Prefrontal mediation of age differences in cognitive reappraisal

Transcription

1 Neurobiology of Aging 33 (2012) Prefrontal mediation of age differences in cognitive reappraisal Philipp C. Opitz, Lindsay C. Rauch, Douglas P. Terry, Heather L. Urry* Tufts University, Medford, MA, USA Received 31 December 2009; received in revised form 29 April 2010; accepted 6 June 2010 Abstract Despite cognitive and physical declines, it has been suggested that older adults remain able to regulate their emotions effectively. However, whether this is true for all emotion regulation processes has not been established. We hypothesized that cognitive reappraisal, a form of emotion regulation requiring intact cognitive control ability, may be compromised in older age, and that this age difference would be mediated by reduced activation in prefrontal cortex (PFC). Sixteen younger and 15 older adults used gaze-directed reappraisal to increase and decrease emotion in response to unpleasant pictures. This was compared with simply viewing the pictures. Relative to younger adults, older adults were less successful using reappraisal to decrease unpleasant emotion but more successful using reappraisal to increase unpleasant emotion. They also exhibited reduced activation in dorsomedial and left ventrolateral prefrontal cortex. Importantly, activation in these regions differentially mediated the effect of age on emotion. This pattern confirms the importance of cognitive control in reappraising unpleasant situations and suggests that older age may (but does not always) confer effective emotion regulation Elsevier Inc. All rights reserved. Keywords: Emotion regulation; Cognitive reappraisal; Attentional deployment; Ventrolateral prefrontal cortex; Dorsomedial prefrontal cortex 1. Introduction Despite experiencing cognitive and physical declines, older adults report higher levels of well-being compared with younger adults (Mroczek, 2001). This paradoxical enhancement of well-being has been suggested to spring from an increase in motivation to regulate emotions in older age (Mather and Carstensen, 2003, 2005). Studies assessing age differences in the neural underpinnings of emotion regulation and its experiential consequences may therefore provide valuable clues as to the neural mechanisms underlying enhancement of well-being in older age. One of the best studied forms of emotion regulation is cognitive reappraisal (CR). CR refers to reworking the meaning of a situation in ways that change its emotional impact. As recently summarized by Kalisch (2009) in a meta-analysis of 13 neuroimaging studies, CR for which the goal is to decrease unpleasant emotion is accompanied by * Corresponding author at: Department of Psychology, Tufts University, 490 Boston Avenue, Medford, MA 02155, USA. Tel.: ; fax: address: heather.urry@tufts.edu (H.L. Urry). greater activation of dorsolateral, ventrolateral, and dorsomedial regions of prefrontal cortex (PFC) relative to a control condition. CR can also be used to increase unpleasant emotional states, which is accompanied by activation of similar PFC regions (e.g., Ochsner et al., 2004; Urry et al., 2009; van Reekum et al., 2007). This pattern of greater activation in PFC when trying to decrease or increase unpleasant emotion compared with a control condition is consistent with the idea that, regardless of the regulatory goal, both forms of CR reflect the needs to generate and maintain a strategy for reframing the meaning of the emotional situation in question, monitor interference between emotion generative and regulative processes, and monitor changes in emotional state (Ochsner et al., 2002). These forms of information processing represent aspects of cognitive control. Cognitive control refers to the selection, planning, coordination, and execution of goal-driven thoughts and actions (Miller and Cohen, 2001). Cognitive control is evident in tasks that demand top-down intervention, for example to inhibit dominant response patterns in favor of nondominant ones (Vanderhasselt and de Raedt, 2009). Of interest, the /$ see front matter 2012 Elsevier Inc. All rights reserved. doi: /j.neurobiolaging

2 646 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) ability to exert cognitive control declines in the later stages of adulthood (Craik and Salthouse, 2007). Thus, because CR draws on cognitive control ability, older adults may be less able than younger adults to regulate their emotions using CR. On the contrary, two studies have offered a direct, performance-based comparison between younger and older adults, and have suggested that older age is actually associated with better use of positive reappraisal, i.e., thinking about the positive aspects of a situation (Shiota and Levenson, 2009), or positive refocusing, i.e., picturing an event or person that makes the person feel good (Phillips et al., 2008), to regulate subjective emotional experience, autonomic physiology, and/or expressive behavior. However, in both cases, age differences in the control condition ( just watch or spontaneous expression, respectively) make it difficult to interpret this pattern. Moreover, it is possible that the observed changes in emotion reflected a different family of emotion-regulatory processes, namely attentional deployment. As described in the process model of emotion regulation (Gross and Thompson, 2007), attentional deployment (AD) refers to directing attention toward or away from emotional information in ways that change its emotional impact. Positive refocusing of attention away from the upsetting material at hand to positive autobiographical experiences, the strategy studied by Phillips et al. (2008), is thus an example of AD. In fact, older adults may naturally engage in AD to positive over negative information more than younger adults. For example, as demonstrated in recent eye-tracking studies, older adults tend to deploy their attention toward happy and away from angry faces (Isaacowitz et al., 2006). Furthermore, when put in a negative mood, older adults tend to direct their attention to mood-incongruent positive information (Isaacowitz et al., 2008), suggesting that AD toward positive information may serve an emotion regulatory function for older adults. The fact that older adults deploy attention toward positive and away from negative information may be problematic in the context of studies assessing CR. For example, van Reekum and colleagues (2007) examined the neural correlates of reappraisals that increase and decrease unpleasant emotion, evoked using pictures, in a sample of older adults. They spent less time looking at emotional information in the pictures when using reappraisals to decrease negative emotion. Moreover, the variance in brain activation putatively explained by reappraising was substantially reduced when taking gaze behavior into account. This suggests that changes in neural activation may actually have been caused by AD instead of (or in addition to) CR. For the researcher interested in making inferences about CR, this is problematic and suggests a need to control the use of AD. The present study was designed to examine age differences in the neural correlates of CR. We used a novel gaze-directed reappraisal paradigm validated by Urry (2010) to hold AD constant across reappraisal conditions. Our expectation was that, relative to younger adults, older adults would have difficulty using CR to increase and decrease their emotional states when we controlled their use of AD. We expected that this age difference would be particularly pronounced for reappraisals whose goal is to decrease negative emotion because, in that case, emotion generative processes are directly opposing emotion regulative processes. As a result, successfully using CR to decrease negative emotion should require greater cognitive control compared with using CR to increase negative emotion (see, e.g., Ochsner et al., 2004). We further expected that this difficulty using CR would be accompanied by differences in activation of prefrontal circuits involved in cognitive control. We recruited younger (ages years) and older (ages years) adults to complete the gaze-directed reappraisal task during which we used functional magnetic resonance imaging to measure neural activation, with a focus on lateral and medial prefrontal brain regions. In addition, at the end of each trial, we collected subjective ratings of emotion to measure CR success. We hypothesized that younger adults would exhibit greater CR success than older adults, as reflected in self report ratings, particularly when the goal was to decrease unpleasant emotion. Furthermore, we hypothesized that younger adults would exhibit greater reappraisal-related activation (i.e., both increase and decrease greater than view) in lateral and medial prefrontal regions than older adults. 2. Methods 2.1. Participants Sixteen younger (8 female; years; mean, 19.25; SD, 1.43) and 15 older (9 female; years; mean, 59.87; SD, 3.14) adults were recruited via internet (e.g., and local newspaper advertisements. Participants endorsed the following nonexclusive racial categories: American Indian or Alaska native: 1 (3%); Asian: 3 (9%); white: 27 (87%); and decline to state: 1 (3%). Two participants (6%) were of Hispanic origin. All but 5 (3 younger) were right-handed, and none reported any history of psychiatric or neurological disorders, or current use of any psychoactive medications. Removing left-handed participants from the analyses did not change the essential pattern of results thus all analyses include all participants. The Massachusetts Institute of Technology s Committee on the Use of Humans as Experimental Subjects approved all procedures. Participants provided written informed consent before participating and received $20 (US) per hour in compensation for the 3-hour session. Participants were screened for contraindications for magnetic resonance imaging (MRI) scanning and, when necessary, were issued MRI-safe glasses with prescription-specific lenses. Demographics are shown in Supplementary Table 1.

3 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) Materials and procedures Stimuli Participants viewed a set of digital color pictures ( pixels) selected from the well-validated International Affective Picture System (IAPS; Lang et al., 2008), the same pictures described by Urry (2010). Twenty-seven unpleasant pictures were selected on the basis of normative data across men and women, to be highly arousing (mean, 5.93; SD, 0.28 on a1to9arousal scale where 9 signifies completely aroused) and unpleasant (mean, 2.10; SD, 0.31 on a 1 to 9 valence scale where 9 signifies completely happy). Twelve neutral pictures were selected to be low in arousal (mean, 3.06; SD, 1.09) and neither pleasant nor unpleasant (mean, 4.99; SD, 0.05) Stimulus presentation Stimuli were presented using E-Prime software (version , Psychology Software Tools, Inc., Pittsburgh, PA, USA), and were delivered via a rear projection system in which mirror images of the stimuli were projected with a liquid crystal display (LCD) projector onto a 17.5 (horizontal measurement) screen placed at the back of the scanner bore. Participants viewed these images using a set of mirrors mounted on the head coil. Images subtended a horizontal visual angle of approximately 21 degrees Cognitive reappraisal Participants were trained to follow one of three prerecorded auditory CR instructions during each of 78 picture trials. The instruction to increase served as a cue to actively try to feel more emotion. For example, in response to a picture of a girl curled up on the floor, participants were told they could imagine that her mother had just died. Conversely, the instruction to decrease signified the cue to actively try to feel less emotion. For example, participants were told they could imagine that the girl in the picture had fallen asleep. Alternatively, on view trials, participants were instructed to view the picture without changing the way they feel. Immediately before completing the reappraisal task, participants were trained outside of the scanner to reappraise using a standardized set of instructions in which example reappraisals were provided for the increase and decrease instructions in response to sample pictures. Participants then practiced generating their own reappraisals of new sample pictures. Afterward, the experimenter asked about the strategies participants had used, and coached as needed. Participants uniformly demonstrated the ability to generate reappraisals that conformed to our instructions. The reappraisal task was presented in 3 blocks of 26 trials of unpleasant and neutral pictures. The unpleasant pictures were paired with increase, decrease, and view instructions (1 instruction per trial) whereas neutral pictures were only paired with the view instruction (because there was no emotion to be regulated). Each picture was presented in 2 trials, once with gaze directed to an arousing area, and once to a nonarousing area as described below. At the end of each trial, participants made a valence rating on the following scale: 1 ( positive ), 2 ( neutral ), or 3 ( negative ). In addition, subjective ratings of emotional intensity were provided on a scale from 1 ( mildly intense ) to 4 ( very intense ). The trial structure is presented in Fig Gaze direction During the reappraisal task, gaze was directed to a specific area of interest (AOI) in each picture half way through the trial. This was accomplished by fading out all but one square area ( pixels) (see Fig. 1). For unpleasant pictures, the saturated square illuminated an area of the picture judged to be emotionally arousing (arousing focus) or an area judged to be neutral (nonarousing focus). For neutral pictures, the saturated square illuminated an area of the picture judged to include a central feature (arousing focus) or an area judged to be peripheral (nonarousing focus). Participants were instructed to keep their gaze focused within the AOI as they followed the reappraisal instruction. To evaluate whether participants complied, they responded by button press to a green dot. It was presented for 33 ms on every trial, 7 seconds after picture onset in a randomly determined location within the AOI. Urry (2010) validated the selection of arousing and nonarousing AOI as well as compliance with the gaze direction manipulation using looking time data acquired via eye tracking in a separate sample Design of the gaze-directed cognitive reappraisal task There were a total of 8 design cells in the 78-trial gazedirected reappraisal task. For 6 cells (9 unpleasant pictures in each), participants viewed unpleasant pictures. These cells were described by the cross between reappraisal goal (increase, view, decrease) and gaze direction (arousing focus, nonarousing focus). For an additional 2 cells (12 neutral pictures in each), participants viewed neutral pictures paired with the view instruction. More neutral than unpleasant pictures were presented to somewhat offset the preponderance of unpleasant images in each block of trials. Cells were presented in random order for each participant. Pictures were randomly assigned to each trial. Gaze was directed to either an arousing or nonarousing focus on every trial Functional magnetic resonance imaging (FMRI) data acquisition FMRI data were collected using a Siemens MAGNETOM Trio 3T MRI scanner (Siemens Healthcare, Erlangen, Germany) at the Athinoula A. Martinos Imaging Center at the McGovern Institute for Brain Research at the Massachusetts Institute of Technology. This scanner has 32 radiofrequency (RF) receiver channels equipped with TQ Engine gradients (amplitude, 45 mt/m [z-axis], 40 mt/m [x, y-axes]; slew rate, 200 T/m/s) and a 12-channel head coil. Participants lay

4 648 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) Fig. 1. Trial structure for the gaze-directed cognitive reappraisal task. Trials began with a white fixation cross presented in the center of a black screen for 1 second followed by the presentation of a picture for a total of 10 seconds. For the first 5 seconds, the entire picture was visible. For the last 5 seconds, most of the picture was faded out except for 1 square area to which gaze was directed. The cognitive reappraisal (CR) instruction was delivered 4 seconds after picture onset. Photo of mourning family by Mikhail Evstafiev used with permission; retrieved from on 18 August supine on the scanner bed. Pillows were used to prevent head movements. Blood oxygenation level-dependent (BOLD) signals were collected in 3 scan runs. Automatic slice prescription, based on alignment of localizer scans to a multisubject atlas, was used to achieve a consistent head position across subjects and scan runs along the anterior commissure-posterior commissure line. In each run, 320 whole-brain T2*-weighted echoplanar images (EPIs) were collected (5 mm slice thickness; 33 interleaved axial slices; repetition time [TR], 2 seconds; echo time [TE], 30 ms; flip angle, 90 ; matrix size, 64 64; field of view [FOV], 200; in-plane resolution, mm). Prior to each run, 5 images were recorded and discarded to allow longitudinal magnetization to reach equilibrium. A 2-volume version of this functional scan, collected after the localizer and scout sequences, was inspected for data quality and coverage before proceeding. For registration purposes, 2 scans were collected. The first, collected before the functional EPIs, was a short T1- weighted EPI scan collected with the same slice prescription as the functional EPIs (TR, 10 seconds; TE, 34 ms). The second, collected after the functional EPIs at the end of the session, was a high-resolution 3-dimensional T1-weighted magnetization-prepared rapid gradient echo (MPRAGE) anatomical scan (TR, 2.53 seconds; TE, 3.39 ms; FOV, 256 mm; matrix, ; sagittal plane; slice thickness, 1.33 mm; 128 slices). For susceptibility artifact correction, field maps were acquired with a gradient-echo sequence using the same slice prescription as the functional EPIs (TR, 500 ms; TE 1, 2.84 ms; TE 2, 5.3 ms; flip angle, 55 ). These were collected after the short, T1-weighted EPI scan FMRI preprocessing and registration FMRI preprocessing and registration were carried out using FSL (FMRIB s Software Library, uk/fsl; version 4.1.0). The functional time series were first realigned to compensate for small head motions. They were then subjected to field map-based B0 unwarping. Residual motion artifacts and physiological noise were then removed using independent component analysis. The data were then spatially smoothed using a full-width half-maximum Gaussian kernel of 5 mm. They were also grand-mean intensity normalized across the entire time series by a single multiplicative factor. Finally, the data were high pass-filtered in the time domain using Gaussian-weighted least squares straight line fitting (sigma 50.0 seconds). Prior to across-subjects analyses, the functional data were registered to Montreal Neurological Institute (MNI) standard space in 3 steps, with voxels resampled to 2 mm 3. Each run of functional data were first linearly registered to the participant s own T1-weighted EPI scan (6 degrees of freedom) and then to the participant s own high-resolution MPRAGE anatomical scan (12 degrees of freedom). In the third step, the functional data were registered to MNI 152

5 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) standard space using a nonlinear transformation matrix (10 mm nonlinear warp field resolution), which handles local changes around the ventricles and sulci caused by atrophy (common in older adults) FMRI analyses FMRI analyses were carried out in three levels using FSLs FEAT (FMRI Expert Analysis Tool) version Ten original explanatory variables (EVs) were modeled at the first level of analysis. Six of these were described by the cross between the 3 reappraisal conditions (increase, view, decrease) and 2 gaze directions (arousing, nonarousing) for unpleasant pictures. Two additional original EVs were constructed for neutral pictures, 1 for each of the 2 gaze directions (both paired with the view instruction). The final 2 original EVs modeled button press responses indicating detection of or failure to detect the 33-ms green dot as covariates of no interest. For each run, we convolved the 10 original EVs with a constrained set of 3 optimal hemodynamic response functions that roughly characterized BOLD signal response amplitude, delay, and width. This yielded 30 real EVs that were submitted to time series statistical analysis under the General Linear Model (GLM) framework with local autocorrelation correction. The first level estimates of response amplitude (and their variances) were passed to the second level analysis, in which the mean effect for each experimental condition was calculated across the 3 functional runs for each voxel using a fixed effects model with the random effects variance set to 0. At this second level, we tested one contrast of interest to identify brain regions in which the 2 active reappraisal conditions, increase ( 1) and decrease ( 1), exhibited greater activation when compared with the view ( 2) condition across gaze directions. Larger numbers indicate greater reappraisal-related activation. At the third level, we conducted across-subjects analyses of reappraisal-related activation from the second-level analyses (and their variances) using FMRIBs Local Analysis of Mixed Effects (FLAME stage 1). Variances were estimated separately for the younger and older groups. In order to obtain the most reliable estimate of brain regions that exhibit reappraisal-related activation, clusters of activation were identified across age groups by thresholding the Z statistic images with a voxelwise Z 2.58 (i.e., p ) and a corrected cluster significance threshold of p 0.05 using Gaussian random field theory. Estimates of activation in the regions that emerged were extracted separately for each participant for each reappraisal goal and gaze direction and then tested for age differences in SPSS (version 16, SPSS Inc., Chicago, IL) Questionnaires Following the scan, participants provided demographic information and reported their impression about the purpose of the experiment and the strategies they used to reappraise. They also completed questionnaires assessing mood and anxiety symptoms (Mood and Anxiety Symptoms Questionnaire [MASQ]) and attentional control (Attentional Control Scale [ACS]). Details of these measures and a table summarizing age differences therein are presented in our Supplementary Material. 3. Results 3.1. Overview The results are structured in three sections. First, we report manipulation checks testing whether we effectively elicited unpleasant emotion and held attentional deployment constant across reappraisal conditions. Second, we examine the main effects of cognitive reappraisal on ratings of emotional experience and neural activation across the whole sample. Finally, we test our a priori hypotheses about age differences in the effects of reappraisal. Results were considered statistically significant at 0.05 (2-tailed). Multivariate F statistics are reported for the GLMs with partial eta squared ( P 2 ) as an index of effect size. Fisher s least significant difference (LSD) was used for follow-up analyses Manipulation checks Have we effectively elicited unpleasant emotion? Mean ratings of emotional intensity for view unpleasant and view neutral trials (averaged across gaze direction conditions) were submitted to a multivariate GLM to assess the main and interactive effects of picture valence (unpleasant, neutral; within-subjects) and age group (younger, older; between-subjects). The GLM revealed a main effect of picture valence, F(1,29) , p 0.001, P , which confirmed that unpleasant pictures (mean, 2.49, SD, 0.62) were rated as more intense than neutral pictures (mean, 1.22; SD, There was no significant effect of age group, F(1,29) 0.14, p 0.71, P , and no significant interaction between valence and age group, F(1,29) 0.44, p 0.51, P Mean ratings of valence revealed the same pattern (see Supplementary Material). Overall these results confirm that the intended unpleasant emotional state was elicited equally in both age groups Have we effectively held attentional deployment constant across reappraisal conditions? The proportion of trials on which participants successfully detected the green dot displayed within the arousing or nonarousing AOI was calculated for each participant for each of the 6 cells in the design involving unpleasant picture trials. Participants detected the dot on the majority of unpleasant trials (mean, 0.77; SD, 0.18). Mean dot detection proportions for unpleasant trials were submitted to a multivariate GLM to assess the main and interactive effects of reappraisal goal (increase, view, decrease; within-subjects), gaze direction (arousing, nonarousing; within-subjects), and

6 650 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) Fig. 2. Panel (a) depicts the dorsal medial prefrontal cortex (PFC) cluster in a series of sagittal slices. Panel (b) depicts the left ventrolateral PFC cluster in a series of axial slices. The image at the far right indicates with blue lines the position of the 6 previous slices in each panel. Voxels that are yellow met a 1-tailed threshold of Z 2.58 whereas those that are red met a less stringent threshold of Z 2.3 (to show the full extent of activation). age group. Older adults (mean, 0.72; SD, 0.12), detected the dot less than younger adults (mean, 0.81; SD, 0.22), although there was no main effect of age group, F(1,29) 2.17, p 0.15, P There was also no main effect of reappraisal condition, F(2,28) 0.53, p 0.59, P , but there was a trend for the interaction between reappraisal goal and age group, F(2,28) 2.67, p 0.09, P Younger adults detected the dot with equal frequency across reappraisal conditions. Older adults detected the dot more in the increase than in the view condition, p 0.04 (see Supplementary Fig. 1). This indicates that AD was held constant across reappraisal conditions in the younger group but there was some variation in the older group Effects of cognitive reappraisal in the sample as a whole Does gaze-directed reappraisal produce changes in subjective emotion experience? Mean ratings of emotional intensity for the unpleasant trials were submitted to a multivariate GLM to assess the main effects of reappraisal goal, gaze direction, and age group. The main effect of reappraisal goal was significant, F(2,28) 26.38, p 0.001, P Ratings of intensity were higher during the increase (mean, 2.87; SD, 0.58), compared with the view (mean, 2.49; SD, 0.63, p 0.001) condition. The difference between the view and decrease (mean, 2.54; SD, 0.65) conditions was not significant, p Gaze direction did not interact with reappraisal goal, F(2,28) 0.14, p 0.87, P Mean ratings of valence for the unpleasant trials revealed the same pattern (See Supplementary Material). These results indicate that the sample as a whole was able to use CR to increase but not decrease unpleasant emotion when AD was held constant. Because valence and intensity ratings revealed the same pattern, we present intensity ratings only in the remainder of the manuscript due to greater possible variation with the 4-point scale Does gaze-directed reappraisal produce reappraisal-related activation in PFC? Across age groups, higher activation when engaging in increase and decrease reappraisals compared with the view condition was observed in two brain regions: dorsal medial PFC (DMPFC), primarily in anterior cingulate cortex (Fig. 2a; maximum Z at MNI x 6, y 26, z 24), and left ventrolateral PFC (VLPFC), primarily in the left inferior frontal gyrus extending into anterior insula (see Fig. 2b; maximum Z at MNI x 50, y 10, z 4). Mean estimates of % BOLD signal change were extracted and submitted to a multivariate GLM to assess the main effects of reappraisal goal, gaze direction, and age group. Gaze direction did not interact with reappraisal goal in either region, DMPFC F(2,28) 0.39, p 0.68, P and left VLPFC F(2,28) 0.11, p 0.90, P These two clusters served as regions of interest (ROI) for hypothesis testing below Hypothesis testing Do younger adults exhibit greater CR success than older adults? As expected, the multivariate GLM assessing the effects of reappraisal goal, gaze direction, and age group on ratings of emotional intensity (described previously) revealed a significant interaction between age group and reappraisal goal, F(2,28) 4.57, p 0.04, P As shown in Table 1, younger adults rated increase trials as more intense than view trials across gaze directions, p Although younger adults rated decrease trials as less intense than view trials, this difference was not reliable, p Older adults also rated increase trials as more intense than view trials, p However, older adults also rated decrease trials as more intense than view trials, an unexpected difference that was nearly significant, p We next calculated (increase view) and (view decrease) contrast scores for each participant across gaze directions for which a larger contrast score signifies greater

7 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) Table 1 Means (SD) for ratings of emotional experience and neural activation as a function of reappraisal goal, gaze direction, and age group Across gaze directions Arousing Nonarousing Increase View Decrease Increase View Decrease Increase View Decrease Ratings of emotional intensity Younger adults 2.65 a (0.65) 2.43 a (0.67) 2.37 (0.62) 2.71 b (0.72) 2.45 b (0.72) 2.42 (0.59) 2.59 c (0.60) 2.41 c (0.66) 2.32 (0.69) Older adults 3.09 a (0.49) 2.56 a,d (0.57) 2.71 d * (0.58) 3.09 b (0.56) 2.62 b (0.58) 2.76 (0.66) 3.09 c (0.49) 2.50 c (0.62) 2.67 (0.54) Ratings of valence Younger adults 2.83 a * (0.21) 2.74 a (0.28) 2.73 (0.25) 2.84 (0.29) 2.75 (0.30) 2.76 (0.28) 2.83 (0.24) 2.74 (0.29) 2.70 (0.31) Older adults 2.92 a (0.21) 2.72 a (0.28) 2.78 (0.21) 2.95 b (0.07) 2.79 b,e (0.25) 2.77 e ( 0.23) 2.88 c (0.22) 2.64 c (0.33) 2.78 (0.26) Dorsal medial PFC activation Younger adults 0.24 a (0.15) 0.12 a,d (0.12) 0.19 d (0.13) 0.25 b (0.16) 0.12 b (0.14) 0.19 (0.16) 0.22 c (0.16) 0.13 c (0.13) 0.18 (0.14) Older adults 0.04 (0.18) (0.14) 0.03 (0.14) 0.04 (0.21) (0.19) 0.06 (0.15) 0.03 (0.16) 0.01 (0.12) (0.15) Left ventrolateral PFC activation Younger adults 0.26 a (0.14) 0.10 a,d (0.12) 0.21 d (0.11) 0.28 b (0.15) 0.11 b,e (0.13) 0.21 e (0.18) 0.25 c (0.17) 0.09 c,f (0.14) 0.21 f (0.13) Older adults 0.12 a (0.16) 0.08 a (0.12) 0.12 (0.15) 0.13 (0.19) 0.08 (0.15) 0.14 (0.15) 0.11 (0.15) 0.08 (0.11) 0.10 (0.17) Neural activation is expressed in units of percent signal change relative to baseline. Key: Arousing, gaze directed to an emotionally relevant area of the picture; Nonarousing, gaze directed away from the emotionally relevant areas of the picture; PFC, prefrontal cortex. a f For the two planned comparisons of interest (increase vs. view, decrease vs. view), means sharing the same letter superscript within a row differ at p 0.05 according to Fisher s least significant difference (LSD). * p 0.07 for this comparison. reappraisal success. Younger adults had marginally greater success with decrease reappraisals compared with older adults, t(29) 1.91, p However, older adults had significantly greater success with increase reappraisals compared with younger adults, t(29) 2.93, p Effect sizes are presented in Supplementary Table Do younger adults exhibit greater reappraisalrelated activation in PFC than older adults? Estimates of mean % BOLD signal change were extracted from the left VLPFC and DMPFC regions identified above. As predicted, younger adults exhibited higher mean reappraisal-related activation than older adults in left VLPFC, t(29) 2.86, p 0.008, with a similar but marginally significant effect evident in DM PFC, t(29) 1.71, p (see Table 1). These results suggest that younger adults recruit these regions of PFC to a greater degree during CR than older adults. Notably, although effect sizes were generally lower (see Supplementary Material), entering dot detection proportions (increase view, decrease view), ACS scores, and MASQ scores into the model as covariates simultaneously did not eliminate these effects Are age differences in CR success mediated by prefrontal activation? The results above presented a bit of a paradox. Although both groups reported greater emotional intensity for increase compared with view conditions, older adults exhibited a larger effect of increase reappraisals than younger adults. At the same time, older adults activated both PFC regions less than younger adults. We therefore computed two simple mediation analyses using a product-of-coefficients strategy (Preacher and Hayes, 2008) to determine whether these two regions of PFC might differentially account for effects of age on reappraisal success. Our mediation analyses provided estimates of the direct effect of age (dummy-coded with older 1) on contrast scores (increase and decrease view) for left VLPFC and DMPFC activation (a) and the direct effect of activation in these two brain regions on contrast scores (increase view or view decrease) for ratings of emotional intensity (b). They also estimated the indirect effect of age on ratings of emotional intensity by way of the two PFC regions (ab) and the direct effect of age on ratings of emotional intensity (c=) when accounting for ab. When increase view ratings of emotional intensity served as the criterion variable, younger age was associated with higher left VLPFC activation, a 0.10, standard error (SE) 0.04, and marginally higher DMPFC activation, a 0.06, SE 0.03, results that recapitulate the age differences reported above. Importantly, higher left VLPFC activation was associated with lower increase view ratings of intensity, b 2.18, SE 1.02, t(29) 2.15, p 0.04, but higher DMPFC activation was associated with higher increase view ratings of intensity, b 2.51, SE 1.05, t(29) 2.40, p 0.02.

8 652 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) Notably, the indirect effect of age on ratings of intensity by way of left VLPFC activation was positive, ab 0.22, SE 0.12, meaning that older age is associated with a bigger increase view effect by way of left VLPFC, 95% confidence interval to The indirect of age on ratings of intensity by way of DMPFC, by contrast, was negative, ab 0.14, SE 0.10, meaning that older age is associated with a smaller increase effect by way of DMPFC, although the 95% confidence interval ( to 0.006) overlapped with 0. The direct pathway linking older age to higher ratings of intensity for increase view was significant, c= 0.24, SE 0.11, t (29) 2.07, p 0.048, thus there was only partial mediation. The size of these effects was similar when entering a dot detection contrast score (increase view) as a covariate. When view decrease ratings of emotional intensity served as the criterion variable, the b and ab paths were not significant. 4. Discussion Compared with younger adults, older adults had greater difficulty decreasing the intensity of unpleasant emotion using cognitive reappraisal. However, older adults were actually better at increasing unpleasant emotion. This asymmetry was accompanied by significant age-related reductions in the activation of two regions of prefrontal cortex that had previously been implicated in cognitive control, namely left ventrolateral (VLPFC) and dorsal medial PFC (DMPFC). We had initially believed that greater activation in these regions would uniformly be associated with greater reappraisal success, regardless of reappraisal goal. Instead, we found that DMPFC activation predicted greater success of reappraisals whose goal was to increase unpleasant emotion. By contrast, left VLPFC activation predicted lower success of reappraisals whose goal was to increase unpleasant emotion. In the paragraphs that ensue, we discuss these results, consider their clinical and theoretical implications, and address the limitations of this work and directions for future research Age differences in reappraisal success Cognitive reappraisal successfully transforms the subjective experience of unpleasant emotion in younger adults (Ochsner et al., 2002). However, it was largely unknown before this study whether older adults would be similarly capable because reappraisal engages regions of prefrontal cortex that enable cognitive control (e.g., Kalisch, 2009), a set of abilities that declines with age. Importantly, unlike recent studies assessing age differences in emotion regulation (Phillips et al., 2008; Shiota and Levenson, 2009), we used a reappraisal task that was designed to hold an alternate emotion regulatory process, namely attentional deployment, constant across cognitive reappraisal conditions. We did so because older adults are prone to deploying their attention to emotional information differently than younger adults (e.g., Allard and Isaacowitz, 2008), and this seems to have an emotion-regulatory effect for them (Isaacowitz et al., 2008). Moreover, a previous study of older adults showed that controlling for attentional deployment during cognitive reappraisal reduced or eliminated variance in neural activation that was putatively explained by cognitive reappraisal (van Reekum et al., 2007). In our gaze-directed cognitive reappraisal paradigm, older adults were less successful using cognitive reappraisal to decrease unpleasant emotion compared with younger adults. In fact, older adults actually rated their experience as marginally more intense in the decrease condition compared with the view condition. By contrast, older adults were more successful using cognitive reappraisal to increase unpleasant emotion compared with younger adults. This represents a striking contrast in older and younger adults ability to use cognitive reappraisal to modulate their subjective experience of emotion, in this case experience of emotional intensity. At the crux of these age differences in the ability to use cognitive reappraisal to modulate experienced emotion is a difference in the information processing taking place in DMPFC and VLPFC, two regions traditionally implicated in cognitive control. This conclusion is supported by the neuroimaging data, to which we now turn Age differences in reappraisal-related prefrontal cortical activation There have been a few neuroimaging studies examining reappraisal in older adults (e.g., Urry et al., 2006, 2009; van Reekum et al., 2007) but none of those studies included a younger comparison group. Thus, it was unknown before this study whether cognitive reappraisal to increase and decrease unpleasant emotion would activate regions of prefrontal cortex implicated in cognitive control to a lesser degree in older than younger adults. In fact it did: younger adults exhibited greater reappraisal-related activation than older adults in left VLPFC and DMPFC, both of which are regions that have previously been implicated in cognitive control. As reported in our Supplementary Material, the age differences we observed in DMPFC and left VLPFC were not due to differences in the spatial extent of activation within these two regions because the age differences remained significant when examining peak signal in each region of interest instead of mean activation. The present results converge with previous studies that have examined effects of reappraisals that both increase and decrease unpleasant emotion in the same study (Eippert et al., 2007; Kim and Hamann, 2007; Ochsner et al., 2004; Urry et al., 2006, 2009; van Reekum et al., 2007). Most of these studies, like the present study, demonstrated both DMPFC and VLPFC activation during both the increase and decrease conditions compared with a control condition. Interestingly, the authors of the studies of only older adults (Urry et al., 2006, 2009; van Reekum et al., 2007) did not collect self-reports of emotional experience, a purposeful

9 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) omission taken to reduce demands on participants. The present results thus raise an important question: were the older adults in these previous studies actually successful in reducing experienced unpleasant emotion? It is impossible to say, although it is notable that one of those studies showed changes in skin conductance that confirmed successful reduction in physiological arousal (Urry et al., 2009). Importantly, the direction of the effect of greater activation in the DMPFC and VLPFC regions of activation on ratings of emotional intensity diverged when participants sought to increase their unpleasant emotion. On the one hand, greater activation in DMPFC was associated with greater success at increasing unpleasant emotion, which is conceptually similar to the finding reported by Urry and colleagues (Urry et al., 2009) in a standard (i.e., nongazedirected) reappraisal task. They found that higher activation in a similar region of dorsal cingulate cortex and more superiorly in supplementary motor area was correlated with heart rate acceleration during reappraisal, which was interpreted as evidence of effort related to somatic preparation for action. This region thus has a facilitative effect, producing changes in emotion that are consistent with the desired goal. On the other hand, greater activation in left VLPFC was associated with lower success at increasing unpleasant emotion. In fact, left VLPFC partially mediated the effect of older age on ratings of emotional intensity. This finding fits nicely with work by Swick, Ashley, and Turken (2008), who argued that left inferior frontal gyrus is necessary for response inhibition, a claim based on Go-No Go data in patients with lesions to this region. With respect to reappraisals that increase unpleasant emotion, this region may be exerting an inhibitory effect, producing changes in emotion that contradict the increase negative emotion goal. Because older adults exhibited lower activation in VLPFC, this inhibitory effect was reduced, leading them to be better at increasing negative emotion using cognitive reappraisal compared with younger adults. Alternatively, greater activation in VLPFC may reflect a selection process (Wager et al., 2008). This interpretation would suggest that older adults, who show reduced VLPFC activation, engage less in selection of increase reappraisal strategies than younger adults. It may be that older adults are more efficient in selecting effective increase reappraisals. Why did we not observe a similarly cogent set of brainbehavior associations to explain the age difference in reported emotional intensity when participants used reappraisal to decrease unpleasant emotion? The younger sample studied by Urry (2010) in a psychophysiology laboratory provided ratings of emotional intensity for the view condition (mean, 2.67) that were lower than those collected from our younger adults studied in the scanner (mean, 2.43). This represents an almost 9% decrement in experienced emotional intensity. It is perhaps not surprising, then, that significantly lower ratings of emotional intensity in the decrease compared with view condition were found by Urry (2010) but not in the present study using the same 4-point scale. There may have been a floor effect that reduced our sensitivity to detect decreases in the experienced intensity of emotion. As a result, the relative difference in ratings of intensity for view versus decrease conditions may have captured less meaningful variation in decrease reappraisal success to be explained by activation in DMPFC and VLPFC regions Clinical and theoretical implications Some forms of psychotherapy like cognitive behavioral therapy (CBT) rely heavily on the principles of cognitive reappraisal (Hofmann et al., 2009). However, our results suggest that, to the extent that a goal of CBT is to teach clients how to reappraise situations to decrease (and not increase) negative emotions, these treatments may not be optimal for treating older adults who struggle with frequent negative emotion. It has previously been shown, in fact, that CBT may be less effective in older adults than younger adults (Mohlman et al., 2003). The current study provides the novel observation that CBT may be less effective in older adults because of age-related reductions in the activation of regions that support cognitive control, specifically those that support somatic preparation for action and response inhibition or selection. For older adults who exhibit such reductions in neural activation, alternative interventions, such as cognitive bias modification, in which participants are trained to deploy their attention away from unpleasant information over repeated trials (see, e.g., MacLeod et al., 2009), should be explored. Finally, our results have important implications for socioemotional selectivity theory (SST) (Carstensen et al., 1999). SST proposes that as we grow older, we perceive time as more limited, which in turn prompts increased motivation to pursue goals that lead to emotional wellbeing. It has been suggested in the context of SST that older adults might achieve emotional well-being in part by virtue of increased skill at emotion regulation, particularly forms of emotion regulation that operate on emotion antecedents (e.g., the situations we are in, the things we pay attention to, and the meanings we attach to those situations) (Carstensen et al., 1998). Our data suggest, however, that older adults have difficulty using cognitive reappraisal to change the meanings they attach to situations to reduce negative emotional experience. Thus, age differences in emotional wellbeing are less likely to occur by virtue of cognitive reappraisal, at least the form studied herein Limitations and additional directions for future research Although we have contributed important, novel observations about the neural correlates of age differences in the ability to use cognitive reappraisal to change subjective

10 654 P.C. Opitz et al. / Neurobiology of Aging 33 (2012) emotional experience, there were some limitations to our approach. First, dot detection provides an indirect measure of deployment of visual attention to the prescribed areas of interest in the pictures. Real-time eye tracking that directly measures gaze location, as reported by Urry (2010), would have been preferable. In addition to providing a stronger check of the gaze direction manipulation, eye tracking data would also have permitted us to evaluate the possibility that age differences in attentional deployment prior to the gaze direction and reappraisal manipulations might contribute to the effects we observed after the manipulations. For instance, if older adults were attending more to arousing areas of the pictures than younger adults early on, this may explain greater experienced emotional intensity when asked to increase and decrease their emotional response later in the trial. While we collected eye tracking data during this experiment, a technical difficulty with the data files prevented us from analyzing them. A direct assessment of visual attention must be addressed in future experiments. Second, this study was optimized to assess whether there are age differences in the emotion-regulating effects of cognitive reappraisal while holding attentional deployment constant. Without a third control condition for the gaze direction factor (e.g., trials in which gaze is free to vary naturally), we cannot tell whether directing gaze to arousing or nonarousing information in these pictures each has emotion regulatory effects (see our Supplementary Material), or if one or the other accounts for higher ratings of emotional intensity and (marginally) neural activation when directing gaze to arousing relative to nonarousing areas. Third, as detailed in our Supplementary Material, controlling for age differences in dot detection, attentional control, and mood and anxiety symptoms reduced, but did not eliminate, the age differences we observed in ratings of emotional intensity and neural activation. This suggests that other processes that covary with age must account for the age differences. It will be important to identify the internal (e.g., cognitive ability) and external (e.g., help from others) resources that are needed to effectively change emotional responses. These resources will likely vary as a function of specific emotion regulation process (e.g., reappraisal, attentional deployment), and may help explain age differences in emotion regulation. We did not assess other resources beyond those noted here thus this remains a direction for future research. Fourth, the present sample had difficulty using reappraisal to decrease unpleasant emotion. It was noted above that this may be due in part to a floor effect. It might also be due to the relatively short amount of time participants were given to generate and enact reappraisals. Older adults may have been particularly affected by timing constraints given the tendency for older adults to exhibit reduced processing speed relative to younger adults (Salthouse, 2000). That being said, Urry (2010) demonstrated significant effects of decrease reappraisals in a sample of younger adults using exactly the same task. Moreover, a number of studies examining reappraisal in older adults used similar time frames (Urry et al., 2006; van Reekum et al., 2007). Nevertheless, it remains possible that providing more time may have facilitated decrease reappraisal success. Lastly, the experience of emotion as captured using ratings of emotional intensity and valence represents just one component of the multisystem emotional response. Because we did not measure other components (e.g., autonomic physiology or expressive behavior), we cannot say whether the age differences we observed in ratings of emotional experience would be mirrored therein. Older adults may be less emotionally expressive than younger adults, and may show decreased autonomic reactivity to unpleasant pictures (Smith et al., 2005) thus we might expect age differences in these measures during cognitive reappraisal too Concluding remarks Older adults have been shown to exhibit maintained or increased well-being despite otherwise declining abilities. It has been suggested that an enhanced ability to regulate emotional responses may explain this apparent paradox. Yet, there are many ways of regulating our emotions, not all of which may be preserved in older age due to age-related declines in the resources required to carry them out. Cognitive control is a resource that has been shown to be required to successfully use one form of emotion regulation, namely cognitive reappraisal. Thus, we expected that cognitive reappraisal would represent a form of emotion regulation with which older adults might have difficulty. To address this hypothesis, we used a novel gaze-directed reappraisal paradigm that holds constant an alternative form of emotion regulation, namely visual attentional deployment, so that we could isolate the effects of cognitive reappraisal. We found that younger adults were better able to decrease unpleasant emotions than older adults, but older adults were better able to increase unpleasant emotions than younger adults, as evidenced by self-report ratings of emotional intensity. Accompanying these findings, we found that older adults exhibited reduced activation in dorsomedial and left ventrolateral regions of PFC compared with younger adults. Importantly, activation in these two regions, particularly left VLPFC, differentially mediated the effect of age on ratings of emotional intensity in the context of increasing unpleasant emotion. Greater activation in DMPFC, on the one hand, was associated with higher ratings of emotional intensity. Greater activation in left VLPFC, by contrast, was associated with lower ratings of emotional intensity. Linking these effects up with previous studies, this pattern confirms the importance of cognitive control in being able to effectively reappraise unpleasant situations. Of particular importance cognitive control processes related to somatic