On age differences in prefrontal function: The importance of emotional/cognitive integration

Abstract

Evidence of prefrontal cortex decline among healthy older adults has been widely reported, although many questions remain regarding the functional heterogeneity of the prefrontal lobes and the uniformity (or lack thereof) with which discrete regions decline with age. MacPherson, Phillips, and Della Sala (2002) previously reported age differences in tasks associated with dorsolateral prefrontal cortex (DLPFC) function (executive control), but not for tasks associated with ventromedial prefrontal cortex (VMPFC) function (emotional/cognitive integration). The present study, conducted using 39 younger adults and 39 older adults, replicates the MacPherson et al. findings regarding DLPFC functioning. However, and perhaps due to the use of more sensitive tasks, we also find age differences in tasks associated with VMPFC function. Specifically, both univariate and multivariate analyses indicated older adults showed deficits across the DLPFC and VMPFC tasks. Exploratory factor analysis of the task performance scores indicated four underlying dimensions, two related to DLPFC functioning and two related to VMPFC functioning. A set of structural equation models specifying age effects on the four task performance factors was tested, in order to contrast models of process-specific vs. common age effects. Our results suggest that older adults show deficits in emotional/cognitive integration as well as in executive function, and that those effects do include process-specific age deficits.

Introduction

Numerous studies in recent years, including human and animal research, have documented the changes in neurological and neuropsychological status occurring in the aging brain (Denburg et al., 2005, Makris et al., 2007, Peters, 2002, Phillips and Della Sala, 1999, Picq, 2007, Raz et al., 2004, West, 1996). Most notably, amidst the evidence of sometimes subtle, yet widespread anatomical alterations occurring among healthy older adults (e.g., Dennis and Cabeza, 2008, Rabbitt and Lowe, 2000; see also Woodruff-Pak, 1997), the current literature emphasizes changes in frontal-lobe anatomy and physiology as contributing specifically to neuropsychological declines occurring with advancing age (see Cabeza, 2002, Langley and Madden, 2000, Madden et al., 1999, Phillips and Della Sala, 1999, West, 1996). Efforts to characterize the neurocognitive status of healthy older adults, reflecting the convergence of neuroimaging and neuropsychological methods, have helped clarify the nature of performance decrements in various frontal-related skills, including episodic and working memory (Head et al., 2002, Rajah and D’Esposito, 2005), executive processing (Head et al., 2002) and emotional processing (Ruffman, Henry, Livingstone, & Phillips, 2008), while also providing insight into the neural correlates of relatively preserved abilities such as language functions and semantic memory (Dennis & Cabeza, 2008; see Allen et al., 2005).

In a recent example of this work, MacPherson, Phillips, and Della Sala (2002) addressed the effect of aging on the frontal lobes, and focused on the neuropsychological correlates of two different prefrontal cortex regions—the ventromedial prefrontal cortex (VMPFC) and dorsolateral prefrontal cortex (DLPFC). Relative to younger adults, older adults showed diminished performance on tasks associated with DLPFC function, including the Wisconsin Card Sorting Task (WCST, Grant & Berg, 1948), the Self-Ordered Pointing Task (SOPT, Petrides & Milner, 1982), and the Delayed-Response Task (Hunter, 1913). Alternatively, age effects were not found on the Gambler's Task (Bechara, Damasio, Damasio, & Anderson, 1994), the Faux Pas Task (Baron-Cohen & Ring, 1994), or the Emotion Identification Task (Hornak, Rolls, & Wade, 1996)—each of which is believed to tap into emotional processes and/or social decision making associated with the VMPFC. Importantly, these findings suggest a potentially non-uniform decline in the frontal lobes (MacPherson et al., 2002), which contradicts the general frontal-lobe hypothesis of aging (West, 1996).

Cognitive aging has been linked in the literature to alterations within the frontal lobes—principally the prefrontal cortex (PFC, e.g. Garden et al., 2001, MacPherson et al., 2002, West, 1996). According to the frontal aging hypothesis (West, 1996), age-related cognitive deficits are primarily due to general PFC dysfunction, with anatomical and physiological support for this perspective coming from numerous studies underscoring the degree of age-related frontal-lobe atrophy (Head et al., 2002, Raz et al., 1997), alterations in neurotransmitter activity (e.g., dopamine function, Bachman and Farde, 2005, Dennis and Cabeza, 2008), markers of neuronal degeneration and or disruption (Greenwood, 2000; see Phillips & Della Sala, 1999) and declines in functional connectivity with other brain regions (Makris et al., 2007).

Although the frontal lobe is certainly not the only region of the brain to show changes with age (for review, see Dennis and Cabeza, 2008, Greenwood, 2000), the highly integrative prefrontal association cortex is naturally emphasized in age-related functional decline given its involvement in memory, decision making, emotional regulation, and general executive function. Accordingly, the frontal aging hypothesis has provided an important framework for investigating age-related changes in cognition and behavior; however, it has not been uniformly accepted as originally presented (e.g., MacPherson et al., 2002, Phillips and Della Sala, 1999), particularly in light of anatomical and emerging behavioral data reflecting functionally distinct divisions within the PFC (Lamar and Resnick, 2004, Rajah and D’Esposito, 2005). Indeed, the functional heterogeneity of the PFC is an important consideration for contemporary perspectives of this brain region, particularly given the extensive and reciprocal connections with other association cortices, limbic areas, and striatal regions (Grafman, 2002). The work of MacPherson et al. highlights this functional subdivision within the prefrontal region, arguing for the psychometric distinction between the dorsolateral prefrontal cortex (DLPFC) and the ventromedial prefrontal cortex (VMPFC)—each believed to contribute differentially to processes involved in memory, emotion, and decision-making (Denburg et al., 2005, MacPherson et al., 2002).

Inasmuch as the DLPFC and VMPFC are anatomically distinct and have unique patterns of afferent and efferent connectivity, it is helpful to recognize the behavioral distinctions associated with these separate, albeit interrelated regions. First, the DLPFC, regionally identified as Brodmann areas 9 and 46 (see Fig. 1), is part of an association circuit integrating sensory and motor processes (MacPherson et al., 2002, Mesulam, 2002, Phillips and Della Sala, 1999) for purposes of managing ongoing experiences and directing behavior (Zillmer & Spiers, 2001). Indeed, current neuropsychological models distinguish the DLPFC as an important area for cognitive abilities, most notably attention, working memory, and executive functions (Head et al., 2002, Lamar and Resnick, 2004, MacPherson et al., 2002). Neuropsychological tasks which emphasize planning, self-monitoring, and decision-making—such as the Wisconsin Card Sorting Task (WCST)—provide sensitive indicators of prefrontal executive functions, with particular evidence of a significant dorsolateral relationship (Head et al., 2002). Lezak, Howieson, and Loring (2004) underscore the generic function of the DLPFC in managing information not only for working memory, but also for integrating motor output with attention, memory, and sensory input.

Integration of multimodal information is a hallmark of prefrontal association cortex (Ongur & Price, 2000), which is further reinforced by the connectivity and functions attributed to the anatomically separate, albeit related prefrontal region, the VMPFC. Localized along the ventral and medial surfaces of the frontal lobes (refer to Fig. 1), the VMPFC includes Brodmann areas 10–14 and 47, in addition to areas 25 and 32 (subcallosal and anterior cingulate cortices, respectively; for excellent review, see Ongur & Price, 2000; also Bechara, Damasio, & Damasio, 2000). The preponderance of literature emphasizes the role of the VMPFC in multimodal sensory integration, with particular relevance for affective judgments and the direction of emotional behavior (Ongur & Price, 2000). This perspective is supported by the close anatomical association with limbic structures and the amygdala in particular (see Holland and Gallagher, 2004, Quirk and Beer, 2006).

Given the integrative nature of the PFC, it follows that this region would participate in the synthesis and organization of complex cognitions, particularly where emotions and contextual cognitive information intersect. As one example of this multimodal intersection, episodic memory represents a cognitive process dependent on a distributed network of polymodal association cortices including medial temporal lobe structures (e.g., hippocampus) and the PFC itself. Considering the evidence of anatomical declines in PFC with age (Phillips & Della Sala, 1999), coupled with metabolic indicators of prefrontal alterations during episodic memory tasks (e.g., Cabeza, 2002; for excellent review see Rajah & D’Esposito, 2005), it is reasonable to expect decrements in episodic memory among older adults in contrast to preserved performance in skill domains believed to be less dependent on a neural network involving the PFC. In support of this pattern, recent research employing structural equation modeling (Allen et al., 2005) suggests that age-related declines in episodic memory are mediated by emotional processing, both of which are posited to reflect altered PFC functioning with advancing age. In these same participants, no appreciable declines were noted for semantic memory, a dimension of long-term memory viewed as less dependent on the PFC regions involved in episodic processes (Allen et al., 2005, Martin and Chao, 2001). Most importantly, the dissociation of relatively spared and somewhat less preserved skills among older adults provides an opportunity to investigate the differential effects of aging on neuropsychological functions attributed to distinct brain regions.

It is noteworthy that this differential effect of aging on distinct types of long-term memory (e.g., episodic vs. semantic) has been known for some time, yet remains somewhat of a paradox in the cognitive psychology literature (Allen et al., 2004, Allen et al., 2002, Allen et al., 2002b, Light, 1991). Nevertheless, efforts to better understand this difference have been helped greatly by the findings and perspectives offered by neuroscience. Of particular interest here is the notion that the limbic system, with specific reference to the medial temporal lobes and the VMPFC, forms a neural circuit involved in processes such as episodic memory and emotional decision making, whereas somewhat different circuits with less specific emphasis on VMPFC processing are thought to underlie semantic memory and other executive-function skills (see Aggleton and Brown, 2006, Head et al., 2002, Martin and Chao, 2001).

While the Allen et al. (2005) study presented some evidence of age differences in VMPFC function (i.e., inefficient affective contextual markers for older adults during episodic memory processing), the results from other studies have not always observed age differences in tasks more commonly discussed as measures of VMPFC function. In particular, MacPherson et al. (2002) failed to find age differences in the Faux Pas Task, emotional facial discrimination, and emotional decision making—all thought to be indices of VMPFC function. In contrast, Denburg et al. (2005) did find an age decrement in emotional decision making on the Iowa Gambling Task. MacPherson et al. may have failed to find age differences in VMPFC function because of a lack of sensitivity in their experimental design. First, the Faux Pas Task has a very heavy loading on semantic memory that is largely intact in older adults (Light, 1991), and may even improve with increased adult age (Allen et al., 2002a, Lien et al., 2006). Thus, we did not use the Faux Pas Task in the present study. Second, MacPherson et al. used an accuracy, but not a latency, measure of emotional facial discrimination. In the present study, we included both reaction time (RT) and accuracy measures on two versions of this task (employing both abstract and real human faces). Third, the younger participants in the MacPherson et al. study were still showing performance improvement from Block 3 to Block 5 on the Iowa Gambling Task (Bechara et al., 1994), but older adults’ performance appeared to have peaked by Block 3 (see MacPherson et al., Fig. 3). We used a computerized version of the original card set (proposed by Bechara et al., 1994, Bechara et al., 2000a) in an attempt to gain measurement precision over using manual cards.

Denburg et al. (2005) found evidence of age differences in emotional decision making, and we expected similar results. However, we went beyond this to hypothesize that the age difference in VMPFC function encompasses more than just emotional decision making. Indeed, we suspected that older adults would show a more widespread decrement in emotional/cognitive integration that includes emotional perceptual discrimination as well as emotional decision making. To test this hypothesis, we included both emotional decision making and emotional facial discrimination tasks similar to those employed by MacPherson et al. (2002). Furthermore, to determine whether these two task domains reflect something unitary about the nature of VMPFC functioning, or whether they reflect separate VMPFC processes (i.e., emotional decision making and emotional perceptual discrimination), we included multiple performance indicators from both tasks so that this issue could be investigated using exploratory factor analysis.

We additionally hypothesized that older adults would exhibit a decrement for neuropsychological indices typically associated with DLPFC function (i.e., Wisconsin Card Sorting Task [WCST], Trails B, the Stroop Task, and Reverse Digit Span). Thus, we do not disagree with the previous work of MacPherson et al. (2002) showing an age deficit on DLPFC measures. However, we suspect a more general pattern of neuropsychological decline on tasks with a common prefrontal contribution (as hypothesized by West, 1996)—although we suspect that these age differences might not be completely consistent across domains (e.g., Allen et al., 2005, observed no central-process age differences for semantic memory tasks but did observe age differences for central-process, episodic memory tasks). That is, we predict age decrements for both tasks associated with the DLPFC and the VMPFC, but recognize that the extent of age-related declines might not be uniform.

Identifying adult age-related differences in prefrontal-related processing is the principal interest in this study. The nature and specific alterations in neuropsychological performance will be evaluated using three separate statistical approaches. First, the question of age differences on individual tasks associated with DLPFC (WCST, Stroop interference, Trails B, and Reverse Digit Span) and VMPFC function (Iowa Gambling Task and Emotion Identification Tasks) will be considered using Analysis of Variance (ANOVA) and Multivariate Analysis of Variance (MANOVA) to test for age group differences. Second, we will investigate the dimensionality of the DLPFC and VMPFC task performance scores, using exploratory factor analysis (EFA). The hope is that the emergent factor structure will support a DLPFC/VMPFC distinction, while also adding to our understanding of potentially separate, albeit interrelated, task-specific sub-modules. Finally, structural equation models (SEM) of age differences in DLPFC/VMPFC-related task constructs will be considered to determine if age differences in performance can be completely explained using a generalized common factor (e.g., Salthouse, 1998, Salthouse and Czaja, 2000), or whether process-specific age differences persist independent of a common factor (Allen et al., 2001, Sliwinski and Hall, 1998, Zelinski and Lewis, 2003).