Sie sind auf Seite 1von 17

Psychology and Aging 2009, Vol. 24, No. 4, 792– 808

© 2009 American Psychological Association 0882-7974/09/$12.00 DOI: 10.1037/a0017798

Neural Underpinnings of Within-Person Variability in Cognitive Functioning

Stuart W. S. MacDonald

University of Victoria and Karolinska Institutet

Lars Ba¨ckman

Karolinska Institutet

Shu-Chen Li

Max Planck Institute for Human Development

Increased intraindividual variability (IIV), reflecting within-person fluctuations in behavioral perfor- mance, is commonly observed in aging as well as in select disorders including traumatic brain injury, schizophrenia, attention-deficit hyperactivity disorder (ADHD), and dementia. Much recent progress has been made toward understanding the functional significance of IIV in cognitive performance (MacDonald, Nyberg, & Ba¨ckman, 2006) and biological information processing (Stein, Gossen, & Jones 2005), with parallel efforts devoted to investigating the links between older adults’ deficient neuromodulation and their more variable neuronal and cognitive functions (Ba¨ckman, Nyberg, Lindenberger, Li, & Farde, 2006). Despite these advances in the study of IIV, there has been little empirical examination of underlying neural correlates and virtually no synthesis of extant findings. The present review summarizes the accumulating empirical evidence linking age-related increases in IIV in cognitive performance to neural correlates at anatomical, functional, neuromodulatory, and genetic levels. Computational theories of neural dynamics (e.g., Li, Lindenberger, & Sikstro¨m, 2001) are also introduced to illustrate how age-related neuromodulatory deficiencies may contribute to increased neuronal noise and render infor- mation processing in aging neurocognitive systems to be less robust. The potential benefits of stochastic resonance and external noise are also discussed with respect to processing subthreshold stimuli (e.g., Li, von Oertzen, & Lindenberger, 2006). We conclude by highlighting important challenges and outstanding research issues that remain to be answered in the study of IIV.

Keywords: cognition, aging, within-person variability, brain, neuromodulation

Most cognitive and neuropsychological research has focused on mean differences in performance across persons. This orientation has largely overshadowed research on performance variability within persons. This is a serious theoretical and practical oversight because, given large processing fluctuations, behavioral perfor- mance assessed from a single occasion is less accurate and repre- sentative. To the extent that variability in performance is small, mean-level differences provide useful predictive information. However, as within-person variability increases and represents systematic as opposed to random sources of error, performance

Stuart W. S. MacDonald, Department of Psychology, University of Victoria, Victoria, British Columbia, Canada, and Aging Research Center, Karolinska Institutet, Stockholm, Sweden; Shu-Chen Li, Center for Life- span Psychology, Max Planck Institute for Human Development, Berlin, Germany; Lars Ba¨ckman, Aging Research Center, Karolinska Institutet. This research was supported by grants from the Swedish Research Council and Swedish Brain Power to Lars Ba¨ckman, who was also supported by an Alexander von Humboldt Research Award. Shu-Chen Li was supported by grants from the German Ministry of Education and Research and the Max Planck Society, and Stuart MacDonald was supported by a Scholar Career Investigator Award from the Michael Smith Foundation for Health Research. Correspondence concerning this article should be addressed to Stuart W. S. MacDonald, Department of Psychology, University of Victoria, P.O. Box 3050 STN CSC, Victoria BC V8W 3P5, Canada. E-mail:


indexed from a single measurement occasion can result in flawed estimates of mean-group differences with implications for both theory and practice (Hultsch & MacDonald, 2004; Hultsch, Strauss, Hunter, & MacDonald, 2008; Stuss & Binns, 2008). On a related note, when within-person variability is large, it poses considerable challenges for the estimation of long-term cognitive change (e.g., Salthouse, 2007). Further, systematic and sizeable variability across measurements raises questions about the foundations of classical test theory and the utility of true score values (Lindenberger & von Oertzen, 2006). Thus, an exclusive emphasis on mean levels in the absence of considering variability represents a critical oversimplification of patterns of behavior that may lead to erroneous inference (Nesselroade, 2002). Recognition of these shortcomings has given rise to models (e.g., diffusion models) that explain per- formance based on multiple distribution parameters, not just central tendency (e.g., Ratcliff, Van Zandt, & McKoon, 1999). Inclusion of variance parameters confers unique predictive in- formation about cognitive functioning independent of mean perfor- mance (e.g., West, Murphy, Armilio, Craik, & Stuss, 2002), with specific analyses of variability facilitating discrimination among groups along multiple dimensions, such as patients with dementia versus healthy controls or patients with dementia subtypes (Hultsch, MacDonald, Hunter, Levy-Bencheton, & Strauss, 2000; Murtha, Cismaru, Waechter, & Chertkow, 2002).



The recent impetus toward examining variability in cognitive

functioning, as both outcome and predictor, derives from a grow- ing literature showing performance fluctuations that (a) are neither due to random error nor measurement unreliability, (b) often share

a negative association with mean cognitive performance (e.g.,

Hultsch & MacDonald, 2004; Hultsch et al., 2008), and (c) are linked to theories explicating processing dynamics (e.g., Ratcliff

et al., 1999) or the roles and potential neuronal sources of noise in

affecting information processing (e.g., Li, Lindenberger, & Sik- stro¨m, 2001). Although the stability perspective has dominated developmental research for years (Nesselroade, 2002), the current focus on variability is not new (see Cattell, 1957; Fisk & Rice, 1955; Head, 1926; Wohlwill, 1973). For example, intraindividual variability was a primary influence in formulating the state–trait anxiety distinction (Cattell & Scheier, 1961). The renewed vigor for the study of variability likely reflects improved means and methods for exploring performance fluctuations (Nesselroade &

Salthouse, 2004). Within this literature, three primary indices of variability have been operationalized vis-a`-vis the minimum con- ditions necessary to observe it with reference to a three- dimensional data box defined by persons, measures, and occasions (Cattell, 1966; Hultsch, MacDonald, & Dixon, 2002). Between persons, variability has been operationalized as differences in performance measured on a single task on a single occasion (referred to as individual differences). Within persons, variability has been operationalized in two ways. The first concerns variability associated with measuring

a single person on a single occasion across multiple tasks

(referred to as intraindividual differences ). For example, Holtzer, Verghese, Wang, Hall, and Lipton (2008) indexed this form of within-person variability across a battery of neuropsy- chological tests to predict risk of incident dementia. The second form of within-person variability is defined by the minimum condition of measuring a single person on a single task across multiple occasions (referred to as intraindividual variability or IIV ). Such intraindividual fluctuations in performance have been indexed across multiple trials within a session or across multiple sessions spanning longer intervals (e.g., hours, days, weeks) for a single measure (Li, Huxhold, & Schmiedek, 2004).

Analytical Perspectives in the Study of Variability

Developmentalists and aging theorists alike are inherently in- terested in change over time, and thus variability within persons and across occasions (i.e., IIV) has garnered particular attention for researchers in these fields. To be sure, a thorough understanding of the wider field of variability research requires an appreciation of the distinctions between several analytical perspectives. To help orient the reader, we briefly will introduce how time scales, task characteristics, and intraindividual dynamics are key factors to consider in any operationalization of within-person variability.

Time Scales

One dichotomy that is fundamental to the study of variability concerns differentiating ontogenetic versus microgenetic forms of within-person change according to two criteria: permanence of change and time scale (Nesselroade, 1991). Intraindividual change represents a relatively slow and enduring process that unfolds

across months, years, or decades (e.g., progressive changes asso- ciated with development and aging, long-term learning, and skill acquisition) and has been referred to as becoming (Ford, 1987) or developing (Li, Huxhold, & Schmiedek, 2004). In contrast, IIV represents relatively rapid and transient short-term change indexed across minutes, days, or weeks for various types of behavior (e.g., shifts in emotional state, variability in cognitive accuracy or pro- cessing speed, fluctuations in physical performance). The shorter time frame is said to reflect being (Ford, 1987) or functioning (Li, Huxhold, & Schmiedek, 2004), with indices of IIV variously referred to as inconsistency, lack of processing robustness, wobble, and lability (for a review, see Hultsch et al., 2008). IIV can be distinguished from more enduring intraindividual change. Al- though both variants of within-person change represent important developmental phenomena, recent research has been focused on IIV as a common component of aging-related cognitive decline, as well as in relation to behavioral changes associated with neurode- generative diseases (e.g., Alzheimer’s disease) and other brain- related disorders (e.g., traumatic brain injury, schizophrenia). Con- sidering the diverse populations that exhibit increasing IIV and concomitant cognitive deficits, more variable cognitive function- ing likely reflects a behavioral proxy for endogenous neural changes underlying impairment. The time scale for measuring IIV itself is also important. For example, IIV indexed from moment to moment (e.g., fluctuations in reaction time [RT] trials) versus day to day or week to week (e.g., fluctuations in mood) likely reflects different underlying sources (e.g., Martin & Hofer, 2004; Rabbitt, Osman, Moore, & Stollery, 2001; Ram, Rabbitt, Stollery, & Nesselroade, 2005; Ro¨cke, Li, & Smith, in press). In several studies, IIV has even been examined across different retest intervals for the same task in the same sample. Rabbitt and colleagues (2001) found that RT vari- ability from trial to trial versus week to week was related, although individual differences in within-session variability did not account for all the variance in between-session variability. Similarly, Hultsch and colleagues (Fuentes, Hunter, Strauss, & Hultsch, 2001; Hultsch et al., 2000) reported that the magnitude of IIV across trials is approximately twice as large as IIV across weeks. If a theoretical model assumes that endogenous sources underlie cognitive variability (e.g., neural correlates such as changes in the efficiency of neurotransmitters), then IIV may be better captured over short intervals (e.g., trial-to-trial fluctuations in RT tasks). In contrast, exogenous modulators of variability (e.g., fatigue, per- ceived stress) may be better indexed over days or weeks.

Task Characteristics

Recent reviews have chiefly focused on the negative association between IIV and cognitive performance (e.g., Hultsch et al., 2008; Luszcz, 2004; MacDonald, Nyberg, & Ba¨ckman, 2006). However, variability is not always a harbinger of maladaptive functioning; rather, variability can be adaptive depending on specific task characteristics. For example, performance variability in the child development literature is related to cognitive development (not impairment), with such variability reflecting diverse strategy ex- ploration for complex tasks. Children who try a variety of strate- gies for complex tasks exhibit greater success. In contrast, more constrained tasks (including many of the response latency mea- sures examined in the cognitive aging literature on variability) are



less amenable to multiple response strategies, and varying strate- gies to such tasks could be maladaptive. Siegler (1994) has re- ferred to such fluctuations in performance as the ebbing and flowing of new and old ways of thinking. Similar positive associ- ations between increased IIV and cognitive functioning have been reported for older adults. Allaire and Marsiske (2005), for exam- ple, computed IIV across performance accuracy trials for three cognitive tasks that permitted implementation of new performance strategies with practice. Results indicated that increased IIV was positively correlated with level of task performance and the mag- nitude of practice-related gains. In contrast, IIV estimates derived from cognitive tasks that provide little opportunity for strategic processing or practice-related gains (e.g., sensorimotor or percep- tual processing speed) are typically associated with maladaptive outcomes (e.g., Hultsch & MacDonald, 2004, Hultsch et al., 2008). Furthermore, depending on the domains of functioning (e.g., cog- nitive vis a´ vis affective), aging is not always associated with increasing within-person variability. For instance, recently Ro¨cke et al. (in press) reported that older adults exhibited less daily fluctuations in positive and negative affect than young adults.

Subtypes of Variability

It is important to emphasize that variation could be operation- alized in numerous forms, even for the most basic cognitive information processing tasks. Indices of variability can be indexed across trials versus sessions or across all response latency trials versus only correct responses and could even be examined in the

drift rate or diffusion coefficient across or within trials (cf. Ratcliff et al., 1999). To better understand the characteristic nature of variability (e.g., adaptive vs. maladaptive), Li, Huxhold, and Schmiedek (2004) proposed a taxonomy of intraindividual dynamics, reflecting vari- ations in an individual’s processes or performance over time that are due to both endogenous and exogenous influences. For exam- ple, the classification of intraindividual dynamics allows for dis- tinctions to be drawn between adaptive versus maladaptive types of variability (see Figure 1). Adaptive forms of variability are typically observed to the extent that a task is amenable to strategy use. These adaptive subtypes include plasticity (the functional malleability to obtain large learning gains following task expo- sure), diversity (exploratory behaviors and various strategies used for performing a complex task), and adaptability (ability to quickly recover peak functioning in the face of challenging task condi- tions). In contrast, continued fluctuation in performance subse- quent to mastering a given level of functioning indicates a lack of processing robustness, defined by an increased ebb and flow in processing, and diminished stability in performance over brief intervals (Li, Huxhold, & Schmiedek, 2004, Li, Lindenberger, et al., 2004). It is this form of maladaptive IIV that characterizes virtually all findings reported in the present review.

The Present Review

For the purposes of the present review, we focus on within-task IIV reflecting intraindividual change in behavior that is relatively

intraindividual change in behavior that is relatively Figure 1. Varieties of intraindividual dynamics. Adaptive

Figure 1. Varieties of intraindividual dynamics. Adaptive forms of variability are typically observed for tasks amenable to strategy use. Adaptive subtypes include plasticity (the ability to exhibit learning gains following task exposure), diversity (exploratory strategy use when performing a complex task), and adaptability (capacity to quickly recover peak functioning despite challenging task conditions). As a maladaptive form of variability, lack of processing robustness reflects continued performance fluctuations and diminished stability subsequent to mastering a given level of functioning. This form of maladaptive variability characterizes most findings reported in the present review. From “Aging and Processing Robustness: Evidence From Cognitive and Sensorimotor Functioning,” by S.-C. Li, O. Huxhold, and F. Schmiedek, 2004, Gerontology, 50, p. 30. Copyright 2004 by Karger. Adapted with permission.



rapid and transient (e.g., trial-to-trial fluctuations on RT tasks). Irrespective of analytical perspective (e.g., task type, time scale, behavior vs. brain), most of the empirical studies reviewed indexed

a maladaptive form of IIV, largely derived from within-person

standard deviations (or related measures) computed across correct (i.e., accurate) trials of response latency tasks for various cognitive outcomes—a form of compromised processing robustness accord- ing to the Li, Huxhold, and Schmiedek (2004) taxonomy. Such IIV increases characterize performance of older adults as well as populations with various conditions including dementia, head in- jury, schizophrenia, and attention-deficit hyperactivity disorder (ADHD). Although select evidence substantiates the hypothesized neural underpinnings of increased IIV from both behavioral (Hultsch et al., 2008) and neuroimaging (Hedden & Gabrieli, 2004; MacDonald, Nyberg, & Ba¨ckman, 2006; Winterer et al., 2004) perspectives, there has been little examination of underlying neural correlates and virtually no synthesis of extant findings. We begin by reviewing variability findings from the life span litera- ture, as well as discussing potential origins of IIV. Next, we summarize evidence linking age-related increases in IIV for cog- nitive functioning to neural correlates at anatomical, functional, neuromodulatory, and genetic levels. Although these distinct neu- ral correlates imply that increased IIV is multidetermined (for a review, see MacDonald et al., 2006), they may also reflect differ- ent levels of related phenomena. For example, fluctuations at the functional level (e.g., in blood oxygen–level dependent [BOLD] signal) may be directly linked to neuromodulation (e.g., dopamine biomarkers). Finally, we conclude by emphasizing important chal- lenges and identifying key questions that remain to be answered in the study of IIV.

IIV: A Life Span Perspective

IIV in cognitive performance across the life span is character- ized by a U-shaped function. Whereas IIV in cognitive functioning

is initially high in childhood, it decreases through adolescence,

only to reverse course with advancing age, where clear links between increasing IIV and concomitant cognitive impairments are observed (e.g., Hultsch et al., 2002; Li, Lindenberger, et al., 2004; Rabbitt et al., 2001; Williams, Hultsch, Strauss, Hunter, & Tannock, 2005; Williams, Strauss, Hultsch, & Hunter, 2007). This developmental pattern parallels the inverted U-shaped function observed for age-related changes in intellectual functioning (Li, Lindenberger, et al., 2004), with maturation and senescence imposing age-specific constraints, particularly under execu- tively demanding conditions (Kray, Eber, & Lindenberger, 2004; Li, Ha¨mmerer, Mu¨ller, Hommel, & Lindenberger, 2009). The preponderance of data on IIV concerns older adults, with most studies revealing clear age-related increases in IIV and con- comitant impairments in memory, attention, and language (e.g., Anstey, 1999; Hultsch et al., 2000; Nesselroade & Salthouse, 2004; Rabbitt et al., 2001). Such within-person fluctuations in cognitive performance could correspond to as much as several decades of between-person age differences (Nesselroade & Salt-

house, 2004). At cross section, IIV correlates negatively with level

of performance for diverse measures representing both fluid (e.g.,

perceptual speed, working memory) and crystallized (e.g., vocab- ulary) abilities. The magnitude of IIV– cognition correlations is age- and task-dependent, with older adults exhibiting larger cor-

relations between IIV and cognitive performance regardless of task complexity. In contrast, associations for young adults are attenu- ated (albeit still negative) and are typically only observed for select measures characterized by higher cognitive demands (e.g., fron- tally based executive processes; Hultsch et al., 2002). Several more stringent longitudinal assessments of the IIV– cognition association are available. For example, MacDonald, Hultsch, and Dixon (2003) examined longitudinal changes in IIV on four RT tasks administered over a 6-year period (three occa- sions) to 446 older adults from the Victoria Longitudinal Study (VLS). Findings showed that 6-year increases in IIV were associ- ated with 6-year cognitive decline for various cognitive outcomes (e.g., episodic memory, working memory), supporting claims that behavioral IIV is a predictor of cognitive aging. A key feature of this study was the proven link between change in IIV and cogni- tion at the within-person level. In an intriguing study, Lo¨vde´n, Li, Shing, and Lindenberger (2007) examined the supposition that variability may represent both a precursor, signaling impending decline in cognitive performance (see Figure 2a), and outcome of change. Bivariate dual change score models were used to test the dynamic coupling between trial-to-trial RT variability for a mea- sure of perceptual speed (identical pictures) and change in mean levels of performance for both perceptual speed and category fluency. Of particular interest was whether individual differences in IIV at baseline preceded subsequent change in cognitive func- tioning. These models were applied to five-occasion 13-year lon- gitudinal data from the Berlin Aging Study (N 447; 70 –102 years old at baseline). At baseline, individual differences in IIV across RT trials for the perceptual speed task were coded into three groups: individuals with scores greater than 0.5 standard devia- tions ( SDs) above the mean (high-variability group), those with scores 0.5 SDs of the mean (median-variability group), and those with scores more than 0.5 SDs below the mean (low-variability group). Notably, individuals in the high-variability group showed accelerated decline in category fluency relative to those in the low-variability group (see Figure 2b). In contrast, individual dif- ferences in mean levels of category fluency at baseline did not reliably influence subsequent change in IIV. These findings sug- gest that shifts in variability precede changes in mean perfor- mance, provide arguably the most convincing developmental test of the variability hypothesis to date, and imply that theories of cognitive aging should recognize and account for the developmen- tal cascade between senescent changes in variability and central tendency. Increased IIV in cognitive functioning has also been linked to impairments and fluctuations in biomarkers of aging, including sensory acuity and physiological functioning (Li, Aggen, Nessel- roade, & Baltes, 2001; MacDonald, Hultsch, & Bunce, 2006; Strauss, MacDonald, Hunter, Moll, & Hultsch, 2002), in accord with interpretations of IIV as a behavioral indicator of central nervous system (CNS) integrity. Other biomarkers such as forced expiratory volume and grip strength also account for considerable variance in IIV and RT latency (Anstey, Dear, Christensen, & Jorm, 2005). Additional behavioral findings linking variability to cognitive performance in the latter part of the life span indicate that IIV is systematically related to death (e.g., Shipley, Der, Taylor, & Deary, 2006). In one study, IIV uniquely predicted terminal cog-




7 9 6 M AC DONALD, LI, AND BÄCKMAN a b 70 60 50 40 30









Initial level of variability +0.5 SD Initial level of variability Initial level of variability -0.5
Initial level of variability +0.5 SD
Initial level of variability
Initial level of variability -0.5 SD

Assessment Occasion

Figure 2. (a) Hypothetical depiction of level of intraindividual variability (IIV) as a precursor of cognitive change for several individuals. (b) Longitudinal change in category fluency plotted as a function of performance variability. At baseline, three variability groups (high, median, and low) were formed on the basis of individual differences in IIV across reaction time trials of a perceptual speed task. Individuals with IIV scores greater than 0.5 standard deviations (SD s) above the mean were coded in the high-variability group, those with IIV scores 0.5 SD s of the mean were coded in the median-variability group, and those with IIV scores more than 0.5 SD s below the mean were coded in the low-variability group. Individuals who exhibited the most variability in perceptual speed at baseline also showed steeper longitudinal decline in category fluency, relative to individuals in the low- or median-IIV groups. From “Within-Person Trial-by-Trial Variability Precedes and Predicts Cognitive Decline in Old and Very Old Age: Longitudinal Data from the Berlin Aging Study,” by M. Lo¨vde´n, S.-C. Li, Y. L. Shing, and U. Lindenberger, 2007, Neuropsychologia, 45, p. 12. Copyright 2007 by Elsevier. Adapted with permission.

nitive decline as many as 15 years prior to the death, with IIV increasing every year closer to death (MacDonald, Hultsch, & Dixon, 2008). In this study, lower IIV scores also predicted the probability of survival independent of age, cardiovascular health, and cognitive performance level, supporting the view that IIV is an early behavioral marker of neurological dysfunction associated with impending death.

Origins of Increased IIV

Research on the origins of IIV remains sparse. At the cognitive level, postulated mechanisms include momentary lapses of atten-

tion (e.g., Bunce, Warr, & Cochrane, 1993) and failure to maintain executive control (e.g., West et al., 2002). Such accounts imply a disproportionately high number of very slow responses in the RT distribution, a proposition supported by recent findings (e.g., Wil- liams et al., 2005). Although most extant research on IIV has relied on such behavioral data, brain correlates have begun to be delin- eated in recent neuroscientific investigations. Closer inspection of related literatures (life span developmental psychology, neuropsy- chology, neuroscience) reveals that increased IIV shares definitive links to numerous age- and non-age-related conditions, including increasing adult age, cognitive decline, impending death, ADHD,



Parkinson’s disease, frontotemporal dementia, traumatic brain in- jury, and a specific allele (Val) of the catechol-O -methyl trans- ferase (COMT) gene (for reviews, see Hultsch et al., 2008; MacDonald, Nyberg, & Ba¨ckman, 2006). Converging evidence indicates that IIV in cognitive functioning is linked to these afore- mentioned outcomes independent of mean-level performance, un- derscoring the unique importance of variability (e.g., for promot- ing early detection of impending disease and improving differential diagnosis). Thus, IIV is not solely the province of aging at the behavioral level but rather reflects multiple endoge- nous brain correlates. For example, rapid changes in IIV from one moment to the next in a cognitive task may reflect brain mecha- nisms, such as fluctuations in the connectivity of neuronal path- ways (e.g., Kelly, Uddin, Biswal, Castellanos, & Milham, 2008), and the efficacy of neurotransmitter systems (e.g., Ba¨ckman et al., 2006). In the remaining sections of this review, we focus on links between increased IIV in cognitive performance and key brain correlates and speculate as to mechanisms underlying increased variability in behavior and brain alike.

Neural Correlates of IIV Structural Brain Correlates

A growing body of evidence links structural brain correlates to IIV. Structural alterations associated with increased performance variability include lesions to frontal gray matter (e.g., Sowell et al., 2003; Stuss, Murphy, Binns, & Alexander, 2003) as well as changes to white matter, including volumetric decline, demyelina- tion, and hyperintensities that reflect abnormalities on magnetic resonance imaging (MRI) scans due to age-related changes in cerebral blood flow, vascular injury, or neurological conditions (e.g., Anstey et al., 2007; Britton, Meyer, & Benecke, 1991; Bunce et al., 2007; Walhovd & Fjell, 2007). Lesions to frontal gray matter and associated increases in per- formance variability have a rich history in neuropsychology (e.g., Goldstein, 1939; Head, 1926). With regard to structural correlates, an immediate question concerns whether elevated IIV exhibits regional specificity. Neuropsychological evidence suggests that the frontal lobes (particularly the prefrontal cortex [PFC]) share a strong association with IIV. For example, Murtha and colleagues (2002) reported that persons with frontotemporal dementia were more variable than those with Alzheimer’s disease for a similar level of disease severity, implying differential relevance of frontal versus medial temporal regions to elevated IIV. Stuss et al. (2003) examined whether frontal regions shared a stronger association with IIV and, if so, whether there were regional frontal variations. They examined 48 patients between 43 and 67 years old who had brain lesions due to various acute conditions including infarction, hemorrhage, and trauma (25 patients with frontal lesions, 11 patients with nonfrontal lesions, and 12 control participants). Four different RT tasks were employed that required distinct levels of feature discrimination and identification: simple (respond to a single shape); easy (four different shapes with one designated as the target on 25% of the trials); complex (discrimination based on shape, color, and internal texture, with the target defined as a combination of these features); and redundant (also containing three features, but the nontarget trials did not contain any target features). Consistent with expectations, those with circumscribed

frontal lesions were more variable than those with nonfrontal lesions or controls. Moreover, for those with frontal lesions, a different profile of variability was observed depending on location. Individuals with lesions in right and left dorsolateral PFC, as well as in the superior medial frontal cortex, exhibited the most pro- nounced variability, which increased with task difficulty. These findings indicate that damage to the frontal lobes impairs stability of cognitive performance. Given the well-documented associations between white matter integrity and select cognitive deficits (e.g., processing speed, at- tention), potential links between white matter status and IIV have been examined in several recent studies. With increasing age, the corpus callosum (CC) is known to atrophy, with the area of the CC potentially reflecting an important indicator of white matter dis- ease related to demyelination or vascular insult (Pfefferbaum, Rosenbloom, Serventi, & Sullivan, 2002). In light of these asso- ciations, Anstey and colleagues (2007) investigated links among the CC area (traced on the mid-sagittal slice), IIV, and cognitive status. They examined differences in IIV across RT trials for both simple and choice RT tasks for two groups of adults who were 60 – 64 years old: a healthy adult group (n 432) and a group characterized as having mild cognitive disorders (n 57). For the normative sample, no reliable association was observed between the CC area and IIV; IIV accounted for less than 1% of the between-person variance in the CC area. However, individuals in the mild cognitive disorders group who were more variable also had a smaller CC area, with IIV explaining as much as 12.7% of the variance. The fact that a stronger association between the CC area and IIV was observed for the at-risk group suggests a role of white matter alterations in both cognitive impairment and in- creased performance variability. Moreover, dividing the CC into three distinct regions (anterior, intermediate, and posterior) yielded the largest associations between the anterior CC and IIV, consistent with previously reported links among frontal structures, attention, and IIV (e.g., West et al., 2002). Among potential explanations for the increased CC area–IIV link in the mild cognitive disorders group, Anstey and colleagues (2007) invoked a concept based on biological limits of cognitive processing (e.g., Baltes, Staudinger, & Lindenberger, 1999). Such limits may be defined in terms of brain reserve (i.e., the integrity of brain structure and neural circuitry) or in terms of cognitive reserve reflecting the ability to efficiently use existing brain struc- tures or to recruit alternate ones if necessary (Stern, 2002). Con- straints on brain or cognitive reserve may be due to numerous factors including structural or functional changes. The reserve premise is supported by functional imaging studies showing pat- terns of individual differences in neural activation. With increasing task difficulty, pathology-free individuals typically exhibit in- creased activation for the same brain regions that were activated for less demanding versions of the same task. However, there are clear individual differences in terms of these activation patterns; for a given level of task difficulty, individuals with greater reserve capacity exhibit less (i.e., more efficient) neural activation to achieve the same level of performance and could presumably perform better than less skilled individuals at maximum exertion (e.g., Grady et al., 1996; Nyberg et al., 2003; Stern, 2002; Stern et al., 2003). Anstey and colleagues (2007) speculated that indi- vidual differences in the CC area in the mild cognitive disorders group could share a stronger association with deterioration in the



PFC that may, in turn, give rise to attentional deficits and increased IIV (Andres, Parmentier, & Escera, 2006; Raz, Gunning-Dixon, Head, Dupuis, & Acker, 1998). Such an interpretation is consistent with previous research showing that increased executive demands result in increased IIV in response latency (West et al., 2002). Individuals in the mild cognitive disorders group who were the most variable also had the most diminished brain reserve (in the form of a smaller CC area, likely indicating pathological white matter alterations). This group’s increased IIV in response latency for a simple RT task could reflect such constraints to brain reserve but may also reflect lessened cognitive reserve (the greater burden of structural brain changes may lead to failures of task-relevant processing or to deficiencies in basic cognitive resources). Either diminished brain or cognitive reserve could exacerbate IIV in cognitive performance. A related study by Bunce and colleagues (2007) examined the association between white matter hyperintensities (WMHs) in var- ious brain regions (e.g., frontal, temporal, parietal) and increased IIV across RT trials for 469 healthy community-dwelling adults (60 – 64 years old). Findings indicated that frontal WMHs were linked to increased IIV but not to performance on other cognitive measures including global cognition, perceptual speed, and epi- sodic memory. In contrast, WMH in other brain regions did not share a significant association with IIV. These patterns imply that increased WMH and the associated deterioration of neural path- ways in the frontal cortex play a key role in increased IIV observed for older adults and those at risk of cognitive impairment. In summary, lesions to frontal gray and white matter share a robust association with increased IIV. The developmental evolu- tion and involution of gray and white matter corresponds, at least grossly, to increasing and then decreasing intellectual functioning (Kray et al., 2004; Li, Lindenberger, et al., 2004) as well as to decreasing and then increasing IIV (MacDonald et al., 2003; Williams et al., 2005) across the life span. For example, the reported decreases in IIV from early childhood through adoles- cence (cf. Williams et al., 2005) may reflect systematic changes in brain morphology, particularly in the frontal cortex (Gogtay et al., 2004). The reductions in gray matter density and synaptic pruning that occur during adolescence and young adulthood (Sowell et al., 2003) may promote increased neural efficiency and decreased noise in cognitive functioning that underlie the concomitant de- creases in IIV during development (Williams et al., 2005, 2007). Similarly, gray matter atrophy and increased neural noise in older adults may underlie increased IIV (Raz et al., 2004; Sowell et al.,


Life span changes in white matter volume, approximating an inverted U-shaped function, mirror closely the U-shaped life span changes observed for patterns of IIV (Gogtay et al., 2004; Sowell et al., 2003). Disconnectivity in associative pathways, whether caused by immature or degraded white matter tracts, can also increase variability. Such structural changes in performance may result in increased neural noise (e.g., due to white matter discon- nectivity), leading to less distinct cortical representations, poorer cognitive performance, and increased performance variability. Ef- fectively, the neural system matures, levels off, and declines with a direct correspondence between increasing and then decreasing intellectual functioning to decreasing and then increasing IIV in cognitive performance (MacDonald, Nyberg, & Ba¨ckman, 2006).

Functional Brain Correlates

Functional MRI. As interest in performance variability as a marker of CNS integrity has increased (Hultsch & MacDonald, 2004; Hultsch et al., 2008), functional brain imaging studies have begun to emerge. Bellgrove, Hester, and Garavan (2004) con- ducted the first neuroscientific investigation of the functional

neuroanatomy of executive control in relation to IIV. A total of 42 healthy participants, 18 – 46 years old, completed a Go/No-Go response inhibition task in an event-related functional magnetic resonance imaging (fMRI) design. The Go/No-Go task presented

two letters, X and Y, in an alternating pattern (i.e., X Y X Y X

Participants were instructed to press the response button for all Go trials but to withhold responses to No-Go stimuli that interrupted the alternating pattern (e.g., X Y X X). Findings indicated that increased IIV across Go trials of this response-inhibition task was associated with lower inhibitory success, slower responding, and increased brain activity in left and right middle frontal regions. Individuals with increased IIV activated inhibitory regions to a greater extent, likely reflecting greater demands for executive control to maintain task performance. Consistent with research underscoring important connections between the frontal lobes and increased IIV (e.g., Stuss et al., 2003), these findings imply that IIV may result from deficient attentional control (e.g., Bunce et al.,1993; Bunce, MacDonald, & Hultsch, 2004) and could serve to index the efficiency of executive control processes (Bellgrove et al., 2004). Moreover, the findings are notable given that the functional correlates of IIV were robust for even a healthy young to middle-aged sample. In a related fMRI investigation in older adults, the functional correlates of IIV were examined in relation to episodic retrieval success (MacDonald, Nyberg, Sandblom, Fischer, & Ba¨ckman, 2008). IIV was computed across latency trials in a word recogni- tion task and then examined in relation to the magnitude and anatomical location of BOLD activations. The 19 older adults (70 –79 years old) were presented with 80 words during an initial encoding session. During the subsequent retrieval session, both brain scans and response latencies were recorded. Lower variabil- ity across successful word retrieval trials was linked to better word recognition (r 0.37), faster RTs (r .66), and increased BOLD activation (r 0.49) in the supramarginal gyrus in the inferior parietal cortex, a structure implicated in sustained atten- tion, deep semantic encoding, and retrieval success (e.g., Cabeza & Nyberg, 2000; Shannon & Buckner, 2004; Wagner, Shannon, Kahn, & Buckner, 2005). In contrast, more variable individuals did not recruit this area of inferior parietal cortex to the same extent and exhibited lower retrieval success. The observed links between decreasing IIV and parietal activations associated with retrieval success support the hypothesis that behavioral IIV represents a proxy for neural integrity. As a putative indicator of CNS efficiency, IIV may not only share associations with activations in circumscribed brain regions but also be linked to activity across functional networks. In a recent investigation, Kelly et al. (2008) explored how IIV may be modulated by competition between functional networks. One net- work of particular interest, the default-mode network, comprises brain regions that show coherent activity at rest (e.g., stimulus- independent thought) and are routinely deactivated during attention-demanding cognitive tasks (Buckner & Vincent, 2007;




Raichle et al., 2001). An inability to suppress such default-mode activity has been linked to attentional lapses (e.g., Weissman, Roberts, Visscher, & Woldorff, 2006), altered task-relevant acti- vation in PFC and anterior cingulate cortex, and various conditions including Alzheimer’s disease, ADHD, and schizophrenia (e.g., Buckner et al., 2005; Liang et al., 2006; Tian et al., 2006). As these patterns and conditions are also well-known correlates of IIV, Kelly and colleagues surmised that IIV and default mode activity may be directly related. To pursue this idea, they used the Eriksen flankers task to examine whether increased IIV and associated attentional lapses could be explained by competition between the default mode network (activations at rest) and regions of a task- relevant dorsal attentional network (activations during the flankers task). Recent studies have shown that this competitive association is represented in the brain as an antiphase (i.e., negative) correla- tion between activation in these two networks. Larger negative correlations are said to reflect better activity regulation between the two networks, with the expectation that such larger antiphase associations would be linked to less IIV in performance. Active (flankers task) and resting state scans were obtained for 26 adults (mean age 28.1 years). For each target network, the time series of BOLD activations were extracted, and then the correlation between

activations in each network was computed (see Figure 3a). The magnitude of the negative association between spontaneous activ- ity in the default-mode network versus task-positive activity in the attentional network was subsequently correlated with IIV (coeffi- cient of variation) for congruent and incongruent RT trials in the flankers task. The findings indicated that as IIV for incongruent RT trials increased, the strong negative association between acti- vations in the default mode versus those in the attentional networks diminished (see Figure 3b). These results indicate that increasing IIV is not only related to select anatomical regions of the brain but is also associated with compromised regulation and coordination of neural networks. Electroencephalography (EEG)/event-related potentials (ERP). Whole-brain functional activation techniques such as EEG have also been used to link behavioral IIV in cognitive performance to event-related oscillations in various EEG spectra. For example, psychometric intelligence is inversely correlated with event- related potential variability (averaged evoked potentials) in the parietal and temporal lobes (Barrett & Eysenck, 1994). Early research in this area also linked increased variability in P300 latencies, an electrophysiological marker of various cognitive pro- cesses including attention and decision making (e.g., Sutton,


including attention and decision making (e.g., Sutton, a b Figure 3. (a) Correlation between time series


including attention and decision making (e.g., Sutton, a b Figure 3. (a) Correlation between time series

Figure 3. (a) Correlation between time series of blood oxygen level– dependent (BOLD) activations for the default mode network (activations at rest) and the dorsal attentional network (activations during the flankers task). Competition between the default mode and dorsal attentional networks is represented in the brain as an antiphase association, with larger negative correlations reflecting better activity regulation between the two networks. TRs repetition times. (b) Correlation between antiphase time series (see Figure 3a) and intraindi- vidual variability (IIV; coefficient of variation) for incongruent reaction time (RT) trials on the flankers task. As IIV for incongruent RT trials increases, the strong negative association between activations in the default mode versus attentional network is weakened. These results suggest that IIV is associated with compromised regulation and coordination of neural networks. CV coefficient of variation. From “Competition Between Functional Brain Networks Mediates Behavioral Variability,” by C. A. M. Kelly, L. Q. Uddin, B. B. Biswal, F. X. Castellanos, and M. P. Milham, 2008, NeuroImage, 39, p. 11. Copyright 2008 by Elsevier. Adapted with permission.



1969), to old age and morbidity (Ku¨gler, Taghavy, & Platt, 1993). More recently, increased prefrontal broadband noise in the theta- frequency band (6.0 – 8.0 Hz) for patients with schizophrenia was associated with increased BOLD signal in the dorsolateral PFC during a working memory task, a pattern interpreted to reflect inefficient neural processing (Winterer & Weinberger, 2004). The link between increases in prefrontal EEG noise (i.e., activity not time-locked to stimuli) during information processing and im- paired working memory in those with schizophrenia may reflect asynchronous field potential oscillations by cortical pyramidal neurons (Callicott et al., 2000). In a recent study, Fjell, Rosquist, and Walhovd (in press) ex- amined instability at both the electrophysiological and behavioral levels. Variability in P3a and P3b ERPs was indexed for a group of 133 adults (20 – 88 years old) for a simple visual task measuring response latency (a visual oddball task composed of 210 stimuli with a 10% probability for both targets and distractors). The P3a and P3b potentials are putative indicators of different attentional processes; the former is related to attention switching and volun- tary allocation of attention as elicited by distractor stimuli (but see Sutton, 1969, who discussed numerous functional interpretations of P3 potentials), with the latter thought to reflect stimulus eval- uation and controlled attentional processes (elicited by target stim- uli). IIV in response latencies for the simple visual task represents a third distinct source of information characterizing response se- lection and execution. The focus of this investigation was to identify the locus of age-related cognitive instability in the infor- mation processing stream (e.g., at the level of stimulus evaluation vs. response execution). With increasing age of the participants, significant increases in variability were observed across RT trials but not in variability in P3a or P3b latency. In light of the patterns observed, Fjell and colleagues (in press) concluded that age-related increases in IIV seem to be localized to later stages of processing (e.g., response decision or execution, as indicated by age-related increases in IIV) as opposed to earlier stages in the response decision stream (e.g., stimulus evaluation as indexed by P3b). This interpretation is consistent with findings by West and colleagues (2002); in that study, young and older adults showed the same level of IIV for a low executive demand condition of the n-back task, with the older group showing increased IIV as a function of increased executive demands.

Neuromodulatory Correlates

In addition to structural and functional brain changes, alterations in select neurotransmitters, including those in the catecholamine and acetylcholine systems, may give rise to increased neural noise (see Ba¨ckman et al., 2006, for review) that undergirds increased IIV in cognitive performance. In particular, alterations in dopa- mine (DA) neuromodulation have been documented for select populations who also exhibit increased behavioral IIV; these pop- ulations include not only older adults (Hultsch et al., 2002; Rabbitt et al., 2001) but also children with ADHD (Castellanos & Tan- nock, 2002; Bellgrove, Gill, Hawi, Kirley, & Robertson, 2005) and patients with schizophrenia (Manoach, 2003) and Parkinson’s dis- ease (Burton, Strauss, Hultsch, Moll, & Hunter, 2006). One work- ing hypothesis is that reduced DA activity increases neural noise, resulting in less distinct cortical representations manifest as de- creases in cognitive performance and increases in behavioral IIV

(Li, Lindenberger, & Sikstro¨m, 2001). In this section, we discuss evidence from neurocomputational models, as well as recent em- pirical findings, that directly link DA dysregulation to increased IIV. Computational models. Various computational approaches have been advanced to explain the mechanisms whereby dopami- nergic modulation affects cognition (see Rolls, Loh, Deco, & Winterer, 2008, for review). These attempts range from realistic biophysical firing rate models of how D1 and D2 receptors affect the stability of working memory representations (Durstewitz, Seamans, & Sejnowksi, 2000; see also Seamans & Yang, 2004, for a review) to more abstract models of dopaminergic effects on the dynamic connectivity between prefrontal cortex and basal ganglia (e.g., O’Reilly & Frank, 2006). Other models focus on the general computational role of DA in affecting the signal-to-noise ratio of neuronal signal transduction (e.g., Cohen & Servan-Schreiber, 1992; Li, Lindenberger, & Sikstro¨m, 2001) or outcome-based evaluation in reinforcement learning (see Montague, Hyman, & Cohen, 2004, for review). Although the various computational approaches differ in level of analysis and biophysical specificity, one basic assumption shared by most theories is that dopaminergic modulation influ- ences the properties of neuronal representations of cognitive and perceptual events. For instance, a two-stage model of dopaminer- gic modulation of working memory aims at capturing the dynamic interactions between DA and glutamate receptors in affecting the neuronal representations of memory items in the prefrontal cortex (Durstewitz et al., 2000). Specifically, when D2 receptor modula- tion predominates during the first stage, the PFC network is supposed to be in an exploratory state with multiple weak repre- sentations. However, in the second stage when D1 receptor mod- ulation predominates, heightened inhibitory mechanisms weed out weaker representations and enhance the representation of the stronger inputs (Seamans & Yang, 2004). This effect is nicely paralleled by results from other models that aim at explicating the computational effects of dopaminergic modulation on the signal- to-noise ratio of information processing at a more molar level (Cohen & Servan-Schreiber, 1992; Li, Lindenberger, & Sikstro¨m, 2001). In these models, DA is commonly assumed to facilitate the responsivity in activity transmission both within and between modules of a neural network. As the gain parameter of a neural network’s activation function is attenuated or increased to mimic deficient or excessive dopaminergic modulation, respectively, the distinctiveness of internal representations of stored perceptual or memory items is reduced, resulting in impaired cognitive perfor- mance and increased behavioral IIV (e.g., Li, Lindenberger, & Sikstro¨m, 2001; Li & Sikstro¨m, 2002). In vivo positron emission tomography (PET) research. Build- ing on these neurocomputational findings, investigators recently attempted to directly link a PET-derived in vivo marker of dopa- minergic neurotransmission (D2 binding potential) to IIV in cog- nitive performance (MacDonald, Cervenka, Farde, Nyberg, & Ba¨ckman, 2009) in a group of middle-aged adults (N 16; 41– 65 years old). D2 binding potential was derived for striatum and three extrastriatal brain regions (orbitofrontal cortex [OFC], anterior cingulate cortex [ACC], and hippocampus [HC]); selection of the region of interest was based on extant findings suggesting that frontal and medial–temporal regions may differentially influence IIV (Bellgrove et al., 2004; Murtha et al., 2002; Takahashi et al.,



2007). IIV in response latencies was indexed for measures of episodic memory and executive functioning on the basis of the established association these tasks share with IIV (see Hultsch et al., 2008). Finally, relative to other DA receptors (e.g., D1), D2 binding may be especially relevant for the study of IIV. D2 receptors serve a critical role in facilitating rapid shifts between different targets (cf. IIV), in contrast to D1 receptors, which have greater relevance for tonic DA levels involved in maintaining a specific cognitive set (e.g., Bilder, Volavka, Lachman, & Grace, 2004; Cohen, Braver, & Brown, 2002). Given the key role of DA modulation for cognitive performance and IIV in the neurocom- putational models, we hypothesized that decreased D2 binding potential would be linked to increased IIV. Consistent with expectations, lower D2 binding in the ACC, HC, and OFC, but not in the striatum, was systematically linked to increased IIV in both memory and executive functioning. This study is the first in which a direct link was found between in vivo measures of DA function and behavioral measures of IIV in cognitive performance. The findings complement the neuro- computational work of Li and colleagues (Li, Lindenberger, & Sikstro¨m, 2001; Li & Sikstro¨m, 2002), suggesting that DA dys- regulation alters the signal-to-noise ratio of neural information processing, impairing neurons’ sensitivity to afferent signals, lead- ing to less distinct cortical representations, and ultimately resulting in increasing IIV and impaired cognitive functioning. The MacDonald et al. (2009) data are also consistent with regional brain correlates identified in the neuropsychological and neuroscience literatures, reinforcing the importance of frontal re- gions as structural correlates of IIV (e.g., Bellgrove et al, 2004; Murtha et al., 2002; Stuss et al, 2003). The strongest association was observed between D2 binding in the ACC and IIV for an executive task that indexed set shifting. This implies that if the ACC is not functioning optimally (e.g., due to reduced D2 bind- ing), the resulting diminished executive control coupled with the ACC’s central processing role may lead to increased IIV in per- formance, particularly for executively demanding tasks (e.g., West et al., 2002). These findings provide a nice extension of research by Takahashi and colleagues (2007). Whereas they examined D2 binding potential in select extrastriatal regions and found that D2 binding in HC was associated with performance on episodic mem- ory tasks as well as frontally mediated executive tasks, our find- ings clearly show the relation of D2 binding in ACC and HC to measures of IIV in memory and executive tasks.

Genetic Correlates

In addition to neuromodulators and other brain-based correlates, emerging evidence indicates that IIV is under genetic influence as well. The catechol-O -methyl transferase (COMT) enzyme de- grades DA in the frontal cortex, with carriers of the Val allele of the COMT gene having lower extracellular DA levels in prefrontal cortex than Met carriers due to elevated enzymatic activity (Wein- shilboum, Otterness, & Szumlanski, 1999). Stefanis and col- leagues (2005) examined whether IIV in cognitive performance varied as a function of COMT allele (Val vs. Met) in a sample of 521 men (mean age 21 years). The relation of DA signaling (tonic vs. phasic) to performance on cognitive tasks is dependent upon the characteristics of the task itself. The Met allele, with its lower enzymatic activity, may differentially benefit performance

on cognitive tasks requiring cognitive stability and tonic levels of DA activation but hamper performance on tasks that require cog- nitive flexibility and phasic activation. To examine this hypothesis, Stefanis and colleagues administered the continuous performance test (identical pairs version), which requires participants to respond whenever two rapidly flashed identical stimuli (e.g., four-digit numbers), among the 300 stimuli pairs, appear over the course of a 5-min trial. Findings indicated that Val carriers exhibited greater IIV than Met carriers across trials of the continuous performance test, thereby linking the COMT polymorphism and frontal DA activity to IIV. The authors concluded that the greater performance stability for the Met genotype may be a consequence of increased extracellular DA, serving to stabilize neural representations in the prefrontal cortex. The link between genetic and neuromodulatory influences un- derscores that, although multidetermined, the neural and genetic correlates of IIV likely reflect different levels of related phenom- ena. Further emphasizing this point, percentage of change in the BOLD signal in the left PFC is also known to vary as a function of the COMT genotype. Mattay and colleagues (2003) found that for a fixed level of n-back performance, participants who were homozygous for the Val allele with less available DA in the synaptic cleft exhibited increased BOLD activity in the PFC rel- ative to homozygous Met carriers. Moreover, this difference was exacerbated with increasing executive task load (i.e., 2-back vs. 1-back condition). The increased PFC activity in Val carriers was interpreted as a less efficient physiological response, perhaps due to diminished neuronal signaling in the PFC (Li & Sikstro¨m, 2002; Stefanis et al., 2005). This interpretation is supported by other neuroscientific findings showing that older adults nonselectively recruit neural tissue in task-irrelevant brain regions during episodic and working memory performance (e.g., Cabeza, 2001; Logan, Sanders, Snyder, Morris, & Buckner, 2002). In synthesizing these findings, we note that a less efficient frontal response during cognitive processing characterizes patients with schizophrenia, COMT Val carriers, and older adults—all groups displaying in- creased IIV in cognitive performance as well as alterations in the DA system. Although the aforementioned findings reveal some promising associations between COMT and IIV, there is a clear need for future research on the links between IIV and other genes that influence DA functions, as well as other systems (e.g., neuro- transmitter, immune) altogether.

Key Issues and Future Research

Despite explosive growth in variability research in the past decade, a number of key challenges remain. In this final section, we address several important themes including measurement is- sues related to IIV, the potential adaptivity of neural noise, and several intriguing avenues for future research.

Measurement Issues and IIV

At the outset of this review, we introduced several analytical perspectives (e.g., task characteristics, time scale) known to influ- ence observed patterns of performance variability. Although often an afterthought, the manner in which IIV is assessed may impact both the magnitude and valence of associations between IIV and select outcome measures. For example, there are several reasons



that age differences are more likely to be found for measures of IIV indexed by response speed versus those indexed by response accuracy. Relative to accuracy-based measures, chronometric re- sponse latencies permit a more precise index of rapidly changing internal processes (Gazzaniga, Ivry, & Mangun, 2002). If such moment-to-moment fluctuations represent the best time frame to assess efficacy of age-related cognitive functioning, then measures of IIV over brief intervals likely confer the purest index. More- over, accuracy measures are often derived from complex tasks, which may be quite sensitive to strategy use (Hultsch et al., 2008); as noted, IIV for tasks amenable to strategy use may be positively related to other cognitive outcomes. Finally, latency measures should provide a more sensitive scale compared with indicators of accuracy due to measurement reliability. Accuracy-based cogni- tive measures often contain a limited number of items, potentially leading to unreliable estimates of variability. Recent findings sug- gest that, regardless of metric, a sufficient number of trials ( 7) are required to reliably index IIV (Schmiedek, 2006). It is also important to keep in mind that relative to measures of IIV, the arithmetic mean has a greater proportion of systematic variance available for association with cognitive outcomes. Thus, observed differences in the strength (but not valence) of association between mean versus IIV may reflect differences in reliability. If the reliability of IIV is lower than that of the mean, it is not surprising that it would share a comparatively weaker association with the same cognitive outcome. Such measurement properties of variabil- ity represent important considerations as numerous researchers collectively attempt to understand patterns observed in the extant literature.

The Adaptive Potential of Noise: Stochastic Resonance

Although our focus has rarely strayed from the negative aspects of IIV, the point that variability or noise is not always suboptimal is perhaps best illustrated in a phenomenon called stochastic res- onance (SR). SR denotes the fact that optimal levels of input (i.e., externally introduced) noise are beneficial for detecting and trans- mitting weak signals in nonlinear physical and biological systems (see Moss, Ward, & Sannita, 2004, for review). A basic aspect of SR (threshold SR) results from the interaction between a response (or activation) threshold, a subthreshold stimulus, and input noise. Consider an example of a hardly recognizable image of Big Ben, with most of its pixels degraded below a certain threshold of gray scale. In this case, with an optimal amount of random noise added to the gray scale, the subthreshold image gradually emerges above the threshold and becomes visible. However, if more noise is added, the image blurs again. Returning to the example of the brain and models of spike-generating neurons, a subthreshold signal cannot cross the threshold to produce action potentials, thus losing the information content of the weak signal. However, the benefit of adding noise in this case is to permit some spikes to cross the threshold, thereby transmitting information about the signal.

Aging, Neuromodulation, and SR

At first glance, the notion of variability as a prime indicator or cause of age-related cognitive deficits conflicts with ample empir- ical and theoretical evidence showing the functional relevance and benefit of SR noise. This apparent paradox leads to a pivotal

question: To what extent are the functions of synaptic noise and characteristics of SR altered in aging neurobiological systems as well as in other conditions marked by suboptimal neuromodula- tion? Formal theories are helpful in guiding empirical research on the interactive dynamics between synaptic noise as a regulatory mechanism of cortical neurons’ response properties and other mechanisms of neuronal gain control, such as neuromodulation and neuronal synchronization (e.g., Diesmann, Gewaltig, & Aert- sen, 1999). Models comparing systems with deficient neuromodu- lation (e.g., in the case of senescence or pathology) to intact systems have generated insights about the interactions between neuromodulatory and synaptic noise gain control. Using the stochastic gain-tuning model of neuromodulatory aging (Li, Lindenberger, & Sikstro¨m, 2001) to simulate experi- mental effects observed in paradigms of subthreshold perceptual processing, we recently investigated how interactions between the benefit of input noise and deficient neuromodulation jointly affect SR in older neurobiological systems (Li et al., 2006). Less efficient neuromodulation in old age was modeled by reducing the gain parameter (G) from optimal to suboptimal levels to yield activation functions with less steep slopes. Neural networks with attenuated system gain modulation reacted to input noise less effectively, resulting in SR functions with lower and right-shifted peaks (see Figure 4). These findings suggest that noise added to weak signals continues to yield beneficial effects in suboptimally modulated systems. However, in aging, the benefits of SR are affected in three major ways: (a) the absolute peak of SR is reduced, (b) the range within which noise operates optimally is reduced, and (c) more external noise is required to reach the peak of the SR function. SR has recently attracted attention in research on aging because of its promising utility in attenuating the functional consequences of older adults’ deficits in sensory and sensorimotor functions. For

deficits in sensory and sensorimotor functions. For Figure 4. Different levels of stochastic gain modulation

Figure 4. Different levels of stochastic gain modulation that simulate age-related differences in neuromodulatory efficiency, result in stochastic resonance (SR) functions that differ in peak magnitude and the extent of right shift of the peak. The bell-shaped SR function depicts the rise and fall of noise-enhanced benefits in signal processing. Neural networks with attenuated system gain parameter (G) react to input noise less effectively, yielding SR functions with lower and right-shifted peaks. Introducing noise can be beneficial for subthreshold signals, but too much noise can obscure the signal. E expected value or mean. From “A Neurocomputational Model of Stochastic Resonance and Aging,” by S.-C. Li, T. von Oertzen, and U. Lindenberger, 2006, Neurocomputing, 69, p. 1558. Copyright 2006 by Elsevier. Adapted with permission.



instance, keeping one’s balance while standing is an important

basic sensorimotor function that declines substantially during ag-

ing (e.g., Huxhold, Li, Schmiedek, & Lindenberger, 2006) and is

often associated with serious negative consequences, such as falls. It has recently been shown that subthreshold electromechanical noise delivered through shoe insoles (Harry, Niemi, Priplata, & Collins, 2005) reduces older adults’ body sway while standing. Important challenges remain if researchers are to fully under-

stand the positive versus negative characteristics of variability and noise. Research on SR clearly indicates that providing external noise into a system can improve signal processing (Li et al., 2006). Even white noise can produce an action potential for an otherwise weak signal. However, as shown in Figure 4, introducing too much noise can also serve to obscure the signal. Additional research is required into how age-related increases in neuronal noise due to endogenous sources may be countered by the introduction of external noise to augment signal processing. In contrast, many examples of the negative repercussions of noise were reviewed, showing clear links between increased variability and disadvan- taged characteristics (e.g., impaired cognitive functioning, smaller

CC area). On the basis of the preponderance of data available, it

seems that the maladaptive forms of IIV reflect changes in endog- enous brain mechanisms, such as white matter disconnectivity or DA dysregulation with increasing age. Such changes in brain structure, function, and neuromodulation likely influence cellular processes that give rise to increased neural noise (i.e., reduced signal to noise) and impaired cognitive representations, which is subsequently manifest as increased performance variability and impaired cognitive functioning. For example, endogenous forms of neural noise underlying increased behavioral variability could be due to a lack of coherence in neuronal firing patterns or to uncoordinated firing within individual cells. Although some data have linked the timing of neuronal firing to behavior (e.g., Johansson & Birznieks, 2004), it remains to be determined whether cellular synchronization patterns map onto IIV in behavior. Also required are single-cell studies in which across-trial variability in firing rate for individual cells is measured directly (see MacDonald, Nyberg, & Ba¨ckman, 2006).

Future Research

With regard to IIV in behavior for select cognitive outcomes, most of the extant research is based upon an evaluation of between-person differences in within-person performance fluctu- ations. Future researchers clearly would benefit from additional within-person analyses of the intraindividual covariation between change in IIV for a given task and mean level change for other cognitive outcomes (cf. Lo¨vde´n et al., 2007). Such analyses will provide a more stringent methodological assessment of whether change in IIV is temporally linked to change in mean performance. Future studies may explore within-person associations between relevant features of a given task, such as the coupling between response accuracy (correct vs. incorrect trials) and the vicissitudes of response latency. As interest for this topic continues to grow, it is imperative to move from univariate to multivariate conceptual- izations of variability, which is akin to examining alterations in a single local function versus alterations in the global organization of numerous functions (e.g., how IIV across trials of a given RT task is affected by, or may influence, shifts in cognitive resource

allocation or functional organization of the brain; Lindenberger & von Oertzen, 2006). More mechanistic evidence on the brain–IIV relationship is needed. Several important avenues of research could be pursued that would provide better understanding of neural mechanisms underlying IIV, including experimental manipulation of IIV, as well as the examination of IIV in the brain itself. With regard to experimental approaches, manipulating brain correlates hypothe- sized to influence performance variability represents one method to directly link IIV to brain function. This could involve pharma- cological manipulations of potential neural substrates of variability such as DA. For example, dopaminergic agonists or antagonists that respectively enhance or hamper neuromodulation could be administered to examine the influence of DA signaling on IIV in cognitive performance. A strong correlative triad is well documented among DA, adult age, and cognitive performance (Ba¨ckman et al., 2006). Given the hypothesized importance of DA to IIV, administering DA might decrease IIV in RT for older adults, whereas an antagonist could increase IIV in the very same measure of RT for young adults. Although recent investigations linking behavioral IIV in cogni- tive functioning to neural indicators represent an important first step, a perhaps even more intriguing avenue of future investigation involves the direct examination of variability in the brain itself. In light of the hypothesis that endogenous sources underlie IIV in cognitive functioning, brain-imaging techniques with excellent temporal resolution should be used to precisely index neural vari- ability. One such direct approach would be to examine fluctuations in the BOLD signal across individual trials. Technological ad- vances in event-related fMRI (e.g. high-field 4-Tesla MRI) make it possible to resolve rapid changes in the hemodynamic signal (Yamaguchi, Hale, D’Esposito, & Knight, 2004). As a conse- quence, it is possible to compute an IIV index for rapid changes across BOLD activation trials within individuals. Note that doing so can also lead to strong developmental tests regarding the nature of variability. Specifically, if increased behavioral IIV is a proxy for impaired neural processes, then it follows that indices of increased brain IIV should be systematically associated with in- creased IIV in cognitive functioning. Following this logic, one could conduct a stringent test of within-person coupling by exam- ining the time-varying covariation between IIV in behavior (e.g., variability across discrete blocks of latency trials) and IIV in the brain (e.g., variability in the BOLD signal for the corresponding behavioral trials). A similar logic applies to examining variability in electrical brain activity as indexed with EEG, which provides even better temporal resolution than fMRI for indexing IIV in the brain. Recently, a novel demonstration of this approach was used to examine the relation between variability of evoked electrical ac- tivity and performance on a cognitive task (face memory) for children (8 –15 years old) and young adults (20 –33 years old; McIntosh, Kovacevic, & Itier, 2008). Relative to the young adults, children exhibited increased behavioral IIV in response latencies as well as poorer face memory. Notably, however, variability in the brain signal itself increased with age, sharing a negative association with both behavioral IIV and accuracy, consistent with previous patterns observed for behavioral data in younger samples (cf. Siegler, 1994). McIntosh and colleagues (2008) argued that as the brain continues to mature, such increased moment-to-moment



variability in EEG signals may reflect enhanced neural complexity and greater functional adaptivity (Gogtay et al., 2004). On the surface, the findings by McIntosh and colleagues (2008) run counter to the expectation that increased variability in the brain and behavior go hand in hand. However, upon closer scrutiny, these results are expected from the perspective of nervous system nonlinear dynamics. Specifically, increased brain variability is adaptive to the extent that it facilitates the parsing of weak incom- ing signals (e.g., Destexhe & Contreras, 2006) or assists in the formation of functional networks (e.g., Fuchs, Ayali, Robinson, Hulata, & Ben-Jacob, 2007). At least from childhood through early adulthood, increased brain variability is indicative of enhanced neural complexity (e.g., increased integration of neuronal circuits, formation of new network connections), with corresponding ben- efits for cognitive adaptability. Clearly, it is of paramount interest to examine patterns of variability in older age groups and to determine in particular whether brain variability for these individ- uals is positively or negatively correlated with cognitive function- ing (including behavioral IIV). As summarized earlier, variability can confer both positive and negative benefits. In future research, investigators should not neglect the potential benefits of variability and the ways in which age-related differences in neuronal noise may interact with processes by which external noise aids or ham- pers signal processing (e.g., SR and how the addition of noise to systems can improve signal processing; Li et al., 2006). In addition, it may be beneficial to include other neuroimaging techniques such as event-related optical signal (EROS) as well as transcranial magnetic stimulation (TMS; see Gazzaniga et al., 2002). In EROS, optical fibers and infrared light are used to index cerebral activity through measurement of the scattering properties of neurons. EROS has good spatial (millimeters) and temporal (milliseconds) measurement properties but cannot be used to index subcortical structures. In TMS, weak electric currents excite neu- rons and facilitate the study of brain connectivity, and thus it is a potentially interesting tool in light of findings linking IIV to network competition (Kelly et al., 2008). Finally, multimodal assessments (e.g., combining fMRI and near-infrared specstros- copy) could help maximize temporal and spatial precision in the study of brain variability.


Extant evidence suggests that there are numerous neural under- pinnings of increased IIV in cognitive performance. Increased IIV may be modulated by changes in gray matter density (e.g., due to development, lesions, or neurodegeneration), changes in the integ- rity of white matter (e.g., disconnectivity in associative pathways, immature or degraded white matter tracts), and neuromodulatory changes (e.g., age-related dysregulation of dopaminergic neuro- modulation). In addition, an emerging literature links IIV to ge- netic factors. Findings from neuropsychology and neuroscience alike underscore that increased IIV shares a robust association with the frontal cortex (e.g., IIV is exacerbated for frontal vs. nonfrontal lesions, and less efficient patterns of BOLD activation in frontal regions are linked to increased IIV). In fMRI research, inefficiency is said to denote excessive activity for a given level of perfor- mance, presumably reflecting unfocused or unstable response cir- cuits (MacDonald, Nyberg, & Ba¨ckman, 2006; Winterer & Wein- berger, 2004). These neurobiological underpinnings of increased

IIV likely reflect different levels of related phenomena, with some levels exhibiting empirical links (e.g., DA, COMT genotype, and fMRI). Examining potential associations among these different sources requires a novel fusion of existing methodologies (e.g., Li, Lindenberger, Nyberg, Heekeren, & Ba¨ckman, in press). Whereas molecular imaging techniques (e.g., PET) are well suited for investigating neuromodulatory correlates of IIV, fMRI techniques are better suited for examining select brain regions and neural circuits linked to IIV. Pursuing multimodal imaging studies com- bining PET-derived DA markers and functional imaging data during cognitive activity for the same individuals will address the question of whether these neurobiological underpinnings jointly influence IIV. An emerging literature suggests a link between DA activity and the magnitude of the BOLD signal during cognitive performance (Landau, Lal, O’Neill, Baker, & Jagust, 2009; Nyberg et al., 2009; Schott et al., 2008). Of particular import in the present context, Landau et al. observed a link between a measure of striatal DA synthesis capacity and the load-dependent BOLD response during working memory in the left PFC, as well as to activations in the supramarginal gyrus (a region frequently coac- tivated with PFC during working memory performance). Mac- Donald et al. (2008) reported that highly variable persons showed underrecruitment of the supramarginal gyrus during episodic rec- ognition. These patterns suggest a linkage between dopaminergic neurotransmission, functional brain activation, and IIV that should be further scrutinized in future research. Moreover, increased IIV in cognitive performance could have a bottom-up origin (e.g., dysregulated DA neurotransmission in old age or in Parkinson’s disease) or may be due to deficient top-down regulation by pre- frontal regions (e.g., following traumatic brain injury). Finally, it is important to emphasize that the concepts of noise and variability are not exclusively negative. Research on SR indicates that white noise added to a system can positively influence performance (Harry et al., 2005; Priplata et al., 2002), and brain variability in early adulthood and behavioral variability for older adults may reflect greater functional variability and neural complexity (McIn- tosh et al., 2008), as well as adaptability (Allaire & Marsiske,



Allaire, J. C., & Marsiske, M. (2005). Intraindividual variability may not always indicate vulnerability in elders’ cognitive performance. Psychol- ogy and Aging, 20, 390 – 401. Andres, P., Parmentier, F. B., & Escera, C. (2006). The effect of age on involuntary capture of attention by irrelevant sounds: A test of the frontal hypothesis of aging. Neuropsychologia, 44, 2564 –2568. Anstey, K. J. (1999). Sensorimotor and forced expiratory volume as correlates of speed, accuracy, and variability in reaction time perfor- mance in late adulthood. Aging, Neuropsychology, and Cognition, 6, 84 –95.

Anstey, K. J., Dear, K., Christensen, H., & Jorm, A. F. (2005). Biomarkers, health, lifestyle, and demographic variables as correlates of reaction time performance in early, middle, and late adulthood. Quarterly Journal of Experimental Psychology Section A, 58, 5–21. Anstey, K. J., Mack, H. A., Christensen, H., Li, S.-C., Reglade-Meslin, C.,

Sachdev, P. (2007). Corpus callosum size, reaction time

speed and variability in mild cognitive disorders and in a normative sample. Neuropsychologia, 45, 1911–1920. Ba¨ckman, L., Nyberg, L., Lindenberger, U., Li, S.-C., & Farde, L. (2006). The correlative triad among aging, dopamine, and cognition: Current

Maller, J.,



status and future prospects. Neuroscience and Biobehavioral Reviews, 39, 791– 897. Baltes, P. B., Staudinger, U. M., & Lindenberger, U. (1999). Lifespan psychology: Theory and application to intellectual functioning. Annual Review of Psychology, 50, 471–507. Barrett, P. T., & Eysenck, H. J. (1994). The relationship between evoked potential component amplitude, latency, contour length, variability, zero-crossings, and psychometric intelligence. Personality and Individ- ual Differences, 16, 3–32. Bellgrove, M. A., Gill, M., Hawi, Z., Kirley, A., & Robertson, I. H. (2005). Dissecting the attention deficit hyperactivity disorder (ADHD) pheno- type: Sustained attention, response variability and spatial attentional asymmetries in relation to dopamine transporter (DAT1) genotype. Neuropsychologia, 43, 1847–1857. Bellgrove, M. A., Hester, R., & Garavan, H. (2004). The functional neuroanatomical correlates of response variability: Evidence from a response inhibition task. Neuropsychologia, 42, 1910 –1916. Bilder, R. M., Volavka, J., Lachman, H. M., & Grace, A. A. (2004). The catechol-O- methyltransferase polymorphism: Relations to the tonic- phasic dopamine hypothesis and neuropsychiatric phenotypes. Neuro-

psychopharmacology, 29, 1943–1961. Britton, T. C., Meyer, B. U., & Benecke, R. (1991). Variability of cortically evoked motor responses in multiple sclerosis. Electroencephalography and Clinical Neurophysiology, 81, 186 –194. Buckner, R. L., Snyder, A. Z., Shannon, B. J., LaRossa, G., Sachs, R.,

Mintun, M. A. (2005). Molecular, structural, and

Fotenos, A. F.,

functional characterization of Alzheimer’s disease: Evidence for a rela- tionship between default activity, amyloid, and memory. Journal of Neuroscience, 25, 7709 –7717.

Buckner, R. L., & Vincent, J. L. (2007). Unrest at rest: The importance of default activity and spontaneous network correlations. NeuroImage, 37,


Bunce, D. J., Anstey, K. J., Christensen, H., Dear, K., Wen, W., & Sachdev, P. (2007). White matter hyper intensities and within-person variability in community-dwelling adults aged 60 to 64 years. Neuro- psychologia, 45, 2009 –2015. Bunce, D. J., MacDonald, S. W. S., & Hultsch, D. F. (2004). Inconsistency in serial choice responding decision and motor reaction times dissociate in younger and older adults. Brain and Cognition, 56, 320 –327. Bunce, D. J., Warr, P. B., & Cochrane, T. (1993). Blocks in choice responding as a function of age and physical fitness. Psychology and Aging, 8, 26 –33. Burton, C. L., Strauss, E., Hultsch, D. F., Moll, A., & Hunter, M. A. (2006). Intraindividual variability as a marker of neurological dysfunc- tion: A comparison of Alzheimer’s disease and Parkinson’s disease. Journal of Clinical and Experimental Neuropsychology, 28, 67– 83. Cabeza, R. (2001). Cognitive neuroscience of aging: Contributions of functional neuroimaging. Scandinavian Journal of Psychology, 42, 277–


Cabeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical

review of 275 PET and fMRI studies. Journal of Cognitive Neuro- science, 12, 1– 47.

Callicott, J. H., Bertolino, A., Mattay, V. S., Langheim, F. J. P., Duyn, J.,

Weinberger, D. R. (2000). Physiological dysfunction of

the dorsolateral prefrontal cortex in schizophrenia revisited. Cerebral Cortex, 10, 1078 –1092. Castellanos, F. X., & Tannock, R. (2002). Neuroscience of attention deficit hyperactivity disorder: The search for endophenotypes. Nature Reviews Neuroscience, 3, 617– 628. Cattell, R. B. (1957). Personality and motivation: Structure and measure- ment. New York, NY: World Books. Cattell, R. B. (1966). The data box: Its ordering of total resources in terms of possible relational systems. In R. B. Cattell (Ed.), Handbook of

Coppola, R.,

multivariate experimental psychology (pp. 67–128). Chicago, IL: Rand McNally. Cattell, R. B., & Scheier, I. H. (1961). The meaning and measurement of neuroticism and anxiety. New York, NY: Ronald Press. Cohen, J. D., Braver, T. S., & Brown, J. W. (2002). Computational perspectives in dopamine function in prefrontal cortex. Current Opinion in Neurobiology, 12, 223–229. Cohen, J. D., & Servan-Schreiber, D. (1992). Context, cortex, and dopa- mine: A connectionist approach to behavior and biology in schizophre- nia. Psychological Review, 99, 45–77. Destexhe, A., & Contreras, D. (2006). Neuronal computations with sto- chastic network states. Science, 314, 85–90. Diesmann, M., Gewaltig, M.-O., & Aertsen, A. (1999, December 2). Stable propagation of synchronous spiking in cortical neural networks. Nature, 402, 529 –533. Durstewitz, D., Seamans, J. K., & Sejnowski, T. J. (2000). Neurocompu- tational models of working memory. Nature Neuroscience, 3, 1184 –


Fisk, D. W., & Rice, L. (1955). Intra-individual response variability.

Psychological Bulletin, 52, 217–250. Fjell, A. M., Rosquist, H., & Walhovd, K. B. (in press). Instability in the latency of P3a/P3b brain potentials and cognitive function in aging. Neurobiology of Aging. Ford, D. H. (1987). Humans as self-constructing living systems: A developmental perspective on behavior and personality. Hillsdale, NJ: Erlbaum. Fuchs, E., Ayali, A., Robinson, A., Hulata, E., & Ben-Jacob, E. (2007). Coemergence of regularity and complexity during neural network de- velopment. Developmental Neurobiology, 67, 1802–1814. Fuentes, K., Hunter, M. A., Strauss, E., & Hultsch, D. F. (2001). Intrain- dividual variability in cognitive performance in persons with chronic fatigue syndrome. The Clinical Neuropsychologist, 15, 210 –227. Gazzaniga, M. S., Ivry, R. B., & Mangun, G. R. (2002). Cognitive neuro- science: The biology of the mind (2nd ed.). New York, NY: Norton. Gogtay, N., Giedd, J. N., Lusk, L., Hayashi, K. M., Greenstein, D.,

Thompson, P. M. (2004, May 25). Dynamic map-

ping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences of the United States of America, 101, 8174 – 8179. Goldstein, K. (1939). The organism. New York, NY: American Book. Grady, C. L., Horwitz, B., Pietrini, P., Mentis, M. J., Ungerleiter, L., Rapoport, S. I., & Haxby, J. (1996). The effect of task difficulty on cerebral blood flow during perceptual matching of faces. Human Brain Mapping, 4, 227–239. Harry, J. D., Niemi, J. B., Priplata, A. A., & Collins, J. J. (2005, April). Balancing act: Noise is the key to restoring the body’s sense of equi- librium. IEEE Spectrum, 36 – 41. Head, H. (1926). Aphasia and kindred disorders of speech. Cambridge, England: Cambridge University Press. Hedden, T., & Gabrieli, J. D. E. (2004). Insights into the ageing mind: A view from cognitive neuroscience. Nature Reviews Neuroscience, 5,

Vaituzis, A. C.,


Holtzer, R., Verghese, J., Wang, C., Hall, C. B., & Lipton, R. B. (2008). Within-person across–neuropsychological test variability and incident dementia. Journal of the American Medical Association, 300, 823– 830. Hultsch, D. F., & MacDonald, S. W. S. (2004). Intraindividual variability in performance as a theoretical window onto cognitive aging. In R. A. Dixon, L. Ba¨ckman, and L.-G. Nilsson (Eds.), New frontiers in cognitive aging (pp. 65– 88). New York, NY: Oxford University Press. Hultsch, D. F., MacDonald, S. W. S., & Dixon, R. A. (2002). Variability in reaction time performance of younger and older adults. Journal of Gerontology, Series B: Psychological Sciences and social Sciences, 57,


Hultsch, D. F., MacDonald, S. W. S., Hunter, M. A., Levy-Bencheton, J.,



& Strauss, E. (2000). Intraindividual variability in cognitive perfor- mance in older adults: Comparison of adults with mild dementia, adults with arthritis, and healthy adults. Neuropsychology, 14, 588 –598.

Hultsch, D. F., Strauss, E., Hunter, M. A., & MacDonald, S. W. S. (2008). Intraindividual variability, cognition, and aging. In F. I. M. Craik &

T. A. Salthouse (Eds.), The handbook of aging and cognition (3rd ed.,

pp. 491–556). New York, NY: Psychology Press. Huxhold, O., Li, S.-C., Schmiedek, F., & Lindenberger, U. (2006). Dual- tasking postural control: Aging and the effect of cognitive demand in conjunction with focus of attention. Brain Research Bulletin, 69, 294 –


Johansson, R. S., & Birznieks, I. (2004). First spikes in ensembles of human tactile afferents code complex spatial fingertip events. Nature Neuroscience, 7, 170 –177. Kelly, C. A. M., Uddin, L. Q., Biswal, B. B., Castellanos, F. X., & Milham,

M. P. (2008). Competition between functional brain networks mediates

behavioral variability. NeuroImage, 39, 527–537. Kray, J., Eber, J., & Lindenberger, U. (2004). Age differences in executive functioning across the lifespan: The role of verbalization in task prepa- ration. Acta Psychologica, 115, 143–165. Ku¨gler, C. F. A., Taghavy, A., & Platt, D. (1993). The event-related P300 potential analysis of cognitive human brain aging: A review. Gerontol- ogy, 39, 280 –303. Landau, S. M., Lal, R., O’Neill, J. P., Baker, S., & Jagust, W. J. (2009). Striatal dopamine and working memory. Cerebral Cortex, 19, 445– 454. Li, S.-C., Aggen, S. H., Nesselroade, J. R., & Baltes, P. B. (2001). Short-term fluctuations in elderly people’s sensorimotor functioning predict text and spatial memory performance: The MacArthur successful aging studies. Gerontology, 47, 100 –116. Li, S.-C., Ha¨mmerer, D., Mu¨ller, V., Hommel, B., & Lindenberger, U. (2009). Lifespan development of stimulus-response conflict cost: Simi- larities and differences between maturation and senescence. Psycholog- ical Research, 73, 777–785. Li, S.-C., Huxhold, O., & Schmiedek, F. (2004). Aging and processing robustness: Evidence from cognitive and sensorimotor functioning. Ger- ontology, 50, 28 –34. Li, S.-C., Lindenberger, U., Hommel, B., Aschersleben, G., Prinz, W., & Baltes, P. B. (2004). Lifespan transformations in the couplings of mental abilities and underlying cognitive processes. Psychological Science, 15,


Li, S.-C., Lindenberger, U., Nyberg, L., Heekeren, H. R., & Ba¨ckman, L. (in press). Dopaminergic modulation of cognition in human aging. In W. Jagust & M. D’Esposito (Eds.), Imaging the aging brain. New York, NY: Oxford University Press. Li, S.-C., Lindenberger, U., & Sikstro¨m, S. (2001). Aging cognition: From neuromodulation to representation. Trends in Cognitive Sciences, 5, 479 – 486. Li, S.-C., & Sikstro¨m, S. (2002). Integrative neurocomputational perspec- tives on cognitive aging, neuromodulation, and representation. Neuro- science and Biobehavioral Reviews, 26, 795– 808. Li, S.-C., von Oertzen, T., & Lindenberger, U. (2006). A neurocomputa-

tional model of stochastic resonance and aging. Neurocomputing, 69,


Liang, M., Zhou, Y., Jiang, T., Liu, Z., Tian, L., Liu, H., & Hao, Y. (2006). Widespread functional disconnectivity in schizophrenia with resting state functional magnetic resonance imaging. NeuroReport, 17, 209 –


Lindenberger, U., & von Oertzen, T. (2006). Variability in cognitive aging:

From taxonomy to theory. In F. I. M. Craik & E. Bialystok (Eds.), Lifespan cognition: Mechanisms of change (pp. 297–314) Oxford: Ox- ford University Press. Logan, J. M., Sanders, A. L., Snyder, A. Z., Morris, J. C., & Buckner, R. L. (2002). Under-recruitment and non-selective recruitment: Dissociable neural mechanisms associated with aging. Neuron, 33, 827– 840.

Lo¨vde´n, M., Li, S.-C., Shing, Y. L., & Lindenberger, U. (2007). Within- person trial-by-trial variability precedes and predicts cognitive decline in old and very old age: Longitudinal data from the Berlin Aging Study. Neuropsychologia, 45, 2827–2838. Luszcz, M. A. (2004). What’s it all about: Variation and aging. Gerontol- ogy, 50, 5– 6. MacDonald, S. W. S., Cervenka, S., Farde, L., Nyberg, L., & Ba¨ckman, L. (2009). Extrastriatal dopamine D2 receptor binding modulates intrain- dividual variability in episodic memory and executive functioning. Neu- ropsychologia, 47, 2299 –2304. MacDonald, S. W. S., Hultsch, D. F., & Bunce, D. (2006). Intraindividual variability in vigilance performance: Does degrading visual stimuli mimic age-related “neural noise”? Journal of Clinical and Experimental Neuropsychology, 28, 655– 675. MacDonald, S. W. S., Hultsch, D. F., & Dixon, R. A. (2003). Performance variability is related to change in cognition: Evidence from the Victoria Longitudinal Study. Psychology and Aging, 18, 510 –523. MacDonald, S. W. S., Hultsch, D. F., & Dixon, R. A. (2008). Predicting impending death: Inconsistency in speed is a selective and early marker. Psychology and Aging, 23, 595– 607. MacDonald, S. W. S., Nyberg, L., & Ba¨ckman, L. (2006). Intra-individual variability in behavior: Links to brain structure, neurotransmission and neuronal activity. Trends in Neurosciences, 29, 475– 480. MacDonald, S. W. S., Nyberg, L., Sandblom, J., Fischer, H., & Ba¨ckman, L. (2008). Increased response-time variability is associated with reduced inferior parietal activation in recognition memory in aging. Journal of Cognitive Neuroscience, 20, 779 –786. Manoach, D. S. (2003). Prefrontal cortex dysfunction during working memory performance in schizophrenia: Reconciling discrepant findings. Schizophrenia Research, 60, 285–298. Martin, M., & Hofer, S. M. (2004). Intraindividual variability, change, and aging: Conceptual and analytical issues. Gerontology, 50, 7–11. Mattay, V. S., Goldberg, T. E., Fera, F., Hariri, A. R., Tessitore, A.,

Weinberger, D. (2003, May 13). Catechol

O-methyltransferase val158 –met genotype and individual variation in the brain response to amphetamine. Proceedings of the National Acad- emy of Sciences of the United States of America, 100, 6186 – 6191. McIntosh, A. R., Kovacevic, N., & Itier, R. J. (2008). Increased brain signal variability accompanies lower behavioral variability in develop- ment. PLoS Computational Biology, 4, 1–9. Montague, P. R., Hyman, S. E., & Cohen, J. D. (2004, October 14). Computational roles for dopamine in behavioural control. Nature, 431, 760 –767. Moss, F., Ward, L. M., & Sannita, W. G. (2004). Stochastic resonance and sensory information processing: A tutorial and review of application. Clinical Neurophysiology, 115, 267–281. Murtha, S., Cismaru, R., Waechter, R., & Chertkow, H. (2002). Increased variability accompanies frontal lobe damage in dementia. Journal of the International Neuropsychological Society, 8, 360 –372. Nesselroade, J. R. (1991). The warp and woof of the developmental fabric. In R. Downs, L. Liben, & D. S. Palermo (Eds.), Visions of aesthetics, the environment, and development: The legacy of Joachim F. Wohlwill (pp. 213–240). Hillsdale, NJ: Erlbaum.

Nesselroade, J. R. (2002). Elaborating the different in differential psychol- ogy. Multivariate Behavioral Research, 37, 543–561. Nesselroade, J. R., & Salthouse, T. A. (2004). Methodological and theo- retical implications of intraindividual variability in perceptual-motor performance. Journal of Gerontology, Series B: Psychological Sciences and Social Sciences, 59, P49 –P55. Nyberg, L., Andersson, M., Forsgren, L., Jakobsson-Mo, S., Larsson, A.,

Ba¨ckman, L. (2009). Striatal dopamine D2 binding is

related to frontal BOLD response during updating of long-term memory representations. NeuroImage, 46, 1194 –1199. Nyberg, L., Sandblom, J., Jones, S., Neely, A. S., Petersson, K. M., Ingvar,

Egan, M. F. ,

Marklund, P.,



M., & Ba¨ckman, L. (2003, November 11). Neural correlates of training- related memory improvement in adulthood and aging. Proceedings of the National Academy of Sciences of the United States of America, 100, 13728 –13733. O’Reilly, R. C., & Frank, M. J. (2006). Making working memory work: A computational model of learning in the prefrontal cortex and basal ganglia. Neural Computation, 18, 283–328. Pfefferbaum, A., Rosenbloom, M., Serventi, K. L., & Sullivan, E. V. (2002). Corpus callosum, pons, and cortical white matter in alcoholic women. Alcoholism: Clinical & Experimental Research, 26, 400 – 406.

Priplata, A., Niemi, J., Salen, M., Harry, J., Lipsitz, L. A., & Collins, J. J. (2002). Noise-enhanced human balance control. Physical Review Let- ters, 89, 238101. Rabbitt, P., Osman, P., Moore, B., & Stollery, B. (2001). There are stable individual differences in performance variability, both from moment to moment and from day to day. Quarterly Journal of Experimental Psy- chology Section A, 54, 981–1003. Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard,

D. A., & Shulman, G. L. (2001, January 16). A default mode of brain

function. Proceedings of the National Academy of Sciences of the United States of America, 98, 676 – 682. Ram, N., Rabbitt, P., Stollery, B., & Nesselroade, J. R. (2005). Cognitive

performance inconsistency: Intraindividual change and variability. Psy- chology and Aging, 20, 623– 633. Ratcliff, R., Van Zandt, T., & McKoon, G. (1999). Connectionist and diffusion models of reaction time. Psychological Review, 106, 261–300. Raz, N., Gunning-Dixon, F. M., Head, D., Dupuis, J. H., & Acker, J. D. (1998). Neuroanatomical correlates of cognitive aging: Evidence from structural magnetic resonance imaging. Neuropsychology, 12, 95–114. Raz, N., Gunning-Dixon, F. M, Head, D., Rodrigue, K. M., Williamson,

A., & Acker, J. D. (2004). Aging, sexual dimorphism, and hemispheric

asymmetry of the cerebral cortex: Replicability of regional differences in volume. Neurobiology of Aging, 25, 377–396.

Ro¨cke, C., Li, S.-C., & Smith, J. (in press). Intraindividual variability in positive and negative affect over 45 days: Do older adults fluctuate less than young adults? Psychology and Aging. Rolls, E. T., Loh, M., Deco, G., & Winterer, G. (2008). Computational models of schizophrenia and dopamine modulation in the prefrontal cortex. Nature Reviews Neuroscience, 9, 696 –709. Salthouse, T. A. (2007). Implications of within-person variability in cog- nitive and neuropsychological functioning on the interpretation of change. Neuropsychology, 21, 401– 411. Schmiedek, F. (2006, April). The dark side of the mean: Plenary on intraindividual variability. Paper presented at the 11th biennial Cogni- tive Aging Conference, Atlanta, GA. Schott, B. H., Minuzzi, L., Krebs, R. M., Elmenhorst, D., Lang, M., Winz,

Bauer, A. (2008). Mesolimbic functional magnetic resonance

imaging activations during reward anticipation correlate with reward-

O. H.,

related ventral striatal dopamine release. Journal of Neuroscience, 28,


Seamans, J. K., & Yang, C. R. (2004). The principal features and mech- anisms of dopamine modulation in the prefrontal cortex. Progress in Neurobiology, 74, 1–58. Shannon, B. J., & Buckner, R. L. (2004). Functional–anatomic correlates of memory retrieval that suggest nontraditional processing roles for multiple distinct regions within posterior parietal cortex. Journal of Neuroscience, 24, 10084 –10092. Shipley, B. A., Der, G., Taylor, M. D., & Deary, I. J. (2006). Cognition and all-cause mortality across the entire adult age range: Health and lifestyle survey. Psychosomatic Medicine, 68, 17–24. Siegler, R. S. (1994). Cognitive variability: A key to understanding cog- nitive development. Current Directions in Psychological Science, 3, 1–5. Sowell, E. R., Peterson, B. S., Thompson, P. M., Welcome, S. E., Henkenius,

A. L., & Toga, A. W. (2003). Mapping cortical change across the human life

span. Nature Neuroscience, 6, 309 –315. Stefanis, N. C., van Os, J., Avramopoulos, D., Smyrnis, N., Evdokimidis,

I., & Stefanis, C. N. (2005). Effect of COMT Val158Met polymorphism

on the continuous performance test, identical pairs version: Tuning rather than improving performance. American Journal of Psychiatry, 162, 1752–1754. Stein, R. B., Gossen, E. R., & Jones, K. E. (2005). Neuronal variability:

Noise or part of the signal? Nature Reviews Neuroscience, 6, 389 –397. Stern, Y. (2002). What is cognitive reserve? Theory and research applica- tion of the reserve concept. Journal of the International Neuropsycho- logical Society, 8, 448 – 460. Stern, Y., Zarahn, E., Hilton, H. J., Flynn, J., DeLaPaz, R., & Rakitin, B. (2003). Exploring the neural basis of cognitive reserve. Journal of Clinical and Experimental Neuropsychology, 25, 691–701.

Strauss, E., MacDonald, S. W. S., Hunter, M. A., Moll, A., & Hultsch, D. F. (2002). Intraindividual variability in cognitive performance in three groups of adults: Cross-domain links to physical status and self- perceived affect and beliefs. Journal of the International Neuropsychol-

ogy Society, 8, 893–906. Stuss, D. T., & Binns, M. A. (2008). The patient as a moving target: The importance to rehabilitation of understanding variability. In D. T. Stuss,

G. Winocur, & I. H. Robertson (Eds.), Cognitive neurorehabilitation:

Evidence and application (2nd ed., pp. 36 – 61). Cambridge, England:

Cambridge University Press. Stuss, D. T., Murphy, K. J., Binns, M. A., & Alexander, M. P. (2003). Staying on the job: The frontal lobes control individual performance variability. Brain, 126, 2363–2380. Sutton, S. (1969). The specification of psychological variables in an average evoked potential experiment. In E. Donchin and D. B. Lindsley (Eds.), Average evoked potentials: Methods, results and evaluations (NASA SP-191, pp. 237–262). Washington, DC: U.S. Government Printing Office. Takahashi, H., Kato, M., Hayashi, M., Okubo, Y., Takano, A., Ito, H., &

Suhara, T. (2007). Memory and frontal lobe functions; possible relations with dopamine D2 receptors in the hippocampus. NeuroImage, 34,


Zhuo, Y.

(2006). Altered resting-state functional connectivity patterns of anterior cingulate cortex in adolescents with attention deficit hyperactivity dis- order. Neuroscience Letters, 400, 39 – 43. Wagner, A. D., Shannon, B. J., Kahn, I., & Buckner, R. L. (2005). Parietal lobe contributions to episodic memory retrieval. Trends in Cognitive Sciences, 9, 445– 453. Walhovd, K. B., & Fjell, A. M. (2007). White matter volume predicts reaction time instability. Neuropsychologia, 45, 2277–2284. Weinshilboum, R. M., Otterness, D. M., & Szumlanski, C. L. (1999). Methylation pharmacogenetics: Catechol O-methyltransferase, thiopu- rine methyltransferase, and histamine N-methyltranferase. Annual Review of Pharmacology and Toxicology, 39, 19 –52. Weissman, D. H., Roberts, K. C., Visscher, K. M., & Woldorff, M. G.

(2006). The neural bases of momentary lapses in attention. Nature Neuroscience, 9, 971–978. West, R., Murphy, K. J., Armilio, M. L., Craik, F. I. M., & Stuss, D. T. (2002). Lapses of intention and performance variability reveal age- related increases in fluctuations of executive control. Brain and Cogni- tion, 49, 402– 419. Williams, B. R., Hultsch, D. F., Strauss, E., Hunter, M. A., & Tannock, R. (2005). Inconsistency in reaction time across the lifespan. Neuropsy- chology, 19, 88 –96. Williams, B. R., Strauss, E., Hultsch, D. F., & Hunter, M. A. (2007). Reaction time inconsistency in a spatial Stroop task: Age-related differ- ences through childhood and adulthood. Aging, Neuropsychology, and Cognition, 14, 417– 439.

Tian, L., Jiang, T., Wang, Y., Zang, Y., He, Y., Liang, M.,



Winterer, G., Coppola, R., Goldberg, T. E., Egan, M. F., Jones, D. W., Sanchez, C. E., & Weinberger, D. R. (2004). Prefrontal broadband noise, working memory, and genetic risk for schizophrenia. American Journal of Psychiatry, 161, 490 –500. Winterer, G., & Weinberger, D. R. (2004). Genes, dopamine, and cortical signal- to-noise ratio in schizophrenia. Trends in Neurosciences, 27, 683–690. Wohlwill, J. F. (1973). The study of behavioral development. New York, NY: Academic Press.

Yamaguchi, S., Hale, L. A., D’Esposito, M., & Knight, R. T. (2004). Rapid prefrontal– hippocampal habituation to novel events. Journal of Neuro- science, 24, 5356 –5363.

Received November 7, 2008 Revision received August 24, 2009 Accepted September 25, 2009

Neuro- science, 24, 5356 –5363. Received November 7, 2008 Revision received August 24, 2009 Accepted September