Abstract

Alterations of the prefrontal cortex (PFC) could contribute to cognitive decline in older adults. We examined the specificity of age-related PFC degeneration and whether cognitive abilities were related to volumetric measurements. Older and younger subjects were tested using a battery of tasks supported by different subregions within the PFC. The cognitive data from older subjects were related to PFC volumetric measurements in order to determine whether cortical morphology was predictive of individual differences in task performance within this age range (72–94 years). Working memory performance best distinguished older from younger subjects. Working memory measures but not other measures were correlated with age in both groups. A larger orbital PFC volume was related to a worse working memory performance and a larger superior PFC volume was related to worse conditional association learning. The volumes of these regions were not related to performance on other tasks. These results suggest that working memory is a sensitive measure of cognitive aging and that regional morphology is associated with specific cognitive abilities in older adults.

Introduction

Normal aging is associated with a decline in a variety of cognitive abilities. This age-related cognitive change prevents functional independence in older adults. Thus, understanding the basis for this decline is a significant health care interest. The present study examined the relationship between age-related changes in cognition and brain volumetric changes in the prefrontal cortex (PFC).

The PFC has received considerable attention in the study of age-related cognitive decline as numerous studies have demonstrated that older adults perform poorly on tasks supported by the PFC (West, 1996). For example, prior studies have shown that older adults decline only slightly in their ability to perform simple working memory tasks of span (Wingfield et al., 1988; Dobbs and Rule, 1989; Wiegersma and Mertse, 1990; Dolman et al., 2000), yet perform significantly worse on tasks that require the manipulation of items within working memory such as in self-ordered choosing (Daigneault and Braun, 1993; West et al., 1998; Janowsky et al., 2000). Functional imaging has shown that age-related changes in dorsolateral PFC function could account for the decline in working memory performance with normal aging (Rypma and D'Esposito, 2000). For example, younger subjects show greater dorsolateral PFC activity compared to older adults during working memory retrieval and reaction times are differentially related to activation patterns in these two groups. Working memory abilities have been related to volumetric measurements in visual processing areas (Raz et al., 1998) and the volume of the PFC has been related to perseverative responses on the Wisconsin Card Sort and other tasks (Raz et al., 1998, 1999; Schretlen et al., 2000). Thus, age-related degeneration of the PFC may contribute to a decline in performance on cognitive tasks that are supported by the PFC. In particular, preferential loss of dorsolateral PFC volume with aging (Raz et al., 1997) could contribute to the age-related decline in working memory performance.

Prior studies have demonstrated age-related changes in prefrontal morphology in both humans (Jernigan et al., 1991; Dickson et al., 1992; Raz et al., 1997; Mueller et al., 1998; Salat et al., 1999a, 2001) and non-human primates (area 9/46) (Peters et al., 1994). Although there is a decline in PFC volume when examining subjects across the age span (Raz et al., 1997), it is less clear that significant prefrontal cortical volumetric change occurs with later aging (Mueller et al., 1998; Salat et al., 1999a, 2001). Regionally preferential changes in cortical morphology do occur with later aging, yet relative measures of morphological change could also be informative in this age range. For example, preservation of the orbital PFC region relative to other PFC regions could be a more prominent age-related phenomenon than the preferential atrophy of a specific region (Salat et al., 2001). Thus, it is possible that either absolute volumes or relative measurements of regional morphology could be predictive of cognitive abilities. For example, relative preservation of the orbital PFC region with later aging (Salat et al., 2001) could be useful in predicting cognitive performance.

The current study examined PFC-supported cognitive processes in well-characterized, older subjects and measured a number of volumetric subregions within the PFC, thus expanding the range of data on cognitive and volumetric changes with aging. We employed cognitive tasks supported by the PFC that have been used in many prior studies in both humans and nonhuman primates, but never in combination and with morphometric measures on the same subjects. This battery included the Conditional Association Task (which tests the ability to form associations between two arbitrary stimuli) (Petrides, 1985), the Self-ordered Pointing Task (Petrides and Milner, 1982) and the N-back Task (Cohen et al., 1997), two widely studied tasks of working memory (the ability to store and manipulate information ‘on-line’ in order to perform a task), and the Object Alternation Task (which tests the ability to shift a cognitive or behavioral set) (Freedman et al., 1998). Conditional association performance activates a superior and posterior PFC region (Brodmann area 8) in functional imaging studies (Petrides et al., 1993). Numerous studies using a variety of techniques have demonstrated a role for dorsolateral and inferior PFC regions (Brodmann areas 9/46 and 47) in the performance of tasks of working memory in both humans and non-human primates (Petrides and Milner, 1982; Petrides et al., 1993; Petrides, 1995; Braver et al., 1997; Cohen et al., 1997; Jahanshahi et al., 1998; Pelosi and Blumhardt, 1999) along with more posterior brain regions. Orbital and ventromedial PFC damage (Brodmann areas 10, 24, 32 and 47) results in increased perseverative responses on object alternation and similar tasks in monkeys and humans (Meunier et al., 1997; Freedman et al., 1998). Consequently, impairments in the performance of these tasks could reflect dysfunction of the PFC subregions that support them. It was expected that PFC regional volumes would be predictive of individual differences in performance on tasks supported by those regions in older adults. Further, it was expected that working memory abilities would differ between younger and older subjects and would be most closely related to age as opposed to other PFC functions because the dorsolateral PFC supports working memory and this region shows a preferential age-related volumetric decline (Raz et al., 1997). Finally, it was expected that degeneration of dorsolateral PFC regions (the middle and inferior PFC) would be more prominent than other PFC regions as demonstrated in a prior study (Raz et al., 1997) and that working memory would be closely related to the volumes of these regions.

Materials and Methods

Subjects

Cognitive data were collected on a group of young subjects (n = 20; 10 men and 10 women of mean age 29.9 years and range 21–43 years) and a group of older subjects (n = 31; 15 men and 16 women of mean age 84.0 years and range 72–94 years). There were similar proportions of men to women in each group (50%/50% young and 48%/52% old). The older and younger subjects were matched for their Weschler Adult Intelligence Scale —Revised (WAIS-R) vocabulary score, which is a measure that correlates well with general intellectual functioning (Weschler, 1981). The older and younger subjects differed marginally in their years of education (1.5 years difference) (young mean = 16.4 years and range 12–22 years and old mean = 14.9 years and range 12–24 years) (P = 0.05). All subjects were right-handed except for two left-handed young subjects. All of the subjects signed informed consent for these studies.

The younger subjects were recruited through flyers displayed at local universities and were screened for a variety of factors related to health and general cognitive function. The young subjects were not taking prescription or non-prescription medications known to affect cognitive function, had no prior neurological problems including seizures, had good vision and had no health concerns.

The older subjects were recruited through the Oregon Brain Aging Study (OBAS), which is a longitudinal study of cognitive, neurological and other aspects of aging. Detailed descriptions of the recruitment procedures and subject criteria for the OBAS as well as extensive medical and cognitive data on older subjects have been published elsewhere (Howieson et al., 1993; Kaye et al., 1994). Healthy older subjects at entry into the OBAS were functionally independent, had English as their principal language, had not sought evaluation for cognitive impairment and scored well on a variety of tests including the Instrumental Activities of Daily Living (Fillenbaum, 1985), the Mini-mental State Examination (MMSE) (Folstein et al., 1975), the Cornell Depression Scale (Yesavage, 1983), the Geriatric Depression Scale (Yesavage et al., 1982) and the Clinical Dementia Rating Scale (Hughes et al., 1982). They did not have significant medical disorders, including diabetes mellitus, hypertension, ischemic heart disease, cardiac arrhythmia, stroke, active cancer, psychiatric disorders and any neurologic disorder, had vision correctable to 20/70 oculus uterque or better, had hearing that did not interfere with speech perception and did not use medicines that affect cognitive function. The OBAS subjects were examined biannually for signs of dementia or changes in their medical status and had annual neurological, neuro-psychological (WAIS-R) (Weschler, 1981), Weschler Memory Scale (Weschler, 1981), verbal fluency (Benton and Hamshire, 1976) and magnetic resonance imaging (MRI) examinations. No subjects had cerebral infarctions. Only minor degrees of periventricular white matter hyperintensity were present in occasional subjects' MRI scans. Thus, these subjects were exceptionally healthy when entered into the OBAS. Thirteen of the 31 older subjects still met these extreme health criteria for entry an average of 4.7 years later when the cognitive tests and MR for this study were administered. The remaining 18 older subjects were deficient in one of the above criteria at the time of this study (a mean of 7.6 years after entry in these subjects). The health concerns of these 18 older subjects were mostly due to heart disease or blood pressure (eight subjects). The other health concerns were cancers outside of the nervous system (three subjects), angina (two subjects), minor depression with stably treated medication (two subjects) and rheumatoid arthritis, trigeminal neuralgia and the presence of an old lacunar infarct (each in one subject). These medical conditions did not contribute to their cognitive abilities (see the Results section) and all subjects were free from any signs of possible dementia as determined by a comprehensive neuropsychological screen (mean MMSE = 28.5) (see Table 1 for a summary of the neuropsychological performance in this group).

Cognitive Measures

All of the subjects were tested with a battery of cognitive tasks associated with PFC function as shown in previous studies using functional imaging and/or human and animal lesion models. The tasks were administered in either the exact same or in a similar manner to tasks described in the literature. The battery included the Conditional Association Task (Petrides, 1985), the Self-ordered Pointing Task (Petrides and Milner, 1982), the N-back Task (Cohen et al., 1997; Braver et al., 1997) and the Object Alternation Task (Freedman et al., 1998) (Fig. 1). All subjects were trained on computer tasks and response procedures prior to task performance in order to assure that the younger and older subjects did not differ in their competency of computer usage.

Conditional Association Task

The Conditional Association Task examined the subjects' ability to learn associations between two arbitrary visual stimuli. Superior and posterior dorsolateral PFC regions (Brodmann area 8) support performance of this task (Petrides, 1985; Petrides et al., 1993; Levine et al., 1997). The subjects were presented with six abstract designs positioned in two rows of three designs in each row on a computer monitor (Fig. 1A). The designs were obtained as part of a larger set from Dr Michael Petrides (see the Self-ordered Pointing Task description below), but digitized for computer display. These were the same stimuli as used in prior studies of PFC function (Petrides, 1985; Petrides et al., 1993). The stimuli were created such that they would be easy to distinguish from one another but were abstract, making them difficult to describe or code verbally.

One of six different colored lines appeared above the six designs on the monitor (Fig. 1A). The subjects were told that each colored line was arbitrarily matched (associated) with one and only one abstract design. The goal of the task was to learn, through trial and error with computer feedback (correct and incorrect beeps), which colored line was associated with each specific abstract design. The subjects chose designs by using designated keys on the computer keyboard that matched the spatial location of the design chosen. If an incorrect response was made, the computer would beep indicating an incorrect response and the subject had to choose again. When the correct design was chosen, the six designs would be cleared from the screen and reappear 1000 ms later in a new spatial arrangement (thereby preventing the use of a spatial strategy for performing the task) with a new colored line at the top of the screen. The subjects would then have to learn the association between the new colored line and a different abstract design.

The subjects were to respond by choosing that same design whenever the associated colored line appeared at the top of the monitor. Thus, the subjects were to choose a particular design conditional upon the color of the line at the top of the screen, ignoring the spatial position of the design on the screen. Stimuli were presented pseudo-randomly such that any colored line could appear on any trial with any one of six spatial arrangements of the same six abstract designs. All six lines and all six spatial arrangements were presented before repeating any one colored line or arrangement. Trials continued until the subjects reached a criterion of 12 consecutive correct responses or until 180 trials had been administered. The data for the Conditional Association Task were analyzed as the average number of errors per trial (the total number of errors divided by the total number of trials).

Self-ordered Pointing Task

The Self-ordered Pointing Task is a self-paced task of non-verbal working memory (the temporary storage and updating of abstract designs held in the mind). Performance on this task is associated with mid-dorsolateral PFC function (Petrides and Milner, 1982). The subjects were presented with a book with pages of 8.5 × 11 inches. Each page displayed a set of abstract designs, with the designs in a different randomly assigned spatial arrangement on each page (Fig. 1B). The stimuli were the same stimuli as used in prior studies of PFC function (Petrides and Milner, 1982). The subjects were told that the goal of the task was to choose one design on each page in the series without choosing the same design more than once. The task was self-ordered in that the subjects could choose designs in any order. Choosing from the same spatial location repeatedly was not advantageous and was discouraged. Thus, the subjects had to keep the chosen designs in working memory and update their memory with each choice in order not to choose the same design on subsequent pages. Four working memory load conditions, differing in the number of designs (6, 8, 10 or 12 designs per page), were presented to all subjects (see the examples in Fig. 1B). Each load condition contained a new group of designs and no design was repeated across load conditions. Each load condition was administered three times (three runs of each of four load conditions). Errors (choosing the same design more than once in a run) were summed for the three runs of each load condition and divided by the total number of trials in order to calculate the percentage of correct choices for each load condition.

N-back Task

The N-back Task is a timed task of verbal working memory (storing, manipulating and updating letters in the mind). Activation of mid-dorsolateral and inferior lateral PFC regions as well as other non-PFC regions occurs with performance of this task (Braver et al., 1997; Cohen et al., 1997). The subjects were presented with a string of random letters, one at a time, at the center of a computer display. The subjects were told to respond by pressing one of two buttons on a button box for each and every letter, dependent on whether the letter was a ‘target’ letter or a ‘non-target’ letter (described below).

The task comprised four different working memory load conditions. The zero-back condition was the control condition and contained all the stimuli and response components of the other conditions, but with no working memory requirements. The subjects were instructed to press the target button every time they saw a specific letter, whether it was in uppercase or lowercase (e.g. respond with the target button every time you see the letter ‘Q’) and to press the non-target button to all other letters (Fig. 1C). In the one-back condition (low working memory load condition) the subjects were told to respond with the target button to any letter that was repeated one letter back (e.g. ‘a–a’), whether it was in upper or lower case and respond with the non-target button to all other letters. No feedback was given as to whether the choice was correct or incorrect. In the two and three-back conditions (moderate and high working memory load conditions respectively) the subjects were to respond with the target button to any letter repeated two and three letters back respectively (e.g. a repeated letter separated by one letter, such as ‘A–Q–a’, in the two-back condition or a repeated letter separated by two letters, such as ‘A–Q–m–a’, in the three back condition). All conditions were matched for stimuli and for response requirements, but differed in the amount of letters to be kept in mind, manipulated and updated in order to perform the task.

Letters were presented for 500 ms with a 2500 ms interstimulus interval between letters. Targets were presented in ~20% of the trials in each load condition. Each load condition of the task was completed three times in random order for a total of 12 blocks (three runs of each of the four load conditions). The data were analyzed using a sensitivity measurement (D′) that takes into account the number of targets correctly identified (responding with the target button to a target letter) and the number of non-targets incorrectly rejected (responding with the target button to a non-target letter). Specifically, the data were analyzed as the Z-score for the percentage of correct targets minus the Z-score for the percentage of incorrect non-targets. These Z-scores were derived from all subjects' data (young and old subjects combined). This measure was used in addition to simple percent correct target identification as the percent correct alone could be quite high even when the subjects made a great number of false-positive errors (responding to a non-target letter as a target).

Object Alternation Task

The Object Alternation Task examined the subjects' ability to shift responses and their cognitive set and is associated with orbital PFC function (Freedman et al., 1998). Errors on this task are perseverations to a specific incorrect response. The subjects were presented with two shapes on the computer monitor, i.e. a red circle and a blue square. The subjects were told that there was a star ‘hidden’ behind one of the shapes (the circle or the square) and that their task was to find the star in each and every trial. The correct strategy was to alternate between the circle and the square regardless of their spatial position. If the subjects chose the incorrect object they would be informed by a beep (incorrect feedback). After a correct response, a correct beep sounded and a star appeared in the object's place for 500 ms (correct feedback). The stimuli then cleared from the screen and a circle and square returned 5 s later, with either shape presented randomly in the left or right position (Fig. 1D). The subjects performed the task until reaching the criterion of 10 consecutive correct responses or until a total of 50 trials was reached. The data for the Object Alternation Task were analyzed as the number of errors per trial for each subject (the total number of trials divided by the total number of errors).

Magnetic Resonance Imaging

MRI scans from 30 of the 31 older subjects described above (n = 30; 14 men and 16 women of mean age 84.2 years) were examined. The scans were only collected in the older subjects in order to determine whether cortical morphology was predictive of cognitive abilities in this age range. The structural scans were collected within 6 months prior to cognitive testing.

Scan Protocol

MRIs were performed using a GE 1.5 T scanner. The brain was visualized with a multi-echo coronal sequence (TR = 3000 ms, TE = 30 and 80 ms and 4 mm slices with no skip). T1-weighted images in the midsagittal plane were used for orienting the coronal plane. The coronal plane was determined as the plane perpendicular to a line drawn from the lowest point of the genu to the lowest point of the splenium of the corpus callosum on the midsagittal image (Mueller et al., 1998).

Region of Interest Analysis

Tissue analysis of PFC MR images was computer assisted using a program called REGION as described previously (Salat et al., 1999a,b, 2001). Briefly, data were first collected from three tissue regions of interest (the total PFC volume, PFC white matter volume and PFC gray matter volume). Structures were outlined with a cursor directly on a computer display (Fig. 2). The sulcal and gyral boundaries were determined by tracing edges around and deep into the cortex. The PFC was defined in the coronal plane beginning with the first slice in which the superior frontal gyrus could be visualized (the tip of the frontal pole) and continued posteriorly, but did not include the first slice in which the anterior tip of the corpus callosum was visualized. This process uses approximately eight slices per subject. The dual-echo scans collected provided both T2- and proton density-weighted contrast of each slice through the PFC. The total PFC volume was defined by tracing all gyri and sulci around the cortical ribbon of each hemisphere in the T2-weighted image (Fig. 2A). The PFC white matter volume was traced in the proton density-weighted image of the same slice after standardized image adjustment in order to maximize the gray matter to white matter contrast and reduce the number of ambiguous pixels (Fig. 2B). The PFC gray matter volume was calculated by subtracting the PFC white matter volume from the total PFC volume. We estimated that these volumes include ~85% of the total PFC, but leave out posterior regions of the PFC, particularly in the orbital region.

After defining the gray matter/white matter boundaries, the PFC was further subdivided into five regions of interest within each cerebral hemisphere: the superior, middle, inferior, orbital and anterior cingulate. Each region was hand traced with the cursor using visual inspection of the image and an atlas-defined protocol (Fig. 2C1–4). The regions were defined using a method modified from regions described in earlier publications (Damasio, 1991, 1995; Salat et al., 2001) in which the gyral and sulcal patterns were used as landmarks in the T2-weighted image. The superior region was defined as beginning at the most ventral portion of the superior frontal sulcus and traced dorso-medial to the dorsal extent of the cortex and then ventral down the interhemispheric fissure to the most lateral portion of the anterior cingulate sulcus. The middle region began at the most ventral portion of the superior frontal sulcus and was traced lateral and ventral down the middle frontal gyrus and continued past the middle frontal sulcus to the most medial portion of the inferior frontal sulcus. The inferior region began at the most medial portion of the inferior frontal sulcus and continued lateral and then ventro-medial to the most dorsal portion of the orbital sulcus. The orbital region began at the most dorsal portion of the orbital sulcus and continued ventro-medial and up the interhemispheric fissure until the most dorsal portion of the gyrus (within the interhemispheric fissure) was reached. Occasionally, images contained indistinct boundaries between the gray and white matter. Data for these ambiguous regions were obtained by tracing through the midpoint of the ambiguous segment to the next clear boundary. The anatomic criteria for regional subdivisions were more difficult to delineate towards the frontal pole. Still, the volume of each region was made up predominantly of the well-defined posterior slices.

Regional pixel areas were first transformed to volumes by multiplying the total pixel counts by a derived constant that transforms pixel size from REGION to cubic centimeters given the MR slice thickness of 4 mm [pixel area × 0.8789 (pixels to mm2) × 4 (mm2 to mm3 by multiplying by slice thickness in mm) × 0.001 (mm3 to cm3)]. The volumes were then adjusted by the subjects' total intracranial volume (ICV) to correct for variations in head size and provide a measure of atrophy as in prior studies (Kaye et al., 1997; Salat et al., 1999) [the use of ICV as a correction measure in studies of aging and Alzheimer's disease is described elsewhere (Jenkins et al., 2000)]. The ICV was defined as all non-bone pixels beginning with the first slice in which the frontal poles were visible and ending at the occipital pole (Kaye et al., 1997). Brainstem and infratentorial structures including the cerebellum were excluded from supratentorial structures by manually tracing their boundaries. The cerebellum and all structures inferior to the cerebellum were excluded by tracing along the superior aspect of the structure below the tectum and quadrigeminal plate. Infratentorial structures were excluded at the level of the pons by tracing a line connecting the most dorsomedial aspects of the middle cerebellar peduncles and the cerebral peduncles on more anterior slices. All non-cortical structures ventral to the mammillary bodies were excluded on the most anterior slices. The total ICV for each subject was determined as the sum of the supratentorial pixel area and transformed to cubic centimeters as described for regional volumes. The examiner was blind to the subjects' age and sex. Five brains were analyzed five times each in order to generate reliability data. The reliability (intraclass correlation) on each subregion was >0.99 for the total PFC, >0.97 for the total gray matter, >0.94 for the total white matter and >0.76 for the superior, >0.84 for the middle, >0.85 for the inferior, >0.89 for the orbital and >0.62 for the anterior cingulate. The anterior cingulate was excluded from future analyses because of poor reliability and because very little of this structure was included in these slices.

Statistical Analyses

Demographic data and the data for the Conditional Association Task, Object Alternation Task and the zero-back condition (control condition) of the N-back Task were compared between the young and old subjects by separate unpaired t-tests. The data for the Self-ordered Pointing and N-back Tasks were analyzed as two-factor (group × working memory load), repeated-measures ANOVAs with working memory load as the repeated measure. Repeated-measures ANOVAs were used for determining whether there was a greater difference between the younger and older subjects at the higher load conditions. An interaction between group and load would suggest that the number of items to be kept in mind is a critical factor in age-related decrement in performance of the working memory task. The relationship between age and cognitive performance was examined using Pearson's correlations in the older and younger subjects separately. Comparisons were considered significant when P < 0.05 (two-tailed). A discriminant function analysis was applied to the cognitive data in order to determine whether any cognitive process best discriminated the older from younger subjects.

The cognitive data were related to regional volumes (ICV corrected) in the elderly with Pearson correlations. Significance values of P < 0.05 (two-tailed) were considered significant.

Most subjects performed all of the cognitive tasks with a few exceptions. Two older subjects were unwilling to complete the Conditional Association Task due to difficulty with task performance. One older and one younger subject were unable to complete the Object Alternation Task because of time constraints. The data reported maximizes the number of subjects for each analysis (all subjects available for each cognitive task).

Results

Subject Characteristics

By design, there was a significant difference in age between the young and older subjects [t(49) = –32.6 and P < 0.001]. There were no differences in the subjects' WAIS-R vocabulary performance and all older subjects performed within the normal range on the MMSE (≥24) (Folstein et al., 1975) (Table 1). Although there was a small difference in years of education between the groups (mean difference 1.5 years) [t(49) = 2.01 and P = 0.05], education was not significantly correlated with the cognitive variables in the older or younger subject groups. We divided the older subjects into ‘criteria’ and ‘non-criteria’ groups based on the health criteria for entry into the OBAS in order to determine whether there were any differences in cognitive performance due to medical co-morbidities. There were no differences in age or any measure of PFC cognitive performance between subjects meeting the OBAS entry criteria at testing and those not meeting the entry criteria at testing. Thus, all subjects were used in the analyses. In addition, the data were analyzed without the two older subjects with depression as it has been suggested that prefrontal function may be altered by depression (Drevets, 1999). These subjects did not alter the data significantly and, thus, they were included in all analyses.

Cognitive Analyses

The older subjects made significantly more errors than the younger subjects in all tasks (Fig. 3).

The older subjects made more errors per trial in the Conditional Association Task [t(47) = 3.3 and P < 0.01] (Fig. 3A) and the Object Alternation Task [t(47) = 4.9 and P < 0.001] (Fig. 3D).

The older subjects made significantly more errors compared to the younger subjects in the Self-ordered Pointing Task [F(1,49) = 32.4 and P < 0.001] (Fig. 3B) as revealed by repeated-measures ANOVA. Errors increased with increasing working memory load [F(3,49) = 105.2 and P < 0.001] (Fig. 3B) and there was a significant group by load interaction [F(3,49) = 11.1 and P < 0.001] (Fig. 3B) with greater differences between the groups at the higher load conditions. This result was not due to the younger subjects performing at ceiling, as the effect is greatest from the 8-picture to 10-picture load condition, when younger subjects are already performing below their performance in the 6-picture load condition.

Using the measure of the percent of trials in which a target was correctly identified in the working memory conditions of the N-back Task, there was a significant effect of group [F(1,49) = 25.2 and P < 0.001] and working memory load [F(2,49) = 61.5 and P < 0.001]. Performance declined with increasing working memory load (Fig. 3C). In contrast to subject-ordered pointing, there was no interaction between group and load. The older subjects performed worse than the young subjects on the zero-back condition of the N-back Task [t(49) = 2.5 and P = 0.02]. The group differences in working memory conditions remained when the data were co-varied for performance on the zero-back condition [Fs(1,49) > 18.2 and Ps < 0.001].

Reaction time data were available for the N-back Task. The younger subjects had significantly faster reaction times for targets and non-targets as compared to the older subjects in all of the working memory conditions combined (one-, two- and three-back conditions) (P < 0.01) (Fig. 4). The older subjects had significantly faster reaction times for incorrect targets as compared to correct targets (P < 0.02). This finding was mainly due to a strong difference between the correct and incorrect target reaction times in the older subjects with performance of the two-back condition and lesser trends in the other conditions. The opposite pattern was found for non-targets as the older subjects had significantly faster reaction times for correct non-targets as compared to incorrect non-targets, suggesting that the incorrect responses to targets were due to disinhibited responses as opposed to a simple loss of information contained within working memory. The younger subjects did not show this pattern of reaction times and did not differ in either their correct or incorrect responses to targets or non-targets.

Data reduction was performed for a portion of the remaining analyses by creating a composite score of all conditions of the two working memory tasks (the N-back and Self-ordered Pointing Tasks). This composite score was created by the sum of the standardized percent correct for all sets of the Self-ordered Pointing Task and the performance score (D′) for all working memory conditions of the N-back Task divided by two in order to obtain a mean standardized score. The use of this composite score was guided by an exploratory factor analysis that demonstrated that performance on various load conditions of the N-back and Self-ordered Pointing Tasks factor together in a factor analysis in older participants. Specifically, all load conditions of the Self-ordered Pointing Task and the two-back load condition of the N-back Task factor together as the factor that best explains all task performance in the older participants. In contrast, the Conditional Association and Object Alternation Tasks did not factor with any load conditions of the N-back or Self-ordered Pointing Tasks, supporting the differential cognitive loading on working memory processes between the N-back and Self-ordered Pointing Tasks as compared to the Conditional Association and Object Alternation Tasks. Although the N-back and Self-ordered Pointing Tasks require some differential cognitive processing for performance, this composite score was created in order to obtain a score that provides a more isolated index of their shared requirement of working memory. Thus, the composite score likely provides an index of the shared cognitive components of the N-back and Self-ordered Pointing Tasks, yet potentially masks other differential properties of each individual task.

A discriminant function analysis was performed in order to determine how well performance on the PFC battery classified the older and younger subjects and determine which cognitive processes were most important in that classification. This analysis showed that the PFC battery correctly classified 95% of the younger subjects and 93% of the older subjects. Performance on the working memory composite was most important for this classification as this measure correlated the most with the standardized canonical discriminant function (r = 0.96 and all other rs < 0.51). The working memory composite alone correctly classified 85% of the young subjects and 94% of the older subjects.

Correlational Analyses

There were significant relationships between age and performance on the two-back condition (r = –0.46 and P < 0.05) (Fig. 5A), the three-back condition (r = –0.64 and P < 0.005) (Fig. 5B) and the total performance of the combined working memory conditions of the N-back Task (r = –0.60 and P = 0.005) (Fig. 5C and Table 2) in the younger subjects. Performance on these working memory measures declined with increasing age. Performance was not related to age in any other task.

There was a significant relationship between age and the Self-ordered Pointing Task percent correct in the six-design load condition (r = –0.51 and P < 0.005), the 12-design load condition (r = –0.40 and P < 0.05) and all sets combined (r = –0.46 and P < 0.01) (Fig. 6A) in the older subjects with the percent correct declining with increasing age. There was a significant relationship between age and performance of the two-back condition of the N-back Task (r = –0.45 and P = 0.01) (Fig. 6B) and all working memory load conditions of the N-back Task combined (r = –0.40 and P < 0.05) (Fig. 6C). Performance was not related to age in any other task.

Correlations Between Regional Volumes, Age and Task Performance

There was a significant increase in the ICV-corrected orbital PFC volume with increasing age (r = 0.39 and P = 0.03) (Fig. 7). No other regional volume was related to age (rs < 0.10 and Ps > 0.10).

There was a significant relationship between performance in the Conditional Association Task and the volume of the ICV-corrected superior PFC region (r = 0.47 and P = 0.01) (Fig. 8A and Table 2). Larger volumes predicted a greater number of errors per trial. Performance in the Conditional Association Task did not correlate with any other regional volume. There was an inverse relationship between the working memory composite and the ICV-corrected volume of the orbital PFC region (r = –0.46 and P = 0.01) (Fig. 8B). Larger volumes predicted worse performance. Performance in the working memory composite did not correlate with any other regional volume. We performed an additional partial correlation between orbital PFC volume and performance on the working memory composite controlling for age in order to determine whether there was an age-independent contribution of orbital volume to working memory performance. This analysis decreased the strength of the relationship between the two factors although a trend was still apparent (r = –0.33 and P = 0.08). Because two subjects had large orbital PFC volumes and poor working memory performance that could affect the correlations obtained (see Fig. 8B), we next analyzed the data without those two subjects (i.e. without subjects with corrected volumes >9). The relationship remained significant with the removal of these two subjects (r = –0.4 and P = 0.03).

We performed one additional analysis in order to determine whether relative preservation of the orbital region of interest compared to all other PFC regions of interest (orbital region of interest/all other regions of interest ratio) was related to cognitive performance. There was a similar relationship between this index of relative change and working memory performance. For example, there were relationships between this relative index and the Self-ordered Pointing Task performance, two-back performance and the working memory composite (all Ps < 0.01) (Fig. 9). In addition, the relative preservation of the orbital region of interest correlated with the Object Alternation Task performance (P < 0.01) (Fig. 10). Greater relative preservation predicted worse object alternation performance.

Discussion

The older subjects performed worse than the younger subjects in the entire battery of cognitive tasks employed in the current study. Thus, there is a generalized decline in the ability to perform tasks supported by the PFC with later aging. The correlations between cognitive performance and the volumetric measurements were task and region specific. Thus, regionally selective alterations in prefrontal morphology could result in a decline of process-specific cognitive abilities with later aging.

Age-related Decline in Cognitive Performance

Measures of effect size and the discriminant analysis showed that the greatest age-related decrements were in the performance of tasks of working memory. A strong decline in working memory abilities with aging was also supported by the number of relationships between working memory performance (two different tasks each at a number of working memory loads) and age in both groups. Similar relationships were not apparent for performance on the other PFC-supported tasks. The smallest effect of aging was found for performance of the zero-back condition of the N-back Task. This finding supports the idea that continuous performance is minimally affected by age (Albert, 1996) and that factors related to a timed task were not responsible for the large differences between the older and younger subjects on the working memory conditions of the N-back Task.

The three-back condition performance correlated with age in the younger subjects and the two-back condition performance correlated with age in the older subjects. These relationships could reflect a staging of working memory capabilities from younger (three-back) to older (two-back) adults. This demonstrates the sensitivity of the N-back Task as a measure of age-related cognitive decline with both early and later aging. Performance on non-working memory PFC-supported tasks did not correlate with age in either group, yet these measures were significantly affected by aging. Conditional association learning and response alternating processes could decline rapidly in the age range separating the younger and older groups of this study, but not show a significant decline after this initial loss. This theory is supported by a trend towards a relationship between age and performance in the younger subjects on the Conditional Association Task (r = 0.39 and P < 0.09) that was absent in the older subjects (r = 0.09 and P = 0.65).

Volumetric Correlations with Task Performance

The volumes of the PFC subregions selectively predicted performance in the cognitive tasks supported by the PFC in older subjects. Specifically, the analyses supported a role for the superior PFC in conditional association learning and the orbital PFC in working memory performance. Larger structures were associated with worse task performance, thereby suggesting that cognitive loss could result from an increased or relatively preserved volume of the region as compared to other regions with late aging. Thus, morphological measurements of the superior and orbital PFC regions selectively predict measures of PFC cognition with aging.

There were relationships between cognitive performance and the superior and orbital PFC volumes, yet only the volume of the orbital region showed a relationship with age. Interestingly, the orbital region showed larger volumes in older subjects. Larger volumes in the region were associated with worse performance on working memory tasks. Larger PFC subregions were associated with worse performance on other tasks as well. These relationships were region (superior and orbital) and task (conditional association and working memory) selective, supporting the speculation that region-specific dysfunction would lead to task-specific attenuation in performance. Still, these results must be considered in relation to more widespread age-related changes in neural function that likely contribute to cognitive decline (Raz et al., 1997).

The finding that the orbital PFC volume was related to working memory abilities was not expected. Still, the organization of working memory processes within the PFC is not completely clear. For example, prior studies have demonstrated a dorsal/ventral organization of spatial/non-spatial working memory processes respectively in non-human primates (Wilson et al., 1993). In addition, the orbitofrontal cortex has been demonstrated to contribute to short-term object memory in monkeys (Meunier et al., 1997). A similar distinction between spatial and non-spatial working memory has been suggested in humans (Courtney et al., 1998). Thus, it is possible that the current study supports this idea as the tasks in the current study assessed non-spatial working memory. The orbital region in our study though was more ventral and medial than the prefrontal regions activated with non-spatial working memory processing described in prior human studies (Courtney et al., 1998). A recent study found orbitofrontal activation for a delayed match-to-sample task using patterns as stimuli (Barrett et al., 2001). This could explain the relationship between the Self-ordered Pointing Task performance and orbitofrontal volume as the Self-ordered Pointing Task uses abstract designs as stimuli. Activation in this prior study was found in a posterior portion of the region described in this study. Interestingly, we found stronger relationships between working memory abilities and orbital volume when using just the four most posterior slices of the region of interest (data not shown).

Prior studies have found a relationship between greater regional volumes and poorer cognitive performance. For example, a study of 70 healthy subjects showed that larger hippocampal volumes are associated with worse explicit memory performance in younger subjects (Chantome et al., 1999). Although the mechanisms of such relationships are unclear, prior studies have found an age-related hypertrophy of astroglia in the frontal cortex that is greater than in the hippocampus (Amenta et al., 1998). This hypertrophy could be a maladaptive compensatory response to white matter degeneration (Peters, 1996; Salat et al., 1999) as astrocytes protect oligodendrocytes from certain types of damage including oxidative stress (Noble et al., 1994). These suggestions are speculative, as it is not known whether these sorts of cellular responses could result in gross morphological changes. Still, age-related hypertrophy of other neural regions including the hypothalamus (Rance et al., 1993) has been reported and neuronal hypertrophy has been suggested to be an early compensatory mechanism with the development of Alzheimer's disease (De Lacalle et al., 1993).

Another potential cause of hypertrophy is the use of medications such as cardiovascular drugs as has been found with neuroleptic treatment in schizophrenic patients (Gur et al., 1998). The subjects that still met entry criteria for the OBAS at the time of testing did not differ from the subjects that did still meet the criteria in orbital volume and there was almost complete overlap in the volumes of these two groups. Similarly, subjects on cardiovascular medication did not differ in volumes compared to the other subjects. Thus, it is unlikely that medication use was a significant factor affecting orbitofrontal volume or the cognitive relationships found.

Another potential interpretation of these data is that relative preservation of a region in the face of other degenerating regions leads to cognitive decline. For example, preservation of orbital PFC function with parallel degeneration of other PFC regions could negatively alter communication among PFC regions and communication of PFC regions with posterior cortical areas. This speculation is supported by the fact that there were significant relationships between cognitive performance and an index of relative preservation within the PFC (the orbital PFC divided by all other PFC regions of interest combined). For example, there was a relationship between relative preservation of the orbital PFC and working memory performance, with higher ratios predicting worse performance. This finding suggests that a combination of greater volume (preservation) of the orbital region along with less volume (degeneration) of the other PFC regions results in a worse working memory performance. In addition, it is possible that there is an age-independent relationship between orbital volume and working memory performance, as there was a trend towards this finding when correlating working memory performance and orbital volume while controlling for age-related variance in these measures.

A recent study in our laboratory found a selective preservation of the orbital PFC volume in relation to the other subregions in older subjects and a trend towards preservation of the same region in a group of subjects with Alzheimer's disease (Salat et al., 2001). This finding is compelling, as the volumetric measurements for the current study were collected from an independent sample of scans (there was very little overlap in the subjects studied in this and the prior study and there was no overlap in the year of the scan analyzed in the same subjects between the two studies). These data suggest that there is a relative preservation, growth or some combination of these factors contributing to the larger volume of the orbital PFC associated with advancing age. The data showing larger volumes with increasing age suggest actual growth of the region does occur with aging and that this growth has a negative consequence on cognitive abilities. Still, the results presented must be considered with caution, as some of the relationships found would not endure with statistical correction for multiple analyses. Such corrections could only be possible with much larger subject samples, which are difficult to obtain in the healthy elderly population. Another potential idea is that these results reflect an effect of the particular cohort studied. Subjects from the OBAS were entered into this study due to their exceptional health and an attractive speculation is that the volume of this region is generally larger in those likely to advance to healthy late aging. These interpretations should be considered with discretion and must be considered within a framework of neural alteration in the entire brain with aging as volumetric decline has been found in numerous brain regions with increasing age including the corpus callosum (Salat et al., 1997), fusiform, inferior temporal and superior parietal cortices (Raz et al., 1997). Still, converging evidence from this study and our prior study of aging and Alzheimer's disease recommend further study of the orbital PFC region and working memory in aging.

Relating Structure to Function in Functional Magnetic Resonance Imaging Studies

The current study demonstrates the potential importance of considering age-related structural changes in the brain when examining data from functional imaging studies. This is particularly true given the finding that structural changes were related to cognitive abilities. There is no consensus on how aging affects the patterns of functional activation for cognitive tasks, but the entire range of possible differences between older and younger subjects has been found. For example, older subjects show both stronger and weaker activations in areas that younger subjects activate as well as differential spatial patterns (Cabeza et al., 2000). These differences in activation patterns between younger and older subjects could be due to differences in hemodynamic coupling between neural activity and the blood oxygen level-dependent fMRI signal in younger and older subjects (D'Esposito et al., 1999; Buckner et al., 2000). Future studies examining the relationship between structural and functional measures could be informative towards reducing some of the age-related variance in the hemodynamic profile of older adults.

A recent study of working memory found differential activation in the dorsolateral PFC between younger and older subjects on working memory tasks (Rypma and D'Esposito, 2000). This study did not examine potential differences in orbitofrontal function or how structural changes relate to functional activity. Thus, it is plausible that both dorsolateral and orbitofrontal dysfunction contribute differentially to age-related decline in working memory performance. The current study suggests that structural measurements must be taken into account when examining functional differences with aging.

Behavioral Inhibition Requirements in Working Memory Tasks

Working memory task requirements typically include inhibition of pre-potent responses and the ability to shift responses. The orbital PFC and ventromedial PFC have been implicated in a variety of similar cognitive processes including behavioral inhibition (Lineberry and Siegel, 1971; Dias et al., 1997) attentional set shifting (Dias et al., 1996), guessing (Elliott et al., 1999) and advantageous decision making (Bechara et al., 1994, 1997). Older subjects make more perseverative errors on the Self-ordered Pointing Task than younger subjects (West et al., 1998) and demonstrate disinhibitory responses with Conditional Association performance (Levine et al., 1997). Thus, the relationship found between the orbital PFC volume and working memory performance in this study could be due to inhibitory deficits or response perseveration. Older subjects had significantly faster reaction times with incorrect responses compared to correct responses to targets on the N-back Task. The younger subjects did not show such a reaction time difference, suggesting that disinhibition contributes to errors on the N-back Task in older subjects. Thus, age-related changes in the orbital PFC could result in attenuated performance on working memory tasks via attenuated inhibitory or impaired response-shifting mechanisms.

Notes

This research was supported by NIA grant AG12611, NIMH grant MH11855 and a VA merit review grant. The authors thank Milar Moore, Dara Wasserman, Alison Dame and Suzanne Lehman for technical assistance and subject coordination for this study and Drs Barry Oken, John Crabbe and Gary Sexton for statistical consultation and valuable comments on this manuscript.

Table 1

Neuropsychological performance in older subjectsa

 Mean ± SE Average score age >70b 
aData obtained within 12 months of MRI scan. 
bMany norms do not cover the full age range of subjects. Norms may be representative of subjects younger than those tested. 
Age  84.3 ± 0.9  
 (71.6–93.5)  
MMSE (max. 30)  28.6 ± 0.2 26.9 
 (25–30)  
GDS (max. 30)  5.0 ± 0.6 
  (0–12)  
Boston Naming Test-Total (max. 60)  55.5 ±0.6 55.8 
 (48–60)  
Digit Span (WAIS-R Raw) (max. 28)  14.1 ± 0.5 13 
  (8–20)  
Block Design (WAIS-R Raw) (max. 51)  26.8 ± 1.6 17 
  (7–43)  
Digit Symbol (WAIS-R Raw) (max. 93)  41.5 ± 1.8 31 
 (25–62)  
Logical Memory I (WMS-3) (max. 50)  28.0 ± 1.2 20.9 
 (10–43)  
Logical Memory II (Delayed) (WMS-3) (max. 50)  29.5 ± 1.3 14.7 
  (9–44)  
Visual Reproduction I (WMS-R) (max. 41)  32.8 ± 0.8 24.2 
 (19–40)  
Visual Reproduction II (WMS-R) (max. 41)  27.6 ± 1.5 16.5 
  (4 – 38)  
Face Recognition I (WMS-3) (max. 48)  34.0 ± 0.8 32 
 (25 – 43)  
Face Recognition II (WMS-3) (max. 48)  34.2 ± 0.9 31 
 (23 – 46)  
 Mean ± SE Average score age >70b 
aData obtained within 12 months of MRI scan. 
bMany norms do not cover the full age range of subjects. Norms may be representative of subjects younger than those tested. 
Age  84.3 ± 0.9  
 (71.6–93.5)  
MMSE (max. 30)  28.6 ± 0.2 26.9 
 (25–30)  
GDS (max. 30)  5.0 ± 0.6 
  (0–12)  
Boston Naming Test-Total (max. 60)  55.5 ±0.6 55.8 
 (48–60)  
Digit Span (WAIS-R Raw) (max. 28)  14.1 ± 0.5 13 
  (8–20)  
Block Design (WAIS-R Raw) (max. 51)  26.8 ± 1.6 17 
  (7–43)  
Digit Symbol (WAIS-R Raw) (max. 93)  41.5 ± 1.8 31 
 (25–62)  
Logical Memory I (WMS-3) (max. 50)  28.0 ± 1.2 20.9 
 (10–43)  
Logical Memory II (Delayed) (WMS-3) (max. 50)  29.5 ± 1.3 14.7 
  (9–44)  
Visual Reproduction I (WMS-R) (max. 41)  32.8 ± 0.8 24.2 
 (19–40)  
Visual Reproduction II (WMS-R) (max. 41)  27.6 ± 1.5 16.5 
  (4 – 38)  
Face Recognition I (WMS-3) (max. 48)  34.0 ± 0.8 32 
 (25 – 43)  
Face Recognition II (WMS-3) (max. 48)  34.2 ± 0.9 31 
 (23 – 46)  
Table 2

Correlation table of relationships between age and cognitive performance in younger and older subjectsa

Task Load Young Old 
CAT, Conditional Association Task; SOP, Self-ordered Pointing Task; N-back, N-back task; OAT, Object Alternation Task. 
aPearson's correlation r-values presented. *P < 0.05, **P < 0.01. 
CAT NA  0.39  0.09 
SOP 6-design –0.23 –0.51* 
 8-design –0.17 –0.35 
 10-design –0.41 –0.08 
 12-design –0.17 –0.41* 
 Total SOP –0.27 –0.46* 
N-back 0-back –0.17 –0.24 
 1-back –0.06 –0.26 
 2-back –0.46* –0.45* 
 3-back –0.64** –0.15 
 Total N-back –0.60** –0.40* 
OAT NA  0.25  0.10 
Task Load Young Old 
CAT, Conditional Association Task; SOP, Self-ordered Pointing Task; N-back, N-back task; OAT, Object Alternation Task. 
aPearson's correlation r-values presented. *P < 0.05, **P < 0.01. 
CAT NA  0.39  0.09 
SOP 6-design –0.23 –0.51* 
 8-design –0.17 –0.35 
 10-design –0.41 –0.08 
 12-design –0.17 –0.41* 
 Total SOP –0.27 –0.46* 
N-back 0-back –0.17 –0.24 
 1-back –0.06 –0.26 
 2-back –0.46* –0.45* 
 3-back –0.64** –0.15 
 Total N-back –0.60** –0.40* 
OAT NA  0.25  0.10 
Table 3

Correlation table of relationships between cognitive task performance and regional volumes in older subjectsa

 Superior Middle Inferior Orbital 
CAT, Conditional Association Task; WM COMP, Working Memory Composite; OAT, Object Alternation Task. 
aPearson's correlation r-values presented. *P < 0.05. 
CAT  0.47* 0.27 –0.14  0.05 
WM COMP –0.18 0.04 0.13 –0.46* 
OAT –0.20 –0.20 –0.26  0.11 
 Superior Middle Inferior Orbital 
CAT, Conditional Association Task; WM COMP, Working Memory Composite; OAT, Object Alternation Task. 
aPearson's correlation r-values presented. *P < 0.05. 
CAT  0.47* 0.27 –0.14  0.05 
WM COMP –0.18 0.04 0.13 –0.46* 
OAT –0.20 –0.20 –0.26  0.11 
Figure 1.

Cognitive tasks differentially supported by regions within the PFC. See the text for full descriptions of the prefrontal tasks. (A) The Conditional Association Task. Subjects were to choose a particular design conditional upon the color of the line at the top of the monitor. For this example, the top center design is highlighted in order to demonstrate the choice of this design every time the associated colored line appeared above the designs. (B) The Self-ordered Pointing Task. Example of the eight-design working memory load condition (top panel) and 10-design working memory load condition (bottom panel). In each condition subjects were to choose a design on each and every page in the set without choosing the same design more than once. (C) The N-back Task. The example shows the two-back working memory load and zero-back (control) conditions of the task. Subjects viewed a string of letters presented sequentially on the screen one at a time. In the two-back condition subjects were to respond to any letter that repeated itself two letters back in the sequence (i.e. any letter that repeated itself separated by one letter in the sequence). In this example, the letter ‘a’ is repeated two letters back, separated by one letter (‘Q’). In the zero-back condition subjects were to respond every time they saw a specific letter and this letter did not have to be repeated. In this example, the subject responds as a target to every ‘q’ in the sequence. (D) The Object Alternation Task. Subjects had to learn that the best strategy for performance of the task was to alternate between choosing the circle and square on each trial, regardless of the object's spatial position (to the left or right of the other object). Thus, for optimal performance, subjects would choose the circle on trials 1, 3, 5, 7, 9, etc. and the square on trials 2, 4, 6, 8, 10, etc.

Figure 1.

Cognitive tasks differentially supported by regions within the PFC. See the text for full descriptions of the prefrontal tasks. (A) The Conditional Association Task. Subjects were to choose a particular design conditional upon the color of the line at the top of the monitor. For this example, the top center design is highlighted in order to demonstrate the choice of this design every time the associated colored line appeared above the designs. (B) The Self-ordered Pointing Task. Example of the eight-design working memory load condition (top panel) and 10-design working memory load condition (bottom panel). In each condition subjects were to choose a design on each and every page in the set without choosing the same design more than once. (C) The N-back Task. The example shows the two-back working memory load and zero-back (control) conditions of the task. Subjects viewed a string of letters presented sequentially on the screen one at a time. In the two-back condition subjects were to respond to any letter that repeated itself two letters back in the sequence (i.e. any letter that repeated itself separated by one letter in the sequence). In this example, the letter ‘a’ is repeated two letters back, separated by one letter (‘Q’). In the zero-back condition subjects were to respond every time they saw a specific letter and this letter did not have to be repeated. In this example, the subject responds as a target to every ‘q’ in the sequence. (D) The Object Alternation Task. Subjects had to learn that the best strategy for performance of the task was to alternate between choosing the circle and square on each trial, regardless of the object's spatial position (to the left or right of the other object). Thus, for optimal performance, subjects would choose the circle on trials 1, 3, 5, 7, 9, etc. and the square on trials 2, 4, 6, 8, 10, etc.

Figure 2.

Cartoon representation of the volumetric method used for calculating prefrontal region of interest volumes. The images are the same as those used in the data collection, but regional demarcations have been smoothed for publication purposes. The total prefrontal volume, prefrontal white matter volume and subregion parcellation for all subjects were determined by edge tracing the cortical ribbon and gray/white boundary with a cursor directly on a computer display. (A) The most posterior slice of the prefrontal region of interest is shown. The total prefrontal volume was defined by tracing all gyri and sulci around the cortical ribbon of each hemisphere in the T2-weighted image. Pixel areas were transformed to volumes as described in the text. (B) The prefrontal white matter volume was traced in the proton density-weighted image of the same exact slice as (A). The prefrontal gray matter volume was calculated by subtracting the prefrontal white matter volume from the total prefrontal volume. (C) T2-weighted image of four posterior coronal PFC slices (C1 to C4: most posterior to most anterior) showing regional delineations for the superior (yellow), middle (pink), inferior (orange), orbital (green) and anterior cingulate (blue) regions. The subregions were defined as described in the text.

Figure 2.

Cartoon representation of the volumetric method used for calculating prefrontal region of interest volumes. The images are the same as those used in the data collection, but regional demarcations have been smoothed for publication purposes. The total prefrontal volume, prefrontal white matter volume and subregion parcellation for all subjects were determined by edge tracing the cortical ribbon and gray/white boundary with a cursor directly on a computer display. (A) The most posterior slice of the prefrontal region of interest is shown. The total prefrontal volume was defined by tracing all gyri and sulci around the cortical ribbon of each hemisphere in the T2-weighted image. Pixel areas were transformed to volumes as described in the text. (B) The prefrontal white matter volume was traced in the proton density-weighted image of the same exact slice as (A). The prefrontal gray matter volume was calculated by subtracting the prefrontal white matter volume from the total prefrontal volume. (C) T2-weighted image of four posterior coronal PFC slices (C1 to C4: most posterior to most anterior) showing regional delineations for the superior (yellow), middle (pink), inferior (orange), orbital (green) and anterior cingulate (blue) regions. The subregions were defined as described in the text.

Figure 3.

Cognitive performance in the PFC cognitive battery in young and older subjects. (A) Older subjects made more errors per trial in the Conditional Association Task. (B) Older subjects performed worse than younger subjects for all working memory loads of the Self-ordered Pointing Task (P < 0.001). Both young and older subjects performed worse with increasing working memory load (P < 0.001). There was a significant interaction, with greater differences between groups at the higher working memory loads (P < 0.001). (C) Younger subjects had a significantly greater percentage of targets correctly identified for all working memory loads of the N-back Task (P < 0.001). Both younger and older subjects decreased the number of correct responses with increasing working memory loads (P < 0.001) and there was no significant interaction. (D) Older subjects made more errors per trial in the Object Alternation Task (*Ps < 0.01).

Figure 3.

Cognitive performance in the PFC cognitive battery in young and older subjects. (A) Older subjects made more errors per trial in the Conditional Association Task. (B) Older subjects performed worse than younger subjects for all working memory loads of the Self-ordered Pointing Task (P < 0.001). Both young and older subjects performed worse with increasing working memory load (P < 0.001). There was a significant interaction, with greater differences between groups at the higher working memory loads (P < 0.001). (C) Younger subjects had a significantly greater percentage of targets correctly identified for all working memory loads of the N-back Task (P < 0.001). Both younger and older subjects decreased the number of correct responses with increasing working memory loads (P < 0.001) and there was no significant interaction. (D) Older subjects made more errors per trial in the Object Alternation Task (*Ps < 0.01).

Figure 4.

N-back Task performance reaction times in younger and older subjects. Older subjects had significantly faster reaction times for incorrect target compared to correct target responses. The opposite pattern was found for non-target responses in older subjects. Younger subjects did not show similar reaction time patterns (*P < 0.05 compared to target correct responses in older subjects).

Figure 4.

N-back Task performance reaction times in younger and older subjects. Older subjects had significantly faster reaction times for incorrect target compared to correct target responses. The opposite pattern was found for non-target responses in older subjects. Younger subjects did not show similar reaction time patterns (*P < 0.05 compared to target correct responses in older subjects).

Figure 5.

Scatter plot of performance in the PFC cognitive battery and age in younger subjects. There was a relationship between age and performance on (A) the two-back condition, (B) the three-back condition and (C) all N-back Task conditions combined in the younger subjects. Other relationships existed, but are not shown in the figure.

Figure 5.

Scatter plot of performance in the PFC cognitive battery and age in younger subjects. There was a relationship between age and performance on (A) the two-back condition, (B) the three-back condition and (C) all N-back Task conditions combined in the younger subjects. Other relationships existed, but are not shown in the figure.

Figure 6.

Scatter plot of performance in the PFC cognitive battery and age in older subjects. There was a relationship between age and percent correct of (A) all load conditions of the Self-ordered Pointing Task combined, (B) the two-back condition of the N-back Task and (C) total performance on all conditions of the N-back Task. Other relationships existed, but are not shown in the figure.

Figure 6.

Scatter plot of performance in the PFC cognitive battery and age in older subjects. There was a relationship between age and percent correct of (A) all load conditions of the Self-ordered Pointing Task combined, (B) the two-back condition of the N-back Task and (C) total performance on all conditions of the N-back Task. Other relationships existed, but are not shown in the figure.

Figure 7.

Scatter plot of the correlation between age (years) and orbital PFC volume (cm3) corrected for subjects' total ICV. Older subjects had larger orbital PFC volumes.

Figure 7.

Scatter plot of the correlation between age (years) and orbital PFC volume (cm3) corrected for subjects' total ICV. Older subjects had larger orbital PFC volumes.

Figure 8.

Scatter plot of the correlations between cognitive task performance and regional volumes. (A) A greater superior PFC volume was associated with more errors per trial on the Conditional Association Task. (B) A greater orbital PFC volume was associated with worse performance (D') on the working memory composite.

Figure 8.

Scatter plot of the correlations between cognitive task performance and regional volumes. (A) A greater superior PFC volume was associated with more errors per trial on the Conditional Association Task. (B) A greater orbital PFC volume was associated with worse performance (D') on the working memory composite.

Figure 9.

Scatter plot of the correlations between working memory task performance and relative preservation of the orbital PFC. Greater preservation was associated with (A) more errors on the Self-ordered Pointing Task, (B) worse performance on the two-back condition of the N-back Task and (C) worse performance on the working memory composite.

Figure 9.

Scatter plot of the correlations between working memory task performance and relative preservation of the orbital PFC. Greater preservation was associated with (A) more errors on the Self-ordered Pointing Task, (B) worse performance on the two-back condition of the N-back Task and (C) worse performance on the working memory composite.

Figure 10.

Scatter plot of Object Alternation Task performance and relative volume of the orbital region of interest to other PFC regions of interest. Greater relative preservation of the orbital region of interest was associated with more errors per trial.

Figure 10.

Scatter plot of Object Alternation Task performance and relative volume of the orbital region of interest to other PFC regions of interest. Greater relative preservation of the orbital region of interest was associated with more errors per trial.

4
Current address: MGH-NMR Center, Department of Radiology, Building 149, 13th Street, Mail Code 149 (2301), Charlestown, MA 02129-2060, USA

References

Albert MS (1996) Neuropsychology of aging: findings in humans and monkeys. In: Handbook of the biology of aging (Schneider EL, Rowe JW, eds), pp. 217–229. San Diego: Academic Press.
Amenta F, Bronzetti E, Sabbatini M, Vega JA (
1998
) Astrocyte changes in aging cerebral cortex and hippocampus: a quantitative immunohisto-chemical study.
Microsc Res Tech
 
43
:
29
–33.
Barrett NA, Large MM, Smith GL, Michie PT, Karayanidis F, Kavanagh DJ, Fawdry R, Henderson D, O'Sullivan BT (
2001
) Human cortical processing of colour and pattern.
Hum Brain Mapp
 
13
:
213
–225.
Bechara A, Damasio AR, Damasio H, Anderson SW (
1994
) Insensitivity to future consequences following damage to human prefrontal cortex.
Cognition
 
50
:
7
–15.
Bechara A, Damasio H, Tranel D, Damasio AR (
1997
) Deciding advantageously before knowing the advantageous strategy.
Science
 
275
:
1293
–1295.
Benton AL, Hamsher K (1976) Multilingual aphasia examination. Iowa City: University of Iowa Press.
Braver TS, Cohen JD, Nystrom LE, Jonides J, Smith EE, Noll DC (
1997
) A parametric study of prefrontal cortex involvement in human working memory.
Neuroimage
 
5
:
49
–62.
Buckner RL, Snyder AZ, Sanders AL, Raichle ME, Morris JC (
2000
) Functional brain imaging of young, nondemented, and demented older adults.
J Cogn Neurosci
 
12
(Suppl 2):
24
–34.
Cabeza R, Anderson ND, Houle S, Mangels JA, Nyberg L. (
2000
) Age-related differences in neural activity during item and temporal-order memory retrieval: a positron emission tomography study.
J Cogn Neurosci
 
12
:
197
–206.
Chantome M, Perruchet P, Hasboun D, Dormont D, Sahel M, Sourour N, Zouaoui A, Marsault C, Duyme M (
1999
) Is there a negative correlation between explicit memory and hippocampal volume?
Neuroimage
 
10
:
589
–595.
Cohen JD, Perlstein WM, Braver TS, Nystrom LE, Noll DC, Jonides J, Smith EE (
1997
) Temporal dynamics of brain activation during a working memory task.
Nature
 
386
:
604
–608.
Courtney SM, Petit L, Maisog JM, Ungerleider LG, Haxby JV (
1998
) An area specialized for spatial working memory in human frontal cortex.
Science
 
279
:
1347
–1351.
Daigneault S, Braun CM (
1993
) Working memory and the Self-ordered Pointing Task: further evidence of early prefrontal decline in normal aging.
J Clin Exp Neuropsychol
 
15
:
881
–895.
Damasio H (1991) Neuroanatomy of the frontal lobe in vivo: a comment on methodology. In: Frontal lobe function and dysfunction (Levin HS, Eisenberg HM, Benton AL, eds), pp. 92–121. New York: Oxford University Press.
Damasio H (1995) Human brain anatomy in computerized images. New York: Oxford University Press.
De Lacalle S, Iraizoz I, Gonzalo LM (
1993
) Cell loss in supraoptic and para-ventricular nucleus in Alzheimer's disease.
Brain Res
 
609
:
154
–158.
D'Esposito M, Zarahn E, Aguirre GK, Rypma B (
1999
) The effect of normal aging on the coupling of neural activity to the bold hemodynamic response.
Neuroimage
 
10
:
6
–14.
Dias R, Robbins TW, Roberts AC (
1996
) Dissociation in prefrontal cortex of affective and attentional shifts.
Nature
 
380
:
69
–72.
Dias R, Robbins TW, Roberts AC (
1997
) Dissociable forms of inhibitory control within prefrontal cortex with an analog of the Wisconsin Card Sort Test: restriction to novel situations and independence from ‘on-line’ processing.
J Neurosci
 
17
:
9285
–9297.
Dickson DW, Crystal HA, Mattiace LA, Masur DM, Blau AD, Davies P, Yen SH, Aronson MK (
1992
) Identification of normal and pathological aging in prospectively studied nondemented elderly humans.
Neurobiol Aging
 
13
:
179
–189.
Dobbs AR, Rule BG (
1989
) Adult age differences in working memory.
Psychol Aging
 
4
:
500
–503.
Dolman R, Roy EA, Dimeck PT, Hall CR (
2000
) Age, gesture span, and dissociations among component subsystems of working memory.
Brain Cogn
 
43
:
164
–168.
Drevets WC (
1999
) Prefrontal cortical-amygdalar metabolism in major depression.
Ann NY Acad Sci
 
877
:
614
–637.
Elliott R, Rees G, Dolan RJ (
1999
) Ventromedial prefrontal cortex mediates guessing.
Neuropsychologia
 
37
:
403
–411.
Fillenbaum GG (
1985
) Screening the elderly. A brief instrumental activities of daily living measure.
J Am Geriat Soc
 
33
:
698
–706.
Folstein MF, Folstein SE, McHugh PR (
1975
) ‘Mini-mental State’. A practical method for grading the cognitive state of patients for the clinician.
J Psychiat Res
 
12
:
189
–198.
Freedman M, Black S, Ebert P, Binns M (
1998
) Orbitofrontal function, object alternation and perseveration.
Cereb Cortex
 
8
:
18
–27.
Gur RE, Maany V, Mozley D, Swanson C, Bilker W, Gur RC (1998) Subcortical MRI volumes in neuroleptic-naive and treated patients with schizophrenia Am J Psychiat 155:1711–1717.
Howieson DB, Holm LA, Kaye JA, Oken BS, Howieson J (
1993
) Neurologic function in the optimally healthy oldest old. Neuropsychological evaluation.
Neurology
 
43
:
1882
–1886.
Hughes CP, Berg L, Danziger WL, Coben LA, Martin RL (
1982
) A new clinical scale for the staging of dementia.
Br J Psychiat
 
140
:
566
–572.
Jahanshahi M, Profice P, Brown RG, Ridding MC, Dirnberger G, Rothwell JC (
1998
) The effects of transcranial magnetic stimulation over the dorsolateral prefrontal cortex on suppression of habitual counting during random number generation.
Brain
 
121
:
1533
–1544.
Janowsky JS, Chavez B, Orwoll E (
2000
) Sex steroids modify working memory.
J Cogn Neurosci
 
12
:
407
–414.
Jenkins R, Fox NC, Rossor AM, Harvey RJ, Rossor MN. Intracranial volume and Alzheimer disease: evidence against the cerebral reserve hypothesis.
Arch Neurol
 
57
:
220
–224.
Jernigan TL, Archibald SL, Berhow MT, Sowell ER, Foster DS, Hesselink JR (
1991
) Cerebral structure on MRI, part I: localization of age-related changes.
Biol Psychiat
 
29
:
55
–67.
Kaye JA, Oken BS, Howieson DB, Howieson J, Holm LA, Dennison K (
1994
) Neurologic evaluation of the optimally healthy oldest old.
Arch Neurol
 
51
:
1205
–1211.
Kaye JA, Swihart T, Howieson D, Dame A, Moore MM, Karnos T, Camicioli R, Ball M, Oken B, Sexton G (
1997
) Volume loss of the hippocampus and temporal lobe in healthy elderly persons destined to develop dementia.
Neurology
 
48
:
1297
–1304.
Levine B, Stuss DT, Milberg WP (
1997
) Effects of aging on conditional associative learning: process analyses and comparison with focal frontal lesions.
Neuropsychology
 
11
:
367
–381.
Lineberry CG, Siegel J (
1971
) EEG synchronization, behavioral inhibition, and mesencephalic unit effects produced by stimulation of orbital cortex, basal forebrain and caudate nucleus.
Brain Res
 
34
:
143
–161.
Meunier M, Bachevalier J, Mishkin M (
1997
) Effects of orbital frontal and anterior cingulate lesions on object and spatial memory in rhesus monkeys.
Neuropsychologia
 
35
:
999
–1015.
Mueller EA, Moore MM, Kerr DC, Sexton G, Camicioli RM, Howieson DB, Quinn JF, Kaye JA (
1998
) Brain volume preserved in healthy elderly through the eleventh decade.
Neurology
 
51
:
1555
–1562.
Noble PG, Antel JP, Yong VW (
1994
) Astrocytes and catalase prevent the toxicity of catecholamines to oligodendrocytes.
Brain Res
 
633
:
83
–90.
Pelosi L, Blumhardt LD (
1999
) Effects of age on working memory: an event-related potential study.
Brain Res Cogn Brain Res
 
7
:
321
–334.
Peters A (
1996
) Age-related changes in oligodendrocytes in monkey cerebral cortex.
J Comp Neurol
 
371
:
153
–163.
Peters A, Leahu D, Moss MB, McNally KJ (
1994
) The effects of aging on area 46 of the frontal cortex of the rhesus monkey.
Cereb Cortex
 
4
:
621
–635.
Petrides M (
1985
) Deficits on conditional associative-learning tasks after frontal- and temporal-lobe lesions in man.
Neuropsychologia
 
23
:
601
–614.
Petrides M (
1995
) Impairments on nonspatial self-ordered and externally ordered working memory tasks after lesions of the mid-dorsal part of the lateral frontal cortex in the monkey.
J Neurosci
 
15
:
359
–375.
Petrides M, Milner B (
1982
) Deficits on subject-ordered tasks after frontal- and temporal-lobe lesions in man.
Neuropsychologia
 
20
:
249
–262.
Petrides M, Alivisatos B, Meyer E, Evans AC (
1993
) Functional activation of the human frontal cortex during the performance of verbal working memory tasks.
Proc Natl Acad Sci USA
 
90
:
878
–882.
Rance NE, Uswandi SV, McMullen NT (
1993
) Neuronal hypertrophy in the hypothalamus of older men.
Neurobiol Aging
 
14
:
337
–342.
Raz N, Gunning FM, Head D, Dupuis JH, McQuain J, Briggs SD, Loken WJ, Thornton AE, Acker JD (
1997
) Selective aging of the human cerebral cortex observed in vivo: differential vulnerability of the prefrontal gray matter.
Cereb Cortex
 
7
:
268
–282.
Raz N, Gunning-Dixon FM, Head D, Dupuis JH, Acker JD (
1998
) Neuroanatomical correlates of cognitive aging: evidence from structural magnetic resonance imaging.
Neuropsychology
 
12
:
95
–114.
Raz N, Briggs SD, Marks W, Acker JD (
1999
) Age-related deficits in generation and manipulation of mental images: II. The role of dorsolateral prefrontal cortex.
Psychol Aging
 
14
:
436
–444.
Rypma B, D'Esposito M (
2000
) Isolating the neural mechanisms of age-related changes in human working memory.
Nat Neurosci
 
3
:
509
–515.
Salat D, Ward A, Kaye JA, Janowsky JS (
1997
) Sex differences in the corpus callosum with aging.
Neurobiol Aging
 
2
:
191
–197.
Salat DH, Kaye JA, Janowsky JS (
1999
) Prefrontal gray and white matter volumes in healthy aging and Alzheimer disease.
Arch Neurol
 
56
:
338
–344.
Salat DH, Stangl PA, Kaye JA, Janowsky JS (
1999
) Sex differences in prefrontal volume with aging and Alzheimer's disease.
Neurobiol Aging
 
20
:
591
–596.
Salat DH, Kaye JA, Janowsky JS (
2001
) Selective preservation and degeneration of the prefrontal cortex in aging and Alzheimer disease.
Arch Neurol
 
58
:
1403
–1408.
Schretlen D, Pearlson GD, Anthony JC, Aylward EH, Augustine AM, Davis A, Barta P (
2000
) Elucidating the contributions of processing speed, executive ability, and frontal lobe volume to normal age-related differences in fluid intelligence.
J Int Neuropsychol Soc
 
6
:
52
–61.
Weschler D (1981) WAIS-R manual. San Antonio, TX: The Psychological Corp.
West RL (
1996
) An application of prefrontal cortex function theory to cognitive aging.
Psychol Bull
 
120
:
272
–292.
West R, Ergis AM, Winocur G, Saint-Cyr J (
1998
) The contribution of impaired working memory monitoring to performance of the selfordered pointing task in normal aging and Parkinson's disease.
Neuropsychology
 
12
:
546
–554.
Wiegersma S, Meertse K (
1990
) Subjective ordering, working memory, and aging.
Exp Aging Res
 
16
:
73
–77.
Wilson FAW, O'Scalaidhe SP, Goldman-Rakic PS (
1993
) Dissociation of object and spatial processing domains in primate prefrontal cortex.
Science
 
260
:
1955
–1958.
Wingfield A, Stine EA, Lahar CJ, Aberdeen JS (
1988
) Does the capacity of working memory change with age?
Exp Aging Res
 
14
:
103
–107.
Yesavage JA (
1983
) Bipolar illness: correlates of dangerous inpatient behaviour.
Br J Psychiat
 
143
:
554
–557.
Yesavage JA, Brink TL, Rose TL, Lum O, Huang V, Adey M, Leirer VO (
1982
) Development and validation of a geriatric depression screening scale: a preliminary report.
J Psychiat Res
 
17
:
37
–49.