Abstract

A fundamental issue in cognitive neuroscience is the nature of developmental changes in human cerebral functional organization for higher cognitive functions. Event-related functional magnetic resonance imaging was used to measure developmental changes in the functional neuroanatomy subserving controlled lexical association. First, brain regions showing significant differences in activity between school-age children and young adults, despite equivalent task performance, were identified. Then, activity in these regions was more fully characterized in individuals spanning the ages of 7–32 years old. Cross-sectional and regression analyses showed systematic increases and decreases in levels of activity over age, by region. Age-related increases in activity were primarily newly recruited, later-stage processing regions, such as in left frontal and left parietal cortex. Decreases, on the other hand, were all positive activations that attenuated with age and were found across a wider neuroanatomical range, including earlier processing regions such as bilateral extrastriate cortex. The hemodynamic magnitude, neuroanatomical location and maturational timecourse of these progressive and regressive changes have implications for models of the developing specialization in human cerebral functional organization.

Introduction

The human brain likely undergoes dramatic changes in its functional organization in relation to cognitive development. The nature of these changes is a fundamental issue in cognitive neuroscience. A better understanding of maturational changes in human cerebral functional organization might inform our view of cognitive development, give insight into the mature organization, and could potentially lead to new ways of identifying, remedying and possibly preventing developmental cognitive disorders.

Theories of the relationship between brain maturation and cognitive development have emphasized both progressive and regressive neural mechanisms of change — i.e. regions of the brain becoming increasingly or decreasingly involved in cognitive tasks with age. Scientists have proposed competing views that either regressive or progressive neurobiological changes are the principal mechanism underlying cognitive growth. Often referred to as ‘selectionist’ and ‘constructivist’ traditions, several such models have been outlined in detail (e.g. Hebb, 1949; Jerne, 1967; Changeux and Danchin, 1976; Gottlieb, 1976; Mehler, 1985; Edelman, 1987; Greenough et al., 1987; Quartz and Sejnowski, 1997; Bourgeois, 2001; see also Thelen and Smith, 1994; Elman et al., 1996).

For example, according to the selective stabilization hypothesis proposed by Changeux and Danchin (1976), cortical connectivity progresses from greater to fewer connections by way of regressive phenomena. Mechanisms that are largely intrinsic control the initial formation of large pools of synaptic contacts. Extrinsic, evoked activity largely determines the subsequent stabilization of the most active synapses and the elimination of the less active ones. In the ‘constructivist manifesto’ proposed by Quartz and Sejnowski (1997), on the other hand, connectivity progresses from fewer to greater connections. According to this model, spontaneous activity and patterns of evoked activity cause the formation of new synapses (through dendritic and axonal arborization and other mechanisms) and bias elaboration of the structural organization. The losses of neurons and synapses, in their view, are simply epiphenomenal. These theories, though often highly specified in terms of purported underlying mechanisms, have not drawn heavily on human developmental evidence that links brain measures with behavior.

Studies of human neuroanatomical growth have also focused on both progressive and regressive changes and have provided evidence that different brain structures mature at different rates. Although limited in humans, available evidence suggests frontal brain regions develop more slowly than other regions, maturing even into late adolescence. For example, measures of myelination (Yakovlev and Lecours, 1967; Dobbing and Sands, 1973; Klingberg et al., 1999), synaptic density and pruning (Huttenlocher and Dabholkar, 1997), and gray to white matter proportions (Jernigan et al., 1991; Pfefferbaum et al., 1994; Giedd et al., 1999; Sowell et al., 2003) all show protracted development of frontal cortex.

There is still relatively little empirical evidence demonstrating developmental changes in human cerebral functional organization. Positron emission tomographic (PET) studies have corroborated structural evidence, showing delayed maturation of frontal resting metabolic rates relative to other cortical regions (Chugani et al., 1987). Functional magnetic resonance imaging (fMRI) studies comparing healthy school-age children and young adults during controlled processing tasks have largely confirmed the general belief that frontal maturation is involved (Casey et al., 1995, 1997, 2002; Thomas et al., 1999; Gaillard et al., 2000, 2003; Holland et al., 2001; Luna et al., 2001; Bunge et al., 2002; Klingberg et al., 2002; Kwon et al., 2002; Schlaggar et al., 2002b; Tamm et al., 2002; Booth et al., 2003). These studies, to varying degrees, have found both progressive and regressive changes in the developing functional neuroanatomy. Nevertheless, they have not concentrated on a functional distinction between these two kinds of change.

Regardless of the specific underlying mechanisms, fMRI can be used to characterize and compare progressive and regressive changes over development in several ways. Hemodynamic response timecourses can be used to distinguish different kinds of involvement among brain regions that change over development. For example, among regions that show increases with age in levels of activity for the same tasks, regions that are not activated by children but activated by adults would be interpreted to play a different functional role than regions that are deactivated by children and not activated by adults. Likewise, information on differences between progressive and regressive developmental changes in neuroanatomical location and in developmental timing (i.e. the rate at which activity matures) would help to develop a functional distinction between them.

The purpose of the present study was to explore this functional distinction by providing an empirical characterization of progressive and regressive developmental changes in the functional brain organization underlying controlled lexical association. Because of their simplicity and flexibility for use in a variety of experimental paradigms, controlled lexical association or word generation tasks have been commonly used with healthy subjects and diverse clinical groups, across many ages (e.g. Petersen et al., 1988, 1990; Posner et al., 1988; Wise et al., 1991; McCarthy et al., 1993; Petersen and Fiez, 1993; Raichle et al., 1994; Buckner et al., 1995; Passingham, 1996; Seger et al., 1997; Thompson-Schill et al., 1998; Barch et al., 2000; Holland et al., 2001; Ojemann et al., 2002; Schlaggar et al., 2002b; Burton et al., 2003; Corina et al., 2003; Gaillard et al., 2003). These tasks are among the most frequently studied and perhaps best characterized in functional neuroimaging. They, therefore, provide a particularly reliable context in which to build developmental theory.

In order to test for effects in a way that allows distinctions to be made among various types of developmental change, we combined several conceptual and methodological features. First, levels of brain activity were measured and analyzed as hemodynamic response timecourses in an event-related design, not as relative differences between children and adults. This provided a check for the physiological plausibility of observed effects and allowed for the dissociation of regions that are activated, not activated or deactivated across age — distinctions that are key to the current purpose. Second, we tested for effects that were statistically reliable across several related tasks, allowing for the identification of developmental changes that are general to these controlled lexical task types. The focus of this report is not on differences among these tasks but rather in their functional neuroanatomical and developmental commonalities. Third, we employed an approach for exploring the potentially confounding effects of task performance discrepancy during imaging, thereby increasing the confidence with which we could interpret differences across age as being related to changes in functional brain organization. Finally, we used both cross-sectional and regression analysis approaches in a rather large sample size, providing measures of the rates of functional maturation for all developing regions. By investigating developmental changes in the functional neuroanatomy of word generation in this way, we hoped to advance the current understanding of the functional brain changes associated with the development of controlled cognition.

Materials and Methods

General Methods

Subjects

Ninety-five healthy individuals 7–32 years old served as participants in the study. All were right-handed native English speakers. All adult subjects gave informed consent, and subjects under the age of 18 gave assent with parental informed consent. Individuals with metal implants, heart arrhythmias, claustrophobia or a history of head trauma, neurological or psychiatric illness, including the use of psychotropic medications, were excluded. Adult subjects were screened by interview and questionnaire. Minor subjects were examined by a pediatric neurologist (B.L.S.) and screened for neurological, psychiatric and medical problems that may impact development. To document intellectual level, minor participants were administered the Wechsler Abbreviated Scale of Intelligence (Wechsler, 1999) by a neuropsychologist at St Louis Children's Hospital. Adult participants were college students and graduates, graduate and medical students, and one postdoctoral fellow, mostly from Washington University. Adult intellectual levels were calculated using established full-scale IQ estimates based on educational attainment (Matarazzo, 1972). Children were acclimated to the MRI environment through the use of a mock scanner several days prior to the experiment. The study was approved by the Washington University Human Studies Committee and all subjects were reimbursed for their participation.

Word Generation Tasks

All subjects were scanned while performing three controlled lexical association tasks: (1) verb generation in response to a noun (e.g. stimulus: ‘CAR’, a correct response: ‘drive’); (2) opposite generation (e.g. stimulus: ‘UP’, a correct response: ‘down’); and (3) rhyme generation (e.g. stimulus: ‘HAT’, a correct response: ‘cat’). These tasks were chosen because they are similar to those previously employed in many adult studies, and the neural substrates underlying these tasks are thought to undergo significant development across the ages we studied here.

While in the scanner, stimuli for the three tasks were presented to participants in both visual and auditory modalities. Therefore, all participants completed a total of six task runs. Subjects were asked to generate single word responses aloud to stimulus words that either appeared on a screen or were heard through headphones. They were instructed to give each response as quickly and accurately as possible and to minimize head movement, even while speaking. Subjects' overt verbal responses were recorded during scanning, allowing reaction times and accuracy to be measured, as previously described (Nelles et al., 2003). Each word was presented every second, third or fourth MR frame (TR = 3.08 s; average interstimulus interval = 9.24 s) in pseudo-random fashion. This jittering allowed the event-related timecourse of the response to be extracted (Miezin et al., 2000).

Visual stimulus duration was relatively prolonged to aid reading (1.37 s) and auditory stimuli varied by word length. Each run lasted 3 min 39 s and consisted of 21 stimulus trials (i.e. 21 words were presented), followed by ∼90 s of rest between runs. Each run included only one task type with words presented only in one sensory modality. Task runs were presented in a pseudo-randomized order across subjects. Words were white on black, and each letter subtended ∼0.5°. A fixation crosshair was presented in the center of the screen at the beginning of each run and remained on the screen for the duration of the run, except when replaced by a visual stimulus word. For auditory tasks, the crosshair remained on the screen. Subjects were asked to maintain visual fixation on the crosshair. Word lists were counterbalanced for syllables, and words were presented in random order. Stimulus words were derived from available lists of US children's first-encountered reading words.

Imaging Data Acquisition and Preprocessing

Functional and structural neuroimaging data were collected on a Siemens 1.5 Tesla MAGNETOM Vision system (Erlangen, Germany) as previously described (Miezin et al., 2000). Rapid three-dimensional high-resolution structural images were acquired using a sagittal magnetization-prepared rapid gradient echo (MP-RAGE) sequence (slice TE = 4 ms, TR = 9.7 ms, TI = 300 ms, flip angle = 12°, 128 slices, 1.25 × 1 × 1 mm voxels). Functional data were collected parallel to the anterior commissure–posterior commissure plane using an asymmetric spin-echo echo-planar pulse sequence sensitive to blood oxygenation level-dependent (BOLD) contrast (TR = 3.08 s, T2* evolution time = 50 ms, flip angle = 90°). During each scan, 73 frames of 16 contiguous interleaved 8 mm axial slices were acquired (3.75 × 3.75 mm in-plane resolution), allowing complete brain coverage. Steady state was assumed after three frames (∼ 9 s). Thus, acquisition of functional imaging data began with the fourth frame of each run.

Preliminary automated image data processing was carried out in order to remove noise and artifacts (see Miezin et al., 2000, for detailed procedures). This preprocessing included removal of a single pixel spike caused by signal offset, whole-brain normalization of signal intensity across MR frames, correction for subject movement within and across runs, and slice-by-slice normalization to correct for changes in signal intensity introduced by the acquisition of interleaved slices.

Prior to statistical analysis, functional (BOLD) data were registered to the structural (MP-RAGE) data for a given subject. Importantly, data for all subjects were transformed into the same standard stereotactic space (based on Talairach and Tournoux, 1988), allowing direct voxel-wise statistical comparisons to be made. It has been assumed that the variability in children's structural and functional neuroanatomy and brain size is sufficient to preclude direct statistical comparisons with adults within the same stereotactic space. Recent empirical studies demonstrate, however, that these differences are below a level that would adversely affect results in most current fMRI experiments (Muzik et al., 2000; Burgund et al., 2002; Schlaggar et al., 2002b; Kang et al., 2003).

To encourage minimal movement, subjects were positioned in the scanner using a thermoplastic mask individually fitted to the face and attached to the head coil. Subject motion was corrected and quantified using an analysis of head position based on rigid body translation and rotation. The data derived from the adjustments needed to realign head movement on a frame-by-frame basis were calculated as root mean square (RMS) values for translation and rotation in the x, y and z planes in millimeters. Total RMS values were calculated on a run-by-run basis for each subject derived from deviations from the initial position for each run. For each subject, a median RMS was calculated for the six runs. Median RMS values ranged from 0.19 to 1.37 mm, with an average of 0.47 mm of movement over all tasks and subjects. Median RMS was negatively correlated with age, showing that head motion decreased with increasing subject age (r = −0.57, P < 0.001).

Imaging Data Analyses

Statistical analyses of event-related fMRI data were based on the general linear model (GLM) as previously described (Miezin et al., 2000; Schlaggar et al., 2002b), conducted using in-house software programmed in the Interactive Data Language (IDL; Research Systems, Inc., Boulder, CO). The purpose of the present study was to identify common functional neuroanatomical and developmental aspects of controlled lexical association across the three task types and across visual and auditory stimulus modalities. Studies in adults comparing lexical tasks requiring phonological versus semantic association have shown considerable overlap in their patterns of activation (Klein et al., 1995; Duncan and Owen, 2000; McDermott et al., 2003). Likewise, in direct comparisons among verb-, opposite- and rhyme-generation tasks, we have found similarities in recruited regions across these tasks for both children and adults (Schlaggar et al., 2002b). In addition, preliminary studies have yielded very few regions that show stimulus modality effects in these tasks. In a brain-wide direct statistical comparison, only four regions showed modality specificity — all in extrastriate cortex (Schlaggar et al., 2002a). So, although important for future studies, a characterization of individual task and modality effects is orthogonal to the scope of the current inquiry. Therefore, for the present purposes, we combined imaging data for each subject across verb-, opposite- and rhyme-generation tasks, and across visual and auditory stimuli. Importantly, all analyses for all subjects included functional imaging data only for trials in which correct behavioral responses were made. In this way, all comparisons included only brain responses reflecting successful task performance.

The GLM design included time as a seven-level factor, made up of the seven MR frames following presentation of the stimulus. Since no assumptions were made regarding the shape of the hemodynamic response function (HRF), this allowed detection of a BOLD response with any shape over time. The HRF (% BOLD signal change as a function of time) was modeled over a period of ∼21 s (7 frames, 3.08 s per frame). For all analyses, timecourse values were entered into an ANOVA using a random effects model. A significant effect of time indicated that the detected hemodynamic response was not a flat line across the seven frames.

Timecourses for all subjects across all regions were screened for highly aberrant values, blind to age. Outlier timecourses were defined as those showing any time point with greater than 2% signal change. Regions containing 10 or more outlying timecourses across subjects were removed from subsequent analyses. All regions removed in this way were at or near the edges of brain space. Regions showing fewer than 10 outlying timecourses were retained, but data from those particular subjects were removed for analyses of that region. Regions that showed effects but were determined by close inspection of the group anatomical average to be squarely within white matter or ventricle were also removed.

For all voxel-wise analyses, a correction based on Monte Carlo simulation was implemented to guard against false positives that may result from conducting a large number of statistical comparisons over many images (Forman et al., 1995; McAvoy et al., 2001). To achieve P < 0.05 corrected for voxel clusters, we used a threshold of 24 contiguous voxels with a z-value > 3.5. For all region-wise analyses, Box's sphericity correction was used, adjusting for temporal autocorrelation and possible inhomogeneity of variance over the repeated measure (i.e. time).

Analysis Step 1: Identification of Brain Regions Showing Age/Performance Independent, Performance-related and Age-related effects Using Endpoint Age Groups

Subjects

In order to identify brain regions showing different functional relationships between childhood and adulthood, we began with two ‘endpoint’ age groups made up of only our youngest and oldest subjects. Twenty-six adults (13 male, 13 female; aged 18–32, mean age 24.9) and 32 children (14 male, 18 female; aged 7–10, mean age 9.3) were included. Using an approach similar to that of Schlaggar et al. (2002b), we used performance matching in order to identify regions showing three different kinds of responses: (i) age/performance independent effects; (ii) performance-related effects; and (iii) age-related effects.

In regions showing age/performance independent effects, children and adults show statistically comparable responses regardless of differences in levels of task performance (as measured by reaction time and accuracy). Regions showing performance-related effects are those where children and adults show differences in activity only in subjects who show statistically differing task performance. In other words, these brain regions appear to be used similarly by children and adults when they perform similarly. Performance-related regions are those where differences, if performance matching were not done, would likely be mistakenly attributed to maturational differences in functional organization. Brain regions showing age-related effects, on the other hand, are regions where children and adults show significant differences in activity even when they exhibit statistically indistinguishable behavioral performance. Voxel-level analyses were used to test brain-wide for candidate regions showing these three effect types, and region-level analyses were used to extract precise timecourses of activity and to verify the specific functional relations.

Identification of Age/Performance-independent Effects

In order to identify brain regions that show statistically equivalent activation between children and adults, we first tested for a main effect of time in the ANOVA model. The most highly significant main effects of time constitute timecourses that are likely to be (but are not necessarily) statistically similar between the two age groups. So, a voxel-wise main effect of time image was created. A peak-finding and region-defining algorithm was then used to obtain center of mass and cluster coordinates for statistically significant voxels in the time image. Peaks were identified using a 4 mm hard sphere preblur and a statistical threshold of z > 11.0. Regions were then defined beginning with a radius of 10 mm, searching for significant voxel clusters around the peaks that surpassed the Monte Carlo correction, and then consolidating regions with extremes closer than 10 mm. Regions smaller than 24 voxels (the size of the Monte Carlo correction) were eliminated from further analysis. Negative-going main effects of time were also eliminated; for the present report, we sought regions where children and adults showed equivalent positive activations only. A region-wise ANOVA was then conducted on brain activity within the remaining regions of interest, comparing the endpoint child and adult groups. If there was no statistical difference between the adult and child timecourses, the region was considered to be age/performance independent.

Identification of Performance-related and Age-related Effects

In order to test for overall group differences, age group was included as a between-subjects factor in the GLM. Thus, an age group (2 levels: child/adult) × time (7 levels: BOLD measures every ∼3 s for ∼21 s) ANOVA model was used. An activation difference between children and adults was expressed, by region, as a significant group × time interaction. In other words, regions where adults and children differed were those in which the hemodynamic response over time differed across the two age groups. Peaks were identified and regions were defined in the age group × time image using the same algorithm and parameters applied to the main effect of time image, this time with a statistical threshold of z > 3.0.

We next sought to distinguish regions that differed across age regardless of performance (age-related regions) from regions that showed age differences related to performance discrepancy (performance-related regions). Although the child and adult groups differed on average in their task performance as measured by percent correct (%cor) and reaction time (RT) (t = 7.10, P < 0.001; t = 5.60, P < 0.001, respectively), some subjects from each age group did not differ. From the original two groups, we selected a subset of 10 adults and 10 children who, during spontaneous task performance, were statistically equivalent in both %cor and RT, creating four total groups — performance matched children and adults and non-matched children and adults (See Table 1).

Table 1

Demographic and task performance characteristics of the ‘endpoint’ groups

 Children, ages 7–10
 
 Adults, ages 18–32
 
 

 
Nonmatched
 
Matched
 
Matched
 
Nonmatched
 
n 22 10 10 16 
Gender 12 F/10 M 6 F/4 M 5 F/5 M 8 F/8 M 
Age (years) 9.4 (0.9) 9.2 (1.2) 22.7 (1.8) 26.3 (3.6) 
% correct 69.1 (11.7) 76.2a (8.6) 82.7a (5.1) 92.6 (3.0) 
RT (ms) 2236 (478) 1661a (257) 1556a (265) 1390a (236) 
Verbal IQ 121a (12) 123a (15) – – 
Perf IQ 109a (12) 114a (17) – – 
Full Scale IQ
 
117a (11)
 
121a (17)
 
116a (13)
 
118a (12)
 
 Children, ages 7–10
 
 Adults, ages 18–32
 
 

 
Nonmatched
 
Matched
 
Matched
 
Nonmatched
 
n 22 10 10 16 
Gender 12 F/10 M 6 F/4 M 5 F/5 M 8 F/8 M 
Age (years) 9.4 (0.9) 9.2 (1.2) 22.7 (1.8) 26.3 (3.6) 
% correct 69.1 (11.7) 76.2a (8.6) 82.7a (5.1) 92.6 (3.0) 
RT (ms) 2236 (478) 1661a (257) 1556a (265) 1390a (236) 
Verbal IQ 121a (12) 123a (15) – – 
Perf IQ 109a (12) 114a (17) – – 
Full Scale IQ
 
117a (11)
 
121a (17)
 
116a (13)
 
118a (12)
 
a

Statistically equivalent across groups by ANOVA, P < 0.05.

It is important to note that we did not conduct performance matching at the trial level — i.e. choosing trials from across individuals where performance was the best and comparing it to trials in which performance was the worst. Rather, we simply capitalized on the spontaneous clustering of children and adults whose overall accuracy and reaction times during scanning were statistically indistinguishable and, therefore, suggested similar levels of ability and effort at the time of the study.

A region-wise ANOVA was then conducted to examine, in performance-matched and non-matched children and adults, task-related brain activity in the regions identified as showing significant group differences overall. This investigation of performance allowed us to classify these brain regions as showing effects that were likely driven by either age group or performance. Performance-related regions were identified as those that showed an age difference for performance non-matched groups but showed no significant difference when performance was matched. In other words, performance, not age, appeared to be driving the differences in these regions. Age-related regions, on the other hand, were defined as those regions that exhibited a difference between children and adults for both matched and non-matched groups. These regions showed statistically robust differences in brain activity even in children and adults who showed statistically equivalent task performance.

Analysis Step 2: Characterization of brain activity in age-related regions over development using cross-sectional and regression analyses

Subjects

In order to explore how task-related brain activity in the 40 age-related regions of interest changes over maturation, we then conducted analyses using BOLD data from all 95 subjects across all ages, filling in the age range between our endpoint groups. Using both cross-sectional and regression analysis approaches by age, we sought a more complete characterization of the developing functional neuroanatomy for controlled lexical association.

Cross-sectional Analysis by Age Group

The 95 participants were divided into six groups of similar size based on age: (1) 7–8 year olds; (2) 9–10 year olds; (3) 11–13 year olds; (4) 14–15 year olds; (5) 16–22 year olds; and (6) 23–32 year olds (see Table 2). These six groupings allowed the measurement of finer developmental changes at younger ages and were statistically balanced for gender (χ2 = 4.65, P = 0.46) and IQ (ANOVA, VIQ P = 0.14; PIQ P = 0.07; FSIQ P = 0.30).

Table 2

Forty-three brain regions showing age/performance-independent effects in children and adults

x
 
y
 
z
 
Size (cm3)
 
Location
 
Approx. Brodmann area
 
Left      
−44 26 3.91 frontal 
−40 −4 34 3.63 frontal 
−39 16 −2 3.42 frontal 13 
−29 41 26 3.75 frontal 10 
−25 30 23 3.81 frontal 
−1 10 41 4.22 med. front./ant. cing. 32 
−40 −39 26 3.68 parietal 40 
−19 −35 54 4.14 parietal 
−28 −57 43 3.72 parietal 
−45 41 3.71 med. par./post. cing. 31 
−9 −70 −11 3.70 med. occip./med. temp. 18 
−9 −45 −4 3.44 med. occip./med. temp. 30 
−6 −76 3.55 med. occip./med. temp. 18 
−54 −22 4.22 occipital/temporal 41 
−52 −41 4.02 occipital/temporal 22 
−49 3.88 occipital/temporal 22 
−38 −35 11 3.87 occipital/temporal 41 
−36 −59 −21 3.90 occipital/temporal 37 
−22 −89 −10 3.88 occipital/temporal 18 
−14 −58 −9 3.55 occipital/temporal 19 
−31 −18 13 3.74 subcortical – 
−27 −23 3.71 subcortical – 
−22 −10 3.20 subcortical – 
−15 −6 13 3.68 subcortical – 
−9 −24 4.07 subcortical – 
Right      
39 −3 33 3.72 frontal 
33 −41 25 4.00 frontal 13 
52 −10 25 3.95 parietal 43 
26 −56 45 3.77 parietal 
16 −35 52 3.93 parietal 
−47 3.41 med. par./post. cing. 29 
−34 45 3.48 med. par./post. cing. 
−68 4.06 med. occip./med. temp. 18 
−39 3.29 med. occip./med. temp. 30 
−77 19 3.35 med. occip./med. temp. 18 
55 −35 11 3.29 occipital/temporal 42 
52 −24 3.34 occipital/temporal 41 
47 −33 3.14 occipital/temporal 22 
24 −85 −8 3.30 occipital/temporal 18 
30 −16 −2 4.02 subcortical – 
19 −3 3.99 subcortical – 
17 −1 29 4.18 subcortical – 
9
 
−21
 
6
 
4.14
 
subcortical
 

 
x
 
y
 
z
 
Size (cm3)
 
Location
 
Approx. Brodmann area
 
Left      
−44 26 3.91 frontal 
−40 −4 34 3.63 frontal 
−39 16 −2 3.42 frontal 13 
−29 41 26 3.75 frontal 10 
−25 30 23 3.81 frontal 
−1 10 41 4.22 med. front./ant. cing. 32 
−40 −39 26 3.68 parietal 40 
−19 −35 54 4.14 parietal 
−28 −57 43 3.72 parietal 
−45 41 3.71 med. par./post. cing. 31 
−9 −70 −11 3.70 med. occip./med. temp. 18 
−9 −45 −4 3.44 med. occip./med. temp. 30 
−6 −76 3.55 med. occip./med. temp. 18 
−54 −22 4.22 occipital/temporal 41 
−52 −41 4.02 occipital/temporal 22 
−49 3.88 occipital/temporal 22 
−38 −35 11 3.87 occipital/temporal 41 
−36 −59 −21 3.90 occipital/temporal 37 
−22 −89 −10 3.88 occipital/temporal 18 
−14 −58 −9 3.55 occipital/temporal 19 
−31 −18 13 3.74 subcortical – 
−27 −23 3.71 subcortical – 
−22 −10 3.20 subcortical – 
−15 −6 13 3.68 subcortical – 
−9 −24 4.07 subcortical – 
Right      
39 −3 33 3.72 frontal 
33 −41 25 4.00 frontal 13 
52 −10 25 3.95 parietal 43 
26 −56 45 3.77 parietal 
16 −35 52 3.93 parietal 
−47 3.41 med. par./post. cing. 29 
−34 45 3.48 med. par./post. cing. 
−68 4.06 med. occip./med. temp. 18 
−39 3.29 med. occip./med. temp. 30 
−77 19 3.35 med. occip./med. temp. 18 
55 −35 11 3.29 occipital/temporal 42 
52 −24 3.34 occipital/temporal 41 
47 −33 3.14 occipital/temporal 22 
24 −85 −8 3.30 occipital/temporal 18 
30 −16 −2 4.02 subcortical – 
19 −3 3.99 subcortical – 
17 −1 29 4.18 subcortical – 
9
 
−21
 
6
 
4.14
 
subcortical
 

 

We then performed region-wise ANOVAs across all 40 age-related regions of interest and extracted average timecourses of brain activity for each of the six age groups. All of the regions were characterized according to the specific changes they exhibited in the direction and hemodynamic shape of their BOLD responses over development. Using a combination of statistical reliability and a measure of signal magnitude, we classified the responses at each region for each age group as statistically positive, negative or flat over time. In order for a group's regional timecourse to be considered an activation or deactivation, it was required that activity show a significant effect over time (by ANOVA, corrected for multiple tests and nonsphericity), and that the timecourse exhibit a peak >0.1 or <−0.1 percent signal change. Group timecourses not meeting both of these criteria were deemed to be statistically flat over time.

Across all subjects and all age-related regions, time point three (representing BOLD signal within the period from 6.17 to 9.24 s) showed the greatest variability and was almost invariably the point of peak magnitude for the BOLD response. Therefore, we used time point three to explore changes in peak brain activity and variability over age. Using Levene's test of equality of variance across the six age groups, we tested whether the variability in peak activity changes significantly over development.

Regression Analysis by Age

In order to fully utilize the information provided by this large sample size in the service of characterizing developmental changes, we also used regression analysis with curve fitting to show trends in regional brain activity over maturation. Data from all 95 subjects were entered into a regression model for each of the 40 age-related regions. Peak BOLD activity was modeled as a function of age to tenths of a year. Two nonlinear regression functions were found to optimally explain the data (maximize the coefficients of determination) without being over-parameterized (i.e. no dependencies equaled 1.0). For regions showing systematic increases in activity over age, a three-parameter, single-exponent rise to max function was used: [y = y0 + a(1 − ebx) ]. For regions showing age-related decreases in activity, a complementary three-parameter, single-exponent decay function was used: [y = y0 + aebx ]. Curves for these functions were fitted to the data, showing maturational trends in brain activity for each of the 40 developing regions.

These fitted curves also permitted a calculation of the rate at which brain activity becomes ‘adult-like’ within each of the maturing regions. Using the regression model for each region, we determined the ages at which peak activity became 50% adult-like (i.e. reached a level that is 50% of the total range observed over the youngest and oldest subjects) and 75% adult-like.

Results

Analysis Step 1

Our approach successfully identified and dissociated brain regions showing statistically reliable age/performance-independent effects, performance-related effects and age-related effects (see Fig. 1). Forty-three regions showed similar activation by children and adults, regardless of performance. These age/performance-independent regions were evident throughout the brain, in cortical and subcortical locations bilaterally (see Fig. 2). Using both inspection of the group-averaged anatomy and the Talairach Daemon atlas (Lancaster et al., 2000), we classified the 43 regions according to their broad neuroanatomical locations and approximate Brodmann areas (see Table 2). Age/performance-independent effects were seen mostly in left and right frontal cortex, anterior and posterior cingulate, left, right and medial occipital and temporal cortex, and thalamus.

Figure 1.

Three developmental functional relationships. The hemodynamic response function timecourses for performance matched and nonmatched adults and children and the P-values derived from the group × time interactions are shown for four example regions derived from the ANOVA. Three of the regions are derived from the interaction image and one region is derived from the main effect of time image. The percentage of MR signal change is plotted against time in seconds (see legend icon in lower left). The numeric coordinates (e.g. −49, 3, 39) indicate the location of the center of mass of the region in standard atlas space. (A) Age/performance-independent effects. Regions that exhibited age/performance-independent effects are shown in green. The example region is indicated with an oval. (B) Performance-related effects. Regions that exhibited performance-related effects are shown in yellow, with the example region indicated by an oval. (C) Age-related regions. Regions that exhibited age-related effects are shown in red and blue. Red indicates regions where adults showed greater levels of activity than children and blue indicates regions where children showed greater activity than adults. One of each region type is indicated by ovals. All surface-rendered images were created using CARET software and surface-based atlases (Van Essen et al., 2001; Van Essen, 2002).

Figure 1.

Three developmental functional relationships. The hemodynamic response function timecourses for performance matched and nonmatched adults and children and the P-values derived from the group × time interactions are shown for four example regions derived from the ANOVA. Three of the regions are derived from the interaction image and one region is derived from the main effect of time image. The percentage of MR signal change is plotted against time in seconds (see legend icon in lower left). The numeric coordinates (e.g. −49, 3, 39) indicate the location of the center of mass of the region in standard atlas space. (A) Age/performance-independent effects. Regions that exhibited age/performance-independent effects are shown in green. The example region is indicated with an oval. (B) Performance-related effects. Regions that exhibited performance-related effects are shown in yellow, with the example region indicated by an oval. (C) Age-related regions. Regions that exhibited age-related effects are shown in red and blue. Red indicates regions where adults showed greater levels of activity than children and blue indicates regions where children showed greater activity than adults. One of each region type is indicated by ovals. All surface-rendered images were created using CARET software and surface-based atlases (Van Essen et al., 2001; Van Essen, 2002).

Figure 2.

Age/performance-independent effects. Regions that exhibited positive age/performance-independent effects are shown in green (43 regions). These regions were activated by children and adults to a statistically equivalent degree, regardless of performance. All effects shown surpassed a z-statistic of 11.0, and negative-going timecourses (deactivations) were excluded.

Figure 2.

Age/performance-independent effects. Regions that exhibited positive age/performance-independent effects are shown in green (43 regions). These regions were activated by children and adults to a statistically equivalent degree, regardless of performance. All effects shown surpassed a z-statistic of 11.0, and negative-going timecourses (deactivations) were excluded.

Table 3

Thirty-seven brain regions showing performance-related differences in children and adults

x
 
y
 
z
 
Size (cm3)
 
Location
 
Approx. Brodmann area
 
Left      
    −32 22 0.23 frontal 13 
    −8 44 40 1.72 med. front./ant. cing. 
    −6 56 21 2.99 med. front./ant. cing. 
    −1 52 2.73 med. front./ant. cing. 
    −22 −62 27 0.44 parietal 
    −9 −56 40 0.56 med. par./post. cing. 
    −8 −62 29 0.98 med. par./post. cing. 31 
    −6 −52 29 1.40 med. par./post. cing. 31 
    −5 −55 12 1.00 med. par./post. cing. 29 
    −29 −58 1.06 med. occip./med. temp. 30 
    −28 −70 0.99 med. occip./med. temp. 30 
    −41 −68 1.51 occipital/temporal 19 
    −33 −82 15 0.34 occipital/temporal 19 
    −32 −93 0.62 occipital/temporal 18 
    −24 −44 −5 1.98 occipital/temporal 37 
    −20 −84 1.16 occipital/temporal 17 
    −18 −55 −7 0.44 occipital/temporal 19 
    −17 −98 1.17 occipital/temporal 18 
Right      
51 −13 31 0.61 frontal 
50 27 −2 0.21 frontal 47 
44 −4 49 0.97 frontal 
41 −16 44 0.25 frontal 
37 11 −13 0.22 frontal 13 
33 −21 15 0.91 frontal 13 
26 −5 49 1.31 frontal 
19 49 11 1.45 frontal 10 
48 39 2.74 med. front./ant. cing. 
10 40 2.53 med. front./ant. cing. 32 
43 −56 30 1.55 parietal 39 
−54 29 1.18 med. par./post. cing. 31 
61 −6 −14 0.42 occipital/temporal 21 
38 −69 −4 0.51 occipital/temporal 19 
31 −91 −2 1.45 occipital/temporal 18 
22 −77 −4 0.31 occipital/temporal 18 
31 1.74 subcortical – 
27 20 0.86 subcortical – 
24
 
16
 
3
 
1.78
 
subcortical
 

 
x
 
y
 
z
 
Size (cm3)
 
Location
 
Approx. Brodmann area
 
Left      
    −32 22 0.23 frontal 13 
    −8 44 40 1.72 med. front./ant. cing. 
    −6 56 21 2.99 med. front./ant. cing. 
    −1 52 2.73 med. front./ant. cing. 
    −22 −62 27 0.44 parietal 
    −9 −56 40 0.56 med. par./post. cing. 
    −8 −62 29 0.98 med. par./post. cing. 31 
    −6 −52 29 1.40 med. par./post. cing. 31 
    −5 −55 12 1.00 med. par./post. cing. 29 
    −29 −58 1.06 med. occip./med. temp. 30 
    −28 −70 0.99 med. occip./med. temp. 30 
    −41 −68 1.51 occipital/temporal 19 
    −33 −82 15 0.34 occipital/temporal 19 
    −32 −93 0.62 occipital/temporal 18 
    −24 −44 −5 1.98 occipital/temporal 37 
    −20 −84 1.16 occipital/temporal 17 
    −18 −55 −7 0.44 occipital/temporal 19 
    −17 −98 1.17 occipital/temporal 18 
Right      
51 −13 31 0.61 frontal 
50 27 −2 0.21 frontal 47 
44 −4 49 0.97 frontal 
41 −16 44 0.25 frontal 
37 11 −13 0.22 frontal 13 
33 −21 15 0.91 frontal 13 
26 −5 49 1.31 frontal 
19 49 11 1.45 frontal 10 
48 39 2.74 med. front./ant. cing. 
10 40 2.53 med. front./ant. cing. 32 
43 −56 30 1.55 parietal 39 
−54 29 1.18 med. par./post. cing. 31 
61 −6 −14 0.42 occipital/temporal 21 
38 −69 −4 0.51 occipital/temporal 19 
31 −91 −2 1.45 occipital/temporal 18 
22 −77 −4 0.31 occipital/temporal 18 
31 1.74 subcortical – 
27 20 0.86 subcortical – 
24
 
16
 
3
 
1.78
 
subcortical
 

 

Seventy-seven brain regions showed statistically significant differences between children and adults. Of these, 37 regions showed differences between children and adults that were driven by performance (see Fig. 3 and Table 3). Performance-related regions were found predominantly in right and medial frontal cortex, medial parietal cortex, posterior cingulate, and left and right occipital cortex.

Figure 3.

Performance-related effects. Regions that exhibited performance-related effects are shown in yellow (37 regions). These regions showed statistically significant differences in activity between performance non-matched children and adults, but not between matched children and adults (corrected ANOVAs, P < 0.05).

Figure 3.

Performance-related effects. Regions that exhibited performance-related effects are shown in yellow (37 regions). These regions showed statistically significant differences in activity between performance non-matched children and adults, but not between matched children and adults (corrected ANOVAs, P < 0.05).

Table 4

Forty brain regions showing age-related differences in children and adults

x y z Size (cm3Location Approximate Brodmann area Activity ‘grows’ Pattern of change (7–8 → 23–32) Levene's statistic P Regression
 
  Age Adult-like
 
 

 

 

 

 

 

 

 

 

 

 
F
 
P
 
R2
 
50%
 
75%
 
Left               
−53 −12 40 1.70 frontal up pos → POS 1.1 0.365 20.2 <0.0001 0.31 12.5 15.5 
−49 39 1.47 frontal up flat → POS 0.8 0.540 16.4 <0.0001 0.26 11.8 15.6 
−41 −13 −2 0.36 frontal 13 down POS → flat 0.9 0.484 9.9 <0.0001 0.18 11.6 14.3 
−39 −4 48 2.66 frontal up NEG → POS 3.3 0.010 29.2 <0.0001 0.39 10.4 12.6 
−25 50 2.24 frontal up NEG → POS 4.1 0.002 29.6 <0.0001 0.39 10.2 12.3 
−24 −12 56 1.65 frontal up flat → POS 2.9 0.018 17.6 <0.0001 0.28 11.1 13.5 
−18 55 1.84 frontal 10 down POS → NEG 1.1 0.384 14.5 <0.0001 0.25 11.8 14.7 
−10 32 20 2.55 med. front./ant. cing. 32 down POS → pos 2.3 0.051 23.3 <0.0001 0.34 11.9 15.4 
−8 43 24 2.99 med. front./ant. cing. down POS → pos 2.2 0.060 21.9 <0.0001 0.32 11.3 14.6 
−5 23 19 2.43 med. front./ant. cing. 24 down POS → pos 2.4 0.045 17.2 <0.0001 0.27 13.3 17.3 
−3 52 11 3.06 med. front./ant. cing. 10 down POS → NEG 1.1 0.342 13.0 <0.0001 0.22 10.6 12.8 
12 56 2.56 med. front./ant. cing. up NEG → POS 0.7 0.626 14.0 <0.0001 0.24 10.3 12.0 
−60 −21 33 0.94 parietal up flat → POS 2.6 0.032 6.4 0.0026 0.12 14.8 20.5 
−57 −27 41 1.26 parietal up flat → POS 3.8 0.004 11.3 <0.0001 0.20 16.1 19.8 
−37 −44 37 0.42 parietal 40 up pos → POS 2.3 0.048 3.4 0.0372 0.07 11.8 14.4 
−35 −32 30 0.32 parietal down POS → flat 4.3 0.002 12.0 <0.0001 0.21 9.8 11.2 
−12 −42 12 1.58 med. par./post. cing. 29 down POS → flat 4.0 0.003 18.4 <0.0001 0.29 12.1 15.6 
−5 −41 29 1.50 med. par./post. cing. 31 down POS → pos 3.2 0.010 6.7 0.0020 0.13 14.4 18.5 
−44 −55 27 1.80 occipital/temporal 39 down POS → flat 1.8 0.125 10.3 <0.0001 0.18 13.5 17.5 
−41 −85 0.89 occipital/temporal 19 down POS → NEG 2.0 0.091 14.0 <0.0001 0.23 17.0 21.3 
−37 −18 1.58 occipital/temporal 38 down POS → NEG 0.8 0.579 27.0 <0.0001 0.37 11.2 14.2 
−36 −80 −5 0.36 occipital/temporal 19 down POS → pos 2.6 0.031 5.1 0.0078 0.10 13.6 18.5 
Right               
50 20 0.32 frontal 44 down POS → pos 5.0 <0.001 14.2 <0.0001 0.24 11.2 14.6 
43 −43 15 1.66 frontal 13 down POS → pos 2.0 0.080 13.2 <0.0001 0.22 16.0 20.1 
40 −12 1.25 frontal 13 down POS → flat 2.3 0.056 19.4 <0.0001 0.30 12.1 15.8 
26 52 1.02 frontal up NEG → flat 2.7 0.024 19.6 <0.0001 0.30 10.0 11.9 
10 38 23 3.42 med. front./ant. cing. 32 down POS → pos 4.1 0.002 22.7 <0.0001 0.33 11.8 15.4 
11 20 2.04 med. front./ant. cing. 33 down POS → pos 2.3 0.054 16.7 <0.0001 0.27 13.7 18.2 
40 −69 31 1.44 parietal 39 down POS → flat 2.8 0.021 13.8 <0.0001 0.23 12.0 15.2 
19 −72 27 0.50 med par/post cing 31 down POS → pos 2.9 0.019 14.0 <0.0001 0.23 11.8 15.6 
14 −39 30 1.35 med. par./post. cing. 31 down POS → flat 3.5 0.007 10.4 <0.0001 0.18 13.2 17.4 
11 −61 45 0.45 med. par./post. cing. down POS → pos 1.9 0.097 4.0 0.0213 0.08 15.3 19.6 
52 −66 0.58 occipital/temporal 37 down POS → flat 0.4 0.838 9.2 0.0002 0.17 11.6 16.4 
37 −85 1.74 occipital/temporal 19 down POS → flat 1.1 0.388 16.3 <0.0001 0.26 13.7 18.3 
36 −32 12 0.81 occipital/temporal 41 down POS →pos 2.5 0.036 14.2 <0.0001 0.24 10.8 13.1 
28 −14 0.68 occipital/temporal 34 down POS → flat 1.0 0.404 11.9 <0.0001 0.21 10.9 13.3 
26 −82 26 0.66 occipital/temporal 19 down POS → pos 3.1 0.014 19.6 <0.0001 0.30 14.3 19.4 
26 −52 10 1.50 occipital/temporal 30 down POS → flat 5.0 <0.001 11.7 <0.0001 0.20 12.4 16.2 
26 −43 −5 2.23 occipital/temporal 19 down POS → pos 5.6 <0.001 14.3 <0.0001 0.24 14.3 18.2 
23
 
−95
 
7
 
2.26
 
occipital/temporal
 
18
 
down
 
POS → NEG
 
0.4
 
0.834
 
15.3
 
<0.0001
 
0.25
 
15.5
 
20.8
 
x y z Size (cm3Location Approximate Brodmann area Activity ‘grows’ Pattern of change (7–8 → 23–32) Levene's statistic P Regression
 
  Age Adult-like
 
 

 

 

 

 

 

 

 

 

 

 
F
 
P
 
R2
 
50%
 
75%
 
Left               
−53 −12 40 1.70 frontal up pos → POS 1.1 0.365 20.2 <0.0001 0.31 12.5 15.5 
−49 39 1.47 frontal up flat → POS 0.8 0.540 16.4 <0.0001 0.26 11.8 15.6 
−41 −13 −2 0.36 frontal 13 down POS → flat 0.9 0.484 9.9 <0.0001 0.18 11.6 14.3 
−39 −4 48 2.66 frontal up NEG → POS 3.3 0.010 29.2 <0.0001 0.39 10.4 12.6 
−25 50 2.24 frontal up NEG → POS 4.1 0.002 29.6 <0.0001 0.39 10.2 12.3 
−24 −12 56 1.65 frontal up flat → POS 2.9 0.018 17.6 <0.0001 0.28 11.1 13.5 
−18 55 1.84 frontal 10 down POS → NEG 1.1 0.384 14.5 <0.0001 0.25 11.8 14.7 
−10 32 20 2.55 med. front./ant. cing. 32 down POS → pos 2.3 0.051 23.3 <0.0001 0.34 11.9 15.4 
−8 43 24 2.99 med. front./ant. cing. down POS → pos 2.2 0.060 21.9 <0.0001 0.32 11.3 14.6 
−5 23 19 2.43 med. front./ant. cing. 24 down POS → pos 2.4 0.045 17.2 <0.0001 0.27 13.3 17.3 
−3 52 11 3.06 med. front./ant. cing. 10 down POS → NEG 1.1 0.342 13.0 <0.0001 0.22 10.6 12.8 
12 56 2.56 med. front./ant. cing. up NEG → POS 0.7 0.626 14.0 <0.0001 0.24 10.3 12.0 
−60 −21 33 0.94 parietal up flat → POS 2.6 0.032 6.4 0.0026 0.12 14.8 20.5 
−57 −27 41 1.26 parietal up flat → POS 3.8 0.004 11.3 <0.0001 0.20 16.1 19.8 
−37 −44 37 0.42 parietal 40 up pos → POS 2.3 0.048 3.4 0.0372 0.07 11.8 14.4 
−35 −32 30 0.32 parietal down POS → flat 4.3 0.002 12.0 <0.0001 0.21 9.8 11.2 
−12 −42 12 1.58 med. par./post. cing. 29 down POS → flat 4.0 0.003 18.4 <0.0001 0.29 12.1 15.6 
−5 −41 29 1.50 med. par./post. cing. 31 down POS → pos 3.2 0.010 6.7 0.0020 0.13 14.4 18.5 
−44 −55 27 1.80 occipital/temporal 39 down POS → flat 1.8 0.125 10.3 <0.0001 0.18 13.5 17.5 
−41 −85 0.89 occipital/temporal 19 down POS → NEG 2.0 0.091 14.0 <0.0001 0.23 17.0 21.3 
−37 −18 1.58 occipital/temporal 38 down POS → NEG 0.8 0.579 27.0 <0.0001 0.37 11.2 14.2 
−36 −80 −5 0.36 occipital/temporal 19 down POS → pos 2.6 0.031 5.1 0.0078 0.10 13.6 18.5 
Right               
50 20 0.32 frontal 44 down POS → pos 5.0 <0.001 14.2 <0.0001 0.24 11.2 14.6 
43 −43 15 1.66 frontal 13 down POS → pos 2.0 0.080 13.2 <0.0001 0.22 16.0 20.1 
40 −12 1.25 frontal 13 down POS → flat 2.3 0.056 19.4 <0.0001 0.30 12.1 15.8 
26 52 1.02 frontal up NEG → flat 2.7 0.024 19.6 <0.0001 0.30 10.0 11.9 
10 38 23 3.42 med. front./ant. cing. 32 down POS → pos 4.1 0.002 22.7 <0.0001 0.33 11.8 15.4 
11 20 2.04 med. front./ant. cing. 33 down POS → pos 2.3 0.054 16.7 <0.0001 0.27 13.7 18.2 
40 −69 31 1.44 parietal 39 down POS → flat 2.8 0.021 13.8 <0.0001 0.23 12.0 15.2 
19 −72 27 0.50 med par/post cing 31 down POS → pos 2.9 0.019 14.0 <0.0001 0.23 11.8 15.6 
14 −39 30 1.35 med. par./post. cing. 31 down POS → flat 3.5 0.007 10.4 <0.0001 0.18 13.2 17.4 
11 −61 45 0.45 med. par./post. cing. down POS → pos 1.9 0.097 4.0 0.0213 0.08 15.3 19.6 
52 −66 0.58 occipital/temporal 37 down POS → flat 0.4 0.838 9.2 0.0002 0.17 11.6 16.4 
37 −85 1.74 occipital/temporal 19 down POS → flat 1.1 0.388 16.3 <0.0001 0.26 13.7 18.3 
36 −32 12 0.81 occipital/temporal 41 down POS →pos 2.5 0.036 14.2 <0.0001 0.24 10.8 13.1 
28 −14 0.68 occipital/temporal 34 down POS → flat 1.0 0.404 11.9 <0.0001 0.21 10.9 13.3 
26 −82 26 0.66 occipital/temporal 19 down POS → pos 3.1 0.014 19.6 <0.0001 0.30 14.3 19.4 
26 −52 10 1.50 occipital/temporal 30 down POS → flat 5.0 <0.001 11.7 <0.0001 0.20 12.4 16.2 
26 −43 −5 2.23 occipital/temporal 19 down POS → pos 5.6 <0.001 14.3 <0.0001 0.24 14.3 18.2 
23
 
−95
 
7
 
2.26
 
occipital/temporal
 
18
 
down
 
POS → NEG
 
0.4
 
0.834
 
15.3
 
<0.0001
 
0.25
 
15.5
 
20.8
 

Forty brain regions were classified as showing age-related effects. These regions were divided according to whether adults or children showed greater activity (see Fig. 4 and Table 4). In other words, they could be classified according to whether brain activity ‘grew up’ or ‘grew down’. The majority of regions — 30 out of 40 (75%) — showed decreases in activity over age. These developmentally decreasing regions were distributed bilaterally and were evident most prominently in medial frontal and anterior cingulate cortex, right frontal cortex, medial parietal and posterior cingulate cortex, and bilateral occipitoparietal cortex. On the other hand, 10 of the 40 age-related brain regions (25%) exhibited increases in activity over age. The majority of regions that showed significant developmental increases were found in left lateral and medial dorsal frontal cortex and left parietal cortex, including supramarginal gyrus.

Figure 4.

Age-related effects. Regions that exhibited age-related effects are shown in red and blue (40 regions). Red indicates regions where adults showed greater levels of activity than children and blue indicates regions where children showed greater activity than adults. In all regions, performance-nonmatched children and adults as well as matched children and adults showed statistically significant differences in activity that surpassed a z-statistic of 3.5.

Figure 4.

Age-related effects. Regions that exhibited age-related effects are shown in red and blue (40 regions). Red indicates regions where adults showed greater levels of activity than children and blue indicates regions where children showed greater activity than adults. In all regions, performance-nonmatched children and adults as well as matched children and adults showed statistically significant differences in activity that surpassed a z-statistic of 3.5.

Table 5

Demographic and task performance characteristics of the six cross-sectional groups

 Age in years
 
     

 
7–8
 
9–10
 
11–13
 
14–15
 
16–22
 
23–32
 
n 13 19 17 13 15 18 
Gender 8 F/5 M 10 F/9 M 10 F/7 M 5 F/8 M 11 F/4 M 8 F/10 M 
Age (years) 8.3 (0.5) 10.1 (0.5) 12.4 (1.0) 15.1 (0.3) 19.4 (2.4) 26.6 (2.8) 
% correct 70.9a (11.3) 71.6a (11.4) 80.2a,b (9.9) 85.1b (5.1) 87.3b (4.3) 89.9b (6.7) 
RT (ms) 2089a (455) 2034a,b (535) 1806a,b,c (328) 1617b,c (277) 1551c (310) 1441c (259) 
Verbal IQ 125a (20) 120a (8) 113a (7) 112a (11) 118a (6) – 
Perf IQ 118a (14) 107a (12) 117a (8) 115a (9) 117a (7) – 
Full Scale IQ
 
125a (18)
 
115a (9)
 
117a (7)
 
116a (10)
 
120a (6)
 
118a (5)
 
 Age in years
 
     

 
7–8
 
9–10
 
11–13
 
14–15
 
16–22
 
23–32
 
n 13 19 17 13 15 18 
Gender 8 F/5 M 10 F/9 M 10 F/7 M 5 F/8 M 11 F/4 M 8 F/10 M 
Age (years) 8.3 (0.5) 10.1 (0.5) 12.4 (1.0) 15.1 (0.3) 19.4 (2.4) 26.6 (2.8) 
% correct 70.9a (11.3) 71.6a (11.4) 80.2a,b (9.9) 85.1b (5.1) 87.3b (4.3) 89.9b (6.7) 
RT (ms) 2089a (455) 2034a,b (535) 1806a,b,c (328) 1617b,c (277) 1551c (310) 1441c (259) 
Verbal IQ 125a (20) 120a (8) 113a (7) 112a (11) 118a (6) – 
Perf IQ 118a (14) 107a (12) 117a (8) 115a (9) 117a (7) – 
Full Scale IQ
 
125a (18)
 
115a (9)
 
117a (7)
 
116a (10)
 
120a (6)
 
118a (5)
 
a,b,c

Statistically equivalent across groups by ANOVA, P < 0.05.

In order to ensure that a difference in power was not artificially driving the distinction between age-related and performance-related regions, we conducted an analysis of peak BOLD response magnitudes. For all of the regions showing the three different kinds of effects, we calculated the difference in peak BOLD signal magnitude between nonmatched children and adults and subtracted from it the difference in peak BOLD signal magnitude between matched children and adults as follows: |NONmatched children – NONmatched adults| – |Matched children – Matched adults|. This index captures, for all regions, the degree to which the age difference in BOLD signal magnitude changes with performance. If our statistical classification is working reasonably well (i.e. is capturing hemodynamic differences as opposed to being driven merely by power differences), age-related regions should show a stable BOLD magnitude relationship across performance, and performance-related regions should show BOLD magnitude differences across levels of performance.

Indeed, age- and performance-related regions showed precisely this relationship (see online Supplementary Material). The index of magnitude difference for age-related regions centered on a value near zero. Performance-related regions, in contrast, showed an index of magnitude difference that centers around 0.2% BOLD signal change, differing significantly from both age-related and age/performance-independent regions, which did not differ significantly from each other.

Analysis Step 2

Cross-sectional Analysis by Age Group

Age-related regions showed notable regularity in the specific patterns of change they exhibited over maturation (see Table 4 and Fig. 5). Among the regions decreasing over age, the most common pattern of change was activity that went from statistically positive in the youngest age group to significantly lower but still positive in the oldest age group (e.g. −36, −80, −5, left extrastriate cortex, ∼BA 19). Fourteen of the 30 decreasing regions (47%) demonstrated this pattern over age. The next most common pattern of developmental change in decreasing regions was activity that, again, began as significantly activated in the youngest children but diminished over age into a statistically flat line (e.g. −44, −55, 27, left occipital-temporal-parietal cortex, ∼BA 39). These two developmental patterns — positive to smaller positive (POS → pos) and positive to flat (POS → flat) — together accounted for 25 (or 83%) of the regions showing decreases in activity over maturation. The five remaining regions growing down over age began as statistically positive activations in the youngest group and were deactivations in the oldest group (POS → NEG).

Figure 5.

Age-related patterns of change in activity over development. Age-related effects took the form of several different patterns of change in the hemodynamic response function timecourses over development. Six example regions are shown, with the numeric coordinates (e.g. −49, 3, 39) indicating the location of the center of mass of the region in standard atlas space. For each region, timecourses of activity are shown for all six age groups (blue = 7–8 year olds, purple = 9–10, green = 11–13, red = 14–15, orange = 16–22, and yellow = 23–32). The percentage of MR signal change is plotted against time in seconds. Using a combination of statistical reliability and signal magnitude, the timecourse for each age group was characterized as showing significant positive change (activation), negative change (deactivation), or no change (a ‘flat’ line) over time. Regions are labeled according to the changes in activity that were observed between the youngest and oldest age groups (7–8 and 23–32 year olds, respectively). These include activation becoming deactivation (POS → NEG), no activation becoming activation (flat → POS), deactivation becoming activation (NEG → POS), activation becoming significantly more activated (pos → POS), activation becoming no activation (POS → flat), and activation becoming significantly less activated (POS → pos) over maturation.

Figure 5.

Age-related patterns of change in activity over development. Age-related effects took the form of several different patterns of change in the hemodynamic response function timecourses over development. Six example regions are shown, with the numeric coordinates (e.g. −49, 3, 39) indicating the location of the center of mass of the region in standard atlas space. For each region, timecourses of activity are shown for all six age groups (blue = 7–8 year olds, purple = 9–10, green = 11–13, red = 14–15, orange = 16–22, and yellow = 23–32). The percentage of MR signal change is plotted against time in seconds. Using a combination of statistical reliability and signal magnitude, the timecourse for each age group was characterized as showing significant positive change (activation), negative change (deactivation), or no change (a ‘flat’ line) over time. Regions are labeled according to the changes in activity that were observed between the youngest and oldest age groups (7–8 and 23–32 year olds, respectively). These include activation becoming deactivation (POS → NEG), no activation becoming activation (flat → POS), deactivation becoming activation (NEG → POS), activation becoming significantly more activated (pos → POS), activation becoming no activation (POS → flat), and activation becoming significantly less activated (POS → pos) over maturation.

Among regions showing developmental increases, the most common pattern of change was activity that began in the youngest subjects as a statistically flat timecourse (i.e. not reliably activated during the tasks) and incremented over age into a significant positive activation for the oldest subjects (e.g. −49, 3, 39, left frontal cortex, ∼BA 6). Four regions (out of 10 increasing over age) showed this maturational pattern (flat → POS). Three of the increasing regions changed over age from statistically negative-going activity to positive-going activity (NEG → POS, e.g. −39, −4, 48, left dorsal frontal cortex, ∼BA 6). Two regions began as activations that grew up with age (pos → POS), and one region showed a change in its pattern of activity that began as negative and matured into no change from baseline (NEG → flat).

Left frontal cortex showed the most age-related increases in brain activity for these word generation tasks (see Fig. 6). Five of the seven left frontal regions showed significantly increasing involvement. In contrast, right frontal cortex showed mostly decreased involvement over the ages studied here. Only one right frontal region exhibited a reliable age-related increase in activity relative to baseline. This region, however, was not positively activated by any age group; it began as a deactivation and matured into a flat response. Medial frontal cortex and anterior cingulate mostly showed age-related decreases for these tasks. Only one medial frontal region showed increases in activity with age.

Figure 6.

Timecourses by age group in 40 age-related regions. For all 40 regions showing age-related effects, hemodynamic response function timecourses are shown for each of the six cross-sectional age groups (blue = 7–8 year olds, purple = 9–10, green = 11–13, red = 14–15, orange = 16–22, and yellow = 23–32). Numeric coordinates (e.g. −49, 3, 39) indicate the location of the center of mass of the region in standard atlas space. The percentage of MR signal change is plotted against time in seconds.

Figure 6.

Timecourses by age group in 40 age-related regions. For all 40 regions showing age-related effects, hemodynamic response function timecourses are shown for each of the six cross-sectional age groups (blue = 7–8 year olds, purple = 9–10, green = 11–13, red = 14–15, orange = 16–22, and yellow = 23–32). Numeric coordinates (e.g. −49, 3, 39) indicate the location of the center of mass of the region in standard atlas space. The percentage of MR signal change is plotted against time in seconds.

In brain regions outside frontal cortex, the majority of developmental changes took the form of age-related decreases in activity. Out of all non-frontal regions that showed age-related effects for controlled lexical processing, only three exhibited increases over development. All three of these regions were found in left parietal locations. The remaining regions all showed significant decreases in levels of activity over maturation.

The variability of peak activity within many but not all brain regions changed significantly over age (see Table 4). Using Levene's test, a test of homogeneity of variance across groups, 47% of decreasing age-related regions and 70% of increasing age-related regions showed statistically significant differences across the six age groups in BOLD signal variability (P < 0.05). Generally, brain activity showed the greatest variability in the youngest age group and variability decreased over development.

Regression Analysis by Age

For all 40 regions showing age-related differences, the nonlinear regression functions explained a statistically significant proportion of the variance in peak brain activity over development (see Table 4 and Fig. 7). Coefficients of determination (R2) ranged from 0.08 to 0.39 across all regions. Regions growing up and regions growing down showed evidence of becoming adult-like at different ages. Activity in decreasing age-related regions on average became 50% adult-like at age 12.8 and 75% adult-like at age 16.5. Regions showing maturational increases, on the other hand, matured somewhat earlier showing peak activity that became 50% adult-like by the age of 11.9 and 75% adult-like by the age of 14.8.

Figure 7.

Peak brain activity as a function of age in 40 age-related regions. For all 40 regions showing age-related effects, fitted nonlinear regression curves are shown. For each region, peak brain activity in percentage of MR signal change is plotted as a function of age to tenths of a year for all 95 subjects. Regions showing increasing activity over age were modeled using a three-parameter, single-exponent rise to max function [y = y0 + a(1 − ebx) ] and are shown as red curves. Regions showing decreasing activity over age were modeled using a complementary three-parameter, single-exponent decay function [y = y0 + aebx ] and are shown as blue curves. Numeric coordinates (e.g. −49, 3, 39) indicate the location of the center of mass of the region in standard atlas space.

Figure 7.

Peak brain activity as a function of age in 40 age-related regions. For all 40 regions showing age-related effects, fitted nonlinear regression curves are shown. For each region, peak brain activity in percentage of MR signal change is plotted as a function of age to tenths of a year for all 95 subjects. Regions showing increasing activity over age were modeled using a three-parameter, single-exponent rise to max function [y = y0 + a(1 − ebx) ] and are shown as red curves. Regions showing decreasing activity over age were modeled using a complementary three-parameter, single-exponent decay function [y = y0 + aebx ] and are shown as blue curves. Numeric coordinates (e.g. −49, 3, 39) indicate the location of the center of mass of the region in standard atlas space.

According to these measures, brain activity in frontal age-related regions matured at a faster rate than regions outside frontal cortex. Peak brain activity approached adult-like levels at systematically older ages, going generally from polar and dorsal frontal cortex to anterior and posterior cingulate, to temporal and parietal cortex, and then extrastriate cortex (see Fig. 8). Across all 40 regions showing age-related effects, there was a statistically significant relationship between measures of the age at which brain activity within a region matured and its neuroanatomical location in the posterior-anterior dimension. The ages at which activity became 50% adult-like and 75% adult-like were both significantly negatively correlated with region position along the horizontal axis (r = −0.40, P < 0.01 and r = −0.44, P < 0.004, respectively), indicating that, generally speaking, frontal regions were the earliest to reach mature levels of activity for these tasks.

Figure 8.

Maturation rates of 40 age-related regions. All 40 regions that exhibited age-related effects are shown with region color varying by the age at which brain activity became 50% “adult-like” (i.e. the age at which brain activity in that region equals 50 percent of the total range of activity exhibited across the youngest and oldest subjects).

Figure 8.

Maturation rates of 40 age-related regions. All 40 regions that exhibited age-related effects are shown with region color varying by the age at which brain activity became 50% “adult-like” (i.e. the age at which brain activity in that region equals 50 percent of the total range of activity exhibited across the youngest and oldest subjects).

Discussion

The purpose of this study was to provide a detailed test and empirical characterization of maturational changes in human cerebral functional organization, focusing on a distinction between progressive and regressive developmental changes underlying controlled cognition. Regions ‘growing up’ and regions ‘growing down’ showed neuroanatomical segregation, different developmental changes in the magnitude of the hemodynamic response, and evidence of differences in maturational timing. The pattern of results we found suggests that progressive and regressive changes in the functional neuroanatomy may relate to two different kinds of developmental specialization: the recruitment of top-down control mechanisms and the tuning of lower-level mechanisms.

Progressive developmental changes were found exclusively in frontal and parietal cortex, including regions reliably demonstrated in adults to be involved in top-down cognitive control (Luria, 1966/1980; Stuss and Benson, 1984; Posner and Petersen, 1990; Kolb and Whishaw, 1996; Miller and Cohen, 2001; Corbetta and Shulman, 2002). In addition, these developmental increases in activity primarily began as statistically ‘flat’ responses in children (i.e. were not activated) and ‘grew up’ into robust activations in adults. This finding suggests that regions showing age-related increases in activity for these tasks are mostly ‘new recruits’ in the maturing functional organization.

Regressive developmental changes in activity, on the other hand, were more frequent and were found across a wider range of locations. Showing clear neuroanatomical dissociation from progressive effects, regressive changes were found mostly outside lateral frontal cortex and included earlier processing regions such as bilateral occipital and temporal cortex. All regions showing significant developmental decreases began as positively activated in children. The large majority declined over age into either smaller but still positive activation or into statistically flat responses (i.e. were not activated by adults). This finding suggests that the maturing functional organization also includes regions or sets of regions that become more selectively activated with age.

Evidence from functional neuroimaging studies of adults supports the notion that ‘new’ regions of the brain can be recruited over the course of learning, practice, and skill acquisition to support specialized control systems. Raichle et al. (1994) found that regions of the brain most active during initial performance of a verb generation task (anterior cingulate, left prefrontal and left posterior temporal cortex, and the right cerebellar hemisphere), compared to reading nouns, were all significantly less active during practiced performance. These changes, however, were accompanied by increases in activation of the insular cortex bilaterally and left medial extrastriate cortex. Petersen et al. (1998) found similar changes in functional organization related to the practice of verb generation and maze tracing. For both tasks, practice induced a shift of activity away from regions in frontal cortex, anterior cingulate and cerebellum to greater activity in insular cortex and superior medial frontal cortex respectively, including newly recruited regions.

There is also evidence that specialization can take the form of ‘tuning’ brain regions — i.e. regions or sets of regions of the brain become more selectively activated over age for the same tasks (see Johnson, 2000). Using event-related potentials (ERPs), several studies have demonstrated that electrophysiological components sensitive to the processing of faces change significantly between the ages of 3 and 12 months old (de Haan, 1998; Johnson, 2001; Halit et al., 2003). The developmental changes in topography and amplitude that have been found suggest that the regions adults use for processing faces are selected from an initially larger set and may respond to a wider range of stimuli at younger ages. Similar tuning effects have been observed in studies of infant speech perception. For instance, ERPs were measured while infants (13–20 months old) listened to words they understood, words they did not understand and words spoken backwards (Mills et al., 1997). Differences between known and unknown words showed a wide spatial distribution of activity before the age of 17 months, evoking responses in broad bilateral anterior and posterior regions. In contrast, these effects were much more spatially limited in 20 month olds and were found only in temporal and parietal regions of the left hemisphere. All of these findings are consistent with the general notion that the input domain for cognitive and perceptual systems generally narrows over the course of development (Gauthier and Nelson, 2001).

We found that activity in frontal regions approached adult-like levels at relatively younger ages than activity in other regions. In fact, there was a significant relationship between the age at which brain activity within a region became adult-like and its neuroanatomical location in the posterior–anterior dimension, with brain regions generally showing adult-like activity in anterior regions first with more posterior regions becoming adult-like at older ages. These results, though counterintuitive to a simple caudal to rostral maturational view of functional brain development, are in accord with other perspectives.

One alternative view of how the mature functional neuroanatomy for a particular cognitive task might emerge is by the use of a ‘progressive neural scaffolding’. This view is similar to that which has been suggested for changes that accompany the acquisition of skilled task performance in adults (Petersen et al., 1998). In such a scenario, ‘immature’ performance might be carried out by a relatively large set of lower level brain regions recruited to cope with novel task demands; this set of regions may sustain task performance at early developmental levels. Then, higher-level control mechanisms (potentially in frontal and parietal cortex) might be recruited to provide top-down support, and to guide the selection of the lower-level mechanisms for specific task performance.

Our results are also strikingly consistent with predictions made by the interactive specialization framework of functional brain development articulated by Johnson and his colleagues (Johnson, 2000, 2001; Johnson et al., 2002). The specific kinds of progressive and regressive changes we have seen support the notions that generally more pathways are recruited at younger ages for the same tasks, that there are significant changes in functional localization during development, and that some prefrontal regions show precocial task involvement.

Several conceptual and methodological aspects of the current study warrant further discussion. The changes we observed in functional organization over development for these controlled lexical tasks could presumably be associated with age-related changes in the specific information-processing operations that are employed for task performance. For example, children might accomplish these tasks by using a general object recognition and manipulation approach. These operations may transition into more specialized lexical processing in adulthood. Such a change would be consistent with the decreasing reliance on extrastriate regions and the increasing reliance on left frontal regions over development that we found. Our study does not address these cognitive strategy differences that may exist between children and adults. Our study also does not provide evidence that speaks to the relative contributions of biological maturation versus individual task experience per se. This issue is of great interest in functional brain development and could be addressed by research that closely tracks and manipulates variables such as task exposure and practice and explores their relation to age.

Approximately half of the brain regions we found to show differences between children and adults showed differences in activity that were related to task performance discrepancy. These regions appeared to be functionally similar across age in that levels of activity were equivalent in adults and children who performed the tasks at the same level of proficiency. This observation demonstrates the importance of measuring and accounting for behavioral performance levels in functional neuroimaging studies. Indeed, this issue is not unique to child studies; it applies with equal importance to virtually any study that aims to compare two groups where one performs generally better than the other on cognitive tasks (e.g. see Weinberger and Berman, 1998, regarding similar issues in functional neuroimaging research on schizophrenia). For developmental studies, there are clear implications. Without some way of addressing performance differences, the differences in brain activity between children and adults could be misinterpreted as maturational differences in functional neuroanatomy. Although regions showing performance-related effects are interesting in their own right, they maintain an important functional distinction from regions showing effects that appear to be driven by age.

It is noteworthy that the youngest and oldest participants in our study showed statistically equivalent activation of much of the brain for these controlled lexical tasks. Similar activation was found in expected regions such as primary visual, auditory and motor cortex, but also in regions such as prefrontal cortex. Although we found age-related effects in left dorsal and posterior frontal regions for word generation, comparable brain regions have shown functional similarities between similarly aged adults and children for other tasks, such as visuospatial working memory (Klingberg et al., 2002). It would appear, then, that there is no simple mapping of maturational effects onto broad brain regions, such as the mere ‘coming online’ of left frontal cortex. Instead, task-specific developmental changes appear to occur in a complex functional mosaic.

Because we set out to explore only regions that show reliable differences between our youngest and oldest subjects, it is possible that we have missed regions that show other kinds of nonlinear developmental effects over this age range. For example, this approach would fail to identify brain regions exhibiting ‘U-’ and ‘inverted U-’shaped effects over age, where brain activity is the same in the endpoint age groups but shows significant differences at ages in between. Human cognition and behavior has long been known to exhibit changes of these types over development (e.g. Mehler, 1982a; Mehler, 1982b). Recent functional imaging studies suggest that some brain regions show these effects for certain tasks as well (Luna et al., 2001).

Regression analysis approaches are increasingly being used to characterize the rate and course of changes in regional brain activity over development (e.g. Kwon et al., 2002; Shaywitz et al., 2002; Turkeltaub et al., 2003). Our results provide evidence that these changes may be largely nonlinear and heteroscedastic, and may, therefore, require statistical methods that do not assume linearity and homogeneity of variance.

With increasing sophistication in conceptual and methodological approaches, functional neuroimaging holds promise for moving beyond a confirmatory role for existing knowledge about development. The accumulation of more empirical data on developmental changes in cerebral functional organization should create a richer theoretical context and lead to greater insight into the neural underpinnings of human cognitive development.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oupjournals.org/.

The authors gratefully thank the children, teens and adults who participated in this study; Mark McAvoy and Avi Snyder for neuroimaging application development; David Van Essen and his colleagues for the use of CARET; the Petersen and Schlaggar groups and other members of the NeuroImaging Laboratories at Washington University for thoughtful comments and suggestions. This work was supported in part by a National Science Foundation Graduate Research Fellowship to Tim Brown and by NIH NSADA (B.L.S.), NS32979 (S.E.P.), NS41255 (S.E.P.), NS46424 (S.E.P.), The McDonnell Center for Higher Brain Function (S.E.P., B.L.S.) and The Charles A. Dana Foundation (B.L.S.). Brad Schlaggar is a Scholar of the Child Health Research Center of Excellence in Developmental Biology at Washington University School of Medicine (HD01487). Portions of this work were presented at the 9th Annual Meeting of the Cognitive Neuroscience Society, San Francisco, CA and the 32nd Annual Meeting of the Society for Neuroscience, Orlando, FL.

References

Barch DM, Braver TS, Sabb FW, Noll DC (
2000
) Anterior cingulate and the monitoring of response conflict: Evidence from an fMRI study of overt verb generation.
J Cogn Neurosci
 
12
:
298
–309.
Booth JR, Burman DD, Meyer JR, Lei Z, Trommer BL, Davenport ND, Li W, Parrish TB, Gitelman DR, Mesulam MM (
2003
) Neural development of selective attention and response inhibition.
Neuroimage
 
20
:
737
–751.
Bourgeois JP (
2001
) Synaptogenesis in the neocortex of the newborn: the ultimate frontier for individuation? In: Handbook of developmental cognitive neuroscience (Nelson C, Luciana M, eds), pp. 23–34. Cambridge, MA: MIT Press.
Buckner RL, Raichle ME, Petersen SE (
1995
) Dissociation of human prefrontal cortical areas across different speech production tasks and gender groups.
J Neurophysiol
 
74
:
2163
–2173.
Bunge SA, Dudukovic NM, Thomason ME, Vaidya CJ, Gabrieli JD (
2002
) Immature frontal lobe contributions to cognitive control in children: evidence from fMRI.
Neuron
 
33
:
301
–311.
Burgund ED, Kang H-SC, Kelly JE, Buckner RL, Snyder AZ, Petersen SE, Schlaggar BL (
2002
) The feasibility of a common stereotactic space for children and adults in fMRI studies of development.
Neuroimage
 
17
:
184
–200.
Burton H, Diamond JB, McDermott KB (
2003
) Dissociating cortical regions activated by semantic and phonological tasks: a fMRI study in blind and sighted people.
J Neurophysiol
 
90
:
1965
–1082.
Casey BJ, Cohen JD, Jezzard P, Turner R, Noll DC, Trainor RJ, Giedd J, Kaysen D, Hertz-Pannier L, Rapoport JL (
1995
) Activation of prefrontal cortex in children during a nonspatial working memory task with functional MRI.
Neuroimage
 
2
:
221
–229.
Casey BJ, Trainor RJ, Orendi JL, Schubert AB, Nystrom LE, Giedd JN, Castellanos FX, Haxby JV, Noll DC, Cohen JD, Forman SD, Dahl RE, Rapoport JL (
1997
) A developmental functional MRI study of prefrontal activation during performance of a go–no-go task.
J Cogn Neurosci
 
9
:
835
–847.
Casey BJ, Thomas KM, Davidson MC, Kunz K, Franzen PL (
2002
) Dissociating striatal and hippocampal function developmentally with a stimulus–response compatibility task.
J Neurosci
 
22
:
8647
–8652.
Changeux JP, Danchin A (
1976
) Selective stabilisation of developing synapses as a mechanism for the specification of neuronal networks.
Nature
 
264
:
705
–712.
Chugani HT, Phelps ME, Mazziotta JC (
1987
) Positron emission tomography study of human brain functional development.
Ann Neurol
 
22
:
487
–497.
Corbetta M, Shulman GL (
2002
) Control of goal-directed and stimulus-driven attention in the brain.
Nat Rev Neurosci
 
3
:
201
–215.
Corina DP, San Jose-Robertson L, Guillemin A, High J, Braun AR (
2003
) Language lateralization in a bimanual language.
J Cogn Neurosci
 
15
:
718
–730.
de Haan M (
1998
) Electrophysiological correlates of face processing by adults and 6-month-old infants. J Cogn Neurosci (Ann Meeting Suppl) 36.
Dobbing J, Sands J (
1973
) Quantitative growth and development of human brain.
Arch Dis Child
 
48
:
757
–767.
Duncan J, Owen AM (
2000
) Common regions of the human frontal lobe recruited by diverse cognitive demands.
Trends Neurosci
 
23
:
475
–483.
Edelman GM (
1987
) Neural Darwinism. New York: Basic Books.
Elman JL, Bates EA, Johnson MH, Karmiloff-Smith A, Parisi D, Plunkett K (
1996
) Rethinking innateness: a connectionist perspective on development. Cambridge, MA: MIT Press.
Forman SD, Cohen JD, Fitzgerald M, Eddy WF, Mintun MA, Noll DC (
1995
) Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): use of a cluster-size threshold.
Magn Reson Med
 
33
:
636
–647.
Gaillard WD, Hertz-Pannier L, Mott SH, Barnett AS, LeBihan D, Theodore WH (
2000
) Functional anatomy of cognitive development: fMRI of verbal fluency in children and adults.
Neurology
 
54
:
180
–185.
Gaillard WD, Sachs BC, Whitnah JR, Ahmad Z, Balsamo LM, Petrella JR, Braniecki SH, McKinney CM, Hunter K, Xu B, Grandin CB (
2003
) Developmental aspects of language processing: fMRI of verbal fluency in children and adults.
Hum Brain Mapp
 
18
:
176
–185.
Gauthier I, Nelson CA (
2001
) The development of face expertise.
Curr Opin Neurobiol
 
11
:
219
–224.
Giedd JN, Blumenthal J, Jeffries NO, Castellanos FX, Liu H, Zijdenbos A, Paus T, Evans AC, Rapoport JL (
1999
) Brain development during childhood and adolescence: a longitudinal MRI study.
Nat Neurosci
 
2
:
861
–863.
Gottlieb G (
1976
) The role of experience in the development of behavior and the nervous system. In: Neural and behavioral specificity (Gottlieb G, ed.), pp. New York: Academic Press.
Greenough WT, Black JE, Wallace CS (
1987
) Experience and brain development.
Child Dev
 
58
:
539
–559.
Halit H, de Haan M, Johnson MH (
2003
) Cortical specialisation for face processing: face-sensitive event-related potential components in 3- and 12-month-old infants.
Neuroimage
 
19
:
1180
–1193.
Hebb DO (
1949
) The organization of behavior. New York: John Wiley.
Holland SK, Plante E, Weber Byars A, Strawsburg RH, Schmithorst VJ, Ball WS, Jr (
2001
) Normal fMRI brain activation patterns in children performing a verb generation task.
Neuroimage
 
14
:
837
–843.
Huttenlocher PR, Dabholkar AS (
1997
) Regional differences in synaptogenesis in human cerebral cortex.
J Comp Neurol
 
387
:
167
–178.
Jerne NK (
1967
) Antibodies and learning: Selection versus instruction. In: The neurosciences (Quarton G, Melnechuk T, Schmitt F, eds), pp. 200–205. New York: The Rockefeller University Press.
Jernigan T, Hesselink J, Sowell E, Tallal P (
1991
) Cerebral structure of magnetic resonance imaging in language- and learning- impaired children.
Arch Neurol
 
48
:
539
–545.
Johnson MH (
2000
) Functional brain development in infants: elements of an interactive specialization framework.
Child Dev
 
71
:
75
–81.
Johnson MH (
2001
) Functional brain development in humans.
Nat Rev Neurosci
 
2
:
475
–483.
Johnson MH, Halit H, Grice SJ, Karmiloff-Smith A (
2002
) Neuroimaging of typical and atypical development: A perspective from multiple levels of analysis.
Dev Psychopathol
 
14
:
521
–536.
Kang H-SC, Burgund ED, Lugar HM, Petersen SE, Schlaggar BL (
2003
) Comparison of functional activation foci in children and adults using a common stereotactic space.
Neuroimage
 
19
:
16
–28.
Klein D, Milner B, Zatorre RJ, Meyer E, Evans AC (
1995
) The neural substrates underlying word generation: a bilingual functional-imaging study.
Proc Natl Acad Sci USA
 
92
:
2899
–2903.
Klingberg T, Vaidya CJ, Gabrieli JD, Moseley ME, Hedehus M (
1999
) Myelination and organization of the frontal white matter in children: a diffusion tensor MRI study.
Neuroreport
 
10
:
2817
–2821.
Klingberg T, Forssberg H, Westerberg H (
2002
) Increased brain activity in frontal and parietal cortex underlies the development of visuospatial working memory capacity during childhood.
J Cogn Neurosci
 
14
:
1
–10.
Kolb B, Whishaw IQ (
1996
) Fundamentals of human neuropsychology. New York: W. H. Freeman.
Kwon H, Reiss AL, Menon V (
2002
) Neural basis of protracted developmental changes in visuo-spatial working memory.
Proc Natl Acad Sci U S A
 
99
:
13336
–13341.
Lancaster JL, Woldorff MG, Parsons LM, Liotti M, Freitas CS, Rainey L, Kochunov PV, Nickerson D, Mikiten SA, Fox PT (
2000
) Automated Talairach atlas labels for functional brain mapping.
Hum Brain Mapp
 
10
:
120
–131.
Luna B, Thulborn KR, Munoz DP, Merriam EP, Garver KE, Minshew NJ, Keshavan MS, Genovese CR, Eddy WF, Sweeney JA (
2001
) Maturation of widely distributed brain function subserves cognitive development.
Neuroimage
 
13
:
786
–793.
Luria AR (
1966
) Higher cortical functions in man. New York: Basic Books.
Matarazzo JD (
1972
) Wechsler's measurement and appraisal of adult intelligence. Baltimore: Williams & Wilkins.
McAvoy MP, Ollinger JM, Buckner RL (
2001
) Cluster size thresholds for assessment of significant activation in fMRI.
Neuroimage
 
13
:
S198
.
McCarthy G, Blamire AM, Rothman DL, Gruetter R, Shulman RG (
1993
) Echo-planar magnetic resonance imaging studies of frontal cortex activation during word generation in humans.
Proc Natl Acad Sci USA
 
90
:
4952
–4956.
McDermott KB, Petersen SE, Watson JM, Ojemann JG (
2003
) A procedure for identifying regions preferentially activated by attention to semantic and phonological relations using functional magnetic resonance imaging.
Neuropsychologia
 
41
:
293
–303.
Mehler J (
1982
a) Studies in the development of cognitive processes. In: U-shaped behavioral growth (Strauss S, ed.), pp. 271–293. New York: Academic Press.
Mehler J (
1982
b) Unlearning: dips and drops–a theory of cognitive development. In: Regressions in development: basic phenomena and theoretical alternatives (Bever T, ed.), pp. 133–152. Hillsdale, NJ: Erlbaum.
Miezin F, Maccotta L, Ollinger J, Petersen S, Buckner R (
2000
) Characterizing the hemodynamic response: effects of presentation rate, sampling procedure, and the possibility of ordering brain activity based on relative timing.
Neuroimage
 
11
:
735
–759.
Miller EK, Cohen JD (
2001
) An integrative theory of prefrontal cortex function.
Annu Rev Neurosci
 
24
:
167
–202.
Mills, DL, Coffey-Corina S, Neville, HJ (
1997
) Language comprehension and cerebral specialization from 13 to 20 months.
Dev Neuropsychol
 
13
:
397
–445.
Muzik O, Chugani DC, Juhasz C, Shen C, Chugani HT (
2000
) Statistical parametric mapping: Assessment of application in children.
Neuroimage
 
12
:
538
–549.
Nelles JL, Lugar HM, Coalson RS, Miezin FM, Petersen SE, Schlaggar BL (
2003
) An automated method for extracting response latencies of subject vocalizations in event-related fMRI experiments.
Neuroimage
 
20
:
1865
–1871.
Ojemann JG, Ojemann GA, Lettich E (
2002
) Cortical stimulation mapping of language cortex by using a verb generation task: effects of learning and comparison to mapping based on object naming.
J Neurosurg
 
97
:
33
–38.
Passingham RE (
1996
) Attention to action.
Philos Trans R Soc Lond B Biol Sci
 
351
:
1473
–1479.
Petersen SE, Fiez JA (
1993
) The processing of single words studied with positron emission tomography.
Annu Rev Neurosci
 
16
:
509
–530.
Petersen SE, Fox PT, Posner MI, Mintun M, Raichle ME (
1988
) Positron emission tomographic studies of the cortical anatomy of single-word processing.
Nature
 
331
:
585
–589.
Petersen SE, Fox PT, Snyder AZ, Raichle ME (
1990
) Activation of extrastriate and frontal cortical areas by visual words and word-like stimuli.
Science
 
249
:
1041
–1044.
Petersen SE, van Mier H, Fiez JA, Raichle ME (
1998
) The effects of practice on the functional anatomy of task performance.
Proc Natl Acad Sci USA
 
95
:
853
–860.
Pfefferbaum A, Mathalon DH, Sullivan EV, Rawles JM, Zipursky RB, Lim KO (
1994
) A quantitative magnetic resonance imaging study of changes in brain morphology from infancy to late adulthood.
Arch Neurol
 
51
:
874
–887.
Posner MI, Petersen SE (
1990
) The attention system of the human brain.
Annu Rev Neurosci
 
13
:
25
–42.
Posner MI, Petersen SE, Fox PT, Raichle ME (
1988
) Localization of cognitive operations in the human brain.
Science
 
240
:
1627
–1631.
Quartz SR, Sejnowski TJ (
1997
) The neural basis of cognitive development: a constructivist manifesto.
Behav Brain Sci
 
20
:
537
–56; discussion 556–596.
Raichle ME, Fiez JA, Videen TO, MacLoed A-MK, Pardo JV, Fox PT, Petersen SE (
1994
) Practice-related changes in human brain functional anatomy during nonmotor learning.
Cereb Cortex
 
4
:
8
–26.
Schlaggar BL, Brown TT, Lugar HM, Coalson RS, Petersen SE (2002a) fMRI in performance-matched children and adults: modality dependent and independent age-related differences in lexical processing. Paper presented at the Society for Neuroscience, Orlando, FL
Schlaggar BL, Brown TT, Lugar HM, Visscher KM, Miezin FM, Petersen SE (
2002
b) Functional neuroanatomical differences between adults and school-age children in the processing of single words.
Science
 
296
:
1476
–1479.
Seger CA, Rabin LA, Zarella M, Gabrieli JD (
1997
) Preserved verb generation priming in global amnesia.
Neuropsychologia
 
35
:
1069
–1074.
Shaywitz BA, Shaywitz SE, Pugh KR, Mencl WE, Fulbright RK, Skudlarski P, Constable RT, Marchione KE, Fletcher JM, Lyon GR, Gore JC (
2002
) Disruption of posterior brain systems for reading in children with developmental dyslexia.
Biol Psychiatry
 
52
:
101
–110.
Sowell ER, Peterson BS, Thompson PM, Welcome SE, Henkenius AL, Toga AW (
2003
) Mapping cortical change across the human life span.
Nat Neurosci
 
6
:
309
–315.
Stuss DT, Benson DF (
1984
) Neuropsychological studies of the frontal lobes.
Psychol Bull
 
95
:
3
–28.
Talairach J, Tournoux P (
1988
) Co-planar stereotaxic atlas of the human brain. New York: Thieme.
Tamm L, Menon V, Reiss AL (
2002
) Maturation of brain function associated with response inhibition.
J Am Acad Child Adolesc Psychiatry
 
41
:
1231
–1238.
Thelen E, Smith L (
1994
) A dynamic systems approach to the development of cognition and action. Cambridge, MA: MIT Press.
Thomas KM, King SW, Franzen PL, Welsh TF, Berkowitz AL, Noll DC, Birmaher V, Casey BJ (
1999
) A developmental functional MRI study of spatial working memory.
Neuroimage
 
10
:
327
–338.
Thompson-Schill SL, Swick D, Farah MJ, D'Esposito M, Kan IP, Knight RT (
1998
) Verb generation in patients with focal frontal lesions: a neuropsychological test of neuroimaging findings.
Proc Natl Acad Sci USA
 
22
:
15855
–15860.
Turkeltaub PE, Gareau L, Flowers DL, Zeffiro TA, Eden GF (
2003
) Development of neural mechanisms for reading.
Nat Neurosci
 
6
:
767
–773.
Van Essen DC (
2002
) Windows on the brain: the emerging role of atlases and databases in neuroscience. Curr Opin Neurobiol 12:574–579. See also http://pulvinar.wustl.edu:8081/sums/search.do?filename=ATLAS&filetype=tar.gz
Van Essen DC, Drury HA, Dickson J, Harwell J, Hanlon D, Anderson CH (
2001
) An integrated software suite for surface-based analyses of cerebral cortex. J Am Med Inform Assoc 8:443–459. See also http://brainmap.wustl.edu/caret
Wechsler D (
1999
) Wechsler abbreviated scale of intelligence. San Antonio: The Psychological Corporation.
Weinberger DR, Berman KF (
1998
) Prefrontal function in schizophrenia: Confounds and controversies. In: The prefrontal cortex: executive and cognitive functions (Roberts AC, Robbins TW, Weiskrantz L, eds), pp. 165–180. New York: Oxford.
Wise R, Chollet F, Hadar U, Friston K, Hoffner E, Frackowiak R (
1991
) Distribution of cortical neural networks involved in word comprehension and word retrieval.
Brain
 
114
:
1803
–1817.
Yakovlev PI, Lecours AR (
1967
) The myelogenetic cycles of regional maturation of the brain. In: Regional development of the brain in early life (Minkowski A, ed.), pp. 3–70. Oxford: Blackwell Scientific.

Author notes

1Department of Psychology, Washington University in St Louis, St Louis, MO 63130, USA, 2Department of Neurology, Washington University School of Medicine, St Louis, MO 63110, USA, 3Department of Radiology, Washington University School of Medicine, St Louis, MO 63110, USA, 4Department of Anatomy and Neurobiology, Washington University School of Medicine, St Louis, MO 63110, USA, 5Department of Pediatrics, Washington University School of Medicine, St Louis, MO 63110, USA