**1. Introduction**

248 Neuroimaging – Cognitive and Clinical Neuroscience

Rahman, F, Smith K, Bullitt E, Marks B. (2008). Relationships between exercise and cerebral

Raji CA, Ho AJ, Parikshak NN, Becker JT, Lopez OL, Kuller LH, Hua X, Leow AD, Toga

Ross JS, Tkach J, Ruggieri PM, Lieber M, Lapresto E. (2003). The Mind's Eye: Functional MR

Scheibel AB. (2009). Aging of the Brain, In: *Handbook of the Neuroscience of Aging, Ed: PR Hoff & CV Mobbs*, pp. 5-9, Academic Press, ISBN: 978-0-12-374898-0, New York:NY Soreca I, Rosana C, Jennings R, et al. (2009). Gain in adiposity across 15 years is associated with reduced gray matter volume in healthy women. *Psychosom Med,* 71:485-490 Spirduso WW, Francis KL, MacRae PG. (2005). *Physical Dimensions of Aging*, 2nd Edition.

Stebbins GT, Murphy CM. (2009). Diffusion tensor imaging in Alzheimer's disease and mild

US Department of Health and Human Services (USDHHS). 2008 Physical Activity Guideline

US Department of Health and Human Services (USDHHS). National Institutes of Health

Ward M, Carlsson CM, Trivedi MA, Sager MA, Johnson SC. (2005). The effect of body mass

 In collection MIDAS/National Alliance for Medical Image Computing (NAMIC)/ NAMIC: Public Data Repository; NIH Neuroscience Roadmap Initiative van Praag H, Christie BR, Sejnowski TJ, Gage FH. (1999). Running enhances neurogenesis, learning, and long-term potentiation in mice. *Proc Natl Acad Sci,* 96:13427-13431

index on global brain volume in middle-aged adults: a cross-sectional study. *BMC Neurol*, 5:23 doi:10.1186/1471-2377-5-23. [Published Online 2005 December 2] Vachet C, Bullitt E, Katz L, Marks B, Davis B, Styner M. (2009). UNC Elderly Brain Atlas,

http://projectreporter.nih.gov/reporter.cfm) Last accessed April 25, 2011

Published Online First: 6 August 2009; doi: 10.1002/hbm.20870

Orleans, May 31 – June 5. Abstract ID: 6009535

24:1036-1044

Human Kinetics, Champaigne:IL

cognitive impairment. *Behav Neurol*, 21(1):39-49

for Americans. ODPHP Pub. No. U0036, October 2008 www.health.gov/paguidleines last accessed April 23, 2011

Research Portfolio Online Reporting Tools (RePORT)

http://www.insight-journal.org/midas/item/view/2330 and

http://www.insight-journal.org/midas/gallery

blood flow in older adults. American Society of Neuroradiology Conference, New

AW, Thompsom PM. (2009). Brain structure and obesity. *Human Brain Mapp*,

imaging evaluation of golf motor imagery. *American Journal of Neuroradiology,* 

We can flexibly process and make decisions regarding multiple types of information in daily situations such as driving and cooking. However, human error is increased in complex or combined tasks (relative to simple tasks) because our information processing capacity is limited. This limited cognitive function is associated with working memory (WM), which is proposed to be a higher-level human ability to memorize, maintain, and manipulate mental representations in the mind for a short time (Baddeley, 1986). Most theorists think that WM function includes active manipulation as well as passive short-term maintenance. An oftenused metaphor for working memory is "the blackboard of the mind." For example, imagine that you are rearranging the furniture in your room. You can move around the furniture in your mind, that is, transform the imagination any number of times. To guide behavior and make decisions about what to do next, WM temporarily selects and retains task-relevant information such as recently processed sensory input, retrieved information from long-term memory, or mentally manipulated images. Thus, WM is directly linked to any and all other brain functions, including perception, movement, emotion, and problem solving.

Baddeley & Hitch (1974) proposed a basic psychological model in which WM is divided into separate components, the "storage system" and the "central executive". The "storage system" consists of 2 temporary storage buffers for visual information (visuospatial sketch pad, i.e., visual working memory) and auditory-verbal information (phonological loop, i.e., verbal working memory) and an episodic buffer for long-term memory, whereas the "central executive" controls the allocations of attention, selects relevant information, and manipulates information held in the storage systems (Baddeley, 1986; Baddeley & Hitch, 1974; Phillips, 1974; Baddeley, 2000). Extensive experimental evidence from behavioral performance of normal subjects, lesion studies, and neuroimaging studies supports this view. For example, performance in dual tasks requiring 2 separate perceptual domains (i.e., a visual and a verbal task, or a mental processing task and a maintenance task) is nearly as efficient as performance of individual tasks (for a review, Cowan, 2001; Della Sala & Logie, 1993). These findings indicate that the visual and verbal WM are separated.

Both visual and verbal WM have 3 phases: encoding, which imports the relevant information in memory; maintenance, which stores the encoded information; and retrieval (or rehearsal), which briefly uses the information for a task. To investigate the neural substrate for WM, previous electrophysiological studies in nonhuman primates and human

Human Oscillatory EEG Activities Representing Working Memory Capacity 251

over retention intervals. The number of items within the sample display is manipulated. Following the retention interval, one probe item (test display) or multiple probe items (whole display) are presented at one location within the sample array, and participants are then required to judge whether a change has occurred or not. These 2 tests have shown different performance scores, since VWM storage capacity is vulnerable to visual interference created during the encoding period (Wheeler & Treisman, 2002). Therefore, many behavioral and neuroimaging studies have applied the single-probe test. To avoid the possibility of using verbal strategies, most studies involving the DMS task used very short exposure duration for the sample display (about 150 ms), and require participants to engage in phonological tasks simltaneously, e.g., repeating a word during the sample display and

Many previous studies have proposed a VWM capacity of 3 or 4 items (Luck & Vogel, 1997) because the accuracy rates for many DMS tasks systematically decrease as the number of items increases beyond 3 or 4. More recently, one study demonstrated that VWM capacity decreases as object complexity increases, and proposed that VWM capacity varies by the type of features (Alvarez & Cavanagh, 2004). The authors used complex items, Chinese characters, which are thought to be a combination of simple shapes. Although the issue retains some controversy, many studies have demonstrated consensus on the existence of

To estimate the capacity of VWM in terms of objects stored in DMS tasks, Cowan (2001) has proposed a model that takes both hit rates (accurately detecting a change) and correct rejection rates (accurately reporting no change when none occurred) into account. The

*K NK* ( ) *H g N N*

( ) (1 ) *K NK CR <sup>g</sup> N N*

where *K* denotes the estimated number of items stored in VWM, *N* is the total number of items presented in the sample display, *H* is the probability of a hit rate, *CR* is the probability of a correct rejection rate, and *g* is the guessing rate for coincidentally giving a correct answer. The theory assumes that when one of the items within the VWM capacity (*K*/*N*; Fig. 1 purple area) changed, subjects could detect whether the change occurred. In contrast, they could not detect whether a change occurred in objects exceeding the capacity ((*N – K)*/*N*;

However, in some cases subjects happened to answer correctly on some portion of the trials (*g*) under an alternative forced-choice paradigm or, in another portion of the trials (1 *– g*), coincidently report correctly that no change occurred in the no-change trial, although they could not detect this. This guessing rate could not be estimated from the performance of the DMS tasks. Thus, given the hit rates and correct rejection rates for a particular set size, these

(1)

(2)

*K N H CR* ( 1) (3)

model estimates hit rates and correct rejection rates with the following equations:

retention intervals (Baddeley, Lewis & Vallar, 1984).

large individual differences in VWM capacity.

equations (1) and (2) can be solved for the set size:

Fig. 1 green area).

neuroimaging studies have shown sustained neural activity over the retention interval in distributed brain regions including frontal, parietal, occipital, and temporal areas during maintenance of relevant information (e.g., Chafee & Godman-Rakic, 1998). If these brain regions are actually involved in maintaining mental representations, their activities are thought to be correlated with WM capacity. In fact, brain activity has been reported to increase with increasing number of objects to be remembered and saturated below the limited WM capacity (Todd & Marois, 2004; Vogel & Machizawa, 2004). Frontal regions also represent the limitation of executive functions, since activity there is increased during engagement in dual tasks (Marois & Ivanoff, 2005). These results suggest that frontal regions are associated with executive functions and posterior regions are involved in maintenance of mental representations. Thus, although much is known concerning the brain areas involved in various WM functions, understanding how these brain areas temporally communicate is more difficult.

To address this issue, measuring electrophysiological (EEG) data during WM tasks and analyzing the synchronizations in local areas and between different areas has proved particularly useful (Varela et al., 2001). Our previous EEG studies used mental calculation as the auditory WM task and mental spatial manipulation as the visual WM task (Kawasaki et al., 2010). The EEG results clearly demonstrated that the frontal theta (4–6 Hz) activity increased during the manipulation periods on both WM tasks, and the parietal and temporal alpha activities were enhanced only during the maintenance periods on the auditory and visual WM task, respectively. Phase synchronization analysis revealed significant theta synchronizations between the frontal and parietal regions for visual WM and between the frontal and temporal regions for auditory WM. These results indicated that long-range theta synchronizations could connect the different brain regions to manipulate task-relevant representations. Interestingly, the concurrent theta and alpha phases were significantly synchronized in task-relevant storage areas, which suggests the presence of gating mechanisms to extract stored information. Theta and alpha activities thus play an important role in several WM functions; however little is known regarding how these oscillations represent WM limitations.

This chapter describes investigations into the neural dynamics of EEG oscillatory activities that underlie the capacity limitations for executive functions and storage buffers in WM, particularly for visual infomation. To advance understanding of the detailed brain networks involved, the use and interpretation of EEG time-frequency analyses such as wavelet analysis and the role of each EEG oscillatory activity in WM functions is discussed, and 2 experiments are described. Visual storage systems were investigated using delayedmatching-to-sample tasks with visual stimuli, and a dual WM task with visual and auditory representations was used to identify the bottleneck of the central executive function. These EEG findings may contribute to understanding the causes of human error.

#### **2. Capacity limitations of working memory**

To investigate the limitation of visual WM (VWM) storage capacity, previous behavioral and neuroimaging studies used a change detection paradigm, namely, delayed matching to sample (DMS) tasks with a visual stimulus. In this paradim, multiple visual items are presented (sample display) and participants are required to memorize and retain these items

neuroimaging studies have shown sustained neural activity over the retention interval in distributed brain regions including frontal, parietal, occipital, and temporal areas during maintenance of relevant information (e.g., Chafee & Godman-Rakic, 1998). If these brain regions are actually involved in maintaining mental representations, their activities are thought to be correlated with WM capacity. In fact, brain activity has been reported to increase with increasing number of objects to be remembered and saturated below the limited WM capacity (Todd & Marois, 2004; Vogel & Machizawa, 2004). Frontal regions also represent the limitation of executive functions, since activity there is increased during engagement in dual tasks (Marois & Ivanoff, 2005). These results suggest that frontal regions are associated with executive functions and posterior regions are involved in maintenance of mental representations. Thus, although much is known concerning the brain areas involved in various WM functions, understanding how these brain areas temporally communicate is

To address this issue, measuring electrophysiological (EEG) data during WM tasks and analyzing the synchronizations in local areas and between different areas has proved particularly useful (Varela et al., 2001). Our previous EEG studies used mental calculation as the auditory WM task and mental spatial manipulation as the visual WM task (Kawasaki et al., 2010). The EEG results clearly demonstrated that the frontal theta (4–6 Hz) activity increased during the manipulation periods on both WM tasks, and the parietal and temporal alpha activities were enhanced only during the maintenance periods on the auditory and visual WM task, respectively. Phase synchronization analysis revealed significant theta synchronizations between the frontal and parietal regions for visual WM and between the frontal and temporal regions for auditory WM. These results indicated that long-range theta synchronizations could connect the different brain regions to manipulate task-relevant representations. Interestingly, the concurrent theta and alpha phases were significantly synchronized in task-relevant storage areas, which suggests the presence of gating mechanisms to extract stored information. Theta and alpha activities thus play an important role in several WM functions; however little is known regarding how these

This chapter describes investigations into the neural dynamics of EEG oscillatory activities that underlie the capacity limitations for executive functions and storage buffers in WM, particularly for visual infomation. To advance understanding of the detailed brain networks involved, the use and interpretation of EEG time-frequency analyses such as wavelet analysis and the role of each EEG oscillatory activity in WM functions is discussed, and 2 experiments are described. Visual storage systems were investigated using delayedmatching-to-sample tasks with visual stimuli, and a dual WM task with visual and auditory representations was used to identify the bottleneck of the central executive function. These

To investigate the limitation of visual WM (VWM) storage capacity, previous behavioral and neuroimaging studies used a change detection paradigm, namely, delayed matching to sample (DMS) tasks with a visual stimulus. In this paradim, multiple visual items are presented (sample display) and participants are required to memorize and retain these items

EEG findings may contribute to understanding the causes of human error.

more difficult.

oscillations represent WM limitations.

**2. Capacity limitations of working memory** 

over retention intervals. The number of items within the sample display is manipulated. Following the retention interval, one probe item (test display) or multiple probe items (whole display) are presented at one location within the sample array, and participants are then required to judge whether a change has occurred or not. These 2 tests have shown different performance scores, since VWM storage capacity is vulnerable to visual interference created during the encoding period (Wheeler & Treisman, 2002). Therefore, many behavioral and neuroimaging studies have applied the single-probe test. To avoid the possibility of using verbal strategies, most studies involving the DMS task used very short exposure duration for the sample display (about 150 ms), and require participants to engage in phonological tasks simltaneously, e.g., repeating a word during the sample display and retention intervals (Baddeley, Lewis & Vallar, 1984).

Many previous studies have proposed a VWM capacity of 3 or 4 items (Luck & Vogel, 1997) because the accuracy rates for many DMS tasks systematically decrease as the number of items increases beyond 3 or 4. More recently, one study demonstrated that VWM capacity decreases as object complexity increases, and proposed that VWM capacity varies by the type of features (Alvarez & Cavanagh, 2004). The authors used complex items, Chinese characters, which are thought to be a combination of simple shapes. Although the issue retains some controversy, many studies have demonstrated consensus on the existence of large individual differences in VWM capacity.

To estimate the capacity of VWM in terms of objects stored in DMS tasks, Cowan (2001) has proposed a model that takes both hit rates (accurately detecting a change) and correct rejection rates (accurately reporting no change when none occurred) into account. The model estimates hit rates and correct rejection rates with the following equations:

$$H = \frac{K}{N} + \frac{(N-K)}{N} \times g \tag{1}$$

$$CR = \frac{K}{N} + \frac{(N-K)}{N} \times (1-g) \tag{2}$$

where *K* denotes the estimated number of items stored in VWM, *N* is the total number of items presented in the sample display, *H* is the probability of a hit rate, *CR* is the probability of a correct rejection rate, and *g* is the guessing rate for coincidentally giving a correct answer. The theory assumes that when one of the items within the VWM capacity (*K*/*N*; Fig. 1 purple area) changed, subjects could detect whether the change occurred. In contrast, they could not detect whether a change occurred in objects exceeding the capacity ((*N – K)*/*N*; Fig. 1 green area).

However, in some cases subjects happened to answer correctly on some portion of the trials (*g*) under an alternative forced-choice paradigm or, in another portion of the trials (1 *– g*), coincidently report correctly that no change occurred in the no-change trial, although they could not detect this. This guessing rate could not be estimated from the performance of the DMS tasks. Thus, given the hit rates and correct rejection rates for a particular set size, these equations (1) and (2) can be solved for the set size:

$$K = N \times (H + CR - 1)\tag{3}$$

Human Oscillatory EEG Activities Representing Working Memory Capacity 253

may reflect the representation of multiple items to be maintained, and have indeed shown that a subset of the distributed network demonstrated delay-period activity sensitive to the number of items in the sample display (Diwadkar et al., 2000; Glahn et al., 2002; Jha & McCarthey 2000; Linden et al., 2003). The VWM load-sensitive network includes the frontal, parietal, and visual cortices. Notably, some studies have revealed that activity in the posterior parietal cortex is correlated with the number of items to be remembered (Cowan's *K* value) and indicated that this area actually stored the representations (Kawasaki et al.,

In contrast to the posterior parietal and visual cortices, anterior regions including the frontal cortex have also been associated with executive processes such as attentional selection and manipulation of information (Curtis & D'Esposito, 2003). For instance, in studies using a spatial WM task that requires participants to memorize the spatial locations of simultaneous or sequentially presented items and, after a delay, select one relevant location, the prefrontal cortex has been reported to show transient activity during the selection period and no sustained activity during the retention interval (Rowe et al., 2000). Furthermore, the frontal cortex is particularly sensitive to the number of listed items to be maintained in VWM in the n-back task, which requires participants to maintain a series of items and their order, select a relevant item from VWM, and compare it with the earlier item (Smith & Jonides, 1999). Moreover, the frontal cortex is proposed to serve in maintaining task-specific goals (Miller & Cohen, 2001; Passingham & Sakai, 2004) and assist in maintaining high loads and/or long

Although, thus far, many neuroimaging studies have identified the neural substrate for the storage systems and central executive of WM, they have not dealt with how these brain areas temporally communicate. To address this issue, some studies have investigated the dynamic relationships governing brain activity by focusing on electroencephalograph (EEG) oscillations, which are closely related to synchronization of a large number of neurons underlying a particular function (Varela et al., 2001). Previous human scalp-recorded EEG studies have revealed modulated theta (about 4–8 Hz) and alpha (about 9–12 Hz) rhythms in distributed brain regions and phase synchronization between them during various WM tasks (Jensen & Tesche, 2002; Kawasaki & Watanabe, 2007; Klimesch et al., 2008; Mizuhara et al., 2004; Sauseng et al., 2005). Frontal theta activity in particular has been associated with the mental manipulation of WM, because these oscillations were enhanced in tasks such as mental calculation and image transfomation (Kawasaki et al., 2010). In contrast, posterior alpha activities are thought to be involved in the WM storage systems, because these oscillations are mainly observed in the retention intervals of many WM tasks. However, whether these oscillatory activities are increased or decreased during each WM period remains controversial. Furthermore, little is known regarding how these oscillations represent WM limitations; therefore, their detailed mechanisms have not yet been identified. To clarify the functional role of the theta and alpha oscillations in WM, the study described in the following 2 sections used EEG data measured during DMS and dual WM tasks to demonstrate 2 types of EEG activity that were correlated with the WM capacities for visual

This section describes the investigation of EEG oscillatory activity correlated with VWM capacity, which aimed to identify the roles of different oscillations in the VWM storage

2008; Todd & Marois, 2004, 2005; Vogel & Machizawa, 2004; Xu & Chun, 2006).

retention intervals (Braver et al., 1997; Linden et al., 2003).

storage and central executive systems.

**4. EEG oscillations for visual storage capacity** 

Participants' response VWM capacity

The Cowan's *K* value is obtained from the set size of each sample display as each subject's VWM capacity for a given material.

Fig. 1. Combination of participants' response and trial type (change or not) in change detection paradigm (left) and a model of Cowan's formula (right).

Unlike VWM, the WM capacity for executive functions has been evaluated using dual WM tasks. Although no interference exists between independent storage components such as visual and verbal storage, simultaneously processing more than 2 information sources that require mental manipulation and reactions is thought to be difficult. Previous studies have revealed a psychological refractory period, in which a second task elicits a longer reaction time when the interval between the first and second tasks (i.e., stimulus onset asynchrony; SOA) is short (Marois & Ivanoff, 2005). That is to say, if the 2 tasks seem to be processed simultaneously, the performance is degraded. This phenomenon is known to be a bottleneck of the central executive function.
