C

Ga

pb

Sc

Z

Tfthsbeislt“menrwatrs

“wctpummwepcbeccfipt

t

Neuroscience 139 (2006) 1–3

0d

OGNITIVE NEUROSCIENCE OF WORKING MEMORY: A PROLOGUE

mtpcasAltakbi“iclo(

c1pqwsaticd

A

Ttaitstitptrttnap

. REPOVŠa,b AND M. BRESJANACc

Department of Psychology, Washington University in St. Louis, Cam-us Box 1125, One Brookings Drive, St. Louis, MO 63130, USA

Department of Psychology, University of Ljubljana, Askerceva 2,I-1000, Ljubljana, Slovenia

Institute of Pathophysiology, Medical Faculty, University of Ljubljana,aloska 4, SI-1000 Ljubljana, Slovenia

he roots of the concept of working memory and the basesor its exploration within cognitive neuroscience can beraced all the way back to the realization that human be-avior cannot be fully explained through specifying thetimulus-response relations, as purported by the prevalentehavioral paradigm at the time. Rather, to be able toxplain human behavior and experience one has to take

nto account the ways that information provided by outsidetimuli gets processed, manipulated, combined and re-ated to the previous experiences before it is actually usedo guide behavior. Tolman (1948) put that down very clearly:We assert that the central office itself is far more like aap control room than it is like an old-fashioned telephonexchange. The stimuli, which are allowed in, are not con-ected by just simple one-to-one switches to the outgoingesponses. Rather, the incoming impulses are usuallyorked over and elaborated in the central control room intotentative, cognitive-like map of the environment. And it is

his tentative map, indicating routes and paths and envi-onmental relationships, which finally determines what re-ponses, if any, the animal will finally release.”

To try to figure out and explain, what goes on in thecontrol room,” a whole new set of methods and metaphorsas needed. The developments in the areas of communi-ation technologies, information theory and computing ledo the establishment of the information-processing (IP)aradigm. The mind and its substrate, the brain, werenderstood as a system for processing and storing infor-ation, similar to a computer, which became the prevailingetaphor. Adopting a decompositional strategy the goal ofhat became known as the cognitive paradigm was toxplain the capacities and properties the human mindossesses in terms of characteristics of the parts of itsognitive system. The basic assumption has been that therain enables a person to have these capacities and prop-rties by virtue of having specific information-processingomponents operating in a specific way. This approachan be compared with reverse engineering trying togure out the design of a machine by studying its com-onent parts, their properties, connections and interac-ions (Atkinson, 1998).

To be able to process it and have it available at a later

ime, the information has to be stored. The concept of f306-4522/06$30.00�0.00 © 2006 Published by Elsevier Ltd on behalf of IBRO.oi:10.1016/j.neuroscience.2005.12.007

1

emory therefore became an important one within cogni-ive paradigm and a number of theories of memory wereroposed (e.g. Atkinson and Shriffin, 1968). For the exe-ution of complex cognitive abilities and for the control ofn ongoing goal-directed behavior, a general-purposehort-term store seemed to be an especially important one.ny kind of information processing system that manipu-

ates, integrates or analyzes information, needs to havehe ability to store the relevant information, the intermedi-te products and the results of manipulation at hand. Anyind of a system that needs to flexibly choose and executeehavioral plans needs to have the ability to hold that

nformation online. Miller et al. (1960) explicitly wrote:When we have decided to execute some particular Plan,t is probably put into some special state or place where itan be remembered while it is being executed. We wouldike to speak of the memory we use for the execution ofur Plans as a kind of quick-access, ‘working memory.’”p. 65), and so marked the first use of the term.

The term working memory was subsequently used inomputational modeling approaches (Newell and Simon,972), in animal learning studies that required the partici-ant animals to hold information across a number of se-uential trials (Olton, 1979), and in cognitive psychology,here it has been applied to a short-term memory (STM)tore (Atkinson and Shiffrin, 1968). But it was Baddeleynd Hitch (1974) who first explicitly explored the role ofemporary information storage as a common system that isnvolved in the performance of a wide range of complexognitive tasks, and with that started the scientific en-eavor of working memory research.

multidisciplinary research framework gets adopted

he study of human mind and brain has never been limitedo one scientific discipline. The subject has been tackled by

number of disciplines, each addressing the subject fromts specific point of view, using the methods, conceptualools and research paradigms that were developed andhaped through the history of the discipline and adjusted tohe specific questions it tried to answer. Though they werenitially difficult to relate to one another, the rise of cogni-ive paradigm and adoption of information processing ap-roach provided a much-needed common conceptual sys-em that enabled the findings of individual disciplines to beelated and combined. Cognitive psychologists were ableo build computational models and computer simulations ofheir theories. Neuropsychologists could use models oformal brain function to explain cognitive dysfunctionsfter brain damage and to plan appropriate rehabilitationrograms. Neuroscientists were able to use models of

unctional architecture to guide their research efforts in

fipcs

iactcpcna

asrhadwptttkiabnttbbtntf

T

Tiactrswtubitsp

T

Tomapdtfipqn

dimaopc

yramid a

G. Repovš and M. Bresjanac / Neuroscience 139 (2006) 1–32

nding relevant neuroanatomical structures and neuro-hysiological mechanisms, and at the same time provideognitive psychologists with important ideas and con-traints for their theories.

The existence of a bridge, however, does not ensurets actual or most efficient use. Even though combinationnd confrontation of evidence from different fields ofognitive neuroscience enabled some important break-hroughs, it is not always practiced or even readily ac-epted. A systematic research framework that would pro-ose a way to coordinate research efforts and encourageonsideration of findings from different fields of cognitiveeuroscience when developing theories of human brainnd mind might therefore be helpful.

Extending Kosslyn’s (1996) cognitive neuroscience tri-ngle framework we propose a pyramid approach to re-earch in cognitive neuroscience in which four lines ofesearch have to be considered when developing compre-ensive models of mind/brain (Fig. 1). First, the capacitiesnd properties that the model is to explain need to beescribed in detail and a functional architecture by virtue ofhich the mind/brain enables the person to have thoseroperties and capacities needs to be put forward. This ishe role of cognitive psychology. Second, working compu-ational models of cognitive functions that specify in detailhe proposed theoretical assumptions and are based onnown neurobiology of the brain need to be built. Thats the domain of computational science. Third, the actualnatomical structure and physiology of the brain need toe taken into account to be able to explain how the cog-itive functions are actually instantiated in the brain. That ishe purview of neuroscience. Fourth, cognitive dysfunc-ions following specific brain damage caused either byrain injury, neurological or psychiatrical disease, need toe studied to provide further constraints and tests of theheory. That is the realm of cognitive neuropsychology andeuropsychiatry. Each of the stated lines of inquiry con-ributes unique and necessary information. This is the

cognitive neurops

cognitive psychology

computationa

Fig. 1. The fields of research constituting the p

ramework we adopted as the backbone of the issue. a

he aims of the issue get explained

here are two aims that generated the idea of this specialssue and guided its development. First, the special issueims to provide an overview of the current research inognitive neuroscience of working memory, both in theerms of the recent findings, as well as in terms of theesearch methods being used. To present a comprehen-ive overview, the cognitive neuroscience pyramid frame-ork (Fig. 1) was followed. Second, since it is the integra-

ion of findings in a multidisciplinary framework that willltimately enable us to unravel the mysteries of the humanrain, and the mind that it gives rise to, the present special

ssue hopes to encourage a purposeful, systematic mul-idisciplinary approach to research in cognitive neuro-cience by illustrating it with the application of such ap-roach to working memory research.

he overview of the issue is presented

he issue is organized in three parts. The first part consistsf broad review papers that introduce the main theoreticalodels and views of working memory and some of thepproaches to the study of working memory. The secondart provides problem-oriented review papers that intro-uce individual research questions and possible answershat cognitive neuroscience offers today. The third andnal part of the issue gives space to a number of empiricalapers that further illustrate the variety of specific researchuestions being addressed and methods used by cognitiveeuroscientists.

The issue starts off with a paper by Repovš and Bad-eley that presents the current state of the arguably most

nfluential model of working memory, the multi-componentodel of working memory proposed initially by Baddeleynd Hitch (1974), and reviews recent empirical findingsffered by behavioral studies. The issue continues with aaper by Postle, arguing that rather than postulating spe-ialized brain subsystems that act as buffers for storing

logy / neuropsychiatry

nce

neuroscience

pproach to research in cognitive neuroscience.

ycho

l scie

nd manipulating information in working memory, working

mrt

aidsvpootcupitwhmtttSmft

coaacobtCatptcb

ctcwons(eL

sMsptitivimGmmRpdm

p(bMf

T

Dtmbphw“fm

A

A

B

K

M

N

OT

G. Repovš and M. Bresjanac / Neuroscience 139 (2006) 1–3 3

emory should be understood as a flexible system thatecruits individual brain subsystems as needed to performhe working memory tasks.

Six subsequent review papers introduce some of thepproaches used to study working memory, and their find-

ngs. Jarold and Towse show how the study of individualifferences can shed light on the structure of a cognitiveystem such as working memory. Mueller and Knight pro-ide examples of how human brain lesion studies canrovide important data for testing and developing modelsf working memory. Honey and Fletcher continue with anverview of advantages and pitfalls to consider when ex-

ending the brain lesion approach of cognitive neuropsy-hology to cognitive dysfunctions found in psychiatric pop-lation, specifically patients with schizophrenia, and showossible solutions such as the use of psychopharmacolog-

cal models of disease in healthy human subjects. Barchhen follows with a more detailed analysis of how studies oforking memory in schizophrenia compare with existinguman and non-human primate data and bear on theodels of working memory. Looking at another alternative

o brain lesion studies, Mottaghy introduces the use ofranscranial magnetic stimulation (TMS) as a method ofemporary virtual lesioning of the brain. Finally, a paper bychloesser, presenting structural equation modeling as aethod of assessing the brain network employed in per-

orming working memory tasks, concludes the first part ofhe issue.

The second part of the issue starts by focusing on theomputational approaches to the study of working mem-ry. A paper by Hazy et al. demonstrates how importantdvances in understanding cognitive systems can bechieved by using computational modeling to specifyomputational and neural mechanisms involved in itsperation. While the first paper focuses on how largerrain systems interact in performing complex cognitiveasks, the following papers, by Durstewitz and Seamans,ompte, and Tanaka explore how neurocellular propertiesnd their interaction in neural networks translate to func-ional characteristic of the nervous system enabling com-utational mechanisms that give rise to cognition. Takenogether these papers provide an exciting view of howomputational modeling can build bridges between neuro-iology and cognition.

The remaining papers of the second part focus on un-overing individual neural mechanisms and representationshat jointly form working memory. Reflecting its popularity inognitive neuroscience research due to relative ease withhich it relates cognitive performance to brain activity, mostf the papers in this section are based on functional mag-etic resonance imaging (fMRI) applied to research onpatial working memory (Curtis), proactive interferenceJonides and Nee), coding of order (Marshuetz and Smith),xecutive functions (Collette, Hogge, Salmon and Van del

inden), and delayed task performance (Rypma). The

trength of electro- and magnetoencephalography (EEG,EG) in elucidating brain activity and large-scale neuronal

ynchronization on a finer time-scale is demonstrated in aaper by Jensen, exploring the use of temporal segmen-ation as a possible mechanism of maintenance of multipletems in working memory. Funahashi continues with elec-rophysiology, taking it farther to multiple single cell record-ngs in monkeys, enabling the exploration of populationectors of activity, functional interaction among neighbor-

ng neurones and their dynamic modulation in workingemory tasks. Turning the focus to neuropharmacology,raham and Castner explore the fine balance between theultiple mechanisms of dopamine modulation in workingemory. The second part concludes with a paper byanangath that integrates findings from multiple ap-roaches in a true multidisciplinary manner to explore howifferent cortical areas interact to form and maintain visualemories.

The issue closes with a wide variety of empirical pa-ers covering issues from cross-modal working memoryOhara, Lenz and Zhou) to modulation of working memoryy social and affective factors (Park, Gibson and Mc-ichael), employing a variety of methods including MEG,

MRI, neuropharmacology and behavioral studies.

he guest-editors bid farewell

esigning and editing this special issue has been an in-eresting journey and learning experience. If anything, itade evident that the extent and diversity of researcheing conducted in working memory make it impossible toresent it comprehensively in a single journal issue. We doope, however, that it provides the Neuroscience readersith a broad sketch of the area, offers them a chance to

dip their toes” in the subject, and encourages them tourther explore the multidisciplinary research in workingemory.

REFERENCES

tkinson AP (1998) Wholes and their parts in cognitive psychology:Systems, subsystems, and persons. http://cogprints.org/337/.

tkinson RC, Shiffrin RM (1968) Human memory: A proposed systemand its control processes. In: The psychology of learning andmotivation: Advances in research and theory (Spence KW, ed)pp 89–195. New York: Academic Press.

addeley AD, Hitch GJ (1974) Working memory. In: Recent advancesin learning and motivation, Vol. 8. (Bower GA, ed), pp 47–90. NewYork: Academic Press.

osslyn SH (1996) Image and brain: The resolution of the imagerydebate. Cambridge: MIT Press.

iller GA, Galanter E, Pribram KH (1960) Plans and the structure ofbehavior. New York: Holt, Rinehart & Winston.

ewell A, Simon HA (1972) Human problem solving. Englewood Cliffs,NJ: Prentice Hall.

lton DS (1979) Mazes, maps and memory. Am Psychol 34:583–596.olman EC (1948) Cognitive maps in rats and man. Psychol Rev

55:189–208.

(Accepted 12 December 2005)(Available online 19 January 2006)