the encoding of posi
Donna R. Addisa, Christina M
aDepartment of Psychology, University of Auc
bDepartment of Psychology, Boston College, C
edical
related changes in the interactions between affect-processing regions and the hippo-
Though many aspects of memory change with aging, within
the realm of emotional memory, one of the most intriguing
effects of age has been a ‘‘positivity effect’’ in memory.
Compared to younger adults, proportionally more of what
Carstensen, 2005). Though there is active debate about the
generality of this effect (e.g., Murphy and Isaacowitz, 2008), it
also is clear that there are some circumstances that reliably
elicit a positivity effect (reviewed by Mather, 2006). The posi-
tivity effect generally has been interpreted within the
* Corresponding author. Department of Psychology, Boston College, 140 Commonwealth Avenue, Chestnut Hill, MA 02467, USA.
ava i lab le at www.sc ienced i rec t . com
.e l
c o r t e x 4 6 ( 2 0 1 0 ) 4 2 5 – 4 3 3
E-mail address: elizabeth.kensinger@bc.edu (E.A. Kensinger).
campus during the encoding of positive information.
ª 2009 Elsevier Srl. All rights reserved.
1. Introduction older adults remember is positive (reviewed by Mather and
fMRI
Memory
Structural equation modeling
Results: Aging did not impact the connectivity among regions engaged during the encoding
of negative information, but age differences did arise during the encoding of positive
information. Most notably, in older adults, the ventromedial prefrontal cortex and amyg-
dala strongly influenced hippocampal activity during the encoding of positive information.
By contrast, in young adults, a strong thalamic influence on hippocampal activity was
evident during encoding.
Conclusions: These findings suggest that older adults’ ‘‘positivity effect’’ may arise from age-
cAthinoula A. Martinos Center for Biom
a r t i c l e i n f o
Article history:
Received 13 November 2008
Reviewed 13 February 2009
Revised 10 March 2009
Accepted 13 April 2009
Published online 18 May 2009
Keywords:
Aging
Connectivity
Emotion
0010-9452/$ – see front matter ª 2009 Elsevi
doi:10.1016/j.cortex.2009.04.011
tive, but not negative, information
. Leclercb, Keely A. Muscatellb and Elizabeth A. Kensingerb,c,*
kland, New Zealand
hestnut Hill, MA, USA
Imaging, Massachusetts General Hospital, Charlestown, MA, USA
a b s t r a c t
Introduction: Older adults often show sustained attention toward positive information and
an improved memory for positive events. Little is known about the neural changes that
may underlie these effects, although recent research has suggested that older adults may
show differential recruitment of prefrontal regions during the successful encoding of
emotional information. In the present study, effective connectivity analyses examined the
network of regions that college-age and older adults recruited during the encoding of
positive and negative images.
Methods: Participants viewed positive and negative images while undergoing a functional
magnetic resonance imaging (fMRI) scan. Structural equation modeling was used to
compare young and older adults’ connectivity among regions of the emotional memory
network while they encoded negative or positive items.
There are age-rel
d changes in neural connectivity during
ate
Special issue: Research report
journa l homepage : www
er Srl. All rights reserved
sev ier . com/ loca te / cor tex
.
tion and process it more deeply than young adults or because
c o r t e x 4 6 ( 2 0 1 0 ) 4 2 5 – 4 3 3426
framework of Socioemotional Selectivity Theory (Carstensen
et al., 1999); this theory states that when adults view time as
limited (as occurs with advancing age), they are more likely to
prioritize their emotional wellbeing. The positivity effect has
been hypothesized to reflect the increased importance that
older adults place on emotional gratification: older adults are
more motivated than young adults to process information in
a manner that will provide them with emotional fulfillment,
and because of this motivational change, older adults are
more likely to attend toward, and to remember, positive
information (reviewed by Carstensen and Mikels, 2005;
Mather and Carstensen, 2005).
The hypothesis that older adults’ positivity effect in
memory is tied to motivational changes in the processing of
emotional information is a plausible one, particularly because
the effect is noted most readily when older adults have
extensive cognitive resources available to devote toward pro-
cessing emotional information (Mather and Knight, 2005). The
hypothesis is also consistent with a number of studies that
have suggested an age-related shift away from amygdala-
based emotion processing and toward prefrontal-based pro-
cessing (where this prefrontal-based processing may reflect
a more deliberative, controlled type of processing; e.g., Satpute
and Lieberman, 2006). Most of the studies that have revealed
these age-related changes in neural activity have examined
the processing of facial expressions that convey negative
emotion. In these studies, older adults tend to under-activate
the amygdala and to over-activate the prefrontal cortex (as
compared to young adults) when processing negative facial
expressions (e.g., Gunning-Dixon et al., 2003; Fischer et al.,
2005; Tessitore et al., 2005; Williams et al., 2006). A recent set of
studies has suggested that this shift from amygdala-driven to
prefrontal-based processing may also occur when older adults
view negative photographs (rather than facial expressions),
and that these changes may be tied to older adults’ poor
memory for negative information (see also Mather et al., 2004;
St Jacques et al., 2009; St Jacques et al., 2010).
Because these studies only examined the processing of
negative information, they could not clarify whether the age-
related changes in neural activity are specific to negative
valence or whether they extend to positive information as
well. There is reason to believe that valence may be an
important factor to consider, with a few studies suggesting
that older adults’ recruitment of the amygdala may be
stronger during the processing of positive as compared to
negative information (e.g., Mather et al., 2004; Leclerc and
Kensinger, 2008) and that their prefrontal recruitment may
also be heightened during the presentation of positive as
compared to negative information (e.g., Leclerc and Ken-
singer, 2008; Leclerc and Kensinger, in press). In fact, old age
sometimes results in a valence reversal: whereas young adults
often activate the amygdala and medial prefrontal cortex
more strongly in response to negative as compared to positive
items, older adults sometimes show the opposite pattern of
activity (e.g., Mather et al., 2004; Leclerc and Kensinger, 2008).
These studies suggest that older adults’ processing of both
positive and negative information may be altered, but that the
alterations may diverge for the two valences of information: The
strength of older adults’ recruitment of the amygdala and of
medial prefrontal regions may be stronger for positive
they direct less attention and devote fewer processing resources
toward negative information? Neuroimaging can provide
a viable way to address this issue, by examining whether aging
primarily impacts the processes recruited during the encoding of
positive information or if it also influences the processes
recruited during the encoding of negative information.
The present study used effective connectivity analyses of
functional magnetic resonance imaging (fMRI) data to reveal
how aging affects the connections among brain regions
recruited during encoding of negative and positive items. We
focus on age-related connectivity changes within a network of
brain regions including the amygdala, the prefrontal cortex,
and the hippocampus. These regions have been linked to the
successful encoding of emotional information by neuro-
imaging studies that have used a ‘‘subsequent memory’’
paradigm (reviewed by Paller and Wagner, 2002). In such
paradigms, neural activity during the encoding phase is sorted
based upon whether the items are later remembered or later
forgotten, and regions are considered to have a link to
successful encoding if their activity is greater for items that
are later remembered than for those that are later forgotten. A
large literature has revealed a consistent network of regions
that are implicated in the successful encoding of emotional
information. Some of these regions seem to be recruited
specifically when information is emotional (e.g., the orbito-
frontal cortex and amygdala), whereas others serve a general
role in the encoding of information with or without emotional
content, though their activity may be modulated by the
emotional salience of information (e.g., the hippocampus; see
Hamann, 2001; LaBar and Cabeza, 2006; Kensinger, 2009 for
reviews). Our primary question was how aging would affect
the connections between the amygdala, the prefrontal
regions, and the hippocampus, a region that is essential for
the successful encoding of long-term memories (see Postle,
2009 for recent review). We were especially interested in
whether age-related changes in connectivity would be
comparable whenever participants were processing
emotional information (regardless of its valence) or whether
the age-related differences would be more pronounced for one
valence of information.
2. Methods
2.1. Participants
The participants comprised 17 younger adults (12 women)
between 19 and 31 years of age and 20 older adults (13 women)
between 61 and 80 years of age (see Table 1 for addition
information than for negative information. It is not clear from
this prior research whether, or how, these age-related changes
connect to older adults’ ‘‘positivity effect’’ in memory. In
particular, it is widely debated whether older adults’ ‘‘positivity
effect’’ is best characterized as a shift toward the positive or as
a shift away from the negative (see Murphy and Isaacowitz,
2008). For example, do older adults remember proportionally
more positive information because they attend to that informa-
participant characteristics). All participants were right-
handed, native English speakers, with no history of depression
c o r t e x 4 6 ( 2 0 1 0 ) 4 2 5 – 4 3 3 427
or other psychiatric or neurological disorder. All participants
had scores on the Mini Mental Status Examination (Folstein
et al., 1975) of greater than 27. No participant was taking any
medication that would affect the central nervous system.
Informed consent was obtained from all the participants in
a manner approved by the Boston College and Massachusetts
General Hospital Institutional Review Boards. Participants
were compensated $75 for their participation.
2.2. Materials
The photo objects used in the present study were a subset of
those previously rated by a separate group of young and older
adults on valence and arousal dimensions, using 9-point Lik-
ert-type scales (1¼negative valence or low arousal and
9¼ positive valence or high arousal). One-third of the images
selected were negative and high arousal (valence ratings less
than 3.5, arousal ratings higher than 5), one-third were posi-
tive and high arousal (valence higher than 5.5, arousal higher
than 5), and one-third were neutral (valence ratings between 3
and 6, arousal less than 5; see Leclerc and Kensinger, 2008 for
more details on the stimuli). The participants in this study
also rated the stimuli for valence and arousal, and their clas-
sifications of the stimuli generally agreed with those based
upon prior ratings. In the rare instances where classifications
did not agree (fewer than 2% of all items), those items were not
Table 1 – Participant characteristics.
Measure Young Older
Years Education 15.2 (.38) 16.8 (.32)
Digit Symbol Substitution 71.4 (3.3) 49.1 (2.2)
Vocabulary .82 (.02) .89 (.02)
Digit Span – Forward 5.9 (.45) 5.3 (.25)
Digit Span – Backward 5.12 (.25) 4.65 (.33)
Note: Digit Symbol Substitution, Digit Span Backward, & Arithmetic
from Wechesler Adult Intelligence Scale (WAIS)-III Wechesler
Adult Intelligence Scale; Wechsler, D. Technical Manual for the
Wechsler Adult Intelligence and Memory Scale, 3rd ed. New York:
The Psychological Corporation, 1997. Vocabulary reflects the
percentage correct on the Shipley (1986) measure. All values
represent the mean (SE) raw, nonstandardized scores.
included in the analyses.
2.3. Procedure
While in the fMRI scanner, participants viewed 108 positive,
108 negative, and 108 neutral photo objects for 1000 msec
apiece. Images from each emotion category were pseudor-
andomly intermixed with one another. After each image was
presented, an intertrial fixation cross was presented for
a variable duration, ranging between 5 and 13 sec, to provide
jitter (Dale and Buckner, 1997). The order of the images, and
the length of each intertrial fixation interval, was determined
using a program that optimizes for detection of the hemody-
namic response associated with each image (the optseq
program available from http://www.nitrc.org/projects/
optseq). While viewing the photos, participants were asked
to indicate whether each object would fit inside of a file
cabinet drawer. This task ensured that participants were
attending to each item but did not direct them to process its
emotional relevance nor to memorize the items for a later
memory task (no participant indicated that they were aware
that a memory assessment would follow the scan), a set of
conditions that facilitates the revelation of a positivity effect
(see Mather, 2006). Once participants were settled outside of
the scanner (after approximately a 20 min delay), they per-
formed a recognition memory task. Participants were pre-
sented with a series of studied items intermixed with novel
items; items were presented one at a time, and for each item,
participants were asked to indicate whether it had been
studied. The recognition test was self-paced, so that as soon
as a participant made a response, the next item was pre-
sented. The present analyses focus only on the neural activity
during the encoding of items that were later recognized
successfully, so that age differences in connectivity could not
be attributed to differences in the proportion of items that
were successfully encoded (see Leclerc and Kensinger, 2008
for presentation of the neuroimaging data for all items,
regardless of later memory performance). Because successful
encoding was operationalized as the encoding of items that
were later recognized, it is possible that a small proportion of
items was mistakenly considered to be ‘‘successfully enco-
ded’’ because participants correctly guessed about those
items’ presentations.
2.4. Image acquisition and data analysis
Images were acquired on a 3.0 T Siemens Allegra magnetic
resonance imaging (MRI) scanner. Stimuli were back-
projected onto a screen in the scanner bore, and the partici-
pants viewed the images through an angled mirror attached to
the head coil. Detailed anatomic images were acquired using
a multiplanar rapidly acquired gradient echo sequence.
Functional images were acquired using a T2*-weighted echo
planar imaging sequence (repetition time – TR¼ 3000 msec,
time to echo – TE¼ 30 msec, field of view – FOV¼ 200 mm; flip
angle¼ 90�). Twenty-eight axial-oblique slices (3.2 mm thick-
ness, .6 mm skip between slices), aligned in a plane along the
axis connecting the anterior commissure and the posterior
commissure, were acquired in an interleaved fashion.
All preprocessing and data analysis were conducted within
SPM2 (Wellcome Department of Cognitive Neurology). Stan-
dard preprocessing was performed on the functional data,
including slice-timing correction, rigid body motion correc-
tion, normalization to the Montreal Neurological Institute
template (resampling at 3-mm cubic voxels), and spatial
smoothing (using a 7.6-mm full-width half maximum
isotropic Gaussian kernel).
2.5. Effective connectivity analyses
In order to examine the interactions between the amygdala
and other limbic and prefrontal regions during the processing
of negative and positive items, structural equation modeling
(SEM) was carried out using Lisrel software (Joreskog and
Sorbom, 1993). Unlike simple correlations, SEM allows for
a consideration of connections across multiple nodes of
a network, and it provides information about the direction-
ality of influences between different regions. The first step of
the SEM analysis was to specify the anatomical model.
Regions were included in the model if they were of theoretical
relevance to emotional memory (see reviews by Hamann,
2001; Phelps and LeDoux, 2005; LaBar and Cabeza, 2006) and
were revealed in a whole-brain analysis comparing activity
elicited during the successful encoding of positive or negative
information to the activity elicited during the unsuccessful
encoding of that emotional information. This analysis was
conducted collapsing across the age groups, so as to define the
regions in an unbiased fashion with regard to age. Because of
the contrast used to define the regions, all regions included in
the model were related to memory for emotional items rather
than to more general emotion processing. Note, however, that
we did not require regions in this model to be implicated only
in memory for emotional (and not neutral) information,
because we expect there should be overlap between nodes of
the emotional memory network and nodes of a more general-
purpose mnemonic network that is not specific to emotional
information (and see Hamann, 2001; LaBar and Cabeza, 2006
for discussion). The regions selected were: ventromedial
prefrontal cortex (Talairach coordinates, x, y, z¼ 0, 39, �5;
Brodmann area – BA 10/32), dorsomedial prefrontal cortex (x,
y, z¼ 8, 61, 15; BA 10), left orbitofrontal cortex (x, y, z¼�36, 44,
�8; spanning BA 10/11/47), left amygdala (x, y, z¼�28, �3,
�12), left hippocampus (x, y, z¼�36, �7, �23), thalamus (x, y,
z¼�6,�20,�2), and left fusiform gyrus (x, y, z¼�46,�48,�18;
BA 37; see Fig. 1 for the anatomic model). Table 2 presents
c o r t e x 4 6 ( 2 0 1 0 ) 4 2 5 – 4 3 3428
additional information on the pattern of activity revealed
within these regions; this information was determined by
analysis of variances (ANOVAs), computed separately for each
region, which compared the maximum signal change reached
Fig. 1 – Anatomical model used in the SEM.
(summing across 4–8 sec post-stimulus onset) as a function of
subsequent memory (remembered, forgotten), valence
(positive, negative, neutral), and age (young, older adult).
On the basis of anatomical research in nonhumans (e.g.,
Swanson and Petrovich, 1998; Patterson and Schmidt, 2003),
an anatomical connectivity model was created to specify the
anatomically plausible connections (including multi-synaptic
connections) between the specified network nodes and the
potential directions of those connections (McIntosh, 1999;
Addis et al., 2007).
Next, a functional model was constructed for each group.
Signal change was extracted from all active voxels within
a 5 mm sphere (centered on the peak voxel) in each region of
interest, using the MarsBar toolbox implemented within SPM2
(Brett et al., 2002). The sum of the signal change across the
4–8 sec period post-stimulus onset was extracted for remem-
bered items only, and this signal change was extracted sepa-
rately for each person, for each valence type, and for each
region. We focus on the successful encoding of information
(i.e., on signal change to subsequently remembered items) so
that differences in the models cannot be accounted for by age
differences in encoding effectiveness. Correlation matrices
were then cr