Neuropsychology

Asynchrony in Executive Networks Predicts CognitiveSlowing in Multiple SclerosisNicholas A. Hubbard, Joanna L. Hutchison, Monroe P. Turner, Saranya Sundaram, LarryOasay, Diana Robinson, Jeremy Strain, Travis Weaver, Scott L. Davis, Gina M. Remington,Hao Huang, Bharat B. Biswal, John Hart, Jr., Teresa C. Frohman, Elliot M. Frohman, and BartRypmaOnline First Publication, July 6, 2015. http://dx.doi.org/10.1037/neu0000202

CITATIONHubbard, N. A., Hutchison, J. L., Turner, M. P., Sundaram, S., Oasay, L., Robinson, D., Strain,J., Weaver, T., Davis, S. L., Remington, G. M., Huang, H., Biswal, B. B., Hart, J., Jr., Frohman,T. C., Frohman, E. M., & Rypma, B. (2015, July 6). Asynchrony in Executive Networks PredictsCognitive Slowing in Multiple Sclerosis. Neuropsychology. Advance online publication.http://dx.doi.org/10.1037/neu0000202

Asynchrony in Executive Networks Predicts Cognitive Slowing inMultiple Sclerosis

Nicholas A. HubbardUniversity of Texas at Dallas

Joanna L. HutchisonUniversity of Texas at Dallas and University of Texas

Southwestern Medical Center

Monroe P. Turner, Saranya Sundaram, Larry Oasay,Diana Robinson, Jeremy Strain, and Travis Weaver

University of Texas at Dallas

Scott L. DavisSouthern Methodist University

Gina M. Remington and Hao HuangUniversity of Texas Southwestern Medical Center

Bharat B. BiswalNew Jersey Institute of Technology

John Hart Jr.University of Texas at Dallas and University of Texas

Southwestern Medical Center

Teresa C. Frohman and Elliot M. FrohmanUniversity of Texas Southwestern Medical Center

Bart RypmaUniversity of Texas at Dallas and University of Texas Southwestern Medical Center

Objective: Cognitive slowing is a core neuropsychological symptom of Multiple Sclerosis (MS). Weaimed to assess the extent to which cognitive slowing in MS was predicted by changes in dorsolateralprefrontal networks. Method: We assessed patients with relapsing-remitting MS and healthy controls(HCs) on measures of processing speed. Participants underwent a functional MRI while performing aprocessing speed task to allow assessment of task-based connectivity. Results: Patients were slower thanHCs on the processing speed tasks. Patients showed attenuated connectivity between right and leftdorsolateral prefrontal cortex (DLPFC) and task-relevant brain regions compared to HCs during pro-cessing speed task performance. Patients’ connectivity with DLPFC in these group-disparate networksaccounted for significant variability in their performance on processing speed measures administeredboth in and out of the imaging environment. Specifically, patients who had stronger functional connec-tions with DLPFC in group-disparate networks performed faster than patients with weaker connectionswith DLPFC in group-disparate networks. Conclusion: Results suggest that MS-related cognitiveslowing can be accounted for by systemic alterations in executive functional networks.

Keywords: executive networks, prefrontal cortex, processing speed, fMRI, multiple sclerosis

Supplemental materials: http://dx.doi.org/10.1037/neu0000202.supp

Nicholas A. Hubbard, School of Behavioral and Brain Sciences, TheCenter for BrainHealth, University of Texas at Dallas; Joanna L. Hutchi-son, School of Behavioral and Brain Sciences, The Center for BrainHealth,University of Texas at Dallas and Department of Psychiatry, University ofTexas Southwestern Medical Center; Monroe P. Turner, SaranyaSundaram, Larry Oasay, Diana Robinson, Jeremy Strain, and TravisWeaver, School of Behavioral and Brain Sciences, The Center for Brain-Health, University of Texas at Dallas; Scott L. Davis, Department ofApplied Physiology and Wellness, Southern Methodist University; GinaM. Remington, Department of Neurology and Neurotherapeutics and De-partment of Ophthalmology, University of Texas Southwestern MedicalCenter; Hao Huang, Advanced Imaging Research Center and Departmentof Radiology, University of Texas Southwestern Medical Center; Bharat B.Biswal, Department of Biomedical Engineering, New Jersey Institute ofTechnology; John Hart Jr., School of Behavioral and Brain Sciences, The

Center for BrainHealth, University of Texas at Dallas and Department ofPsychiatry and Department of Neurology and Neurotherapeutics, Uni-versity of Texas Southwestern Medical Center; Teresa C. Frohman,Department of Neurology and Neurotherapeutics, University of TexasSouthwestern Medical Center; Elliot M. Frohman, Department of Neu-rology and Neurotherapeutics and Department of Ophthalmology, Uni-versity of Texas Southwestern Medical Center; Bart Rypma, School ofBehavioral and Brain Sciences, The Center for BrainHealth, Universityof Texas at Dallas and Department of Psychiatry, University of TexasSouthwestern Medical Center.

Dr. Hao Huang is now at the Department of Radiology, University ofPennsylvania, Philadelphia, Pennsylvania.

Correspondence concerning this article should be addressed to NicholasA. Hubbard, School of Behavioral and Brain Sciences, The Center forBrainHealth, University of Texas at Dallas, 2200 West Mockingbird Lane,Dallas, TX 75235. E-mail: nicholas.hubbard@utdallas.edu

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

Neuropsychology © 2015 American Psychological Association2015, Vol. 29, No. 4, 000 0894-4105/15/$12.00 http://dx.doi.org/10.1037/neu0000202

1

Neuropsychological dysfunction accompanies approximately65% of cases of Multiple Sclerosis (MS; see Genova, Hillary,Wylie, Rypma, & DeLuca, 2009). It is frequently observed thatprocessing speed, the speed with which individuals can executeelementary cognitive operations (Rypma et al., 2006; Salthouse,1992), is compromised in MS patients (see Rao et al., 2014).Deficits in this basic neuropsychological ability exert deleteriouseffects upon day-to-day functioning and are reflected in morewidespread, higher-order cognitive deficits (e.g., working memoryand reasoning; see Ackerman, Beier, & Boyle, 2002; Rypma et al.,2006; Rypma & Prabhakaran, 2009; Salthouse, 1996; Salthouse &Babcock, 1991; Vernon, 1983). To date, however, only two studieshave examined the neurofunctional basis of processing speed inMS patients (Genova, Sumowski, Chiaravalloti, Voelbel, & De-Luca, 2009; Leavitt, Wylie, Genova, Chiaravalloti, & DeLuca,2012), and it is not known whether MS-related processing speeddeficits are associated with alterations in functional neural net-works. In the current study, we used functional neuroimaging toassess (a) whether MS-related changes existed in dorsolateralprefrontal networks during processing speed task performance and(b) whether MS-related changes in these networks predicted cog-nitive slowing.

In fMRI research, functional connectivity techniques permitobservation of interregional fluctuations of blood-oxygen-leveldependent (BOLD) activity. Anatomical regions, or nodes, are saidto form a functionally connected network when BOLD activityacross nodes shows a high degree of temporal coherence (i.e.,synchrony; Joel, Caffo, van Zijl, & Pekar, 2011). Functional con-nectivity analyses have shown reduced synchronous activity in MSpatients compared to healthy controls (HCs) during both restingbrain states (e.g., Bonavita et al., 2011; Cruz-Gómez, Ventura-Campos, Belenguer, Ávila, & Forn, 2014; Janssen, Boster, Patter-son, Abduljalil, & Prakash, 2013; Roosendaal et al., 2010; Saini etal., 2004) and during task engagement (e.g., Au Duong et al., 2005;Cader, Cifelli, Abu-Omar, Palace, & Matthews, 2006; Passamontiet al., 2009). For instance, Cruz-Gómez and colleagues (2014)demonstrated that cognitively impaired MS patients showed lessfunctional connectivity in resting networks than their unimpairedcohorts. Similarly, Cader and colleagues (2006) showed that, dur-ing working memory task performance, dorsolateral prefrontalcortex (DLPFC) was less functionally connected with other task-relevant regions in MS patients relative to HCs

DLPFC has been characterized as an information-processingcontrol center that mediates executive cognitive processes (e.g.,Curtis & D’Esposito, 2003; Hillary, Genova, Chiaravalloti,Rypma, & DeLuca, 2006; Hubbard, Hutchson et al., 2014; Rypmaet al., 2006; Rypma & Prabhakaran, 2009). This executive centerinteracts with memory, motor, and sensory structures to directthought and action in accordance with internal goal-states. Assuch, DLPFC acts as a hub within a rich network of connections;it receives input from sensory projections and has direct connec-tions with regions relevant to behavioral response (e.g., premotorregions, frontal eye fields, cerebellum, and basal ganglia; seeMiller & Cohen, 2001). Thus, both the magnitude of activity inDLPFC and connectivity with DLPFC have been implicated inprocessing speed and higher order cognitive processes (Hillary etal., 2006; Rypma et al., 2006; Rypma & Prabhakaran, 2009).

One study has examined processing speed performance and themagnitude of BOLD activity in MS patients (Genova, Sumowski

et al., 2009). During processing speed performance, MS patientshad significantly less BOLD activity than HCs in Brodmann’sArea 9 (BA 9), the superior portion of DLPFC. The two groupswere similar in task accuracy, but MS patients were slower torespond compared to HCs (Genova, Sumowski et al., 2009). It isnot yet known whether MS patients show systemic alterations infunctional connectivity with DLPFC during processing speed per-formance. Such alterations could slow information processing inMS (cf. Rypma et al., 2006; Rypma & Prabhakaran, 2009). Thus,we sought to evaluate whether MS-related disparities in DLPFCfunctional connectivity existed and whether these disparities couldpredict MS patients’ processing speed performance.

Processing speed performance is associated with BOLD activityand connectivity in DLPFC (e.g., Rypma et al., 2006; Rypma &Prabhakaran, 2009). MS patients have shown attenuated BOLDactivity in DLPFC compared to controls. However, it is not yetknown whether there are relative connectivity deficits withDLPFC in MS patients compared to HCs during processing speedperformance.

In the current study, we examined processing speed perfor-mance of MS patients and HCs both during scanning and outsideof the imaging environment. We assessed DLPFC (i.e., BA 9)connectivity during DSST performance at the group level. We thenexamined whether MS patients and HCs differed in their extent ofconnectivity with DLPFC. We predicted that MS patients wouldshow reduced functional connectivity with DLPFC during process-ing speed performance compared to HCs. Further, we assessed theextent to which connectivity with DLPFC in group-disparate re-gions (i.e., regions where MS patients and HCs significantly dif-fered) could predict processing speed performance both in and outof the imaging environment. We predicted that MS patients with apattern of DLPFC connectivity less like that of HCs, reflectingreduced executive connections, would show more cognitive slow-ing compared to MS patients with a more typical pattern ofDLPFC connectivity. Finally, several analyses (random seed anal-yses, DLPFC BOLD activity analyses, lesion burden and otherMS-related factor associations with connectivity) were undertakento ensure that our results were specific to DLPFC network dys-function and not due to other MS-related factors.

Method and Materials

Participants and Procedure

Thirty patients with relapsing-remitting MS were recruited forthis study. One patient was excluded for taking psychostimulantmedication immediately prior to scanning (n � 29). We requiredthat patients were at least 1 month past a previous exacerbation anduse of corticosteroids (see Mezei et al., 2013). Twenty-five age-,education-, and sex-matched HCs who responded to advertise-ments distributed throughout the Dallas-Fort Worth Metroplexwere recruited. One HC was excluded for a history of seizures andone was excluded for image registration failure (n � 23). Thus,there were 52 participants in total included in this study (N � 52;see Table 1 for characteristics). Participants were screened forpossible imaging contraindications. Participants included in thestudy had normal or corrected-to-normal vision. Those included inthe study were administered portions of the brief neuropsycholog-ical battery (Rao, 1990) and several other measures as part of a

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

2 HUBBARD ET AL.

larger series of studies (see Table 2), and underwent structural andfunctional imaging. Participants gave informed consent beforeundergoing any procedure. All procedures were approved by thegoverning Institutional Review Board.

Behavioral Measurement

Before scanning, participants completed practice runs of theprocessing speed task (i.e., digit-symbol substitution task [DSST];

Rypma et al., 2006). Practice was discontinued after several suc-cessive, accurate trials. For the actual task, participants completedthree runs (75 trials per run) of an event-related DSST (Rypma etal., 2006). In each trial, participants viewed a key of nine digit–symbol pairs and one probe digit–symbol pair for 4,000 ms (seeFigure 1). Participants were instructed to indicate as quickly andaccurately as possible via left- or right-thumb button-press whetherthe probe digit–symbol pair matched one of the digit–symbol pairsin the key. Intertrial intervals were jittered at 0, 2, 4, and 6 sintervals. Accuracy and reaction time (RT) were recorded. RT wascalculated for both groups only for correct responses. Further, as ameasure of external validity, we examined processing speed per-formance outside of the fMRI environment.

Participants completed a symbol–digit modalities task (Smith,1991) outside of the imaging environment. This task was similar tothe DSST in that participants were given a 9-item symbol–digitkey and asked to complete probe symbol–digit pairs. However, incontrast to the jittered trials and button-press responses of theDSST, the symbol–digit modalities task required oral completionof as many symbol–digit probes as possible within a 90-secondwindow. Two MS patients were not administered this measure(n � 27).

Image Acquisition

Imaging was performed using a Phillips 3Tesla scanner (PhilipsMedical Systems, Best, The Netherlands) with an 8-channelSENSE head coil. Anatomical data were acquired using a T1-weighted magnetization-prepared rapid acquisition of gradientecho (MPRAGE) pulse sequence (Brant-Zawadzki, Gillan, & Nitz,1992). MPRAGE scans were acquired using the following param-eters: 160 slices/volume, sagittal slice orientation, 1 � 1 � 1 mmvoxel, 12° flip angle, echo time (TE) � 3.7 ms, repetition time(TR) � 8.1 ms 256 � 204 matrix, 237-s scan duration. Three runsof functional data were acquired using single-shot, gradient echoplanar sequence with the following parameters: BOLD signal type,TE � 30 ms, TR � 2,000 ms, 39 slices/volume, transverse sliceorientation, 3.43 � 3.43 � 4 mm voxel, 70° flip angle, 64 � 64matrix, 300-s run duration. One T2-fluid attenuated inversionrecovery (FLAIR) image was acquired for T2-FLAIR data wereacquired for 48 participants (nMS � 26; nHC � 22) with thefollowing parameters: TE � 125 ms, TR � 11,000 ms, 33 slices,0 mm slice gap, transverse slice orientation, 1.00 � 1.00 � 5.00mm voxel, 120° refocusing angle, 352 � 212 matrix.

Functional Image Processing

Data were processed in AFNI (Analysis of Functional Neuro-Images; Cox, 1996) using the align_epi_anat.py program. The

Figure 1. Example of digit–symbol substitution task item. Participantswere asked to identify quickly and accurately whether the probe digit-symbol pair (bottom) matched the digit-symbol pair in the key (top).

Table 1Demographics and Patient Characteristics

HCs MS p-value

DemographicsAge 42.13 (2.56) 47.52 (1.64) .084a

Education .820b

High school 5.88% 11.54%Some college 23.53% 23.08%Baccalaureate degree 29.41% 38.46%Graduate degree 23.53% 11.54%Doctorate 17.65% 15.38%

MFIS 19.61 (3.22) 24.40 (2.76) .266c

Sex (% female) 73.91% 82.76% .438d

Patient characteristicsDisease duration — 154.27 (18.63) —EDSS — 2.77 (.36) —Time since last exacerbation — 44.04 (8.50) —Immunomodulation therapy — 82.75% —

Interferon beta — 37.93% —Glatiramer acetate — 10.35% —Natalizumab — 37.93% —

Note. Mean (SEM). Age in years; HCs � healthy controls; MS �multiple sclerosis; MFIS � total Modified Fatigue Impact Score; Diseaseduration in months; EDSS � Expanded Disability Status Scale score; Timesince last exacerbation in months; Immunomodulation therapy � Percentreporting immunomodulation therapy during disease course.a p-value derived from Welch-corrected independent samples t-test with38.69 DOFs (nMS � 29; nHC � 23 reporting). b p-value derived fromPearson �2-test with 4 DOFs (nMS � 26; nHC � 17 reporting). c p-valuederived from Welch-corrected independent samples t-test with 41.47 DOFs(nMS � 25; nHC � 21 reporting). d p-value derived from Pearson �2-testwith 1 DOF (nMS � 29; nHC � 23 reporting).

Table 2Group Neuropsychological Characteristics

MS Patients HC p-value

BC 46.74 (12.49) 55.65 (9.81) .006COWAT 40.00 (12.16) 40.78 (10.77) .810PASAT-3 46.14 (11.98) 49.96 (8.12) .190PASAT-2 36.77 (10.40) 37.68 (9.34) .750a

SRT 40.35 (14.55) 46.23 (12.54) .136b

SPART 19.37 (6.45) 22.17 (5.09) .093TMT-A 27.01 (10.57) 21.07 (5.22) .010TMT-B 63.37 (48.39) 40.98 (9.37) .026

Note. Mean (SEM) group scores on neuropsychological tests. MS �multiple sclerosis; HC � healthy controls; BC � Box Completion Task;COWAT � Controlled Oral Word Association; PASAT � Paced SerialAddition Task; SRT � Selective Reminding Test long-term retentioncomponent; SPART � 10/36 Spatial Recall Test; TMT � Trail MakingTest. p-value attained from Welch-corrected independent samples t-tests.Due to data loss and attrition, the number of observations differed for sometests. For all tests unless otherwise noted observations � 23 HCs and 27MS patients.a � 22 HCs and 26 MS patients; b � 23 HCs and 26 MS patients.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

3MS, PROCESSING SPEED, EXECUTIVE NETWORKS

functional data were time-shifted. Motion was corrected using arigid-body (6 degrees of freedom), least-squares transformation.This procedure attenuated motion effects by coregistering all vol-umes to the first volume of the first run of the DSST, ensuring thatpossible movements did not result in incongruent voxel matricesbetween separate volumes. Groups did not significantly differ inany of the six rigid-body head motion parameters or total frame-wise displacement (see online Supplemental Material A; Power,Barnes, Snyder, Schlaggar, & Petersen, 2012). The skull wasremoved from the MPRAGE image and the motion-correctedfunctional volumes were aligned to the MPRAGE image using alocal Pearson correlation cost functions with 45 degrees of free-dom, a 45-mm search parameter, and a center of mass adjustment.The functional data were high-pass filtered (0.015625 Hz) remov-ing a large portion of the noise spectrum (� .008 Hz), increasingthe signal-to-noise ratio of these data. Data were also spatiallysmoothed (full width at half maximum � 6 mm) to increase thesignal-to-noise ratio. Further, each MPRAGE was warped to theColin TTN27 template using the AFNI @auto_tlrc program. Sub-sequently, each participant’s functional data were warped to his orher MPRAGE within Colin space using the @auto_tlrc program.Spatial normalization allowed for demarcations of regions of in-terests using standard stereotaxic coordinates (Talairach &Tournoux, 1988) while also maintaining our functional voxel size(3.43 � 3.43 � 4.00 mm).

Regions of interest were drawn automatically from the Colinbrain using the AFNI Talairach Daemon, which provided an un-filled, cortical map of left BA 9 and right BA 9, the superiorconstituent of DLPFC (Brodmann, 1909/2006). These maps weresubsequently fractionized (3dfractionize) to fit our functional im-age voxel size (left BA 9 � 647 voxels, right BA 9 � 663). Thefunctional data were then averaged across the three runs and theaverage time-series was extracted from two seed regions: left andright BA 9.

We examined the extent to which each voxel covaried with theaverage left and right BA 9 time series using voxelwise Pearsonproduct–moment correlations (Biswal, Yetkin, Haughton, & Hyde,1995). These voxelwise correlations were transformed to z-scoresusing Fisher’s z transformation (Fisher, 1915). This procedureyielded right and left BA 9 connectivity z-scores for each voxel.

Lesion Burden Quantification

The skull was completely removed from each participant’sFLAIR image using a semiautomated procedure. Hyperintensitieswere defined as exhibiting FLAIR signal intensity greater than 2SDs above the mean on each slice. These hyperintensities werethen manually delineated as lesions by ruling out spurious voxelsowing to fat signal, motion, ventricular edge effects, or coil sen-sitivity inhomogeneities (Hart Jr. et al., 2013). Hyperintensities incontact with a ventricle were qualified as periventricular lesions(PVLs). Hyperintensities not confluent with the margins of theventricles or a PVL were considered deep white matter lesions(DWMLs). Lesion burden was acquired by adding the number ofvoxels demarcated as PVLs, DWMLs, and the addition of the two(Total white matter lesion [TWML] burden). Interrater agreementof lesion burden was calculated using the Dice ratio of PVL,DWML, and TWML volume estimates by two independent raters(L.O. and J.S.; Dice, 1945):

� �2 | R1R2 |�R1��R2�

where � was determined as twice the overlapping voxels deter-mined as lesions from raters 1 and 2 divided by the sum ofindividual voxels demarcated as lesions by each reviewer (Zhanget al., 2007). � � .70 indicates excellent interrater agreement(Zhang et al., 2007). For PVLs, DWMLs, and TWMLs � equaled.91, .86, .89, respectively.

Measures and Statistical Analyses

Behavioral analyses. Neuropsychological measures werescored according to standard protocols associated with each test(e.g., Rao, 1990). Cognitive impairment status was assessed fromthese tests based upon 2 SDs outside of published norms (seeonline Supplemental Material B; Chiaravalloti et al., 2005; Bor-inga et al., 2001; Tombaugh, 2004). DSST accuracy was calcu-lated as the percent of correct trials. DSST RT per trial averagewas calculated as the time in milliseconds, following stimulusonset, to produce a correct response (see Rypma et al., 2006;Rypma & Prabhakaran, 2009). Processing speed on the symbol-digit modalities test was measured based upon the total number ofcorrect items completed within the 90 s administration time.

DLPFC within-group task-based connectivity. We used asingle-sample t test to assess voxelwise connectivity with left andright BA 9 for MS patients and HCs. To avoid redundant corre-lations, voxels contained within left and right BA 9 were removedfrom their respective connectivity maps in which these regionswere the seeds. Voxel-wise maps were thresholded so that onlyvoxels with a large-effect relationship (i.e., z-score � .549, r �.500; Cohen, 1988) with left or right BA 9 were retained (p �8.0 � 10�11). Clusters were required to be comprised of at least 10contiguous voxels to ensure each node extracted had meaningfulmass (� 470 mm3). This procedure was undertaken for tworeasons. The first was that standard connectivity methods do notnormally filter out the physiologic noise spectrum (e.g., cardiac [1 Hz] and respiratory pulsations [ .25 Hz]), which could result inlow, but statistically significant spurious correlations (e.g., Shmu-eli et al., 2007). Such correlations might also occur from headmotion (e.g., Power et al., 2012). The second reason was that thesethresholds were applied so as to only show top nodes within theDLPFC network (i.e., those that share at least 25% of their time-series variance with DLPFC).

We further assessed whether motion effects, as measured byframewise displacement (see Power et al., 2012 for detailed de-scription), affected voxelwise connectivity with left and right BA9. We utilized the AFNI 3dTcorrMap program to derive averageconnectivity of each voxel with all other voxels in the brain. Thisallowed for an average coefficient of connectivity for left and rightBA 9 with all other voxels in the brain.

DLPFC between-groups task-based connectivity. MS is as-sociated with neural and vascular changes (e.g., D’haeseleer, Cam-bron, Vanopdenbosch, & De Keyser, 2011; De Keyser, Steen,Mostert, & Koch, 2008; Marshall et al., 2014; Trapp & Nave,2008). Further, prior research has shown attenuated BOLD activityin BA 9 for MS patients compared to HCs (Genova, Sumowski etal., 2009). Thus, it is possible that MS patients in this study couldshow attenuated voxelwise correlations with BA 9 compared to

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

4 HUBBARD ET AL.

HCs due to restriction of hemodynamic activity across these timecourses.

Before undertaking group comparisons we sought to ensure thatboth groups had similar BOLD variability within BA 9 time series.Briefly, association measures such as the Pearson product–momentcorrelation rely upon sufficient variability in order to show sub-stantive relationships (i.e., restriction of range; Alexander,Allinger, & Hanges, 1984). Decreases in variability for one groupcompared to another might artificially attenuate these associations(i.e., the estimates of connectivity) for one group compared toanother. Thus, to ensure that possible changes in BOLD connec-tivity were not an artifact of possible MS-related changes in BOLDvariability, we compared changes in BOLD activity across all timepoints, between groups. This BOLD method (cf. Kannurpatti &Biswal, 2008; Kannurpatti, Rypma, & Biswal, 2012) characterizesfluctuations of BOLD activity across the time series and is calcu-lated using the standard deviation of each time series of interest.

For between-groups connectivity analyses, we utilized voxel-wise independent samples t tests to examine the extent to whichMS patients and HCs differed in connectivity with right and leftBA 9. To establish between-groups effects we used cluster-extentthresholding via AFNI’s 3dClustSim program to determine theprobability of finding a significant noise-only cluster given astandard (64 � 64) grid, our specified smoothing kernel (full-width at half maximum � 6 mm), and a p value of p � .005. Giventhese parameters, we used a cluster size of at least 23 contiguousvoxels (k � 23) to correct for familywise error rate (p � .05).

DLPFC network disparity and processing speedperformance. We assessed whether z-connectivity scores ingroup-disparate networks could significantly predict processingspeed performance. To accomplish this, we averaged each indi-vidual’s z-connectivity score for each group-disparate node sepa-rately for left and right BA 9. Thus, each participant had two“group-disparate network” z-connectivity scores, one z-scorequantifying connectivity with left BA 9 and one z-score quantify-ing connectivity with right BA 9. At the group-level, we hadsmaller sample sizes (e.g., nMS � 29; nHC � 23). Least-squaresregression tests can be influenced to a great extent by multivariateoutliers, particularly for such smaller sample sizes. Thus, in testingsubsequent hypotheses we selected a procedure for generating pvalues and rejecting null hypotheses that is more robust thanstandard hypothesis testing procedures to smaller sample sizes andoutlier influences. We used p values derived from a 10,000 itera-tion (B � 10,000) bootstrap resampling procedure (see Effron,1982) for group examinations.

Random seed connectivity. Because MS is associated withconnectivity disparities in general, it is possible that any groupeffects are not unique to connections with BA 9, but rather areindexing brain-wide asynchrony associated with the degree ofpathological insult. Thus, it might be that (a) task-related connec-tivity disparities are not unique to BA 9 networks and (b) thedegree of asynchrony between any brain regions is reflective ofneurological insult in MS, resulting in processing speed deficits(cf. Rao et al., 2014). We tested these alternative hypotheses byselecting a region of interest at random and examining groupdifferences in connectivity with the random seed and whetherpossible group-disparate connections with the random seed couldpredict performance. We used a random number generator to selectour seed region from a pool of 23 BAs not found in the prior

analyses to connect to BA 9. This procedure produced BA 34 (left85 voxels, right 87 voxels), a portion of the entorhinal cortex, forour random seed region.

Results

Behavioral Performance

DSST performance accuracy was not significantly differentbetween MS patients and HCs (MMS � 93.24% [SEM � .010] vs.MHC � 94.97% [.004]), t(36.90) � �1.51, p � .141. MS patientswere significantly slower on the DSST compared to HCs (MMS �1804.00 [61.64] vs. MHC � 1595.11 [57.64]), t(49.86) � 2.42, p �.019. Further, MS patients completed significantly fewer digit–symbol pairs on the symbol–digit modalities task (MMS � 52.22[2.49] vs. MHC � 58.26 [1.44]), t(40.87) � �2.10, p � .042.

DLPFC Within-Group Task-BasedConnectivity Analysis

Table 3 and Figure 2, and Table 4 and Figure 3 illustratesignificant nodes for MS patients and HCs, respectively. Frame-wise displacement was not significantly related to voxelwise con-nectivity with left (r � �.01) or right (r � �.04) BA 9 at thesample level. This relationship further failed to reach significance

Table 3Digit-Symbol Substitution Task Multiple Sclerosis Patients’Network Connectivity With BA 9

Anatomical region (BA) x y z Voxels t-value

Left BA 9Right medial frontal gyrus (9) �02 �47 �20 1,827 20.45Left caudate �05 �02 �00 283 17.69Left posterior cingulate (23) �02 �36 �24 49 11.78Right precuneus (7) �02 �67 �40 39 14.75Right inferior parietal lobule (40) �43 �39 �52 35 13.59Left uvula �19 �73 �24 28 12.39Right inferior frontal gyrus (46) �46 �40 �08 20 14.73Right inferior parietal lobule (40) �53 �32 �44 18 12.13Right inferior frontal gyrus (47) �53 �16 �04 16 12.58Right postcentral gyrus (3 and 2) �50 �18 �48 15 12.42Left inferior frontal gyrus (47) �46 �26 �12 12 12.89Right middle frontal gyrus (10) �33 �50 �04 12 12.03Left inferior parietal lobule (40) �43 �53 �40 11 12.86

Right BA 9Right middle frontal gyrus (6) �50 �02 �40 1258 18.06Left caudate �12 �16 �04 105 14.09Right caudate �15 �16 �00 90 15.83Left superior parietal lobule (7) �33 �63 �48 80 14.04Left precentral gyrus (4) �53 �05 �44 28 12.63Right inferior frontal gyrus (47) �50 �16 �04 23 13.17Right inferior parietal lobule (40) �46 �36 �40 23 12.28Left precuneus (7) �02 �56 �32 20 12.55Left pyramis �09 �80 �24 14 11.63Right postcentral gyrus (5) �33 �43 �60 14 11.40Right superior frontal gyrus (6) �12 �15 �68 14 12.98Right cingulate gyrus (23) �02 �25 �28 12 11.85

Note. Listed are anatomical regions (Nearest Brodmann’s Area within 5mm), Talaraich coordinates of the peak voxel in each cluster, voxel countsof each cluster, and the peak voxel single-sample t-score of the peak voxelof each cluster. BA 9 � Brodmann’s Area 9. Significance was assessed atp � 8.0 � 10-11, k � 10.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

5MS, PROCESSING SPEED, EXECUTIVE NETWORKS

at the group level (left rMS � �.17; right rMS � � .20; left rHC �.11; right rHC � .10; ps � .05).

DLPFC Between-Groups Task-BasedConnectivity Analysis

MS patients and HCs did not significantly differ in BOLD ineither left BA 9 (MMS � 1.32 [.12] vs. MHC � 1.46 [.13],t(50) � �.81, p � .424) or right BA 9 (MMS � 1.36 [.17] vs.MHC � 1.35 [.18], t(50) � .05, p � .959). Between-groupsconnectivity analyses revealed widespread connectivity disparitiesfor MS patients compared to HCs with both right and left BA 9(Figure 4 and Table 5). In all group-disparate nodes, MS patientsshowed significantly less connectivity with BA 9 compared toHCs. Participant movement, as measured by framewise displace-ment, was not significantly related to group-disparate networkz-connectivity in left (r � �.01) or right (r � .03) BA 9 (ps �.05). This relationship also failed to reach significance at the grouplevel (left rMS � �.12; right rMS � � .20; left rHC � �.02; rightrHC � .20; ps � .05).

DLPFC Network Disparity and ProcessingSpeed Performance

For the entire sample, group-disparate network z-connectivitywith left and right BA 9 significantly predicted DSST RT,�Left � �570.58, t(51) � �2.48, p � .017, �Right � �708.55,t(51) � �2.98, p � .005. We further examined these effects at thegroup level. HCs showed no substantive relationship betweendisparate network z-connectivity in left and right BA 9 and DSSTRT, �Left � 116.92, t(22) � .295, bootstrap p � .754, �Right �45.53, t(22) � 0.10, bootstrap p � .935. However, disparate

network z-connectivity for left and right BA 9 significantly pre-dicted DSST RT for MS patients, �Left � �756.07,t(28) � �1.89, bootstrap p � .029, �Right � �939.72,t(28) � �2.35, bootstrap p � .033 (Figures 5A and B). We furthersought to assess whether disparate network z-connectivity pre-dicted symbol-digit modalities task performance.

Disparate network z-connectivity with left and right BA 9 sig-nificantly predicted symbol–digit modalities task performance inthe entire sample, �Left � 21.55, t(51) � 2.64, p � .011, �Right �29.97, t(51) � 3.78, p � .001. We also assessed these effects at thegroup level. HCs did not show a significant relationship betweendisparate network z-connectivity in left and right BA 9 andsymbol-digit modalities task performance, �Left � �2.09,t(22) � �.21, bootstrap p � .846, �Right � �7.19, t(22) � �0.63,bootstrap p � .542. However, consistent with the DSST RTresults, disparate network z-connectivity with left and right BA 9significantly predicted symbol-digit modalities task perfor-mance for MS patients, �Left � 39.73, t(28) � 2.27, bootstrapp � .033, �Right � 57.40, t(28) � 4.22, bootstrap p � .001(Figures 5C and D).

Random Seed Task-Based Connectivity and ProcessingSpeed Performance

We examined whether there was group-disparate connectivitywith left and right BA 34 using the same methods described forBA 9. MS patients showed attenuated connectivity with left BA 34in a single node compared to HCs (see Table 6). No significantgroup differences were found in connectivity with right BA 34.Using similar procedures as those used with left and right BA 9,connectivity in the group-disparate node with left BA 34 failed tosignificantly predict performance on either DSST or symbol-digitmodalities task performance across the whole sample (ps � .05)and at the group-level (all bootstrap ps � .05).

DLPFC BOLD Activity and ProcessingSpeed Performance

We sought to characterize the extent to which the groups dif-fered in DLPFC BOLD activity and examined the extent to whichMS-related DLPFC BOLD activity could predict processing speedperformance (see online Supplemental Material C for expandedmethods). Our methods yielded task-related peak BOLD activity inpercent signal change from baseline. We found that peak BOLDactivity was significantly attenuated for MS patients compared toHCs in left DLPFC (MMS � .061 [.00] vs. MHC � .136 [.02],t(34.69) � �3.73, p � .001) and right DLPFC (MMS � .056 [.01]vs. MHC � .120 [.02], t(28.42) � �2.67, p � .012). For the entiresample, left and right DLPFC peak BOLD activity significantlypredicted DSST performance �Left � �1184.97, t(50) � �2.07,p � .044, �Right � �1187.45, t(50) � �2.28, p � .027. However,left and right DLPFC peak BOLD activity did not significantlypredict symbol-digit modalities performance (ps � .05). At thegroup level, neither left nor right DLPFC peak BOLD activitysignificantly predicted HCs’ or MS patients’ DSST or symbol-digit modalities performance (all bootstrap ps � .05). These resultsindicated that DLPFC BOLD activity changes in MS patientscould not have mediated the observed relationship between group-disparate DLPFC connectivity and processing speed performance(see Baron & Kenny, 1986).

Figure 2. Multiple sclerosis significant connectivity with left and rightBrodmann’s Area (BA 9). Clusters of association with left (right) and right(left) BA 9. Line represents significant connection with BA 9 (p � 8.0 �10�11, k � 10). Node size reflects voxels retained in node. Number in nodereflects Brodmann’s Area. Node placement centered on proximal peakvoxel, X and Y Talaraich coordinates. c � caudate, py � pyramidis, u �uvula. � right BA 9, � left BA 9, � caudate, � cingulategyrus, � inferior frontal gyrus, � inferior parietal lobule, � me-dial frontal gyrus, � middle frontal gyrus, � precuneus, � post-central gyrus, � posterior cingulate, � pyramis, � superior pari-etal lobule, � uvula.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

6 HUBBARD ET AL.

DLPFC Network Disparity, Lesion Burden, and OtherMS Factors

MS patients had significantly greater DWML (MMS � 2649.33[579.94] vs. MHC � 226.84 [642.47], t(27.44) � 2.79, p � .008),PVL (MMS � 12624.27 [1993.98] vs. MHC � 1541.37 [318.00],t(27.32) 5.49, p � .001), TWML burden (MMS � 15264.60[2568.19] vs. MHC � 1768.21 [403.06], t(27.28) 5.19, p � .001).Lesion burden did not significantly predict group-disparate net-work z-connectivity with left and right BA 9 for HCs or MSpatients (all bootstrapped ps � .05). Clinical associations (i.e.,Disease Duration, EDSS, Time Since Last Exacerbation, or Im-munomodulatory Therapy Status) with left and right BA 9 failed toreach significance (see online Supplemental Material D).

Discussion

In this study, we compared MS patients and HCs on measures ofprocessing speed (DSST and symbol-digit modalities task) and task-based functional connectivity. Consistent with prior research (see Rao

et al., 2014), MS patients had slower processing speed performanceon tasks both in (i.e., DSST) and out (i.e., symbol-digit modalitiestest) of the imaging environment. Within-group functional connectiv-ity results showed that MS patients and HCs both had significantconnections with left and right DLPFC during DSST performance.However, between-groups connectivity analyses revealed that MSpatients had significantly attenuated functional connections with leftand right DLPFC compared to HCs. These nodes of decreased con-nectivity corresponded to regions whose functions are known to beimportant for execution of processing speed tasks: secondary visual(BAs 18 and 19), visual search (BA 7), object recognition (BA 37),motor initiation (declive), motor execution (putamen), pre- and sup-plementary motor (BA 6), and motor (BA 4) regions. Indeed, lessconnectivity with DLPFC in these group-disparate networks resultedin slower processing speed performance for MS patients. These re-sults supported our hypothesis that MS patients with DLPFC connec-tivity patterns more similar to those of HCs (i.e., stronger connectionswith DLPFC in task-relevant regions) would exhibit less cognitiveslowing.

Table 4Digit-Symbol Substitution Task Healthy Controls’ Network Connectivity With BA 9

Anatomical region (BA) x y z Voxels t-value

Left BA 9Left middle frontal gyrus (8) �40 �26 �40 468 22.22Left middle frontal gyrus (46) �40 �26 �20 258 22.05Left precuneus (7) �09 �70 �00 132 22.91Right middle frontal gyrus (6) �26 �06 �00 131 16.59Left middle occipital gyrus (19) �46 �73 �04 62 18.31Right caudate �09 �01 �04 57 15.83Right lingual gyrus (17) �09 �80 �04 45 16.47Right middle frontal gyrus (9) �40 �30 �28 45 18.08Left thalamus �12 �08 �08 44 18.47Left inferior parietal lobule (40) �50 �46 �40 39 15.92Right middle frontal gyrus (10) �33 �50 �04 38 15.82Right precuneus (7) �26 �63 �40 30 14.87Left putamen �23 �16 �04 28 17.95Left middle frontal gyrus (10) �46 �43 �04 28 17.33Left superior temporal gyrus (13) �57 �39 �16 18 19.54Right cingulate gyrus (23) �02 �29 �24 15 14.57

Right BA 9Left middle frontal gyrus (8) �49 �04 �40 564 19.07Right middle frontal gyrus (46) �43 �19 �20 155 20.59Right middle frontal gyrus (6) �36 �06 �52 112 20.93Right superior frontal gyrus (10) �22 �50 �24 95 18.54Left lingual gyrus (17) �02 �84 �04 73 17.12Right putamen �15 �09 �00 70 18.89Left superior temporal gyrus (13) �57 �43 �16 33 19.70Right supramarginal gyrus (40) �57 �46 �32 28 21.45Left precuneus (7) �09 �70 �40 24 16.01Left superior temporal gyrus (38) �50 �16 �12 23 17.48Left superior frontal gyrus (10) �22 �43 �24 21 16.42Left inferior occipital gyrus (19) �46 �77 �04 20 14.86Left middle frontal gyrus (10) �29 �57 �08 20 15.05Right posterior cingulate (23) �02 �32 �24 19 15.93Right middle temporal gyrus (20) �53 �32 �08 16 13.43Left thalamus �12 �05 �08 14 13.30Left inferior parietal lobule (40) �36 �39 �40 12 15.63Left caudate �05 �13 �04 10 14.83Right precuneus (19) �29 �70 �36 10 13.77

Note. Listed are anatomical regions (Nearest Brodmann’s Area within 5 mm), Talaraich coordinates of thepeak voxel in each cluster, voxel counts of each cluster, and the peak voxel single-sample t-score of the peakvoxel of each cluster. BA 9 � Brodmann’s Area 9. Significance was assessed at p � 8.0 � 10-11, k � 10.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

7MS, PROCESSING SPEED, EXECUTIVE NETWORKS

Random seed analyses suggested that MS-related processingspeed deficits were uniquely related to group-disparate networkconnectivity with DLPFC. Results showed that the relationshipbetween MS-related cognitive slowing and DLPFC network con-

nectivity was not mediated by DLPFC BOLD activity. Lesionburden and other MS-related factors (see online SupplementalMaterial) were not significantly related to group-disparate networkconnectivity with DLPFC. These results preclude mediation of therelationship between MS-related cognitive slowing and DLPFCnetwork connectivity by these factors (see Baron & Kenny, 1986).Taken together, these results suggest that the relationship betweenDLPFC network connectivity and processing speed was not con-founded by the MS-related variables assessed here.

DLPFC controls the flow of neural activity through task-specificnetworks in order to execute task goals (cf. Miller & Cohen, 2001).Functional connections with DLPFC are known to predict process-ing speed performance (Biswal, Eldreth, Motes, & Rypma, 2010;Rypma et al., 2006; Rypma & Prabhakaran, 2009). For example,Rypma and colleagues (2006) showed that the number of efferentconnections from DLPFC to task-related regions accounted for asignificant proportion of the variance in participants’ processingspeed performance on the DSST. They postulated that these func-tional connections reflected the integrity of the anatomical (i.e.,white matter) connections from DLPFC and represented the degreeto which participants could engage executive control over task-related regions (Rypma et al., 2006; Rypma & Prabhakaran, 2009).The present results support this hypothesis (see also Zhu, Johnson,Kim, and Gold, 2015).

During both the DSST and symbol– digit modalities tasks,participants visually scanned and processed a set of digits andsymbols and adjudicated a correct response as quickly as pos-sible. In order for DLPFC to mediate efficient performance onthese tasks, it must receive input from relevant visual areas andin turn modulate activity in relevant association and motor areas(see Miller & Cohen, 2001). In the current study, this was notthe case for MS patients. Compared to HCs, MS patients

Figure 3. Healthy controls significant connectivity with left and rightBrodmann’s Area (BA 9). Clusters of association with left (right) and right(left) BA 9. Line represents significant connection with BA 9 (p � 8.0 �10�11, k � 10). Node size reflects voxels retained in node. Number in nodereflects Brodmann’s Area. Node placement centered on proximal peakvoxel, X and Y Talaraich coordinates. c � caudate, p � putamen, t �thalamus, � right BA 9, � left BA 9, � caudate, � cingulategyrus, � inferior occipital gyrus, � inferior parietal lobule, � lin-gual gyrus, � middle frontal gyrus, � middle occipital gyrus,

� middle temporal gyrus, � precuneus, � posterior cingulate,� putamen, � supramarginal gyrus, � superior frontal gyrus,� superior temporal gyrus, � thalamus.

Figure 4. Group differences in connectivity with left and right Brodma-nn’s Area (BA 9). Significant differences in group connectivity z-scoreswith left (right) and right (left) BA 9. In all clusters multiple sclerosis �healthy controls. Red dashed lines represent MS � HC, p � .005, k � 23,familywise error rate � .05. Node size reflects voxels retained in node.Number in node reflects Brodmann’s Area. Node placement centered onproximal peak voxel, X and Y Talaraich coordinates. d � declive, p �putamen. � right BA 9, } � left BA 9, � cuneus, � declive,

� inferior frontal gyrus, � inferior occipital gyrus, � inferiortemporal gyrus, � lingual gyrus, � middle occipital gyrus, � mid-dle temporal gyrus, � precuneus, � precentral gyrus, � putamen,

� superior frontal gyrus, � superior parietal lobule.

Table 5Digit-Symbol Substitution Task Group-Disparate NetworkConnectivity With BA 9

Anatomical region (BA) x y z Voxels t-value

Left BA 9Right superior parietal lobule (7) �29 �60 �44 96 4.16Left superior frontal gyrus (6) �15 �06 �64 76 6.28Left middle occipital gyrus (19) �40 �73 �00 57 3.95Right lingual gyrus (18) �05 �70 �00 51 3.43Right middle temporal gyrus (37) �46 �63 �08 37 3.61Right inferior occipital gyrus (18) �43 �84 �12 30 3.87Left declive �05 �73 �20 25 3.66Right precentral gyrus (4) �40 �15 �56 25 3.39

Right BA 9Left superior frontal gyrus (6) �15 �06 �64 193 5.84Right putamen �26 �06 �04 93 4.36Right precentral gyrus (4) �40 �15 �56 63 3.84Left inferior temporal gyrus (37) �53 �56 �04 58 4.25Left inferior frontal gyrus (47) �53 �26 �00 56 5.16Left inferior occipital gyrus (18) �40 �84 �08 45 4.22Right precuneus (7) �05 �56 �48 31 3.82Left cuneus (19) �29 �80 �32 25 3.95

Note. Listed are anatomical regions (Nearest Brodmann’s Area within 5mm), Talaraich coordinates of the peak voxel in each cluster, voxel countsof each cluster, and the peak voxel independent-samples t-score of the peakvoxel of each cluster. BA 9 � Brodmann’s Area 9. Significance wasassessed at p � .005, k � 23, familywise error rate � .05.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

8 HUBBARD ET AL.

showed attenuated functional connections between DLPFC anda network of nodes which are essential for efficient perfor-mance on the DSST and symbol-digit modalities tasks. Thefunctions of these nodes included visual processing, visualsearch, object recognition, movement planning/initiation, motorexecution, pre- and supplementary motor operations, and motoroperations. We suggest and our results supported, that func-tional connectivity with DLPFC reflects executive communica-tion with task-relevant nodes. Decreases in this communicationresulted in increased cognitive slowing for MS patients.

Interactions among anatomically and functionally distinct brainregions give rise to cognition at many levels, including faceperception (e.g., Haxby, Hoffman, & Gobbini, 2000), speech per-ception (e.g., Hickok & Poeppel, 2000), visual information pro-cessing (e.g., Rypma et al., 2006), motor learning (e.g., Penhune &

Steele, 2012), working memory (e.g., Curtis & D’Esposito, 2003),and reasoning (e.g., Shokri-Kojori, Motes, Rypma, & Krawczyk,2012).These processes require that spatially remote nodes activatein concert with one another (e.g., Fox et al., 2005; cf. Geshwind,1965; Hebb, 1949; McClellan, 1985). At the systems level, whitematter integrity is known to predict network synchrony (see Uhl-haas, Roux, Rodriguez, Rotarska-Jagiela, & Singer, 2010). Lesionburden was not significantly related to group-disparate connectiv-ity; however, degradation of white matter microstructure, such asthat incurred by patients with MS, is known to decrease theintegrative properties of neural networks (cf. Honey et al., 2009).Future work should examine the extent to which white mattermicrostructural insult (as measured by diffusion weighted imag-ing) might influence connectivity within executive networks inMS.

In the present study we showed that asynchrony within execu-tive networks was predictive of slower processing speed in MS.This implicates executive neural systems in MS-related cognitiveslowing. The extent to which alterations in executive functioningcan account for the relationship between asynchrony within exec-utive networks and slowed cognition for MS patients remainsunknown. For example, Rypma and colleagues (2006) proposedthat slower processing speed might not just be affected by DLPFCconnectivity, but rather might be the product of increased execu-tive monitoring requirements. Future research should address thisquestion by assessing whether altered executive functioning me-diates the relationship between asynchrony within executive net-works and slowed cognition for MS patients.

1000

1200

1400

1600

1800

2000

2200

2400

0 0.2 0.4 0.6 0.8

DSS

T R

T

Left BA 9 Group-disparate Network Connectivity z-score

1000

1200

1400

1600

1800

2000

2200

2400

-0.1 0.1 0.3 0.5 0.7 0.9

DSS

T R

T

Right BA 9 Group-disparate Network Connectivity z-score

25

35

45

55

65

75

85

0 0.2 0.4 0.6 0.8

Com

plet

ed S

DM

T Pa

irs

Left BA 9 Group-disparate Network Connectivity z-score

25

35

45

55

65

75

85

-0.1 0.1 0.3 0.5 0.7 0.9

Com

plet

ed S

DM

T Pa

irs

Right BA 9 Group-disparate Network Connectivity z-score

C D

A B r = -.34 r = -.41

r = .41 r = .64

Figure 5. Individual differences in MS patients’ connectivity with Brodmann’s Area (BA 9) in group-disparatenetworks and processing speed performance. DSST � digit–symbol substitution task; SDMT � symbol–digitmodalities test. Correlation coefficient based on Pearson product–moment correlation.

Table 6Digit-Symbol Substitution Task (DSST) Group-DisparateNetwork Connectivity With Random Seed

Anatomical region (BA) x y z Voxels t-value

Left BA 34Right supramarginal gyrus (40) �46 �29 �24 37 4.24

Note. Listed are anatomical regions (Nearest Brodmann’s Area (BA)within 5 mm), Talaraich coordinates of the peak voxel in each cluster,voxel counts of each cluster, and the peak voxel independent-samplest-score of the peak voxel of each cluster. Significance was assessed at p �.005, k � 23, familywise error rate � .05.

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

9MS, PROCESSING SPEED, EXECUTIVE NETWORKS

The present work is the first to show that network-specificdysfunction is fundamentally linked to cognitive slowing in MSpatients. We suggest that the degree of functional connectivitywith DLPFC indexes the robustness of executive networks in MSpatients, and thus the speed with which communication can occurwithin these networks. Because numerous higher-order cognitivefunctions, including memory, planning, problem solving, and rea-soning are known to depend upon DLPFC function (e.g., Braver,2012; Hubbard, Hutchison, et al., 2014; Miller & Cohen, 2001;Narayanan et al., 2005; Sreenivasan, Curtis, & D’Esposito, 2014;Shokri-Kojori et al., 2012; Prabhakaran, Rypma, & Gabrieli, 2001;Prabhakaran et al., 2011; Rypma & Prabhakaran, 2009), alterationsto DLPFC-central networks might have broadspread consequencesfor cognitive function in individuals with MS.

References

Ackerman, P. L., Beier, M. E., & Boyle, M. O. (2002). Individual differ-ences in working memory within a nomological network of cognitiveand perceptual speed abilities. Journal of Experimental Psychology:General, 131, 567–589. http://dx.doi.org/10.1037/0096-3445.131.4.567

Alexander, R. A., Allinger, G. M., & Hanges, P. J. (1984). Correcting forrange restriction when population variance is unknown. Applied Psy-chological Measurement, 8, 431– 437. http://dx.doi.org/10.1177/014662168400800407

Au Duong, M. V., Audoin, B., Boulanouar, K., Ibarrola, D., Malikova, I.,Confort-Gouny, S., . . . Ranjeva, J. P. (2005). Altered functional con-nectivity related to white matter changes inside the working memorynetwork at the very early stage of MS. Journal of Cerebral Blood Flowand Metabolism, 25, 1245–1253. http://dx.doi.org/10.1038/sj.jcbfm.9600122

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variabledistinction in social psychological research: Conceptual, strategic, andstatistical considerations. Journal of Personality and Social Psychology,51, 1173–1182. http://dx.doi.org/10.1037/0022-3514.51.6.1173

Biswal, B. B., Eldreth, D. A., Motes, M. A., & Rypma, B. (2010).Task-dependent individual differences in prefrontal connectivity. Cere-bral Cortex, 20, 2188–2197. http://dx.doi.org/10.1093/cercor/bhp284

Biswal, B., Yetkin, F. Z., Haughton, V. M., & Hyde, J. S. (1995). Func-tional connectivity in the motor cortex of resting human brain usingecho-planar MRI. Magnetic Resonance in Medicine, 34, 537–541. http://dx.doi.org/10.1002/mrm.1910340409

Bonavita, S., Gallo, A., Sacco, R., Corte, M. D., Bisecco, A., Docimo, R.,. . . Tedeschi, G. (2011). Distributed changes in default-mode resting-state connectivity in multiple sclerosis. Multiple Sclerosis, 17, 411–422.http://dx.doi.org/10.1177/1352458510394609

Boringa, J. B., Lazeron, R. H. C., Reuling, I. E. W., Adèr, H. J., Pfennings,L., Lindeboom, J., . . . Polman, C. H. (2001). The brief repeatable batteryof neuropsychological tests: Normative values allow application in mul-tiple sclerosis clinical practice. Multiple Sclerosis, 7, 263–267. http://dx.doi.org/10.1177/135245850100700409

Brant-Zawadzki, M., Gillan, G. D., & Nitz, W. R. (1992). MP RAGE: Athree-dimensional, T1-weighted, gradient-echo sequence—Initial expe-rience in the brain. Radiology, 182, 769–775. http://dx.doi.org/10.1148/radiology.182.3.1535892

Braver, T. S. (2012). The variable nature of cognitive control: A dualmechanisms framework. Trends in Cognitive Sciences, 16, 106–113.http://dx.doi.org/10.1016/j.tics.2011.12.010

Brodmann, K. (2006). The principles of comparative localization in thecerebral cortex based on cytoarchitectonics. In K. Brodmann (Ed.),Brodmann’s localization in the cerebral cortex (p. 150). New York, NY:Springer. (Original work published 1909)

Cader, S., Cifelli, A., Abu-Omar, Y., Palace, J., & Matthews, P. M. (2006).Reduced brain functional reserve and altered functional connectivity inpatients with multiple sclerosis. Brain: A Journal of Neurology, 129,527–537. http://dx.doi.org/10.1093/brain/awh670

Chiaravalloti, N., Hillary, F., Ricker, J., Christodoulou, C., Kalnin, A., Liu,W.-C., . . . DeLuca, J. (2005). Cerebral activation patterns duringworking memory performance in multiple sclerosis using FMRI. Journalof Clinical and Experimental Neuropsychology, 27, 33–54. http://dx.doi.org/10.1080/138033990513609

Cohen, J. (1988). Statistical power analysis for the behavioral sciences(2nd ed.). Hillside, NJ: Erlbaum.

Cox, R. W. (1996). AFNI: Software for analysis and visualization offunctional magnetic resonance neuroimages. Computers abd BiomedicalResearch, 29, 162–173.

Cruz-Gómez, Á. J., Ventura-Campos, N., Belenguer, A., Ávila, C., & Forn,C. (2014). The link between resting-state functional connectivity andcognition in MS patients. Multiple Sclerosis, 20, 338–348. http://dx.doi.org/10.1177/1352458513495584

Curtis, C. E., & D’Esposito, M. (2003). Persistent activity in the prefrontalcortex during working memory. Trends in Cognitive Sciences, 7, 415–423. http://dx.doi.org/10.1016/S1364-6613(03)00197-9

Dale, A. M., & Buckner, R. L. (1997). Selective averaging of rapidlypresented individual trials using fMRI. Human Brain Mapping, 5, 329–340. http://dx.doi.org/10.1002/(SICI)1097-0193(1997)5:5�329::AID-HBM1�3.0.CO;2-5

De Keyser, J., Steen, C., Mostert, J. P., & Koch, M. W. (2008). Hypoper-fusion of the cerebral white matter in multiple sclerosis: Possible mech-anisms and pathophysiological significance. Journal of Cerebral BloodFlow and Metabolism, 28, 1645–1651. http://dx.doi.org/10.1038/jcbfm.2008.72

D’haeseleer, M., Cambron, M., Vanopdenbosch, L., & De Keyser, J.(2011). Vascular aspects of multiple sclerosis. The Lancet Neurology,10, 657–666. http://dx.doi.org/10.1016/S1474-4422(11)70105-3

Dice, L. R. (1945). Measures of the amount of ecologic associationbetween species. Ecology, 26, 297–302. http://dx.doi.org/10.2307/1932409

Effron, B. (1982). The jackknife, bootstrap, and other resampling plans(1st ed.). Philadelphia, PA: Society for Industrial and Applied Mathe-matics. http://dx.doi.org/10.1137/1.9781611970319

Fisher, R. A. (1915). Frequency distribution of the values of the correlationcoefficient in samples of an indefinitely large population. Biometrika,10, 507–521.

Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C.,& Raichle, M. E. (2005). The human brain is intrinsically organized intodynamic, anticorrelated functional networks. Proceedings of the Na-tional Academy of Sciences of the United States of America, 102,9673–9678. http://dx.doi.org/10.1073/pnas.0504136102

Genova, H. M., Hillary, F. G., Wylie, G., Rypma, B., & DeLuca, J. (2009).Examination of processing speed deficits in multiple sclerosis usingfunctional magnetic resonance imaging. Journal of the InternationalNeuropsychological Society, 15, 383–393. http://dx.doi.org/10.1017/S1355617709090535

Genova, H. M., Sumowski, J. F., Chiaravalloti, N., Voelbel, G. T., &DeLuca, J. (2009). Cognition in multiple sclerosis: A review of neuro-psychological and fMRI research. Frontiers in Bioscience, 14, 1730–1744. http://dx.doi.org/10.2741/3336

Geschwind, N. (1965). Disconnexion syndromes in animals and man. I.Brain: A Journal of Neurology, 88, 237–294. http://dx.doi.org/10.1093/brain/88.2.237

Hart, J., Jr., Kraut, M. A., Womack, K. B., Strain, J., Didehbani, N., Bartz,E., . . . Cullum, C. M. (2013). Neuroimaging of cognitive dysfunctionand depression in aging retired National Football League players. Jour-nal of the American Medical Association Neurology, 70, 326–335.http://dx.doi.org/10.1001/2013.jamaneurol.340

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

10 HUBBARD ET AL.

Haxby, J. V., Hoffman, E. A., & Gobbini, M. I. (2000). The distributedhuman neural system for face perception. Trends in Cognitive Sciences,4, 223–233. http://dx.doi.org/10.1016/S1364-6613(00)01482-0

Hebb, D. O. (1949). The organization of behavior. New York, NY: Wiley.Hickok, G., & Poeppel, D. (2000). Towards a functional neuroanatomy of

speech perception. Trends in Cognitive Sciences, 4, 131–138. http://dx.doi.org/10.1016/S1364-6613(00)01463-7

Hillary, F. G., Genova, H. M., Chiaravalloti, N. D., Rypma, B., & DeLuca,J. (2006). Prefrontal modulation of working memory performance inbrain injury and disease. Human Brain Mapping, 27, 837–847. http://dx.doi.org/10.1002/hbm.20226

Honey, C. J., Sporns, O., Cammoun, L., Gigandet, X., Thiran, J. P., Meuli,R., & Hagmann, P. (2009). Predicting human resting-state functionalconnectivity from structural connectivity. Proceedings of the NationalAcademy of Sciences of the United States of America, 106, 2035–2040.http://dx.doi.org/10.1073/pnas.0811168106

Hubbard, N. A., Hutchison, J. L., Motes, M. A., Shokri-Kojori, E., Bennett,I. J., Brigante, R. M., . . . Rypma, B. (2014). Central executive dysfunc-tion and deferred prefrontal processing in veterans with Gulf WarIllness. Clinical Psychological Science, 2, 319–327. http://dx.doi.org/10.1177/2167702613506580

Hubbard, N. A., Turner, M. P., Robinson, D. M., Sundaram, S., Oasay, L.,Hutchison, J. L., . . . Rypma, B. (2014). Attenuated BOLD hemody-namic response predicted by degree of white matter insult, slows cog-nition in multiple sclerosis. Multiple Sclerosis Journal, 20, 267.

Iannetti, G. D., & Wise, R. G. (2007). BOLD functional MRI in diseaseand pharmacological studies: Room for improvement? Magnetic Reso-nance Imaging, 25, 978–988. http://dx.doi.org/10.1016/j.mri.2007.03.018

Janssen, A. L., Boster, A., Patterson, B. A., Abduljalil, A., & Prakash, R. S.(2013). Resting-state functional connectivity in multiple sclerosis: Anexamination of group differences and individual differences. Neuropsy-chologia, 51, 2918–2929. http://dx.doi.org/10.1016/j.neuropsychologia.2013.08.010

Joel, S. E., Caffo, B. S., van Zijl, P. C., & Pekar, J. J. (2011). On therelationship between seed-based and ICA-based measures of functionalconnectivity. Magnetic Resonance in Medicine, 66, 644–657. http://dx.doi.org/10.1002/mrm.22818

Kannurpatti, S. S., & Biswal, B. B. (2008). Detection and scaling oftask-induced fMRI-BOLD response using resting state fluctuations.NeuroImage, 40, 1567–1574. http://dx.doi.org/10.1016/j.neuroimage.2007.09.040

Kannurpatti, S. S., Rypma, B., & Biswal, B. B. (2012). Prediction oftask-related BOLD fMRI with amplitude signatures of resting-statefMRI. Frontiers in Systems Neuroscience, 6, 7. http://dx.doi.org/10.3389/fnsys.2012.00007

Leavitt, V. M., Wylie, G., Genova, H. M., Chiaravalloti, N. D., & DeLuca,J. (2012). Altered effective connectivity during performance of an in-formation processing speed task in multiple sclerosis. Multiple Sclerosis,18, 409–417. http://dx.doi.org/10.1177/1352458511423651

Marshall, O., Lu, H., Brisset, J-C., Xu, F., Liu, P., Herbert, J., . . . Ge, Y.(2014). Impaired cerebrovascular reactivity in multiple sclerosis. Jour-nal of the American Medical Association Neurology, 71, 1275–1281.http://dx.doi.org/10.1001/jamaneurol.2014.1668

McClelland, J. L. (1985). Distributed models of cognitive processes. Ap-plications to learning and memory. Annals of the New York Academy ofSciences, 444, 1–9. http://dx.doi.org/10.1111/j.1749-6632.1985.tb37576.x

Menon, R. S., & Kim, S. G. (1999). Spatial and temporal limits in cognitiveneuroimaging with fMRI. Trends in Cognitive Sciences, 3, 207–216.http://dx.doi.org/10.1016/S1364-6613(99)01329-7

Mezei, Z., Olah, L., Kardos, L., Kovacs, R. K., Csiba, L., & Csepany, T.(2013). Cerebrovascular hemodynamic changes in multiple sclerosispatients during head-up tilt table test: Effect of high-dose intravenous

steroid treatment. Journal of Neurology, 260, 2335–2342. http://dx.doi.org/10.1007/s00415-013-6977-0

Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontalcortex function. Annual Review of Neuroscience, 24, 167–202. http://dx.doi.org/10.1146/annurev.neuro.24.1.167

Motes, M. A., & Rypma, B. (2010). Working memory component pro-cesses: Isolating BOLD signal changes. NeuroImage, 49, 1933–1941.http://dx.doi.org/10.1016/j.neuroimage.2009.08.054

Narayanan, N. S., Prabhakaran, V., Bunge, S. A., Christoff, K., Fine, E. M.,& Gabrieli, J. D. (2005). The role of the prefrontal cortex in themaintenance of verbal working memory: An event-related FMRI anal-ysis. Neuropsychology, 19, 223–232. http://dx.doi.org/10.1037/0894-4105.19.2.223

Passamonti, L., Cerasa, A., Liguori, M., Gioia, M. C., Valentino, P.,Nisticò, R., . . . Fera, F. (2009). Neurobiological mechanisms underlyingemotional processing in relapsing-remitting multiple sclerosis. Brain: AJournal of Neurology, 132, 3380–3391. http://dx.doi.org/10.1093/brain/awp095

Penhune, V. B., & Steele, C. J. (2012). Parallel contributions of cerebellar,striatal and M1 mechanisms to motor sequence learning. BehaviouralBrain Research, 226, 579–591. http://dx.doi.org/10.1016/j.bbr.2011.09.044

Power, J. D., Barnes, K. A., Snyder, A. Z., Schlaggar, B. L., & Petersen,S. E. (2012). Spurious but systematic correlations in functional connec-tivity MRI networks arise from subject motion. NeuroImage, 59, 2142–2154. http://dx.doi.org/10.1016/j.neuroimage.2011.10.018

Prabhakaran, V., Rypma, B., & Gabrieli, J. D. (2001). Neural substrates ofmathematical reasoning: A functional magnetic resonance imaging studyof neocortical activation during performance of the necessary arithmeticoperations test. Neuropsychology, 15, 115–127. http://dx.doi.org/10.1037/0894-4105.15.1.115

Prabhakaran, V., Rypma, B., Narayanan, N. S., Meier, T. B., Austin, B. P.,Nair, V. A., . . . Gabrieli, J. D. (2011). Capacity-speed relationships inprefrontal cortex. PLoS ONE, 6, e27504. http://dx.doi.org/10.1371/journal.pone.0027504

Rao, S. M. (1990). A manual for the brief, repeatable battery of neuro-psychological tests in multiple sclerosis. New York, NY: NationalMultiple Sclerosis Society.

Rao, S. M., Martin, A. L., Huelin, R., Wissinger, E., Khankhel, Z., Kim, E.,& Fahrbach, K. (2014). Correlations between MRI and informationprocessing speed in MS: A meta-analysis. Multiple Sclerosis Interna-tional, 2014, Article ID 975803. http://dx.doi.org/10.1155/2014/975803

Roosendaal, S. D., Hulst, H. E., Vrenken, H., Feenstra, H. E., Castelijns,J. A., Pouwels, P. J., . . . Geurts, J. J. (2010). Structural and functionalhippocampal changes in multiple sclerosis patients with intact memoryfunction. Radiology, 255, 595– 604. http://dx.doi.org/10.1148/radiol.10091433

Rypma, B., Berger, J. S., Prabhakaran, V., Bly, B. M., Kimberg, D. Y.,Biswal, B. B., & D’Esposito, M. (2006). Neural correlates of cognitiveefficiency. NeuroImage, 33, 969 –979. http://dx.doi.org/10.1016/j.neuroimage.2006.05.065

Rypma, B., & D’Esposito, M. (2001). Age-related changes in brain–behavior relationships: Evidence from event-related functional MRIstudies. European Journal of Cognitive Psychology, 13, 235–256. http://dx.doi.org/10.1080/09541440042000296

Rypma, B., & Prabhakaran, V. (2009). When less is more and when moreis more: The mediating roles of capacity and speed in brain–behaviorefficiency. Intelligence, 37, 207–222. http://dx.doi.org/10.1016/j.intell.2008.12.004

Saini, S., DeStefano, N., Smith, S., Guidi, L., Amato, M. P., Federico, A.,& Matthews, P. M. (2004). Altered cerebellar functional connectivitymediates potential adaptive plasticity in patients with multiple sclerosis.Journal of Neurology, Neurosurgery, and Psychiatry, 75, 840–846.http://dx.doi.org/10.1136/jnnp.2003.016782

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

11MS, PROCESSING SPEED, EXECUTIVE NETWORKS

Salthouse, T. A. (1992). What do adult age differences in the Digit SymbolSubstitution Test reflect? Journal of Gerontology, 47, P121–P128.http://dx.doi.org/10.1093/geronj/47.3.P121

Salthouse, T. A. (1996). The processing-speed theory of adult age differ-ences in cognition. Psychological Review, 103, 403–428. http://dx.doi.org/10.1037/0033-295X.103.3.403

Salthouse, T. A., & Babcock, R. L. (1991). Decomposing adult agedifferences in working memory. Developmental Psychology, 27, 763–776. http://dx.doi.org/10.1037/0012-1649.27.5.763

Shmueli, K., van Gelderen, P., de Zwart, J. A., Horovitz, S. G., Fukunaga,M., Jansma, J. M., & Duyn, J. H. (2007). Low-frequency fluctuations inthe cardiac rate as a source of variance in the resting-state fMRI BOLDsignal. NeuroImage, 38, 306 –320. http://dx.doi.org/10.1016/j.neuroimage.2007.07.037

Shokri-Kojori, E., Motes, M. A., Rypma, B., & Krawczyk, D. C. (2012).The network architecture of cortical processing in visuo-spatial reason-ing. Scientific Reports, 2, 411. http://dx.doi.org/10.1038/srep00411

Smith, A. (1991). Symbol Digit Modalities Test. Los Angeles, CA: WesternPsychological Services.

Sreenivasan, K. K., Curtis, C. E., & D’Esposito, M. (2014). Revisiting therole of persistent neural activity during working memory. Trends inCognitive Sciences, 18, 82–89. http://dx.doi.org/10.1016/j.tics.2013.12.001

Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of thehuman brain. New York, NY: Thieme Medical Publishers.

Tombaugh, T. N. (2004). Trail Making Test A and B: Normative datastratified by age and education. Archives of Clinical Neuropsychology,19, 203–214. http://dx.doi.org/10.1016/S0887-6177(03)00039-8

Trapp, B. D., & Nave, K. A. (2008). Multiple sclerosis: An immune orneurodegenerative disorder? Annual Review of Neuroscience, 31, 247–269. http://dx.doi.org/10.1146/annurev.neuro.30.051606.094313

Uhlhaas, P. J., Roux, F., Rodriguez, E., Rotarska-Jagiela, A., & Singer, W.(2010). Neural synchrony and the development of cortical networks.Trends in Cognitive Sciences, 14, 72–80. http://dx.doi.org/10.1016/j.tics.2009.12.002

Vernon, P. A. (1983). Speed of information processing and general intel-ligence. Intelligence, 7, 53–70. http://dx.doi.org/10.1016/0160-2896(83)90006-5

Ward, B. D. (1998/2006). Deconvolution analysis of fMRI time series data(software manual]). Retrieved from http://afni.nimh.nih.gov/afni/doc/manual/3dDeconvolve

Zhang, D., Snyder, A. Z., Shimony, J. S., Fox, M. D., & Raichle, M. E.(2010). Noninvasive functional and structural connectivity mapping ofthe human thalamocortical system. Cerebral Cortex, 20, 1187–1194.http://dx.doi.org/10.1093/cercor/bhp182

Zhang, L., Dean, D., Liu, J. Z., Sahgal, V., Wang, X., & Yue, G. H. (2007).Quantifying degeneration of white matter in normal aging using fractaldimension. Neurobiology of Aging, 28, 1543–1555. http://dx.doi.org/10.1016/j.neurobiolaging.2006.06.020

Zhu, Z., Johnson, N. F., Kim, C., & Gold, B. T. (2015). Reduced frontalcortex efficiency is associated with lower white matter integrity in aging.Cerebral Cortex, 25, 138–146. http://dx.doi.org/10.1093/cercor/bht212

Received July 21, 2014Revision received January 29, 2015

Accepted March 5, 2015 �

Thi

sdo

cum

ent

isco

pyri

ghte

dby

the

Am

eric

anPs

ycho

logi

cal

Ass

ocia

tion

oron

eof

itsal

lied

publ

ishe

rs.

Thi

sar

ticle

isin

tend

edso

lely

for

the

pers

onal

use

ofth

ein

divi

dual

user

and

isno

tto

bedi

ssem

inat

edbr

oadl

y.

12 HUBBARD ET AL.