Berman SCAN 2010

 

Social Cognitive and Affective Neuroscience Advance Access published September 19, 2010  doi:10.1093/scan/nsq08 0

SCAN (2010) 1 of 8

Depression, rumination and the default network  1

1

2

1

1

Marc G. Berman, Scott Peltier, Derek Evan Nee, Etha Ethan n Kross, Patricia J. Deldin, and John Jonides

1

1

Department of Psychology, University of Michigan, 530 Church Street, Ann Arbor, MI 48109-1043, USA and   2Indiana University, Department of Psychological and Brain Sciences, 1101 E. Tenth St., Bloomington, IN 47405, USA Major depressive depressive disord disorder er (MDD (MDD)) has been characterized characterized by excess excessive ive defaul default-net t-network work activation activation and connectivity connectivity with the subgenual subge nual cingulate. cingulate. These hyper-connect hyper-connectivitie ivities s are often interpreted interpreted as reflec reflecting ting rumination, rumination, where MDDs persev perseverate erate on negative, negat ive, self-r self-referen eferential tial thoug thoughts. hts. However, the relat relationsh ionship ip betwe between en conn connectiv ectivity ity and rumin rumination ation has not been estab established lished.. Furthermore, previous research has not examined how connectivity with the subgenual cingulate differs when individuals are engaged in a task or not. The purpose of the present study was to examine connectivity of the default network specifically in the subgenual cingulate both on- and off-task, and to examine the relationship between connectivity and rumination. Analyses using usi ng a see seed-b d-base ased d con connec nectiv tivity ity app approa roach ch rev reveal ealed ed tha thatt MDD MDDs s sho show w mor more e neu neural ral fun functi ctiona onall con connec nectiv tivity ity bet betwee ween n the posterior-cingulate cortex and the subgenual-cingulate cortex than healthy individuals during rest periods, but not during task  engagement engag ement.. Impor Important tantly, ly, these rest-period rest-period conn connectiv ectivities ities corre correlated lated with behavioral measures of rumi ruminatio nation n and brood brooding, ing, but not reflection.

Keywords:   depression; rumination; default network; subgenual cingulate; functional magnetic resonance imaging  Keywords:

INTRODUCTION

Rumination defined as ‘a mode of responding that involvesisrepetitively and passively focusing to ondistress symptoms of distress and on the possible causes and consequences of these sympt symptoms’ oms’ (Nol (Nolen-H en-Hoekse oeksema ma   et al ., . , 20 2008) 08).. Thi Thiss tendency characterizes depression (Nolen-Hoeksema   et al ., ., 2008), and has been ascribed to deficient control processes that cannot rid memory of negative information (Joormann, 2005 20 05). ). As su such ch,, a gr grow owin ingg li lite tera ratu ture re ha hass ex exam amin ined ed th thee relationsh relat ionship ip betw between een depre depressive ssive rumi rumination nation and cogni cognitive tive control during demanding cognitive tasks. However, it is likely that the most prominent display of  rumination is not when people are engaged in a task, but when they are at rest. Examining how individuals with MDD and healthy controls (HCs) compare during such rest periods is important because interleaved with the ongoing tasks of life are significant periods in which people do not engage in structured tasks, and instead are left to mind-wander or rumina rum inate. te. In fac fact, t, rec recent ent res resear earch ch ind indica icates tes tha thatt peo people ple 1 mind-wander   10 – 15% 1 5% of the their ir wak wakefu efull hou hours rs (Sayette et al ., ., 2009). Neurally, mind-wandering appears to engage   et al ., regions of a ‘default network’ (Christoff  et ., 2009, Mason et al ., ., 2007)a set of neural regions that activate in unison duri du ring ng of offf-ta task sk or ‘r ‘res est’ t’ pe peri riod ods, s, wh whic ich h in incl clud udee th thee

Received 3 May 2010; Accepted 6 August 2010 This work was supported in part by a grant from the National Institute of Health (NIMH) 60655 to J.J., and a National Science Foundation (NSF) graduate fellowship to M.G.B. We thank Courtney Behnke, Melynda Casement Caseme nt and Hyan Hyangg Sook Kim for help with data collection. collection. Corresponde Corre spondence nce should be addre addressed ssed to Marc G. Berman Berman,, Depart Department ment of Psycho Psychology, logy, University University of  Michigan, 530 Church Street, Ann Arbor, MI 48109-1043, USA. E-mail: [email protected] 1

Participants in this study were engaged in a reading task.

posterior poster ior-ci -cingu ngulat late, e, por portio tions ns of lat latera erall par pariet ietal al cor cortex tex as well as portions of the medial temporal lobe (MTL) and medial prefrontal cortex (mPFC; Fox   et al ., ., 2005; Raichle, 2010; Shulman  et al ., ., 1997; Raichle  et al ., ., 2001). Rece Re cent nt re rese sear arch ch ex exam amin inin ingg th thee de defa faul ultt ne netw twor ork k ha hass reveal rev ealed ed str striki iking ng dif differ ferenc ences es bet betwee ween n MDD MDDss and HC HCs. s. MDDs show increased default-network connectivity (compared with HCs) with the subgenual-cingulate cortex (SCC), a region located in the mPFC, which is positively correlated with wi th th thee le leng ngth th of MD MDDs Ds’’ cu curr rren entt de depr pres essi sive ve ep epis isod odes es et al ., (Greicius   et . , 20 2007 07). ). Ot Othe herr re rese sear arch cher erss ha have ve sh show own n abno ab norm rmal alit itie iess in th thee su subg bgen enua uall ci cing ngul ulat atee fo forr MD MDDs Ds (Sheline   et al ., ., 2009; see Drevets   et al ., ., 2008 for a review), and this brain region has also been linked to poor emotion regulation (Abler   et al ., ., 2008) and is activated more when healthy young adults are induced to ruminate (Kross  et al ., ., 2009). Moreover, stimulation of white matter tracts leading to the sub subgen genual ual cin cingul gulate ate in MDD MDDss has bee been n ass associ ociate ated d with remission of depression concomitant with a decrease in hyperactivity of the subgenual-cingulate itself (Mayberg et al ., ., 2005). Even with all of this research, it is not clear what cognitive proces pro cesses ses are ref reflec lected ted by the these se dif differ ferenc ences es in def defaul aulttnetwor net work k con connec nectiv tivity ity (Ra (Raich ichle, le, 20 2010 10). ). Alt Althou hough gh som somee researchers have speculated that this hyper-connectivity reflects flec ts rumi rumination nation (Gre (Greicius icius   et al ., . , 200 2007) 7),, no res resear earch ch has directly tested this hypothesis. The first goal of the present research resea rch was to test wheth whether er defau default-n lt-networ etwork k conne connectivi ctivity, ty, part pa rtic icul ular arly ly in th thee su subg bgen enua uall ci cing ngul ulat ate, e, is re rela late ted d to rumination. Goal two was to examine how differences between MDDs and HCs in default-network connectivity changed when participants were on-task   vs  off-task.   off-task. In everyday life, people’s

 The Author (2010).Published by Oxford Oxford Unive University rsity Press.For Permissions, Permissions, please email: journal journals.permissi s.permissions@oxfordjournal [email protected] s.org

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u rn  r   a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

2 of 8

M. G. Berman et al al..

SCAN (2010)

attention attent ion wav wavers ers bet betwee ween n tas tasks ks and unf unfocu ocused sed tho though ughtt (Sayette   et al ., ., 2009 2009;; Chris Christoff  toff   et al ., ., 200 2009). 9). For exam example, ple, the accountant at work may focus intensely on a balance sheet, but divert attention intermittently to thoughts unrelated to this task. In the present research, the relationship betwee bet ween n def defaul ault-s t-stat tatee con connec nectiv tivity ity and rum rumina inatio tion n was exami ex amined ned usi using ng a sit situat uation ion tha thatt app approx roxima imates tes suc such h rea reallworld wor ld con condit dition ions. s. The rel relati ations onship hip bet betwee ween n rum rumina inatio tion n and default-network connectivity was explored when participants pan ts per perfor formed med a dem demand anding ing sho shortrt-ter term m me memor moryy tas task  k  interleave inter leaved d with perio periods ds of rest. Alternating Alternating betw between een rest and an d ta task sk ep epoc ochs hs in th this is ma mann nner er al allo lowe wed d us to ex expl plor oree whether wheth er defa default-n ult-networ etwork k conne connectiv ctivity ity for MDDs and HCs varied varie d betw between een off-task and on-ta on-task sk perio periods. ds. In addit addition, ion, the relationships between self-report measures of rumination and default-network connectivity could then be compared for rest and task epochs separately. In sum, this work was designed to examine whether measures of rumination predicted connectivity of the default state especi esp eciall allyy in the sub subgen genual ual cin cingul gulate ate for dep depres ressed sed and healthy individuals, connectivity differences at rest persisted for on-task epochs and the relationship between rumination

think ‘‘W think ‘‘Why hy do I alw always ays rea react ct thi thiss way way?’’’ ?’’’ Par Partic ticipa ipants nts responded with a 4-point scale ranging from 1-almost never, to 4-almost always. There were 22 items in total. The RRS can further be subdivided into three components: brooding, reflectio refl ection n and depre depression ssion-rela -related ted item itemss (Tre (Treynor ynor   et al ., 2003). MDDs and HCs differed significantly on brooding, the full RRS and the BDI, but no differences were found between the two groups on reflection scores (see Table S1

and defaul def ault-n t-netw etwork ork con connec nectiv tivity ity var varied ied for onon-tas task k and off-task periods.

gradient-recalled in steady-state imaging (TR ¼ 9 ms, TE ¼acquisition 1.8 ms, flip(SPGR) angle ¼ 15 angle , FOV ¼ 25 –  2 6 mm 26 mm,, slice thickness ¼ 1.2 mm). Each SPGR image was corrected for signal inhomogeneity  and skull-stripped using FSL’s Brain Extraction Tool (Smith et al ., ., 2004). These images were then normalized with SPM5 (Wellcome Department of Cognitive Neurology, London), and nor normal maliza izatio tion n par parame ameter terss wer weree cal calcul culate ated d fro from m the standard MNI template. These parameters were then applied to th thee fu func ncti tion onal al im imag ages es ma main inta tain inin ingg th thei eirr or orig igin inal al 3.44  3.44  3-mm resolution, and were spatially smoothed with a Gaussian kernel of 8  8  8 mm3. Functional images were slice-time corrected using a 4-point sinc-interpolation (Oppenheim et al ., ., 1999) and were corrected for head movement using MCFLIRT (Jenkinson  et al ., ., 2002). To reduce the

MATERIALS AND METHODS Participants and behavioral measures

Default-n Defaul t-netw etwork ork con connec nectiv tivity ity in 15 MDD MDDss (10 fem female ales, s, 5 mal males, es, mean age ¼ 25. 25.77 yea years) rs) and 15 HCs (10 female female,, 5 male, mean age ¼ 23 years years)) was expl explored ored during non-task  non-task  fixation periods at the beginning and end of each run of a short-term memory experiment conducted in a functional magnet mag netic ic res resona onance nce ima imagin gingg (fM (fMRI) RI) env enviro ironme nment. nt. The short-term memory task was a variant of a directed forgetting task (Nee  et al ., ., 2007), in which participants were initially instructed to remember four words, and after a delay  were forget half of the words participants and remember thea otherinstructed half. Aftertoanother delay interval, saw single word and responded yes or no whether that word was one of the words in the to-be-remembered set. Sometimes, the single words were words from the to-be-forgotten set, and an d th thee ab abil ilit ityy of MD MDDs Ds an and d HC HCss to fo forg rget et po posi siti tive vely  ly  valenced   vs   negati negativel velyy val valenc enced ed wor words ds was exa examin mined. ed. A more detailed explanation of the task can be found in the Supplementary materials. materials. The diagnosis diagnosis of MDD was determined determined with a Struc Structured tured Clinical Interview for the DSM-IV (SCID) by an advanced clinical psychology graduate student and was confirmed by  a secon second d indep independen endentt rater rater.. Behav Behavioral ioral rumination rumination score scoress were we re me meas asur ured ed wi with th th thee Ru Rumi mina nati tion on Re Resp spon onse se St Styl yles es (RRS) inventory (Treynor  et al ., ., 2003), and depressive severity was also assessed with the Beck Depression Inventory II  et al ., (BDI; Beck  et ., 1996). The RRS measured rumination sub jectively with questions such as: ‘[How often do you] think  ‘‘What ‘‘Wh at am I doing to deserve this?’’ this?’’ ’ or ‘[Ho ‘[How w often do you]

in the Supplementary Material). Six MDDs were medicated, and two MDDs had co-morbid anxiety. 2 fMRI parameters

Images were acquired on a GE Signa 3-T scanner equipped with wi th a st stan anda dard rd qu quad adra ratu ture re he head ad co coil il.. Fu Func ncti tion onal al T2*-weighted images were acquired using a spiral sequence with wit h 40 con contig tiguou uouss sli slices ces wit with h 3.4 3.444  3.44  3-mm voxel voxelss [repetitio [repe tition n time (TR) ¼ 20 2000 00 ms; echo tim timee (TE (TE)) ¼ 30ms; flip fli p ang angle le ¼ 90 ; fi fiel eld d of vi view ew (F (FOV OV)) ¼ 22mm2] . A T1weight wei ghted ed gra gradie dient nt ech echo o ana anatom tomica icall ove overla rlayy was acq acquir uired ed using the same FOV and slices (TR ¼ 250ms, TE ¼ 5.7ms, flip angle ¼ 90 ). Addit Additional ionally, ly, a 124 124-slic -slicee high-r high-resolu esolution tion T1-weighted anatomical image was collected using spoiled8

8

8

impact of spike artifacts, AFNI’s de-spiking algorithm was implemented. There were 16 TRs of Fixation, eight at the beginning and end of each run, and 168 TRs of task (where participants partic ipants performed performed a shortshort-term term memo memory ry task) task).. There were 12 runs in the experiment. Seed analysis

Default-networ Default-n etwork k conne connectivi ctivity ty was reve revealed aled by sele selecting cting a seed voxel in the posterior-cingulate cortex (PCC),   x ¼ 7,  y ¼ 45,  z ¼ 24, and that voxel’s time-course was correlated within wit hin-su -subj bject ectss for all vox voxels els in the brain. brain. Thi Thiss see seed d was selected anatomically and is similar in location to regions that other authors have used to define the default network  (Greicius   et al ., . , 2003; Fox   et al . 2005; Monk   et al ., ., 2009; Raichle, 2010). The posterior cingulate has been argued to play a central role in the default mode network (Greicius 2

We did not find any differences between medicated and non-medicated MDDs in any rumination scores, BDI or neural connectivity.

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u rn  r   a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

Depression, rumination and de default ne network

SCAN (2010)

3 of 8

Fig. 1   (A) Defaul Default-net t-network work connectivit connectivityy for MDDs and HCs durin duringg fixat fixation ion perio periods ds defin defined ed by conne connectivi ctivity ty with posterior-ci posterior-cingula ngulate te corte cortex, x,   x ¼ 7,   y ¼ 45,   z ¼ 24. Correlations >0.25 (P  <   < 0.001) are displayed. (B) Results of a two-sample   t -test -test comparing comparing MDDs’ and HCs’ default-networ default-networkk conne connectivi ctivity ty durin duringg fixat fixation ion perio periods. ds. MDDs show more connectivity in the subgenual-cingulate than HCs ( P  <  < 0.05 0.05 corr corrected; ected; peak at  x  ¼ 0,  y  ¼ 38,  z  ¼ 9; 46 voxels).

et al ., ., 2003), and has been found to reveal connectivity in the default network most effectively (Greicius  et al ., ., 2003) and is

(containing 978 voxels) which were constructed anatomically based on the WFU PickAtlas. Results for these ROI ana-

a reason why other authors have used the posterior cingulate as a seed to define the default network (Monk  et al ., ., 2009). In add additi ition, on, it is an are areaa of gre greate atest st dea deacti ctivat vation ion dur during ing off-task behavior (Shulman  et al ., ., 1997). Default-network connectivity was calculated separately for fixation and task blocks in which participants performed a shortsho rt-ter term m mem memory ory tas task. k. Tas Task k and fix fixati ation on epo epochs chs wer weree low-pass filtered, de-trended to remove within-run drift in the fMRI signal (de-trending was performed separately for the beginning and end fixations and acts as a high-pass filter) and processed to have a mean of ‘0’ and a standard deviation of ‘1’ separately for each TR, which was done to control for global activation changes that may have occurred over time. These The se run runss wer weree the then n con concat catena enated ted tog togeth ether, er, whi which ch may  have biased the correlations correlations positi positively vely,, but this potent potential ial bias did not interact with group. Thus, there were 192 TRs of fixation/rest fixation/rest (6.4 min) and 201 20166 TRs of task (Inter-Trial (Inter-Trial-Interv Int ervals als were rem remove oved). d). While this may not be an ide ideal al design due to the imbalance in number of TRs, the difference in the number of TRs for task and rest should not interact with wit h gro group, up, which which is of mai main n int intere erest st in thi thiss stu study dy (i. (i.e. e. comparing MDDs   vs   HCs during rest, during task and the difference between rest and task epochs). Correlation coefficients cie nts wer weree con conver verted ted to   z -score -scoress and ent entere ered d int into o our two-sample   t -tests -tests in SPM 5. Results were thresholded at P < 0.00 0.0011 (unco (uncorrec rrected) ted) at the voxel-level voxel-level and corre corrected cted by  using usi ng a clu cluste ster-s r-size ize thr thresh eshold old of 26 vox voxels els to pro produc ducee a P < 0.05 (corrected) (corrected) threshold (Forman (Forman et al ., ., 1995). An additional seed analysis was conducted with the mPFC as the

lyses are reported at  P <a0.05 (uncorrec (uncorrected) ted) of at eight the voxel-level and corrected by using cluster threshold contigu  P  ouss vo ou voxe xells to pr prod oduc ucee a < 0. 0.05 05 cor correc rected ted thr thresh eshold old (Forman  et al ., ., 1995).

 et al ., seed region based on Fox  et ., (2005). We performed region of inter interest est (ROI) analy analyses ses for the connectivity connectivity networks obtained when using the mPFC as the seed. Two ROIs were of  intere int erest, st, the PC PCC C (co (conta ntaini ining ng 45 4555 vox voxels els)) and the MTL

ation scores and connectivity in that MDDs have reliably  higher RRS scores. Therefore, a 10-mm sphere centered on subgenual-cingulate coordinates (x ¼ 6,   y ¼ 36,   z ¼ 4) was constructed from an independent study (Zahn  et al ., ., 2009)

RESULTS Relation between rumination and neural connectivity

As depicted in Figure 1A, both groups showed high connectivity ivi ty in the def defaul aultt net networ work k dur during ing fix fixati ation on per period iods. s. A two-sample   t -tes - testt co comp mpar arin ingg th thee de defa faul ultt ne netw twor orks ks fo forr MDDs vs  HCs  HCs revealed that MDDs had stronger connectivity  with the subgenual cingulate than HCs (Figure 1B) at standard statistical thresholds. At more liberal thresholds, MDDs show more connectivity in other areas as well, which can be see seen n vis visual ually ly in Figure Fig ure 1A 1A.. To examin exa mine how rum rumina inatio tion n scores related to connectivity between thee subgenual cingulatee and the pos lat poster terior ior cin cingul gulate ate,, a fun functi ctiona onall ROI of the subgenual subge nual cingulate cingulate was created based on the two-sample two-sample t -test. -test. Rumination scores (from the RRS) were then correlated with connectivity in this functionally defined region of  interest (ROI) across all participants, which revealed a significant positive relationship (r ¼ 0.68;   P < 0.00 0.001; 1; see Figure 2A). This relationship also held for both MDDs and HCs separately (Figure 2B), though these correlations were smaller due to the reduced range of the RRS measure ( r ¼ 0.30 for MDDs,  r  ¼ 0.23 for HCs). The functionally defined ROI that was used may be biased in that it was based on connectivity that was greater for MDDs, which may inflate the relationship between rumin-

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u r  r  n  a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

4 of 8

SCAN (2010)

M. G. Berman et al al..

Fig. 2   (A) Correlations drawn from the resulting subgenual-cingulate ROI from the two-sample  t -test -test comparing the groups at rest. SCC – PCC PCC connectivity correlates positively with subjective rumination scores across groups ( r ¼ 0.68, 95% confidence interval  r  ¼ 0.44 – 0.85). 0.85). (B) Correlations drawn from the resulting subgenual-cingulate ROI from the two-sample t -test -test comparing the groups at rest. SCC – PCC PCC connectivity correlates positively with subjective rumination scores for both MDDs and HCs. The linear relationship equation is shown in the upper right for MDDs, and lower right for HCs.

and con connec nectiv tivity ity sco scores res wer weree ext extrac racted ted wit within hin thi thiss ROI ROI.. These The se con connec nectiv tivity ity sco scores res wer weree the then n cor correl relate ated d wit with h RRS scoress across all parti score participan cipants. ts. Again Again,, a relia reliable ble correlation correlation (r ¼ 0.53,   P < 0. 0.00 005; 5; Ta Tabl blee 1) wa wass fo foun und d su sugg gges esti ting ng th that at default-ne defau lt-network twork conne connectivi ctivity ty with the subg subgenual enual cing cingulate ulate is related to ruminative tendencies. The correlation between RRS scores and connectivity was not driven by a main effect of group (i.e. the relationship between RRS and PCC – SCC SCC correlations seems to be a continuous positive trend as it was for the functionally defined ROI from the two-sample   t -test). -test). In addition, the relationship between rumination and connectivity did not reliably  differ between groups,  Z (30) (30) ¼ 1.46, n.s. This may not be too surprising in that HCs also ruminate; they just do so to a lesser extent, so the relationship between connectivity and rumination should be the same for MDDs and HCs. One problem with using the full RRS is that it contains items that assess depressive severity; therefore, our rumination results may be driven by depression severity and not rumination (Treynor  et al ., ., 2003). To control for this, connectivity scores from the ROI drawn from (Zahn  et al ., ., 2009)

were correlated with the brooding component of the RRS, which does not contain items related to depression, yielding a significant positive correlation (r ¼ 0.44,  P < 0.05 0.05;; Table 1). Unlike brooding, the reflection sub-component of the RRS did not cor correl relate ate wit with h con connec nectiv tivity ity sco scores res (r ¼ 0.19 0.19,, n.s.; Table 1). Furthermore, when reflection scores were partialed out from the correlation between brooding and connectivity, thee re th rela lati tion onsh ship ip wa wass un unch chan ange ged, d, (r ¼ 0.41,   P < 0.0 0.05). 5). In contrast, when brooding scores were partialed out from the cor correl relati ation on bet betwee ween n ref refle lecti ction on and con connec nectiv tivity ity,, thi thiss relationship became mildly, although not reliably, negative (r ¼ 0. 0.12 12,, n. n.s. s.). ). Th Thes esee tw two o pa part rtia iall co corr rrel elat atio ions ns we were re also al so fo foun und d to be re reli liab ably ly di diff ffer eren entt fr from om on onee an anot othe herr [Z(30) ¼ 2.04,   P < 0.05 0.05], ], which sugge suggests sts that the correlation correlation between PCC and SCC during rest periods is more related to neg negati ative ve for forms ms of rum rumina inatio tion n tha than n to oth other er forms of  self-reflection. Connectivity differences for rest   vs  task epochs

Thesee pa Thes patt tter erns ns of co conn nnec ecti tivi vity ty in th thee de defa faul ultt ne netw twor ork, k, however, howev er, were markedly different different when participants participants were

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u rn  r   a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

Depression, rumination and de default ne network

SCAN (2010)

5 of 8

Bivariate ate correlations correlations for all participants participants for the self-report self-report measu measures res of rumina rumination tion and depres depression sion and the brain conne connectivity ctivity scores during rest and Table 1   Bivari task periods extracted using the ROI from Zahn  et al . (2009) Correlations for all participants

Connectivity rest Pearson R Sig. (two-tailed) Brooding Pearson R Sig. (two-tailed) Reflection Pearson R Sig. (two-tailed) Depression Pearson R Sig. (two-tailed) Full RRS Pearson R Sig. (two-tailed) BDI Pearson R Sig. (two-tailed) Connectivity task Pearson R Sig. (two-tailed)

Connectivity rest

Brooding

Reflection

Depression

Full RRS

BDI

Connectivity Task

1

0.437* 0.016

0.194 0.304

0.554** 0.002

0.518** 0.003

0.557** 0.001

0.562** 0.001

0.437* 0.016

1

0.628** 0.000

0.758** 0.000

0.904** 0.000

0.636** 0.000

0.294 0.115

0.194 0.304

0.628** 0.000

1

0.429* 0.018

0.666** 0.000

0.302 0.105

0.158 0.404

0.554** 0.002

0.758** 0.000

0.429* 0.018

1

0.943** 0.000

0.864** 0.000

0.362* 0.050

0.518** 0.003

0.904** 0.000

0.666** 0.000

0.943** 0.000

1

0.794** 0.000

0.347 0.060

0.557** 0.001

0.636** 0.000

0.302 0.105

0.864** 0.000

0.794** 0.000

1

0.391* 0.033

0.562**

0.294

0.158

0.362*

0.347

0.391*

1

0.001

0.115

0.404

0.050

0.060

0.033

Depression is the subscale of the RRS with the depression-related items. *Correlation is significant at the 0.05 level (two-tailed). **Correlation is significant at the 0.01 level (two-tailed).

engaged in the memory task. A two-way ANOVA conducted on the subgenual-cingulate ROI from Zahn  et al . (2009) was used use d to exp explor loree res restt-   vs   task-relat task-related ed conne connectivi ctivity ty and revealed vea led a sig signif nifica icant nt tas task k (re (rest st   vs   task) task) by gr grou oup p (M (MDD DD vs    HC) HC) int intera eracti ction on [F (1,28) (1,28) ¼ 4.27,   P < 0. 0.05 05;; Fi Figu gure re 3) 3)..

MDDs and HCs did not differ in subgenual-cingulate connectivity,   t (28) (28) ¼ 1.47, n.s., but did differ reliably for rest epochs as expected,   t (28) (28) ¼ 3.15,   P < 0.00 0.005. 5. In addition, the correlation between rumination scores and connectivity between the PCC and SCC during task epochs was not reliable (r ¼ 0.347, n.s.; Table 1) nor was task connectivity reliably  correlated with brooding ( r ¼ 0.29 0.294, 4, n.s.; Table 1), which suggests that being engaged in a task may disrupt the ability  to rumin ruminate ate by distra distracting cting participants participants and there thereby by inter inter-rupting the neural circuit that may mediate rumination. The corre correlatio lation n betw between een resti resting-st ng-state ate conne connectivi ctivity ty and task-related connectivity in the subgenual cingulate was reliably correlated for HCs (r ¼ 0.63,   P < 0.0 0.05) 5),, but was not reliab rel iably ly cor correl relate ated d for MDD MDDss (r ¼ 0.3 0.38, 8, n.s n.s.). .). Whi While le thi thiss intera int eracti ction on was not rel reliab iable le (po (poten tentia tially lly due to a lac lack k of  power), these results suggest that the ruminative connectivity  pattern is more dissimilar for task and rest for MDDs (they  may be ruminating at rest, but may not be able to during task tas k eng engage ageme ment) nt) tha than n HCs (they may be mai mainta ntaini ining ng a similar degree of rumination or mind-wandering throughout) whose conne connectivi ctivity ty patter patterns ns rema remain in simi similar lar for both task and rest epochs.

Compared Compar ed wit with h res rest, t, MDD MDDss dem demons onstra trated ted sig signif nifica icantl ntly  y  reduced connectivity while engaged in a task [ t (14) (14) ¼ 2.87, P < 0.0 0.05] 5].. In con contra trast, st, HCs sho showed wed no rel reliab iable le cha change ngess in connectivi conne ctivity ty for task   vs   rest. rest. Final Finally, ly, durin duringg task epoch epochs, s,

Results could also differ based on medication as Anand et al . (2007) found some differences in connectivity between thee an th ante teri rior or ci cing ngul ulat atee co cort rtex ex (A (ACC CC)) an and d li limb mbic ic ar area eass afte af terr MD MDDs Ds ha had d be been en on me medi dica cati tion on fo forr 6 we week eks. s. Si Sixx of 

Fig. 3  The subgenual-cingulate ROI as defined from Zahn  et al . (2009) demonstrates a task (rest, task)  group (MDD, HC) interaction highlighting a selective difference during periods of quiescence. No differences were found between groups for task epochs. Error bars represent 1 standard error of the group mean.

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u r  n  a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

6 of 8

our MDD MDDss wer weree med medica icated ted,, thu thuss pos poster terior ior-ci -cingu ngulat latee to subgenual subg enual-cin -cingulat gulatee conne connectivi ctivities ties were compa compared red for our medicated and un-medicated groups for task and fixation as well as the interaction between the two for both the independent Zahn  et al . (2009) ROI and the ROI that resulted from our two-sample   t -test. -test. No differences in connectivity  between the medicated and non-medicated participants were found, which mitigates some concerns that our results may  vary based on medication status.

mPFC connectivity

Another analysis was performed to interrogate whether our results may be specific to the PCC as the seed to define connectivity. The other common seed region that is used to define default-mode networks is the mPFC (Fox   et al ., ., 2005;; Grei 2005 Greicius cius   et al ., ., 200 2003). 3). Util Utilizing izing mPFC coordi coordinates nates from fro m Fox   et al . (2 (200 005) 5) fo forr th thee mP mPFC FC (x ¼ 1,   y ¼ 47, z ¼ 4), defau default-n lt-networ etwork k conne connectivi ctivity ty durin duringg fixat fixation ion was calculated with this mPFC seed utilizing the same procedure thatt was imp tha implem lement ented ed wit with h the pos poster terior ior cin cingul gulate ate see seed. d. When the two groups’ connectivity maps were compared at sta standa ndard rd who wholele-bra brain in thr thresh eshold olds, s, no dif differ ferenc ences es wer weree found. At liberal statistical thresholds ( P < 0.0 0.055 for 20 con con-tiguous voxels), MDDs did show more connectivity in other regions implicated in the default network such as the posterior ter ior cin cingul gulate ate.. Whe When n an ROI ana analys lysis is was per perfor formed med within wit hin the PCC PCC,, two cl clust usters ers wer weree dis discov covere ered d in whi which ch MDDs showed more connectivity in this area at corrected thresholds (22 voxels centered at 10, 45, 24 and 11 voxels at 17, 58.3; see ‘Materials and methods’ section). This analysis shows that a similar pattern of results is found using the mPFC as the seed; however, it is not clear whether the same results should be expected with the mPFC as the seed  vs  the posterior cingulate, especially given the connectivity differences that Greicius  et al . (2003) found when using the poster terior ior cin cingul ate as thethe seed see d   vs  showed the mPF mPFC. C. Acc Accord ording ing to Greicius   etgulate al  . (2003), PCC more connectivity  showed d more to hig higher her cor cortic tical al are areas, as, whi while le the mP mPFC FC3 showe connectivity to paralimbic and sub-cortical areas. In fac fact, t, Bar (20 (2009 09)) pro propos poses es tha thatt con constr strain ained ed thi thinki nking ng fosters rumination and may be due to the mPFC exhibiting hyperhyp er-inh inhibi ibitio tion n of the MTL MTL.. A sec second ond ROI analysis analysis was performed within the MTL (consisting of the hippocampus, para pa ra-h -hip ippo poca camp mpus us an and d th thee am amyg ygda dala la). ). Ag Agai ain, n, MD MDDs Ds showed more connectivity in this area at corrected thresholds (20 voxels in the left amygdala centered at 24, 7, 18; 29 vox voxels els in the lef leftt par para-h a-hipp ippoca ocampa mpall gyr gyrus us cen center tered ed at 14,   41, 0; 17 voxels in the right amygdala centered at 14,   3,   21 an and d 18 vo voxe xels ls in th thee ri righ ghtt hi hipp ppoc ocam ampu pus/ s/ parahi par ahippo ppocam campal pal gyr gyrus us cen center tered ed at 31 31,,   24,   18 18;; se seee ‘Materials and methods’ section). This analysis is consistent with wit h the Bar (2 (2009 009)) hyp hypoth othesi esiss tha thatt hyp hyperer-con connec nectiv tivity  ity  3

M. G. Berman et al al..

SCAN (2010)

Greicius et Greicius  et al . (2003) call the mPFC the ventral anterior cingulate cortex (vACC).

(that may be due to hyper-inhibition) between the mPFC and MTL leads to increased rumination. Summary

These data show that the degree of correlation between the posterior poster ior cingu cingulate late and the subge subgenual nual cingulate cingulate is relat related ed to rumination and may distinguish MDDs from HCs, but only  during off-task periods. Furthermore, the relationship between betwe en rumin rumination ation and conne connectiv ctivity ity exist existss   only   during off-ta off -task sk per period iods. s. Las Lastly tly,, bro broodi oding ng sco scores res cor correl relate ated d wit with h off-task off-t ask conne connectivi ctivity, ty, but refl reflectio ection n score scoress did not, which suggests that the thought contents during these off-task periodss wer iod weree neg negati ative ve and not ent entire irely ly dri driven ven by dep depres ressio sion n severity. Therefore, it seems important to explore the relationship between rest and task connectivity when comparing MDDs and HCs. DISCUSSION

Neural hyper Neural hyper-conn -connectiv ectivity ity with the subg subgenual enual-cin -cingulat gulatee seems to exist only during off-task periods for MDDs, and the connectivity of this area with the posterior cingulate is highly related to behavioral assays of depressive rumination. These results build on connectivity differences found in the subgenual cingulate by relating connectivity in that area to psychological processes such as rumination and brooding. These The se res result ultss als also o sho showed wed tha thatt hyp hyperer-con connec nectiv tivity ity at res restt existed when rest periods were joined with task periods, a finding that may closely approximate real-world situations under which people are likely to engage in rumination, e.g. idlee mom idl moment entss at wor work. k. Las Lastly tly,, the these se res result ultss sho showed wed the selectivity of these neural differences (between MDDs and HCs) to off-task periods and that the relationship between rumina rum inativ tivee psy psycho cholog logica icall pro proces cesses ses and con connec nectiv tivity ity is mitigated by engaging in a task. The fact that hyper-connectivity in the subgenual cingulate for MDDs was found only during rest or off-task periods supports research suggestingathat our case, behavioral responsibility for completing task)‘distraction’(in can be effective at temporarily relieving rumination and improving mood (Kross and Ayduk, 2008; for a review see Nolen-Hoeksema et al ., ., 2008). When MDDs engaged in the memory task, they  displayed displ ayed atten attenuated uated leve levels ls of conne connectivi ctivity ty in rumin ruminationationrelated regions. However, when left to their own thoughts, rumi ru mina nati tive ve pr proc oces esse sess we were re en enga gage ged. d. Th This is fi find ndin ingg ha hass important impor tant impli implication cationss for futur futuree neur neuroimag oimaging ing resea research rch and the theory ory-bu -build ilding ing abo about ut the cog cognit nitive ive neu neuros roscie cience nce of  depression. Additionally, when the mPFC was used as a seed, some simila sim ilarr res result ultss wer weree fou found nd whe where re MDD MDDss sho showed wed hyp hypererconnectivi conne ctivity ty in defau default-ne lt-network twork areas (e.g (e.g.. incr increased eased connectivity in the posterior cingulate). It is noteworthy that differences in the MTL were found, where MDDs showed greater connectivity. These hyper-connectivities in the MTL could reflect constrained thinking, which Bar outlines as a mechan mec hanism ism of rum rumina inatio tion. n. Mor Moree wor work k wil willl be nee needed ded to

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u r  n  a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

Depression, rumination and de default ne network

flesh sh out the rel relati ations onship hip bet betwee ween n rum rumina inatio tion n and con con-fle strained thought processing and their relation to neural connectiv nec tivity ity,, but it see seems ms to be a pro promis mising ing ent enterp erpris risee tha thatt could lead to some important therapies. It is tem tempti pting ng try to sep separa arate te dep depres ressio sion n sta status tus fro from m rumination, but this can be quite challenging considering that depression and rumination are highly overlapping constructs. This is evidenced by the high correlation found between rumination scores and BDI scores  r ¼0.79. As such, an alternative explanation of our findings is that the connectivity between the posterior cingulate and the subgenual cingulatee refl gulat reflects ects depre depression ssion sever severity ity more than rumi rumination nation.. While separating depression from rumination is not easily  done, brooding brooding score scoress (whic (which h do not contain depressiondepressionrelated relat ed items items)) corre correlated lated positively positively and relia reliably bly with connectivity, which suggests that this relationship is not driven by depression severity alone. Furthermore, reflection scores did not relate to connectivity scores, which indicate that this networ net work k doe doess not sig signal nal pos positi itive ve tho though ughtt pat patter terns ns in our sample. In sum, depression severity does relate strongly to the connectivity connectivity between the subge subgenual nual cingulate cingulate and the posterior cingulate, but it may do so because of the tight coupling coupli ng bet betwee ween n rum rumina inatio tion n and dep depres ressio sion n as the theyy are highly overlapping constructs. The measure of rumination that was used in our study was a trait measure, measure, namel namelyy how much peopl peoplee rumin ruminate ate in their daily lives. However, it seems reasonable that this trait level measur mea suree wou would ld pre predic dictt sta state te rum rumina inatio tion. n. Fir First, st, as sta stated ted above, our correlations between brooding and connectivity  scores during rest are greater than the correlations between reflection scores and connectivity during rest, thus providing some evidence that the thinking going on during these rest periods is probably not constructive. Second, partialing out reflec ref lectio tion n sco scores res did not aff affect ect the cor correl relati ation on bet betwee ween n broodi bro oding ng and con connec nectiv tivity ity dur during ing res rest, t, ind indica icatin tingg tha thatt other oth er for forms ms of thi thinki nking ng see seem m not to be exp explai lainin ningg the these se data. It would have been difficult for us to ask participants what they were doing during these rest breaks  post hoc , because it would have been difficult for participants to remember the thoughts they were having after the fact. Participants could cou ld hav havee bee been n pro prompt mpted ed wit with h rum rumina inatio tion n que questi stions ons thro th roug ugho hout ut th thee re rest st pe peri riod ods, s, bu butt th then en th thee re rest st pe peri riod odss would not be as unguided. Since our rest periods were unguided (i.e. participants were not induced to ruminate) and interspersed with task epochs, some ecological validity may  have been gained, since participants in the real world are not prompted to ruminate, but do so spontaneously. As a more general point, our data suggest that studying on-task on-tas k behav behavior ior may mitig mitigate ate some differences differences betwe between en MDDs and HCs in that engaging in a task may disrupt rumination. minat ion. Our data showe showed d that the hyper hyper-conn -connectiv ectivities ities that were found at rest disappeared during the task, and the relationship between trait rumination and connectivity  was also eliminated when performing a task. If rumination is a critical component of depression, then studying it may 

SCAN (2010)

7 of 8

require moving to more unguided types of paradigms. This idea is consistent with Raichle’s (2010) suggestion of studying the resting state in both health and disease rather than focus solely on reflexive or ‘on-task’ performance. Importantly, our results build on the results of Greicius et al . (20 (2007 07)) as MDD MDDss dem demons onstra trate te inc increa reased sed def defaul aulttnetwork connectivity with the subgenual cingulate that can be lin linked ked to rum rumina inatio tion, n, but onl onlyy dur during ing ung unguid uided ed res restt periods. Based on these results, ruminative behavioral and psychological processes can be ascribed to these neural differenc fer ences es lin linkin kingg bra brain in and beh behavi avior. or. In sum sum,, sub subgen genual ual-cingul cin gulate ate hyp hyperer-con connec nectiv tivity ity in MDD MDDss was res restri tricte cted d to periods of quiescence and may provide a neural mechanism of ru rumi mina nati tion on/b /bro rood odin ing, g, a de dest stru ruct ctiv ivee fo form rm of mi mind nd-wandering. Conflict of Interest

None declared.

REFERENCES Abler,, B., Hofer, C., Vivi Abler Viviani, ani, R. (200 (2008). 8). Habi Habitual tual emotion regulat regulation ion  –   Neuroreport  strategies baseline brain perfusion. ,  19 of (1),mood (1), 21 4.and deBar, Moshe. and (2009). A cognitive neuroscience hypothesis pression.  Trends in Cognitive Sciences ,  13 (11), 456 – 63. 63. Beck, A.T., Steer, R.A., Brown, G.K. (1996).  Manual for the Beck Depression  Inventory-II . San Antonio, TX: Psychological Corporation. Christoff Chri stoff,, K., Gordon, A.M., Smal Smallwoo lwood, d, J., Smit Smith, h, R., Schooler, Schooler, J.W. (2009). (200 9). Expe Experien rience ce samp sampling ling duri during ng fMRI reve reveals als defau default lt netw network ork and Proceedings of the  executiv exec utivee syst system em cont contribut ributions ions to mind wand wanderin ering. g.   Proceedings National Academy of Sciences ,  106 (21), (21), 8719 – 24. 24. Drevets, W.C., Price, J.L., Furey, M.L. (2008). Brain structural and functional abnormalities in mood disorders: implications for neurocircuitry  models of depression.   Brain Structure & Function ,  213 (1 – 2), 2), 93 – 118. 118. Forman, Form an, S.D., Cohe Cohen, n, J.D., Fitzgerald, Fitzgerald, M., Eddy Eddy,, W.F., Mintun, M.A., Noll,, D.C. (1995). Improved Noll Improved asse assessmen ssmentt of signi significan ficantt acti activatio vation n in functional magnetic-resonance-imaging magnetic-resonance-imaging (fMRI)   –   use of a clust cluster-s er-size ize threshold.   Magnetic Resonance in Medicine ,  33 (5), 636 – 47. 47. Fox, M.D., Snyd Snyder, er, A.Z., Vinc Vincent, ent, J.L., Corbetta, Corbetta, M., van Essen, D.C., Raichle, Raich le, M.E. (2005). The huma human n brain is intri intrinsic nsically ally organized organized into dynamic, anticorrelated functional networks.  Proceedings of the National  Academy of Scien Sciences  ces ,  102 (27), (27), 9673 – 8. 8. Greicius, Greic ius, M.D., Kras Krasnow, now, B., Reiss Reiss,, A.L., Menon, V. (200 (2003). 3). Functional Functional connec con nectiv tivity ity in th thee res restin tingg bra brain: in: A net networ work k ana analys lysis is of th thee de defau fault lt Proceedin dings gs of the Nat Nation ional al Acad Academy emy of Scie Science nces  s , mode hypothesis.   Procee 100 (1), (1), 253 – 8. 8. Greicius, M.D., Flores, B.H., Menon, V., et al. (2007). Resting-state functional connectivity in major depression: Abnormally increased contributions from subgenual cingulate cortex and thalamus.  Biological Psychiatry , 62 (5), (5), 429 – 37. 37. Jenkin Jen kinson son,, M., Ban Bannis nister ter,, P., Br Brady ady,, M., Sm Smith ith,, S. (20 (2002 02). ). Imp Improv roved ed optimization for the robust and accurate linear registration and motion correction of brain images.  NeuroImage ,  17 (2), (2), 825 – 41. 41. Joormann, J. (2005). Inhibition, rumination, and mood regulation in depression. In: Engle, R.W., Sedek, G., von Hecker, U., McIntosh, D.N., editors. Cognitive Cognit ive Limitations Limitations in Aging and Psychopa Psychopathology thology:: Atten Attention, tion, Working  Memory, and Executive Functions . New York: Cambridge University Press,  –  pp.s, 275 Kross, Kros E., 312. Ayduk, Aydu k, O. (200 (2008). 8). Facil Facilitat itating ing adap adaptive tive emot emotional ional analysis: analysis: Distin Dis tingui guishi shing ng dis distan tanced ced-an -analy alysis sis of dep depre ressi ssive ve exp experi erienc ences es fro from m Personality ity and Social Psychol Psychology  ogy  immersed-analysis immersed-an alysis and distraction.   Personal Bulletin ,  34 (7), (7), 924 – 38. 38.

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u r  n  a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0 

 

8 of 8

SCAN (2010)

Kross, E., Davidson, M., Weber, J., Ochsner, K. (2009). Coping with emotions past: the neural bases of regulating affect associated with negative autobiographical autobiographic al memories.  Biological Psychiatry ,  65 (5), (5), 361 – 6. 6. Mason, M.F., Norton, M.I., van Horn, J.D., Wegner, D.M., Grafton, S.T., Macrae Mac rae,, C.N C.N.. (20 (2007 07). ). Wan Wander dering ing min minds: ds: Th Thee def defaul aultt net networ work k and stimulus-independent stimulus-indep endent thought. Science ,  315 (5810), (5810), 393 – 5. 5. Mayberg, H.S., Lozano, A.M., Voon, V., et al. (2005). Deep brain stimulation for treatment-resistant depression.  Neuron ,  45 (5), (5), 651 – 60. 60. Monk, C.S., Peltier, S.J., Wiggins, J.L., et al. (2009). Abnormalities of intrinsic functional connectivity in autism spectrum disorders. NeuroImage , 47 (2), (2), 764 – 72. 72. Nee, Ne e, D.E D.E., ., Jon Jonide ides, s, J., Be Berma rman, n, M.G M.G.. (20 (2007 07). ). Ne Neura urall mec mechan hanism ismss of  proactive interference interference-resolution. -resolution. Neuroimage ,  38 , 740 – 51. 51. Nolen-Ho Nole n-Hoekse eksema, ma, S., Wisc Wisco, o, B.E., Lyubomirsky, Lyubomirsky, S. (200 (2008). 8). Reth Rethinkin inkingg Rumination.  Perspectives on Psychological Science ,  3 (5), 400 – 24. 24. Oppenheim, A.V., Schafer, R.W., Buck, J.R. (1999).   Discrete-Time Signal  Processing,  2nd edn. Upper Saddle River, NJ: Prentice-Hall. Raichle, M.E., MacLeod, A.M., Snyder, A.Z., Powers, W.J., Gusnard, D.A., Shulman, G.L. (2001). A default mode of brain function.  Proceedings of   the National Academy of Scien Sciences  ces ,  98 (2), (2), 676 – 82. 82.

M. G. Berman et al al.. Raichle, M.E. (201 Raichle, (2010). 0). Two views of brain function. function.   Trends Trends in Cogni Cognitive  tive  Sciences ,  14 (4), (4), 180 – 90. 90. Sayette, M.A., Reichle, E.D., Schooler, J.W. (2009). Lost in the sauce: the effec eff ects ts of alc alcoho oholl on min mind d wan wander dering ing..   Psychological Psychological Scienc Science  e ,   20 (6), (6), 747 – 52. 52. Sheline, Y.I., Barch, D.M., Price, J.L., et al. (2009). The default mode network and selfself-refe referent rential ial proc processes esses in depr depressio ession. n.   Proceedi Proceedings ngs of the  National Academy of Sciences ,  106 (6), (6), 1942 – 7. 7. Shulman, G.L., Fiez, J.A., Corbetta, M., et al. (1997). Common blood flow  Journal of    changes chan ges acro across ss visua visuall task tasks. s. 2. Decr Decrease easess in cere cerebral bral cortex.   Journal Cognitive Neuroscience ,  9 (5), (5), 648 – 63. 63. Smith, S.M., Jenkinson, M., Woolrich, M.W., et al. (2004). Advances in functional funct ional and stru structur ctural al MR image analysis and impl implemen ementati tation on as FSL.  NeuroImage ,  23 , S208 – 19. 19. Treynor, W., Gonzalez, R., Nolen-Hoeksema, S. (2003). Rumination reconsidered: A psychometric analysis.   Cognitive Therapy and Research ,  27 (3), (3), 247 – 59. 59. Zahn, R., de Oliveira-Souza, R., Bramati, I., Garrido, G., Moll, J. (2009). Subgenual cingulate activity reflects individual differences in empathic concern.  Neuroscience Letters ,  457 (2), (2), 107 – 10. 10.

D   o w n l    o  a  d   e  d  f   r   o m    s   c   a n .  o x  f    o r   d   j    o  u r  n  a l    s  .  o r   g    a  t    U  n i   v   e r   s  i    t    y   o f   M i    c  h  i    g  a n  o n  S   e  p  t    e m  b   e r  2   0   , 2   0  1   0