Adult studies of working memory have consistently revealed prefrontal and posterior parietal cortical activation in response to intact performance during working memory tasks . In contrast to the large number of fMRI studies in adult populations, very few studies have examined fMRI response to working memory tasks in typically developing adolescents, and most have focused on the development of spatial working memory. The few studies examining fMRI response during verbal and spatial working memory in children and adolescents suggest that, overall, children and adolescents demonstrate similar frontal and parietal patterns of response as adults , but show greater and more widespread activation in these regions with increasing age. To our knowledge, only two of these studies have examined fMRI response to working memory across a sample of typically developing adolescents. One study of spatial working memory among 34 7- to 22-year-olds suggested age-related increases in both the intensity and spatial extent of SWM activation in bilateral dorsolateral prefrontal cortex, left ventrolateral prefrontal cortex, left premotor cortex,ebb and flow trays and bilateral superior and inferior posterior parietal cortices . However, while age was the best predictor of activation in these brain regions, there were significant improvements in SWM performance across the study age range that may have contributed to age-related activation patterns.
Another study examined SWM in thirteen 9- to 18-year-olds and demonstrated increased neural response in bilateral superior frontal and intraparietal cortex and left middle occipital gyrus, and decreased intensity of response with age in right inferior frontal cortex , but no significant relationship between age and the spatial extent of brain response was demonstrated. Thus, while we have some understanding of the developmental changes in the neural systems involved in adolescent working memory, these studies are preliminary and are based on small sample sizes across relatively broad age ranges. Likely due to limited statistical power, to date, no studies have examined gender differences in fMRI response to cognitive tasks across normal adolescent development. Despite this, previous neuroanatomical and cognitive research suggests that developmental gender differences may be present in SWM activation. Specifically, there are established gender differences in the rate of neural development, with females developing earlier than males in frontal and parietal brain regions , which have been consistently implicated in working memory . In addition, gender differences in working memory ability have been identified, specifically for SWM skills. While adult studies have demonstrated a general spatial information processing advantage for males over females that emerges with increasing age , this is primarily due to differences in active spatial processing , which is often not required in traditional nback or delayed matching working memory tasks. Studies of SWM abilities suggest gender differences for accuracy and reaction time in children and adults .
Overall, these studies indicate that adult females demonstrate more accurate SWM performance than adult males , but males tend to show faster reaction times . One investigation suggested a similar profile of gender differences for SWM performance in children that diminishes towards adolescence , and another SWM study in adolescents found no performance differences between the genders . Although the pattern of findings is somewhat difficult to interpret based on the different tasks and samples used across studies, it does suggest that gender discrepancies in SWM performance may vary based on visuo spatial processing demands and stage of development. Given that the majority of developmental working memory research using fMRI has focused on SWM, we chose to further contribute to this literature by utilizing a relatively large sample of normally developing teens to carefully investigate the neural substrates involved in SWM across adolescent development and between the genders using fMRI. Based on the findings from the limited previous research in the area, we predicted that working memory brain activation would increase in frontal and parietal regions as a function of age. In addition, based on known differences in rates of neuromaturation and a potential female advantage in SWM accuracy, we hypothesized that females would demonstrate a more mature pattern of fMRI response than males. Adolescent participants were recruited from local junior high and high schools as part of an ongoing adolescent brain imaging project . This study was approved by the University of California San Diego Institutional Review Board, and written consent and assent were obtained from teens and their guardians. Adolescents were administered a 90-minute telephone screening interview to ascertain eligibility, and a guardian , separately provided corroborative reports.
Exclusion criteria for the study were: use of psychotropic medications; head injury with loss of consciousness >2 minutes; neurological or medical illness; learning disabilities; DSM-IV psychiatric disorder including attention deficit hyperactivity disorder and substance use disorders; significant maternal drinking during pregnancy ; parental history of bipolar I, psychotic disorders or substance use disorders; left handedness; and MRI contraindications. Eligible participants were 49 youth ages 12 to 17, including 24 males and 25 females. Males and females were similar on demographics such as age, ethnicity, and socioeconomic status .The SWM task consisted of 18 21-sec blocks that alternated between experimental and baseline conditions, and three blocks of rest . The task also included six seconds of blank screen at the beginning , allowing the scanner to reach steady state. Total task time was 7 minutes and 48 seconds . Each block started with a one second word cue at the center of the screen to inform the participant of the upcoming block type. Stimuli were presented for 1000 ms, and each interstimulus interval was 1000 ms. During rest blocks, the word “LOOK” appeared at the center of the screen, then a centered fixation cross appeared for 20 seconds. The experimental condition was a memory for locations task in which abstract line drawings were projected one at a time in one of eight locations in a circular array. Locations were chosen to minimize verbal labeling . The word “WHERE” appeared for one second at the beginning of the block, and participants were asked to press a button when a figure appeared in a location in which a design had already appeared during that block. Unbeknownst to participants, target trials were always repeat locations of items displayed two trials prior . In each block, an average of 3 of the 10 stimuli presented were target items. During the vigilance baseline condition, the word “DOTS” appeared at the beginning of the block to alert participants to the block type. Then the same abstract line drawings used in the SWM blocks were presented one at a time in the same eight locations, but a dot appeared above figures on 30% of trials. The purpose of the baseline condition was to control for the motor, sensory,rolling greenhouse benches and attention processes involved in the experimental condition. Given the variability in the onset and timing of pubertal development, chronological age can be an inaccurate indicator of biological maturation. For this reason, all teens completed the Pubertal Development Scale – a 5-item self report measure of pubertal status with demonstrated reliability and validity . The PDS correlates significantly with physician ratings and Sexual Maturation Scale self-ratings of pubertal maturation .To minimize head motion in the scanner, a soft cloth was placed on the participant’s forehead then taped to the head tray, and foam pads were inserted around the head. Task stimuli were projected onto a screen at the foot of the MRI bed, and participants viewed the images from a mirror attached to the head coil. An MRI safe button box collected responses during the task. Anatomical and functional imaging data were acquired on a 1.5 Tesla General Electric Signa LX system.
Structural imaging consisted of a sagittally acquired inversion recovery prepared T1-weighted 3D spiral fast spin echo sequence . During task presentation, functional imaging was collected in the axial plane using T2*-weighted spiral gradient recall echo imaging . Data Analyses SWM task accuracy and reaction time were examined in relationship to age using regression analyses. Gender differences in task performance were analyzed using within-subjects ANOVA with task condition as the within subjects factor and gender as the between subjects factor. Because age did not significantly relate to task performance, it was not used as a covariate in the ANOVA. Significant interactions were followed up with t-tests to examine simple effects. Neuropsychological test performance was analyzed using regression analyses examining age, gender, and their interaction. Imaging data were processed and analyzed using Analysis of Functional Neuro Images . First, we applied a motion-correction algorithm to align each volume in the time series with a base volume, yielding three rotational and three displacement parameters across the time series for each participant. Two independent raters inspected time series data to remove any repetitions on which the algorithm did not adequately adjust for motion; all participants retained at least 80% of repetitions. Using a deconvolution process , the time series data were correlated with a vector representing the design of the task that modeled 1- and 2-TR delays in hemodynamic response , and covaried for estimated degree of motion and linear trends. This process yielded fit coefficients representing the blood oxygen level dependent response contrast between SWM and vigilance, SWM and fixation, and vigilance and fixation in each voxel for every subject. Imaging datasets were transformed into standard Talairach coordinates for structure localization and comparisons among subjects , and functional data were resampled into 3.5mm 3 isotropic voxels. We applied a spatial smoothing Gaussian filter to functional data to account for anatomic variability.Group-level analyses conducted regressions in each voxel of the brain to predict the fit coefficient representing the contrast between SWM and vigilance from gender, age, and their interaction. To control for Type I error when determining clusters that showed significant effects, we used a combination of t-statistic magnitude and cluster volume thresholding by only interpreting clusters exceeding 943 microliters, equal to 22 contiguous significant 3.5mm 3 voxels, yielding a cluster-wise α < .05. In order to understand the nature of these group level clusters, we performed exploratory followup analyses examining contrasts between SWM and fixation and vigilance and fixation in each significant cluster. We utilized the Talairach Daemon and AFNI to confirm gyral labels for significant clusters. To examine the role of pubertal maturation on SWM BOLD response, mean fit coefficients for each participant were computed for each significant activation cluster, and hierarchical regression analyses determined whether PDS scores explained significant variance in brain response above and beyond that explained by chronological age. Planned follow-up analyses examined whether age, gender, and their interaction were related to spatial extent of neural response in brain regions showing significant age-related activation during SWM relative to vigilance. Because we were particularly interested in age-related changes in the spatial extent of activation during SWM , we only examined clusters in which there was greater activation to SWM relative to vigilance, and not regions which were deactivated by the task. Therefore, we created posterior parietal and left prefrontal regions of interest , and counted the number of voxels exhibiting significantly greater activation to SWM relative to vigilance for each participant. In order to best represent functionally important frontal and parietal regions within this sample of adolescents, our ROIs were determined by identifying significant clusters activated by the task, rather than anatomically defined based on specific gyri or Brodmann’s Areas . Because regression analyses indicated a change in location of parietal activation across adolescence , an ROI based on the average activation map for the whole group would not accurately represent regions used for the task in both young and old teens . Therefore, we divided teens on age with median split, and determined significant clusters activated by the task in young and old teens separately using single sample t-tests . This yielded separate posterior parietal clusters for young and old teens. We created a posterior parietal ROI for examining spatial extent of activation by including all voxels that occupied the posterior parietal clusters for young teens and old teens, and all voxels within clusters showing a significant positive or negative relationship to age. A similar procedure determined an appropriate left prefrontal ROI. Since young teens did not demonstrate significant clusters of left prefrontal activation, significant clusters activated by the task in older teens were added to the average left prefrontal cluster showing a significant relationship with age.