitude of the decrease in for serial recall. They prevented rehearsal and its
the auditory secondary task P300 was complementary contributions to performance through articulatory
to the increase in magnitude for the visual primary suppression, a procedure in which a single word
task, consistent with the use of a domain general or sound is continuously repeated (e.g., ‘the, the,
resource for task performance. the. . .’). Capacity was measured in chunks, a name for
Statistical analysis of individual differences has meaningful units of information, which could either
also been used to support theories of a domain be a single word or a pair of words, depending on how
general working-memory component. Kane et al.22 items were presented. In line with past estimates,9,24
performed factor analysis of 12 tasks, 3 from each Chen and Cowan consistently observed recall of about
of the following categories, verbal working memory, 3 chunks across lists of 6–12 words presented as either
visual working memory, verbal short-term memory, single words or word pairs. This study illustrates that,
and visual short-term memory and examined how well when rehearsal and chunking of items are accounted

576  2010 Jo h n Wiley & So n s, L td. Vo lu me 1, Ju ly /Au gu s t 2010

 WIREs Cognitive Science Working memory

for, the contents of working memory are observed to at around 40 Hz seems to define an object32 and that
be a fairly constant number of conceptual items. this pattern of neuronal function seems to be related
Although it is clear that this capacity limit exists, to attention.33 Lisman and Idiart34 reviewed neural
why it exists is an open question without a single evidence suggesting that the synchronous firing for all
answer. Explanations of why we have a capacity limit items in working memory has to be repeated within a
include functional and mechanistic answers. In the period of about a tenth of a second. If too many items
present section, we give brief descriptions of several of are represented within that period then the features of
these explanations that we find most likely. Although one item can be confused with the features of another;
each view will provide a different answer to why we a red circle and blue square can be misremembered as a
have a fundamental capacity limit, they are generally red square and blue circle, for example. Other recent
complementary rather than competing explanations. work has shown that areas of the parietal cortex
The functional explanations treat the capacity demonstrate activation that reflects the number of
limit as a strength. We cannot process everything items that can be kept in mind at any given time.35,36
in the world perfectly so there has to be a decision Perhaps the synchronous firing occurs in relation to
about what to process, and attention is given to some these areas, consistent with some previous findings.33
information at the expense of other information. More research is certainly required to validate these
The capacity limit in working memory may make theories, but they provide interesting leads toward
that decision possible by keeping a limited few items determining why we are constrained in much of our
protected from proactive interference, or confusability thinking.
with previous items, at any given time.25,26 Similarly, Whereas we thus believe that the central
there is a limit in the number of links or bindings component of working memory is limited in the
that can be formed between one item and another number of chunks it can hold at once, we and
or between an item and its context.27 The process others have proposed that modality- and code-specific
of evolution28 could have favored individuals with aspects of working memory are limited in the duration
a working memory size that is well-suited to the for which activation of memory can persist.8,11
tasks that individuals must carry out. Smaller working Perhaps time limits also apply to central memory.
memories would be insufficient for self-preservation The next section explores the possibility of time limits
and larger working memories would be biologically in working memory.
too costly to maintain.
In another sort of functional argument, it has
been suggested that the size of working memory we WHY DOES WORKING MEMORY
have is not only the best that can be done with the APPEAR TO HAVE A TIME LIMIT?
amount of biological energy available; but that it is, in As we have discussed, our working memory is limited
fact, optimal. Dirlam29 and MacGregor30 put forward in capacity, and doing, thinking, and remembering
mathematical arguments based on certain assumptions all require some of this limited capacity in order to
about how people search through memory. The search function. If we do not pay special attention to the items
process involves finding the right group or chunk of we wish to remember, such as the digits in a phone
information and then finding the right item within number we have just been told, they seem to evaporate
the group. With this kind of analysis it turns out from our minds very quickly. Peterson and Peterson37
that the size of grouping that is most efficient is showed that memory for consonant trigrams (such as
within the range of the number of items in working M-K-V) became worse over an 18 s time period while
memory. So assuming that items that are in working participants counted backwards by 3 or 4 s from a
memory at the same time can be grouped together, random three-digit number, eventually resulting in
our working memory could have evolved to allow almost no remaining memory of the trigram. This
maximally efficient groups to be formed and later demonstrates just how quickly we forget even a small
searched. amount of information held in mind while distracted
The mechanistic explanations address the issue by another task. In the present section, we discuss the
of just what it is in the brain that allows a two prominent explanations for why we forget over
certain number of items to be remembered at once time in situations like this one.
and prevents more from being remembered. One One description of why we forget claims that
possibility is based on Milner’s31 proposition that items are represented in the brain as activated
memory items exist because the neurons that represent memory traces and that this activation decays with
their features fire synchronously. This idea has merit, time if the items are not mentally rehearsed or
as several studies have found that synchronous firing refreshed.8,11,20,38,39 At some point, the activation

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 577

 Advanced Review wires.wiley.com/cogsci

level becomes too low and items are no longer time, which leads to quicker loss of memory items
retrievable for conscious recall. This decay theory when a more difficult or more frequent secondary
would explain the Peterson and Peterson results as task must be carried out. These premises lead to the
at least partially due to the passage of time. The predication that the proportion of time occupied by
opposing position is that once items are in memory a distracting nonmaintenance task, compared to total
they remain until some process interferes with the time that the memory must be maintained, predicts
item representation through displacement, feature the amount of forgetting. Evidence in favor of the
overwriting, or some other process,40,41 at which point TBRS model comes primarily from experiments in
information is lost and forgetting occurs. Interference which participants are to remember a series of items,
theories would explain the Peterson and Peterson such as words, digits, or letters. After the presentation
findings as entirely the result of interference from of each item, a processing task is presented, such
the counting task and buildup of item familiarity over as reading numbers off a screen or solving a simple
multiple trials. math equation. Once all items to be recalled and
Traditionally, the strongest evidence supporting related processing events have been presented, the
temporal decay of memory traces as the reason for memory items are all to be recalled. The ratio of
forgetting has been the word length effect (WLE), nonmaintenance time to total time available, called
a term meaning that fewer longer words can be the cognitive load, has a negative linear relationship
remembered concurrently than shorter words.42 This with the accuracy of recall.20,38,52 This supports the
was supposedly because memory traces that are not notion of decay unless another explanation can be
rehearsed will decay beyond the threshold for recovery found as to why recall is linearly related to cognitive
within about 2 s, and longer words require more time load.
per word to rehearse. Although the WLE has been In contrast to memory decay, the existence
replicated many times,43–45 a number of confounds of forgetting due to various forms of interference
eventually discredited it as firm evidence for time- effects is not a controversial topic. In fact, most
based decay. Especially damaging was evidence that decay theorists readily agree that a large amount
increased phonologic complexity, not articulatory of forgetting within working memory is due to
duration, may underlie the WLE.46 Currently there interference. Interference theorists, however, differ
are arguments both supporting the WLE as a result of from decay theorists in that they believe that all
memory decay45 and denying that the WLE provides forgetting within working memory, not just some
evidence for temporal decay,47 leaving its status or most, is due to interference effects. Oberauer
uncertain. and Kliegl,41 for example, mathematically model
More direct tests of decay have also been tried. performance specifically as a function of feature
Cowan and AuBuchon48 presented lists with an overlap between items, speed of processing, and level
irregular timing and sometimes required that recall of nonrelevant concept activation. These parameters
also be in that timing. The theoretical rationale was are allowed to vary both across participants and
that it is difficult to rehearse while remembering the across tasks in order to reflect effects of both
timing. It was found that lists with long intervals near individual differences and varying task requirements.
the beginning impaired recall of items later in the list, This conceptualization appears to be fairly successful,
in keeping with a decay hypothesis. Other studies, explaining more than 80% of observed variance in at
however, have directly manipulated the pace of recall least one verbal and one spatial task. In these models,
and have not found very much impact of the recall working memory capacity, though a domain-general
pace on the amount recalled.49–51 concept, is not constant. The level of overlap between
Perhaps the strongest current argument for a currently held representations will cause a predictable
time-based decay explanation of forgetting comes amount of forgetting, resulting in lower functional
from research by Barrouillet and colleagues20,38 capacity for relatively more similar representations.
using the continuous span task paradigm to support It remains to be seen whether this model can
their time-based resource-sharing (TBRS) model of accommodate the results favoring a constant capacity
working-memory. The TBRS model consists of several (e.g., Refs 9,18,23).
critical assumptions. First, items are activated memory Oberauer and Kliegl also claim that their model
traces which decay quickly and are forgotten. These can explain the findings of Barrouillet and colleagues
items are maintained through an attention-based by positing that free time during memory retention is
refreshing mechanism which is also necessary for spent repairing damaged memory traces. The critical
performing long-term memory retrievals. Attention factor explaining memory performance would be free
may be capable of performing only one task at a time for repair, not the time over which memory

578  2010 Jo h n Wiley & So n s, L td. Vo lu me 1, Ju ly /Au gu s t 2010

 WIREs Cognitive Science Working memory

decays as in the TBRS account. These competing and Carpenter was strongly correlated with verbal
claims were investigated by Portrat, Barrouillet, and scholastic aptitude test (SAT) scores, as well as several
Camos.52 They varied the amount of time between other reading comprehension measures, whereas
items during which distracting processing had to be memory for lists of words was not significantly
carried out but left constant the amount of time correlated with the same comprehension measures.
without distraction. They found that more time Later research also uncovered a strong relationship
processing a distraction still meant more memory between working memory tasks requiring both storage
loss, in keeping with the decay theory. Given that and processing and a large variety of other measures
a single study rarely can resolve a basic issue, more of complex cognition, including, for example, learning
investigation is needed to determine which view of to spell,54 reasoning ability,3 and gF.55 These effects
forgetting is correct, or if some hybrid model should were obtained even when the working memory task
be preferred. included spatial processing along with verbal storage,
In sum, all investigators clearly agree that suggesting that a domain-general working memory
information is lost over time, but some investigators resource is used in these tasks and in complex cognitive
attribute this solely to the effects of interference from functions.22
additional material that is processed during a retention The question remains, what drives these individ-
interval, whereas other investigators believe that there ual differences between people? One possibility is that
are some effects of time regardless of the nature of individuals who demonstrate higher working mem-
interfering stimuli. ory spans have more efficient executive functions, so
The previous sections of this review have that the processing task consumes less attention and
discussed how everyone’s working memories are leaves more for storage.56 A second theory posits that
the same in several fundamental, though debated, individual differences in both processing and stor-
ways. The following two sections review findings age capacity can contribute to overall differences in
of differences across people in working memory working memory performance.57
function and what this may say about the fundamental Various tasks show the role of processing ability.
structure of working memory in general. For example, Conway, Cowan, and Bunting58 used
a dichotic listening task intended to replicate the
‘cocktail party phenomenon’ in which people tend
HOW AND WHY DO PEOPLE DIFFER IN to hear their own name amid a noisy environment.
WORKING MEMORY ABILITIES? They found that only 20% of high span individuals (in
Given that the estimated limits of the working the upper quartile of working memory performance)
memory capacity range from only 3–5 items, it seems noticed that their name had been presented in the
surprising that individual variation would be large unattended auditory channel, compared to 65% of
enough to influence performance on other tasks. low span individuals. This difference is likely due to
Yet this is exactly what we observe, space for one the high spans’ better ability to block the irrelevant
less item in working memory could mean forgetting information and selectively attend to a single auditory
one important point in a conversation or one key channel. High span individuals’ ability to better
part of a problem. We start by describing working selectively attend to, or selectively inhibit, information
memory span tasks and their relationship to other has also been demonstrated in the visual modality.
cognitive abilities. Then we discuss several proposed Vogel, McCollough, and Machizawa59 showed this
explanations of how storage and processing contribute with memory for visual arrays using an event-
to overall working memory capacity. related potential measure of working memory load,
The capacity limited, domain-general form of the contralateral delay activity (CDA) amplitude.
memory we have discussed above appears closely Participants saw a visual array containing target and
linked to cognitive ability. Daneman and Carpenter53 distracter items. In one task, for example, they were to
asked participants to read sets of sentences, while remember the orientations of green bars and ignore the
remembering the last word of each sentence. College orientations of red bars. Low-spans’ CDA amplitude
students in their study were able to remember only 2–5 indicated that they were maintaining both the target
words, depending on the individual. This surprisingly and distracter items in working memory, but the lower
low memory ability relative to the estimates of CDA amplitude of high-span individuals indicated
earlier research6 is understandable because attention that they were maintaining only the target items and
is occupied with processing between the presentation thus were able to inhibit the distracter items from
of items, preventing the grouping of multiple words occupying working memory. Individual differences in
into a chunk. This reading span measure of Daneman working memory performance also predict the ability

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 579

 Advanced Review wires.wiley.com/cogsci

to inhibit automatic responses, such as turning one’s There are several ways in which the processing
eyes toward an item that comes on the screen when the abilities of children differ from those of adults. A
instruction is to turn one’s eyes the opposite direction, prominent difference is that young children lack
an ‘antisaccade’ eye movement.60 strategies, such as covert rehearsal. Children begin
An alternative view is that differences between to rehearse, or silently repeat verbal materials, around
high- and low-span individuals in processing efficiency age seven, but their use of this strategy becomes
are accompanied by true differences in storage more complex over time.63,64 Processing speed also
capacity.5,57 Using a behavioral version of the increases throughout childhood and adolescence, with
Vogel et al.59 distracter task described above, Gold much of the speed increase occurring before the age
et al.5 found that the difference between normal of 12.65
adults and those with schizophrenia was mostly Along with the developing efficiency of pro-
in storage capacity. Individuals with schizophrenia cessing abilities, the storage component of working
ignored distracters about as well as normal adults memory has also been argued to increase in size
but still recalled fewer items overall. Additionally, throughout development.57,66 Cowan, Morey, AuBu-
storage capacity, rather than reading speed or other chon, Zwilling, and Gilchrist67 showed this in a
processing skills, accounts for differences in spelling developmental study similar to Gold et al.’s5 study
ability among good readers.54 When carefully teased of normal and schizophrenic adults. Children were
apart, both storage and the ability to control attention to attend to some items (e.g., colored circles) but
can account for independent variance in measures ignore others (e.g., colored triangles). Usually they
of intelligence.57 Individual differences research
were tested on memory for the color of an attended
supports our understanding of working memory as
item but occasionally they were tested on the color
a combination of the limited content held in an
of an item they were supposed to ignore. Children in
easily accessible state, as well as the processing used
first grade did much better remembering the colors of
to encode and maintain that information. Individual
attended items than ignored items, to the same extent
variance in each of these components contributes to
as older children or adults did, provided that there
differences in overall working memory performance
were only four items in the field. Nevertheless, first-
and high-level cognitive abilities, such as aptitude test-
taking, problem-solving, and reading comprehension. grade children remembered far fewer of the objects of
It appears that in general, everyone is limited in what either shape, indicating a storage deficit despite adult-
they can remember both by the amount of cognitive like filtering out of irrelevant items in this simple task.
space they have to remember items and by how well Taken together, the studies indicate that processing
processing mechanisms can support memory for those failures in children are less likely to occur when the
items. amount of information is within the limits of working
Similar to research on within-group individual memory capacity. Moreover, they suggest that age
differences in working memory, we can also use differences in capacity may cause some age differences
developmental changes to learn about the cognitive in processing, rather than depending upon them.
functions which underlie working memory. In sum, there is evidence for individual and age
group differences in both general working-memory
capacity and the ability to manage this capacity
using central executive processes. There may also
DEVELOPMENTAL CHANGES IN be individual and group differences in other aspects
WORKING MEMORY: EFFICIENCY, of working memory, such as decay and interference
CAPACITY, OR BOTH? effects.49
Agreement exists that older children and adults So far we have relied primarily on behavioral
demonstrate higher working memory span scores data in examining various claims about working
than young children, with adult-level performance memory. The behavioral data remain important even
on simple tasks attained around age 10.61 The debate today but powerful new sources of information are
continues, though, over what developmental changes coming from brain imaging studies as well. Some
account for this difference. This debate mirrors investigators have claimed that brain imaging tells us
the one that persists in the individual differences little about the overall, abstract organization of the
literature, namely whether changes that occur in cognitive system, but the case for the contribution of
working memory during childhood involve processing brain imaging (i.e., neuroimaging) seems stronger in
efficiency alone,60,62 or also include capacity increases the area of working memory than in most other areas
during this timeframe.61 of research.

580  2010 Jo h n Wiley & So n s, L td. Vo lu me 1, Ju ly /Au gu s t 2010

 WIREs Cognitive Science Working memory

WHAT CAN NEUROIMAGING STUDIES representational structure between perception and

TELL US? working memory, with the frontal lobes taking the
role of the central executive that helps to maintain and
The emerging field of neuroimaging holds exciting
manipulate items in storage which are represented by
promise for working memory research, producing new
neural activity elsewhere in the brain.
and interesting constraints on theory. Here we give
The parietal lobes could be the neural correlate
a brief description of findings on working-memory
of the focus of attention. Whereas domain-specific
storage and cognitive processing related to the theories
and findings discussed above. We attempt to show a traces are represented in various areas used for
correspondence between the behavioral literature and sensory and conceptual processing, activation in the
the neuroimaging literature regarding the structure of area of the intra-parietal sulcus (ISP) may underlie the
working memory. actual chunk capacity limits within working memory.
Central to the convergence of behavior-based Using a visual working memory task that required
theory and neuroimaging findings is the observation memory for the spatial location of colored disks, Todd
that items held in working memory tend to activate and Marois35 demonstrated that the IPS increases in
the same representations in the brain as are activated activation as the number of items to be remembered
by perceptual processing. More specifically, the disso- increases, until working-memory capacity is reached,
ciation of spatial and object memory representations around four items. This cannot be explained as a
in the posterior sections of the brain are upheld in perceptual effect because as the number of items to be
working memory,68 mirroring the visual perceptual remembered increased beyond four there was no fur-
differences of the ventral and dorsal pathways, for ther increase in activation. Xu and Chun36 confirmed
what versus where an item is, respectively.69 Spatial this finding and elaborated on it, demonstrating that
memory storage tends to activate dorsal posterior cor- inferior IPS activation increases with the number
tex, whereas object memory tends to activate ventral of items to be remembered until it plateaus at 3 or
posterior cortex.70 Similarly, verbal memory repre- 4 items, whereas superior IPS and lateral occipital
sentations show activation in the same areas active complex (LOC) activation reflect the number of items
in speech perception, and production, primarily the successfully held within working-memory capacity.
left perisylvian cortex.71 These findings are consis- Thus, superior IPS and LOC activation matched
tent with theories positing that items within working inferior IPS activation when remembering simple
memory are composed of activated traces of long-term items, around four items, but was lower, around two
memory representations.39–41 It is important to note items, when remembering more complex figures.
that although domain-specific representations exist in Together, these studies suggest that the inferior
differing portions of the brain, they still all theoreti- IPS contains information about which items are
cally could be within a centralized focus of attention, attended, whereas the superior IPS and LOC index
the neural correlate of which may be elsewhere in the which traces have their features coded and held within
brain.
working memory. Detailed interpretation of these
Some researchers have argued that active
results is difficult given the current understanding
working memory representations are stored within
of mechanisms that govern working-memory capacity
the frontal cortex.72 This frontal storage hypothesis
and performance. Several lines of behavioral research
was directly tested by Postle and colleagues73 with the
are attempting to address precisely how item
use of repetitive transcranial magnetic stimulation,
which temporarily ‘turns off’ a targeted region of complexity effects working memory capacity,74,75 and
cortex. When areas of the dorsolateral prefrontal seem to indicate that complexity affects the resolution
cortex (DLPFC) were stimulated, specifically within of stimulus representations, whereas the number of
the middle frontal gyrus, manipulation of items representations that can be held is unaffected by item
within working memory was disrupted but storage complexity.
of items was not. In contrast to DLPFC stimulation, Executive functions related to working memory
when posterior sections of the brain were stimulated, appear to tap differing brain areas, primarily within
specifically within the superior parietal lobule, both the frontal cortex. Refreshing processes, such as those
manipulation and storage of items within working theorized by Barrouillet and colleagues to be crucial
memory was disrupted. These results suggest that for working memory maintenance in many situations,
posterior cortex is necessary for the storage of items appear to result from activation of the DLPFC.19
within working memory, whereas DLPFC is critical Ventrolateral PFC activation, on the other hand, is
for manipulation but not storage per se. It seems associated with rehearsal of verbal materials,19 the
parsimonious to assume a common or highly related critical maintenance mechanism in Baddeley’s8 model

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 581

 Advanced Review wires.wiley.com/cogsci

working memory. Overall, organization of working over time. Interference accounts can provide power-
memory functions within the brain seem to con- ful descriptions of most, but not yet all, short-term
form to general patterns of frontal refreshing and forgetting. It will be interesting to see if future inter-
manipulation of information and posterior storage of ference models can maintain their simplicity while
the information, which may include both domain- at the same time explaining new findings, especially
specific and domain-general types of storage. Theories those from cognitive load studies. Firmer conclusions
of working memory based on behavioral data now are possible about the reasons for individual and
must be able to explain these and other new findings developmental differences in working memory per-
from brain imaging techniques in addition to data formance. Differences in processing efficiency appear
from their traditional paradigms in order to continue to be responsible for some of the differences in per-
to be viable. formance across individuals and across different age
groups, but differences in working memory storage
capacity also appear to account for some of these
CONCLUSION differences. New questions continue to surface, such
as whether an increase in working memory capac-
Research in the past few decades has come a long ity may influence processing efficiency rather than
way towards answering fundamental questions con- the other way around. Finally, recent research using
cerning the structure and function of that elusive neuroimaging techniques has made valuable contribu-
concept, working memory. For instance, there appears tions to working memory theory. Working memory
to be a constant capacity limit that can be observed storage appears to arise from higher level control
across many experimental paradigms when peripheral functions located in the frontal lobe, and possibly
sources of storage and maintenance (sensory memory, the IPS, maintaining memory traces which consist
chunking, and rehearsal) are eliminated. In order to of perceptual and conceptual features represented in
posit multiple independent working-memory systems, other parts of the brain. This corroborates theoretical
one must now be able to explain cross-modal disrup- descriptions of hierarchically organized models, such
tion of memory across a wide number of contexts. as the embedded process model,11 the TBRS model,20
Research on forgetting has narrowed the range of and some feature models,41 although the existing evi-
plausible processes that may be occurring, but has dence is not specific enough to determine which of
not been able to give a definitive account of how these classes of models are the best fit for describing
fundamental forgetting from working memory occurs working memory function.

REFERENCES
1. Miller GA, Galanter E Pribram KH. Plans and the 6. Miller GA. The magical number seven, plus or minus
Structure of Behavior. New York: Holt, Rinehart and two: some limits on our capacity for processing infor-
Winston; 1960. mation. Psychol Rev 1956, 63:81–97.
7. Atkinson RC, Shiffrin RM. Human memory: a pro-
2. Baddeley AD, Hitch G. Working memory. In: Bower
posed system and its control processes. In: Spence KW,
GH, ed. The Psychology of Learning and Motivation,
Spence JT, eds. The Psychology of Learning and Moti-
vol. 8. New York: Academic Press; 1974, 47–89.
vation: Advances in Research and Theory, vol. 2. New
3. Kyllonen PC, Christal RE. Reasoning ability is (lit- York Academic Press; 1968, 89–195.
tle more than) working-memory capacity?! Intelligence 8. Baddeley AD. Working Memory. Oxford Oxford Uni-
1990, 14:389–433. versity Press, Clarendon Press; 1986.
9. Cowan N. The magical number 4 in short-term mem-
4. Martinussen R, Hayden J, Hogg-Johnson S, Tannock ory: a reconsideration of mental storage capacity. Behav
R. A meta-analysis of working memory impairments in Brain Sci 2001, 24:87–185.
children with attention-deficit/hyperactivity disorder. J
10. Baddeley AD. The episodic buffer: a new component of
Am Acad Child Adolesc Psychiatry 2005, 44:377–384.
working memory? Trends Cogn Sci 2000, 4:417–423.
5. Gold JM, Fuller RL, Robinson BM McMahon RP 11. Cowan N. Evolving conceptions of memory storage,
Braun EL, et al. Intact attentional control of working selective attention, and their mutual constraints within
memory encoding in schizophrenia. J Abnorm Psychol the human information processing system. Psychol Bull
2006, 115:658–673. 1988, 104:163–191.

582  2010 Jo h n Wiley & So n s, L td. Vo lu me 1, Ju ly /Au gu s t 2010

 WIREs Cognitive Science Working memory

12. Cocchini G, Logie RH Della Sala S, MacPherson SE, 28. Darwin C. On the Origin of Species by Means of Nat-
Baddeley AD. Concurrent performance of two memory ural Selection, or the Preservation of Favoured Races
tasks: evidence for domain-specific working memory in the Struggle for Life. London John Murray; 1859.
systems. Mem Cognit 2002, 30:1086–1095.
29. Dirlam DK. Most efficient chunk sizes. Cogn Psychol
13. Friedman NP, Miyake A. Differential roles for visu- 1972, 3:355–359.
ospatial and verbal working memory in situation model
30. MacGregor JN. Short-term memory capacity: limita-
construction. J Exp Psychol Gen 2000, 129:61–83.
tion or optimization? Psychol Rev 1987, 94:107–108.
14. Just MA, Carpenter PA, Keller TA, Emery L Zajac
31. Milner PM. A model for visual shape recognition. Psy-
H, et al. Interdependence of nonoverlapping cortical
chol Rev 1974, 81:521–535.
systems in dual cognitive tasks. Neuroimage 2001,
14:417–426. 32. Gray CM, König P, Engel AK, Singer W. Oscillatory
15. Morey CC, Cowan N. When do verbal and visual mem- responses in cat visual cortex exhibit inter-columnar
ories conflict? The importance of working memory load synchronization which reflects global stimulus proper-
and retrieval. J Exp Psychol Learn Mem Cogn 2005, ties. Nature 1989, 338:334–337.
31:703–713. 33. Tiitinen H, Sinkkonen J, Reinikainen K, Alho K,
16. Sirevaag EJ, Kramer AF, Coles MG, Donchin E. Lavikainen J, et al. Selective attention enhances the
Resource reciprocity: an event-related brain potentials auditory 40-Hz transient response in humans. Nature
analysis. Acta Psychol 1989, 70:77–97. 1993, 364:59–60.
17. Stevanovski B, Jolicoeur P. Visual short-term memory: 34. Lisman JE, Idiart MAP. Storage of 7 ± 2 short-
central capacity limitations in short-term consolidation. term memories in oscillatory subcycles. Science 1995,
Vis Cogn 2007, 15:532–563. 267:1512–1515.
18. Saults JS, Cowan N. A central capacity limit to the 35. Todd JJ Marois R. Capacity limit of visual short-term
simultaneous storage of visual and auditory arrays memory in human posterior parietal cortex. Nature
in working memory. J Exp Psychol Gen 2007, 2004, 428:751–754.
136:663–684. 36. Xu Y, Chu MM. Dissociable neural mechanisms sup-
19. Raye CL, Johnson MK, Mitchell KJ, Greene EJ, John- porting visual short-term memory for objects. Nature
son MR. Refreshing: a minimal executive function. 2006, 440:91–95.
Cortex 2007, 43:135–145. 37. Peterson LR, Peterson MJ. Short-term retention of indi-
20. Barrouillet P, Bernardin S, Camos V. Time constraints vidual verbal items. J Exp Psychol 1959, 58:193–198.
and resource sharing in adults’ working memory spans.
38. Barrouillet P, Bernardin S, Portrat S, Vergauwe E,
J Exp Psychol Gen 2004, 133:83–100.
Camos V. Time and cognitive load in working memory.
21. Polich J. Updating P300: an integrative theory of P3a J Exp Psychol Learn Mem Cogn 2007, 33:570–585.
and P3b. Clin Neurophysiol 2007, 118:2128–2148.
39. Cowan N. Attention and Memory: An Integrated
22. Kane MJ, Hambrick DZ, Tuholski SW, Wilhelm O, Framework. Oxford Oxford University Press; 1995.
Payne TW, et al. The generality of working-memory
capacity: a latent-variable approach to verbal and 40. Nairne JS. A feature model of immediate memory. Mem
visuo-spatial memory span and reasoning. J Exp Psy- Cognit 1990, 18:251–269.
chol Gen 2004, Jun 133:189–217. 41. Oberauer K, Kliegl R. A formal model of capacity limits
23. Chen Z, Cowan N. Core verbal working memory in working memory. J Mem Lang 2006, 55:601–626.
capacity: the limit in words retained without covert 42. Baddeley AD, Thompson N, Buchanan M. Word length
articulation. Q J Exp Psychol 2009, Jul 62:1420–1429. and structure of short-term memory. J Verbal Learn
24. Broadbent DE. The magic number seven after fifteen Verbal Behav 1975, 14:575–589.
years.In: Kennedy A, Wilkes A, eds. Studies in Long- 43. Cowan N, Day L, Saults JS, Keller TA, Johnson T,
term Memory. New York John Wiley & Sons; 1975, et al. The role of verbal output time in the effects of
3–18. word length on immediate memory. J Mem Lang 1992,
25. Cowan N Johnson TD, Saults JS. Capacity limits in list 31:1–17.
item recognition: evidence from proactive interference. 44. Cowan N, Nugent LD, Elliot EM, Geer T. Is there a
Memory 2005, 13:293–299. temporal basis of the word length effect? A response to
26. Halford GS, Maybery MT, Bain JD. Set-size effects in Service (1998). Q J Exp Psychol 2000, 53A:647–660.
primary memory: an age-related capacity limitation? 45. Mueller ST, Seymour TL, Kieras DE, Meyer DE. Theo-
Mem Cognit 1988, 16:480–487. retical implications of articulatory duration, phonolog-
27. Halford GS, Cowan N, Andrews G. Separating cogni- ical similarity and phonological complexity in verbal
tive capacity from knowledge: a new hypothesis. Trends working memory. J Exp Psychol Learn Mem Cogn
Cogn Sci 2007, 11:237–242. 2003, 29:1353–1380.

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 583

 Advanced Review wires.wiley.com/cogsci

46. Service E. The effect of word length on immediate serial role in working memory and cognitive aptitudes. Cogn
recall depends on phonological complexity, not articu- Psychol 2005, 51:42–100.
latory duration. Q J Exp Psychol 1998, 51A:283–304. 62. Case R, Kurland DM, Goldberg J. Operational effi-
47. Lewandowsky S, Oberauer K. The word length effect ciency and the growth of short-term memory span. J
provides no evidence for decay in short-term memory. Exp Child Psychol 1982, 33:386–404.
Psychon Bull 2008, 15:875–888.
63. Cowan N, Saults JS, Morey CC. Development of work-
48. Cowan N, AuBuchon AM. Short-term memory loss ing memory for verbal-spatial associations. J Mem Lang
over time without retroactive stimulus interference. 2006, 55:274–289.
Psychon Bull Rev 2008, 15:230–235.
64. Gathercole SE, Adams AM, Hitch GJ. Do young chil-
49. Cowan N, Elliott EM, Saults JS, Nugent LD, Bomb dren rehearse? An individual-differences analysis. Mem
P, et al. Rethinking speed theories of cognitive devel- Cognit 1994, 22:201–207.
opment: Increasing the rate of recall without affecting
65. Fry AF, Hale S. Processing speed, working memory,
accuracy. Psychol Sci 2006, 17:67–73.
and fluid intelligence. Psychol Sci 1996, 7:237–241.
50. Lewandowsky S, Duncan M, Brown GDA. Time does
not cause forgetting in short term serial recall. Psychon 66. Cowan N, Nugent LD, Elliot EM, Ponomarev I, Saults
Bull Rev 2004, 11:771–790. JS. The role of attention in the development of short-
term memory: age differences in the verbal span of
51. Oberauer K, Lewandowsky S. Forgetting in immedi- apprehension. Child Dev 1999, 70:1082–1097.
ate serial recall: decay, temporal distinctiveness, or
interference? Psychol Rev 2008, 115:544–576. 67. Cowan N, Morey CC, AuBuchon AM, Zwilling CE,
Gilchrist AL. Seven year-olds allocate attention like
52. Portrat S, Barrouillet P, Camos V. Time-related decay
adults do unless working memory is overloaded. Dev
or interference-based forgetting in working memory? J
Sci 2009, 12:1–14.
Exp Psychol Learn Mem Cogn 2008, 34:1561–1564.
68. Jonides J, Lacey SC, Nee DE. Processes of working
53. Daneman M, Carpenter PA. Individual differences in
memory in mind and brain. Curr Dir Psychol Sci 2005,
working memory and reading. J Verbal Learn Verbal
14:2–5.
Behav 1980, 19:450–466.
54. Ormrod JE, Cochran KF. Relationship of verbal abil- 69. Haxby JV, Grady CL, Horwitz B, Ungerleider LG,
ity and working memory to spelling achievement and Mishkin M, et al. Dissociation of object and spatial
learning to spell. Read Res Instruction 1988, 28:33–43. visual processing pathways in human extrastriate cor-
tex. Proc Natl Acad Sci USA 1991, 88:1621–1625.
55. Engle WE, Tuholski SW, Laughlin JE, Conway AR.
Working memory, short-term memory, and general 70. Postle BR, D’Esposito M. ‘‘What’’-then-‘‘where’’ in
fluid intelligence: a latent variable approach. J Exp visual working memory: an event-related fMRI study.
Psychol Gen 1999, 128:309–331. J Cogn Neurosci 1999, 11:585–597.
56. McCollough AW, Vogel EK. Your inner spam filter: 71. Feredoes E, Tononi G, Postle BR. The neural bases of
What makes you so smart? Might be your lizard brain. short-term storage of verbal information are anatom-
Scient Am Mind 2008, 32–35. ically variable across individuals. J Neurosci 2007,
27:11003–11008.
57. Cowan N, Fristoe NM, Elliot EM, Brunner RP, Saults
JS. Scope of attention, control of attention, and intel- 72. Leung HC, Gore JC, Goldman-Rakic PS. Sustained
ligence in children and adults. Mem Cognit 2006, mnemonic response in the human middle frontal gyrus
34:1754–1768. during on-line storage of spatial memoranda. J Cogn
58. Conway ARA, Cowan N, Bunting MF. The cocktail Neurosci 2002, 14:659–671.
party phenomenon revisited: the importance of working 73. Postle BR, Ferrarelli F, Hamidi M, Feredoes E, Massi-
memory capacity. Psychon Bull Rev 2001, 8:331–335. mini M, et al. Repetitive transcranial magnetic stimu-
59. Vogel EK, McCollough AW, Machizawa MG. Neu- lation dissociates working memory manipulation from
ral measures reveal individual differences in con- retention functions in the prefrontal, but not posterior
trolling access to working memory. Nature 2005, parietal, cortex. J Cogn Neurosci 2006, 18:1712–1722.
438:500–503. 74. Awh E, Barton B, Vogel EK. Visual working mem-
60. Kane MJ, Bleckley MK, Conway ARA, Engle RW. A ory represents a fixed number of items regardless of
controlled-attention view of working-memory capacity. complexity. Psychol Sci 2007, 18:622–628.
J Exp Psychol Gen 2001, 130:169–183. 75. Eng HY, Chen D, Jiang Y. Visual working memory for
61. Cowan N, Elliott EM, Saults JS, Morey CC, Mattox S, simple and complex visual stimuli. Psychon Bull Rev
et al. On the capacity of attention: its estimation and its 2005, 12:1127–1133.

FURTHER READING
Baddeley A. Working Memory, Thought, and Action. New York: Oxford University Press; 2007.

584  2010 Jo h n Wiley & So n s, L td. Vo lu me 1, Ju ly /Au gu s t 2010

 WIREs Cognitive Science Working memory

Conway ARA, Jarrold C, Kane MJ, Miyake A, Towse, J, eds. Variation in Working Memory. New York:
Oxford University Press; 2007.
Cowan N. Working Memory Capacity. Hove, East Sussex, UK: Psychology Press; 2005.
Jonides J, Lewis RL, Nee DE, Lustig CA, Berman MG, etal. The mind and brain of short-term memory. Ann
Rev Psychol 2008, 59: 193–224.
Klingberg T. The Overflowing Brain: Information Overload and the Limits of Working Memory. New York:
Oxford University Press; 2009. [Translated by Neil Betteridge.]
Osaka N, Logie RH, D’Esposito M. The Cognitive Neuroscience of Working Memory. New York: Oxford
University Press; 2007.

Vo lu me 1, Ju ly /Au gu s t 2010  2010 Jo h n Wiley & So n s, L td. 585