Functional Connectivity Integrating BehavioralDiffusion Tensor Imaging and Functional Magnetic

Resonance Imaging Data Sets

Abigail A Baird Mary K Colvin John D VanHornSouheil Inati and Michael S Gazzaniga

Abstract

amp In the present study we combined 2 types of magneticresonance technology to explore individual differences on atask that required the recognition of objects presented fromunusual viewpoints This task was chosen based on previouswork that has established the necessity of information trans-fer from the right parietal cortex to the left inferior cortex forits successful completion We used reaction times (RTs) tolocalize regions of cortical activity in the superior parietaland inferior frontal regions (blood oxygen level-dependent

[BOLD] response) that were more active with longer responsetimes These regions were then sampled and their signalchange used to predict individual differences in structuralintegrity of white matter in the corpus callosum (using diffu-sion tensor imaging) Results show that shorter RTs (and as-sociated increases in BOLD response) are associated withincreased organization in the splenium of the corpus callosumwhereas longer RTs are associated with increased organiza-tion in the genu amp

INTRODUCTION

Central to research in neuroscience is the explorationof the relationship between function and structure inthe human brain Advances in magnetic resonance tech-nologies have allowed investigators to collect bothstructural and functional data from human subjectsin vivo In terms of cerebral function blood oxygenlevel-dependent functional magnetic resonance imaging(BOLD fMRI) has been used to measure localized corti-cal activity (Kwong et al 1992 Ogawa et al 1992) fMRIsignal intensity relies upon changes in the balance ofoxyhemoglobin to deoxyhemoglobin in the capillary andvenous vascular beds as a function of nearby neuronalactivity and is derived from the complex interactionbetween cerebral blood flow cerebral blood volumeand oxygen consumption rate This technique has be-come a reliable means by which to localize and studycortical function (Sereno 1998)

More recently diffusion tensor imaging (DTI) hasbeen increasingly used to explore and characterize whitematter structure in vivo (see LeBihan 2003) Statedsimply DTI takes advantage of the fact that bipolarmagnetic field gradient pulses cause 3-D displacementof water molecules within a given area termed diffu-sion Anisotropy is a measure that quantifies the extentto which diffusion varies along the sampled axes Frac-

tional anisotropy (FA) provides information about theshape of the diffusion tensor at each voxel Thus itshows the differences between an isotropic diffusion(where the diffusion tensor is represented by a sphere)and a linear diffusion (cigar-shape ellipsoid) Its rangeis between 0 and 1 0 means an isotropic diffusionand 1 means a highly directional diffusion (Basser ampPierpaoli 1996) FA values are thought to reflect mi-crostructural obstacles that might limit diffusion incertain directions Examples of such obstacles includethe integrity of axonal cell membranes the amount andintegrity of myelin around the axons as well as thenumber and size of axons In sum it has been suggestedthat FA in white matter originates roughly from thepresence and coherence of oriented structures In prac-tice higher FA values have been related to increasesin white matter organizationintegrity (see KlingbergVaidya Gabrieli Moseley amp Hedehus 1999)

To date a small number of studies have combinedfMRI and DTI data to examine the structurendashfunctionrelationship Toosy et al (2004) found that FA values inthe optic radiations correlated with BOLD activity in thevisual cortex (in response to a photic stimulation para-digm) The authors interpreted these results as indi-cating that functional activity within cortical regions isconstrained by the structural properties of the anatom-ical connections subserving those regions The coregis-tration of FA and BOLD data has also been used in smallnumber of studies This method established by WerringDartmouth College

D 2005 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 174 pp 687ndash693

et al (1999) has most recently been used to generateseed points for a fiber tracking study of connectivity inthe human motor cortex (Guye et al 2003) SimplyGuye et al (2003) used statistical mappings of BOLDresponse to place seed points for fiber tracking in thewhite matter adjacent to the gray matter activated by amotor task Using this technique the investigators wereable to generate probabilistic connectivity maps fromM1 to a number of regions in the cortex and subcortexFinally it has been reported that FA values in fronto-parietal white matter correlate strongly with BOLD re-sponse (during a working memory task) in closelylocated gray matter in the superior frontal sulcus andinferior parietal lobe (Oleson Nagy Westerberg ampKlingberg 2003) Taken together these previous inves-tigations underscore the relevance and utility of com-bining these types of data

To test a novel method of combining DTI and fMRIdata sets we chose to examine individual differenceson an object recognition task One remarkable aspect ofthe human visual system is the proficiency with whichit is able to recognize objects even when they are pre-sented from unusual visual perspectives Neuropsycho-logical studies of patients with brain damage (Davidoffamp Warrington 1999 Warrington amp Taylor 1973) as wellas functional neuroimaging studies (Sugio et al 1999Kosslyn et al 1994) have reliably established the role ofthe right hemisphere particularly right parietal cortexin recognizing objects viewed from an unusual orien-tation (Figure 1) In contrast the left parietal cortexhas been shown to be involved in recognizing objectspresented in canonical representations (Warrington ampTaylor 1973) and the left inferior frontal cortex inobject naming (for a review see Humphreys Price ampRiddoch 1999) Thus naming objects presented in aprototypical perspective is mediated by the left hemi-sphere and does not require interhemispheric integra-tion whereas naming objects presented in an unusualorientation does require interhemispheric integration(Warrington amp James 1988 Warrington amp Taylor1978) Although there are likely multiple routes of in-terhemispheric communication involved in object rec-ognition (Kosslyn et al 1994 Warrington amp James1988) the pathway necessary for recognizing objects

presented in unconventional views is thought to passthrough the splenium of the corpus callosum a regionknown to connect right and left parietal regions Pa-tients with selective lesions to this posterior callosalregion can normally name objects when they are pre-sented in conventional views but are impaired when theobjects are presented in unusual orientations (Rudge ampWarrington 1991)

In the current experiment we asked whether thedifference between the time required by participants toname objects presented in either prototypical or un-usual perspective could predict individual differencesin BOLD response in regions previously demonstratedto directly relate to performance on this task Furtherwe investigated whether these individual differences inreaction time (RT) and accompanying BOLD responsecould reveal anything about functional connectivityand information transfer within the intact brain Spe-cifically we expected that naming objects presented atunusual perspectives would produce greater differencesin activity within the left and right superior parietalregions as well as the left and right frontal regions Wealso hypothesized that individual differences in corti-cal activity would relate positively to callosal organiza-tion particularly within the splenial region In sum wepredicted that increased behavioral efficiency on thistask (as measured by RT) would be associated with cor-tical activity (as measured by BOLD) and that increasedactivity in multiple cortical regions would be associatedwith greater callosal organization (as measured by FA)

RESULTS

Cortical regions of interest (ROIs) were defined in 15subjects using a correlational analysis based on the timerequired to name objects presented from unusual per-spectives Four regions were selected based on bothanatomical specificity as per our a priori hypothesesand statistical significance right superior parietal cortex(x = 57 y = 54 z = 33) left superior parietal cortex(x = 39 y = 78 z = 39) right inferior frontal cor-tex (x = 45 y = 12 z = 36) and left inferior frontalcortex (x = 45 y = 30 z = 27) (see Table 1) The areaunder the curve was used to determine a mean activa-

Figure 1 Example of object

used in the present study The

object depicted is a spoon

shown in both (A) unusual and(B) usual perspectives

688 Journal of Cognitive Neuroscience Volume 17 Number 4

tion in each region for each subject These values werethen submitted to a second analysis to examine func-tional connectivity

To isolate white matter regions that might underliethe functional connectivity associated with this task weperformed a voxelwise multiple regression analysis inSPM using the changes in BOLD signal measured in the4 cortical ROIs as the independent variables and the FAas the dependent measure This method was used to ex-plore the idea that individual variation in RT and con-comitant cortical BOLD response during the task wouldpredict the extent of white matter organization withinthe pathways that enable communication between theseregions Our results from this analysis revealed a largeand highly significant region (threshold set to p lt 005uncorrected) within the splenium area of the corpuscallosum a region known to connect the right and leftparietal cortices This region was significantly associatedwith relatively faster RTs and relatively less BOLD re-sponse When the directionality of the regression wasreversed such that significant cortical regions were as-sociated with longer RTs and increased BOLD responsean area in the genu portion of the callosum was observed(threshold set to p lt 005 uncorrected) This region ofthe corpus callosum has been consistently implicatedin information transfer between the frontal lobes (seeFigure 2AB and Table 2)

The time required to recognize objects at unusualviews was also used in a simple regression analysis to in-vestigate regions of white matter integrity that might bepredicted by individual differences in RTs (see Table 2)When the statistical threshold was set at p lt 005 nosignificant regions were observed however when thethreshold was lowered to p lt 1 it became evidentthat RT was related to differences in regions of the cor-pus callosum very similar to those detected by our mul-tiple regression analysis

DISCUSSION

The present study set out to examine the feasibility ofcombining fMRI and DTI data sets Individual differencesin RT on a task requiring subjects to name objects

presented from an unusual perspective were used todetermine cortical activity Longer RTs were found to beassociated with increases in both frontal and parietalcortices Statistical analysis yielded 4 cortical regions ofactivity (in the frontal and parietal cortices) from whichindividual activation values were extracted using an ROIanalysis These values were then used as independentvariables in a voxelwise multiple regression analysiswhere each subjectrsquos FA values were the dependentmeasure This analysis yielded a region in the spleniumof the corpus callosum that was significantly associatedwith decreased BOLD response in the 4 ROIs Addition-ally a region in the genu of the corpus callosum wasfound to significantly relate to increased BOLD responsein the 4 ROIs

Information transfer across the hemispheres was ex-amined by means of a task known to require both rightparietal and left inferior frontal participation Previousresearch has established that the successful namingof objects viewed from an unusual perspective relieson the functional integrity of both the right parietal(Warrington amp Taylor 1973) and left inferior frontalcortex (Humphreys et al 1999) as well as the presenceof the splenial portion of the corpus callosum (Rudge ampWarrington 1991) Although our results affirm thesefindings they significantly extend the previous litera-ture by demonstrating the functionality of the above-described network in the intact brain Furthermorethere appears to be a behavioral continuum for perfor-mance on this task that is neurologically discernableThe current results suggest that an individualrsquos ability toname objects viewed from unusual perspectives can bepredicted by the efficiency of this network measured byregional cortical activity as indexed by BOLD responseand white matter integrity as indexed by FA measure-ments in the splenium of the corpus callosum

Although previous neuropsychological research hassuggested that an intact splenium is necessary fornaming objects viewed in unusual orientations addi-tional research on the object naming abilities of pa-tients with posterior callosal lesions has providedmixed results Some patients with selective spleniallesions can name objects (Intriligator Henaff amp Michel

Table 1 Regions of Interest

Coordinates

Region of Activation Brodmannrsquos Area x y z Z score Volume (mm 3)

Left parietal 7 39 78 39 260 216

Right parietal 7 57 54 33 275 513

Left frontal 9 45 30 27 286 216

Right frontal 9 45 12 36 247 162

Clusters of 5 or more contiguous voxels whose global maxima meet a Z threshold of 291 p lt 005 uncorrected are reported Coordinates are fromthe atlas of Talairach and Tournoux (1988)

Baird et al 689

2000) whereas others demonstrate complete left hemi-anomia which is not limited to objects shown in un-usual perspectives (for a review see Suzuki et al1998) The fact that increased FA in the anterior callosalregion was significantly associated with increasinglypoor performance on the task (see Figure 2) may re-

f lect the existence of multiple callosal channels be-tween the perceptual identification systems of theright hemisphere and the semantic association systemsof the left hemispheres This would be consistent withGeschwindrsquos (1965) proposal that objectsrsquo rich som-esthetic associations enable object naming via anterior

Figure 2 Z scale statistical

maps illustrating regions of

significance based on FA

Positive Z scores are depictedin warm color scale negative in

cool color scale Results from a

multiple regression analysisisolated the splenium of the

corpus callosum as being

associated with relatively faster

RTs and decreased BOLDresponse In contrast the genu

of the corpus callosum was

significantly associated with

longer RTs and increase BOLDresponse in cortical ROIs on a

task that required individuals to

name objects presented fromunusual perspectives Results

are presented in both the (A)

sagittal and (B) axial planes

Table 2 Individual Data

SubjectNumber

Genu FA(Dimensionless Units) SD

Splenium FA(Dimensionless Units) SD

Unusual RT(msec) SD

Usual RT(msec) SD

1 06065 01148 07554 00936 124959 33360 85076 19040

2 07027 00738 06681 00531 110910 39400 81859 19910

3 07301 01206 07950 00506 92717 26420 68787 15060

4 07742 00905 06974 00504 128273 29470 98132 14590

5 07721 00786 07087 00626 92132 22860 69074 8880

6 07960 01335 08176 00693 123676 30120 89166 41260

7 08073 01032 05993 00730 106009 28450 81092 15850

8 06241 00818 04758 01807 133725 33480 99500 30200

9 03223 00805 04500 01646 120221 9580 91694 11960

10 04804 00743 04066 01426 112544 30810 90148 33880

11 07172 00930 08088 00694 71453 15630 59244 6610

12 07988 00935 06820 00704 107052 26630 77610 12230

13 07912 01040 07392 00750 105319 32760 85654 24960

14 03465 01144 07162 00492 147916 26420 139732 30020

15 06558 00709 07282 00512 107933 30190 81989 22050

Total Mean 06617 00952 06699 00837 112323 27705 86584 20433

Total SD 01606 00193 01302 00432 18767 7345 18320 9910

ROI data were collected per individual from both the splenium and genu of the corpus callosum The anterior region was 2160 mm3 and the posteriorregion was 1701 mm3 both regions contained voxels whose global maxima meet a z threshold of 291 p lt 005

690 Journal of Cognitive Neuroscience Volume 17 Number 4

callosal channels when the splenium is damaged Thisidea is further supported by the work of Sidtis VolpeHoltzman Wilson and Gazzaniga (1981) who demon-strated that following partial posterior callosal commis-surotomy transfer of sensory information between thehemispheres was nonexistent however the transfer ofmore cognitive information (ie semantic and episodicinformation) remained intact The authors concludedthat this preservation of function was the result ofsparing of anterior portions of the callosum during thesectioning The present results lend strong support tothis clearly demonstrating that in the intact brain thereexist multiple routes by which information can travelfrom one hemisphere to the other and that these routesvary in their efficiency In the case where performanceon the naming of objects presented at unusual angleswas slower it is conceivable that individuals were re-cruiting greater cognitive resources (as evidenced by therelatively increased BOLD response) and relying princi-pally on the interaction of the frontal cortices to performthe task Conversely it is likely that individuals whodemonstrate relatively faster performance are able to doso because they are relying more heavily on the transferof information between the parietal cortices and thusmaking relatively greater use of relatively faster percep-tual systems to accomplish the task more efficiently

Conclusion

By using behavior to predict cortical activity which wasthen used to predict white matter integrity these datarepresent an important advance in the available methodsfor combining magnetic resonance data sets to examinefunctional connectivity Traditional methods have usedbehavior (eg RT) to predict BOLD response andor FAacross subjects The present study underscores the utilityof a stepwise process whereby more of the varianceacross individuals was accounted for by using an indexof cortical activity (derived from RT) to index FA Theseresults suggest that behavioral performance on a task isrelated to cortical activity in multiple regions as well asthe integrity of the fibers connecting them Although theinterpretation of these results must remain preliminarythe present findings demonstrate the feasibility of usingunique and convergent neuroimaging techniques toexamine specific behavioral phenomena

METHODS

Participants

Fifteen normal right-handed adults (9 women meanage = 259 years) participated in the experiment Allparticipants gave informed consent Handedness wasassessed using the Edinburgh Handedness Inventory(Oldfield 1971) Participants either received monetarycompensation or volunteered as part of an fMRI work-

shop Data from 3 additional participants were excludedbecause of technical difficulties

Experimental Task

Trial onsets were time-locked to the beginning ofvolume acquisition so that each trial was the length of1 TR (3 sec) During each trial a digital photographof a common object was presented in either a prototypi-cal (usual) or an unusual perspective for 1500 msecStimuli were created based on those used by Warringtonand Taylor (1973) The present study replicated thesestimuli in digital photograph form (instead of linedrawings) On each trial participants immediatelypressed the response button upon determining thename of the object and then said the name aloud Tominimize priming effects the unusual orientation alwayspreceded the usual orientation of the same objectBecause our a priori hypotheses were based on responseto the unusual views analyses were based exclusivelyon data from the unusual view condition

During each of the 2 functional runs 25 usual 25unusual and 24 fixation trials were pseudorandomlyintermixed The fixation trials were included to in-troduce jitter into the time series so that unique esti-mates of the hemodynamic responses for the trial typesof interest could be computed (Ollinger Shulman ampCorbetta 2001)

Behavioral Apparatus and Imaging Parameters

Visual stimuli were presented using an Apple G3 laptopcomputer running PsyScope 125 software (CohenMacWhinney Flatt amp Provost 1993) Stimuli were pro-jected to participants with an Epson (model ELP-7000)LCD projector onto a screen positioned at the head endof the bore and viewed through a mirror mounted onthe head coil Participantsrsquo naming times were recordedusing hand-held fiber-optic buttons interfaced with aPsyScope button box (New Micros Dallas TX)

All images were acquired using a 15-T scanner(General Electric Medical Systems Signa CVNvi LX84Waukesha WI) with a standard head coil Anatomicalimages were acquired using a high-resolution 3-D spoiledgradient recovery sequence (128 sagittal slices TR =77 sec TE = 3 msec flip angle = 158 voxel size = 1 1 12 mm) Functional images were collected in runsusing a gradient spin-echo echo-planar sequence sensi-tive to BOLD contrast (T2) (TR = 3 sec TE = 40 msecflip angle = 908 375 375 mm in-plane resolution)During each functional run 88 volumes of axial images(25 slices 45 mm slice thickness 1-mm skip betweenslices) were acquired

Diffusion tensor images (DTI) were acquired usinga diffusion weighted single-shot spin-echo EPI se-quence using the following parameters TR = 10 secTE 88 msec flip angle 908 slice thickness = 25 mm

Baird et al 691

FOV = 240 mm matrix size = 128 128 8 excita-tions acquisition time (9 min 20 sec) Diffusion weight-ing was performed along 6 independent directions withb value 1000 secmm2 A reference image (b = 0 secmm2) was also acquired

Data Analysis

Behavioral Data

For each trial a time to name the presented object wasrecorded These naming times were grouped by objectorientation (usual or unusual) Within each group onlynaming times within 3 SD from the mean were included(see Table 2) The time required to name usual objectswas used in the final analysis

Diffusion Tensor Imaging Data

Using software written by Dr Inati the averaged dif-fusion weighted images and 6 apparent diffusion co-efficients were calculated from which 6 independentelements of the diffusion tensor were determinedfor each voxel Eigenvalues and eigenvectors of thetensor were calculated using a Jacobi transformationFA was calculated from the eigenvalues as describedby Basser and Pierpaoli (1998) and used for all subse-quent analyses

Functional Magnetic Resonance Imaging Data

fMRI data were analyzed using Statistical ParametricMapping software (SPM99 Wellcome Department ofCognitive Neurology London UK) (Friston et al 1995)For each functional run data were preprocessed toremove sources of noise and artifact Functional datawere corrected for differences in acquisition time be-tween slices for each whole-brain volume realignedwithin and across runs to correct for head movementand coregistered with each participantrsquos anatomical dataFunctional data were then transformed into a standardanatomical space (3-mm isotropic voxels) based on theICBM 152 brain template (Montreal Neurological Insti-tute) which approximates the atlas space of Talairachand Tournoux (1988) Normalized data were then spa-tially smoothed (6 mm full width half maximum) usinga Gaussian kernel Analyses took place at 2 levels Firstwithin-subject activity was examined using a fixed-effects model second a random effects model (Holmesamp Friston 1998) was used to explore the data acrosssubjects

Statistical analyses were performed on individual sub-jects using a general linear model incorporating taskeffects modeled with a canonical hemodynamic re-sponse function and its temporal derivative (Fristonet al 1998) as well as mean linear and quadratic trends

for each of the 2 runs This model was used to computeparameter estimates (b) and t contrast images for eachcomparison (usual and unusual views) at each voxel

To identify regions for which the level of activationacross participants was related to the time required toname objects shown at unusual perspectives a simpleregression analysis was performed on the average im-ages for the unusual view condition Individual contrastimages were submitted to a second-level random-effectsregression analysis to create mean t images (thresholdp lt 005 t = 291 uncorrected for multiple compari-sons) Based on previous reports of increased BOLDresponse during sustained attention and cognitive effort(Coull Frackowiak amp Frith 1998 Barch et al 1997)regions that were increasingly active with longer timeon task were selected An automated peak-search algo-rithm identified the location of peak activations basedon voxelwise t values An extent threshold of 5 contig-uous voxels was applied to activated clusters meetingthe voxel-level threshold

For each participant hemodynamic response func-tions (10 frames long) for each trial type were thenestimated across each ROI using a finite impulse re-sponse formulation of the general linear model (Ollingeret al 2001 Burock amp Dale 2000) The parameterestimates for this model (calculated using the least-squares solution to the general linear model) are esti-mates for the temporally evolving response magnitudeat each of the 10 points in peristimulus time selectivelyaveraged across all occurrences of that peristimulustime interval This approach has recently been imple-mented by Poldrack et al as an add-on toolbox to theSPM analysis software (SPM ROI Toolbox sourceforgenetprojectsspm-toolbox ) This method resulted in asingle value for each region per individual These valueswere later used in a regression analysis to predictindividual differences in diffusion anisotropy

Regression Analysis

Single BOLD values in 4 ROIs (left and right frontal leftand right parietal) per subject were entered as indepen-dent variables in a regression analysis using SPM the FAimages of each subject were used as the dependentmeasure This analysis yielded ROIs that were predictedby BOLD response FA values were sampled per subjectfrom these regions

Acknowledgments

This research was supported in part by the Dartmouth BrainImaging Center the National Institutes of Health and theMcDonnell-Pew Foundation MKC is supported by a graduateresearch fellowship from the National Science Foundation Wethank W Kelley G Wolford T Laroche J Woodward andS Ketay for their assistance and the 2002 Summer Institutein Cognitive Neuroscience fellows who participated in thisproject

692 Journal of Cognitive Neuroscience Volume 17 Number 4

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693

et al (1999) has most recently been used to generateseed points for a fiber tracking study of connectivity inthe human motor cortex (Guye et al 2003) SimplyGuye et al (2003) used statistical mappings of BOLDresponse to place seed points for fiber tracking in thewhite matter adjacent to the gray matter activated by amotor task Using this technique the investigators wereable to generate probabilistic connectivity maps fromM1 to a number of regions in the cortex and subcortexFinally it has been reported that FA values in fronto-parietal white matter correlate strongly with BOLD re-sponse (during a working memory task) in closelylocated gray matter in the superior frontal sulcus andinferior parietal lobe (Oleson Nagy Westerberg ampKlingberg 2003) Taken together these previous inves-tigations underscore the relevance and utility of com-bining these types of data

To test a novel method of combining DTI and fMRIdata sets we chose to examine individual differenceson an object recognition task One remarkable aspect ofthe human visual system is the proficiency with whichit is able to recognize objects even when they are pre-sented from unusual visual perspectives Neuropsycho-logical studies of patients with brain damage (Davidoffamp Warrington 1999 Warrington amp Taylor 1973) as wellas functional neuroimaging studies (Sugio et al 1999Kosslyn et al 1994) have reliably established the role ofthe right hemisphere particularly right parietal cortexin recognizing objects viewed from an unusual orien-tation (Figure 1) In contrast the left parietal cortexhas been shown to be involved in recognizing objectspresented in canonical representations (Warrington ampTaylor 1973) and the left inferior frontal cortex inobject naming (for a review see Humphreys Price ampRiddoch 1999) Thus naming objects presented in aprototypical perspective is mediated by the left hemi-sphere and does not require interhemispheric integra-tion whereas naming objects presented in an unusualorientation does require interhemispheric integration(Warrington amp James 1988 Warrington amp Taylor1978) Although there are likely multiple routes of in-terhemispheric communication involved in object rec-ognition (Kosslyn et al 1994 Warrington amp James1988) the pathway necessary for recognizing objects

presented in unconventional views is thought to passthrough the splenium of the corpus callosum a regionknown to connect right and left parietal regions Pa-tients with selective lesions to this posterior callosalregion can normally name objects when they are pre-sented in conventional views but are impaired when theobjects are presented in unusual orientations (Rudge ampWarrington 1991)

In the current experiment we asked whether thedifference between the time required by participants toname objects presented in either prototypical or un-usual perspective could predict individual differencesin BOLD response in regions previously demonstratedto directly relate to performance on this task Furtherwe investigated whether these individual differences inreaction time (RT) and accompanying BOLD responsecould reveal anything about functional connectivityand information transfer within the intact brain Spe-cifically we expected that naming objects presented atunusual perspectives would produce greater differencesin activity within the left and right superior parietalregions as well as the left and right frontal regions Wealso hypothesized that individual differences in corti-cal activity would relate positively to callosal organiza-tion particularly within the splenial region In sum wepredicted that increased behavioral efficiency on thistask (as measured by RT) would be associated with cor-tical activity (as measured by BOLD) and that increasedactivity in multiple cortical regions would be associatedwith greater callosal organization (as measured by FA)

RESULTS

Cortical regions of interest (ROIs) were defined in 15subjects using a correlational analysis based on the timerequired to name objects presented from unusual per-spectives Four regions were selected based on bothanatomical specificity as per our a priori hypothesesand statistical significance right superior parietal cortex(x = 57 y = 54 z = 33) left superior parietal cortex(x = 39 y = 78 z = 39) right inferior frontal cor-tex (x = 45 y = 12 z = 36) and left inferior frontalcortex (x = 45 y = 30 z = 27) (see Table 1) The areaunder the curve was used to determine a mean activa-

Figure 1 Example of object

used in the present study The

object depicted is a spoon

shown in both (A) unusual and(B) usual perspectives

688 Journal of Cognitive Neuroscience Volume 17 Number 4

tion in each region for each subject These values werethen submitted to a second analysis to examine func-tional connectivity

To isolate white matter regions that might underliethe functional connectivity associated with this task weperformed a voxelwise multiple regression analysis inSPM using the changes in BOLD signal measured in the4 cortical ROIs as the independent variables and the FAas the dependent measure This method was used to ex-plore the idea that individual variation in RT and con-comitant cortical BOLD response during the task wouldpredict the extent of white matter organization withinthe pathways that enable communication between theseregions Our results from this analysis revealed a largeand highly significant region (threshold set to p lt 005uncorrected) within the splenium area of the corpuscallosum a region known to connect the right and leftparietal cortices This region was significantly associatedwith relatively faster RTs and relatively less BOLD re-sponse When the directionality of the regression wasreversed such that significant cortical regions were as-sociated with longer RTs and increased BOLD responsean area in the genu portion of the callosum was observed(threshold set to p lt 005 uncorrected) This region ofthe corpus callosum has been consistently implicatedin information transfer between the frontal lobes (seeFigure 2AB and Table 2)

The time required to recognize objects at unusualviews was also used in a simple regression analysis to in-vestigate regions of white matter integrity that might bepredicted by individual differences in RTs (see Table 2)When the statistical threshold was set at p lt 005 nosignificant regions were observed however when thethreshold was lowered to p lt 1 it became evidentthat RT was related to differences in regions of the cor-pus callosum very similar to those detected by our mul-tiple regression analysis

DISCUSSION

The present study set out to examine the feasibility ofcombining fMRI and DTI data sets Individual differencesin RT on a task requiring subjects to name objects

presented from an unusual perspective were used todetermine cortical activity Longer RTs were found to beassociated with increases in both frontal and parietalcortices Statistical analysis yielded 4 cortical regions ofactivity (in the frontal and parietal cortices) from whichindividual activation values were extracted using an ROIanalysis These values were then used as independentvariables in a voxelwise multiple regression analysiswhere each subjectrsquos FA values were the dependentmeasure This analysis yielded a region in the spleniumof the corpus callosum that was significantly associatedwith decreased BOLD response in the 4 ROIs Addition-ally a region in the genu of the corpus callosum wasfound to significantly relate to increased BOLD responsein the 4 ROIs

Information transfer across the hemispheres was ex-amined by means of a task known to require both rightparietal and left inferior frontal participation Previousresearch has established that the successful namingof objects viewed from an unusual perspective relieson the functional integrity of both the right parietal(Warrington amp Taylor 1973) and left inferior frontalcortex (Humphreys et al 1999) as well as the presenceof the splenial portion of the corpus callosum (Rudge ampWarrington 1991) Although our results affirm thesefindings they significantly extend the previous litera-ture by demonstrating the functionality of the above-described network in the intact brain Furthermorethere appears to be a behavioral continuum for perfor-mance on this task that is neurologically discernableThe current results suggest that an individualrsquos ability toname objects viewed from unusual perspectives can bepredicted by the efficiency of this network measured byregional cortical activity as indexed by BOLD responseand white matter integrity as indexed by FA measure-ments in the splenium of the corpus callosum

Although previous neuropsychological research hassuggested that an intact splenium is necessary fornaming objects viewed in unusual orientations addi-tional research on the object naming abilities of pa-tients with posterior callosal lesions has providedmixed results Some patients with selective spleniallesions can name objects (Intriligator Henaff amp Michel

Table 1 Regions of Interest

Coordinates

Region of Activation Brodmannrsquos Area x y z Z score Volume (mm 3)

Left parietal 7 39 78 39 260 216

Right parietal 7 57 54 33 275 513

Left frontal 9 45 30 27 286 216

Right frontal 9 45 12 36 247 162

Clusters of 5 or more contiguous voxels whose global maxima meet a Z threshold of 291 p lt 005 uncorrected are reported Coordinates are fromthe atlas of Talairach and Tournoux (1988)

Baird et al 689

2000) whereas others demonstrate complete left hemi-anomia which is not limited to objects shown in un-usual perspectives (for a review see Suzuki et al1998) The fact that increased FA in the anterior callosalregion was significantly associated with increasinglypoor performance on the task (see Figure 2) may re-

f lect the existence of multiple callosal channels be-tween the perceptual identification systems of theright hemisphere and the semantic association systemsof the left hemispheres This would be consistent withGeschwindrsquos (1965) proposal that objectsrsquo rich som-esthetic associations enable object naming via anterior

Figure 2 Z scale statistical

maps illustrating regions of

significance based on FA

Positive Z scores are depictedin warm color scale negative in

cool color scale Results from a

multiple regression analysisisolated the splenium of the

corpus callosum as being

associated with relatively faster

RTs and decreased BOLDresponse In contrast the genu

of the corpus callosum was

significantly associated with

longer RTs and increase BOLDresponse in cortical ROIs on a

task that required individuals to

name objects presented fromunusual perspectives Results

are presented in both the (A)

sagittal and (B) axial planes

Table 2 Individual Data

SubjectNumber

Genu FA(Dimensionless Units) SD

Splenium FA(Dimensionless Units) SD

Unusual RT(msec) SD

Usual RT(msec) SD

1 06065 01148 07554 00936 124959 33360 85076 19040

2 07027 00738 06681 00531 110910 39400 81859 19910

3 07301 01206 07950 00506 92717 26420 68787 15060

4 07742 00905 06974 00504 128273 29470 98132 14590

5 07721 00786 07087 00626 92132 22860 69074 8880

6 07960 01335 08176 00693 123676 30120 89166 41260

7 08073 01032 05993 00730 106009 28450 81092 15850

8 06241 00818 04758 01807 133725 33480 99500 30200

9 03223 00805 04500 01646 120221 9580 91694 11960

10 04804 00743 04066 01426 112544 30810 90148 33880

11 07172 00930 08088 00694 71453 15630 59244 6610

12 07988 00935 06820 00704 107052 26630 77610 12230

13 07912 01040 07392 00750 105319 32760 85654 24960

14 03465 01144 07162 00492 147916 26420 139732 30020

15 06558 00709 07282 00512 107933 30190 81989 22050

Total Mean 06617 00952 06699 00837 112323 27705 86584 20433

Total SD 01606 00193 01302 00432 18767 7345 18320 9910

ROI data were collected per individual from both the splenium and genu of the corpus callosum The anterior region was 2160 mm3 and the posteriorregion was 1701 mm3 both regions contained voxels whose global maxima meet a z threshold of 291 p lt 005

690 Journal of Cognitive Neuroscience Volume 17 Number 4

callosal channels when the splenium is damaged Thisidea is further supported by the work of Sidtis VolpeHoltzman Wilson and Gazzaniga (1981) who demon-strated that following partial posterior callosal commis-surotomy transfer of sensory information between thehemispheres was nonexistent however the transfer ofmore cognitive information (ie semantic and episodicinformation) remained intact The authors concludedthat this preservation of function was the result ofsparing of anterior portions of the callosum during thesectioning The present results lend strong support tothis clearly demonstrating that in the intact brain thereexist multiple routes by which information can travelfrom one hemisphere to the other and that these routesvary in their efficiency In the case where performanceon the naming of objects presented at unusual angleswas slower it is conceivable that individuals were re-cruiting greater cognitive resources (as evidenced by therelatively increased BOLD response) and relying princi-pally on the interaction of the frontal cortices to performthe task Conversely it is likely that individuals whodemonstrate relatively faster performance are able to doso because they are relying more heavily on the transferof information between the parietal cortices and thusmaking relatively greater use of relatively faster percep-tual systems to accomplish the task more efficiently

Conclusion

By using behavior to predict cortical activity which wasthen used to predict white matter integrity these datarepresent an important advance in the available methodsfor combining magnetic resonance data sets to examinefunctional connectivity Traditional methods have usedbehavior (eg RT) to predict BOLD response andor FAacross subjects The present study underscores the utilityof a stepwise process whereby more of the varianceacross individuals was accounted for by using an indexof cortical activity (derived from RT) to index FA Theseresults suggest that behavioral performance on a task isrelated to cortical activity in multiple regions as well asthe integrity of the fibers connecting them Although theinterpretation of these results must remain preliminarythe present findings demonstrate the feasibility of usingunique and convergent neuroimaging techniques toexamine specific behavioral phenomena

METHODS

Participants

Fifteen normal right-handed adults (9 women meanage = 259 years) participated in the experiment Allparticipants gave informed consent Handedness wasassessed using the Edinburgh Handedness Inventory(Oldfield 1971) Participants either received monetarycompensation or volunteered as part of an fMRI work-

shop Data from 3 additional participants were excludedbecause of technical difficulties

Experimental Task

Trial onsets were time-locked to the beginning ofvolume acquisition so that each trial was the length of1 TR (3 sec) During each trial a digital photographof a common object was presented in either a prototypi-cal (usual) or an unusual perspective for 1500 msecStimuli were created based on those used by Warringtonand Taylor (1973) The present study replicated thesestimuli in digital photograph form (instead of linedrawings) On each trial participants immediatelypressed the response button upon determining thename of the object and then said the name aloud Tominimize priming effects the unusual orientation alwayspreceded the usual orientation of the same objectBecause our a priori hypotheses were based on responseto the unusual views analyses were based exclusivelyon data from the unusual view condition

During each of the 2 functional runs 25 usual 25unusual and 24 fixation trials were pseudorandomlyintermixed The fixation trials were included to in-troduce jitter into the time series so that unique esti-mates of the hemodynamic responses for the trial typesof interest could be computed (Ollinger Shulman ampCorbetta 2001)

Behavioral Apparatus and Imaging Parameters

Visual stimuli were presented using an Apple G3 laptopcomputer running PsyScope 125 software (CohenMacWhinney Flatt amp Provost 1993) Stimuli were pro-jected to participants with an Epson (model ELP-7000)LCD projector onto a screen positioned at the head endof the bore and viewed through a mirror mounted onthe head coil Participantsrsquo naming times were recordedusing hand-held fiber-optic buttons interfaced with aPsyScope button box (New Micros Dallas TX)

All images were acquired using a 15-T scanner(General Electric Medical Systems Signa CVNvi LX84Waukesha WI) with a standard head coil Anatomicalimages were acquired using a high-resolution 3-D spoiledgradient recovery sequence (128 sagittal slices TR =77 sec TE = 3 msec flip angle = 158 voxel size = 1 1 12 mm) Functional images were collected in runsusing a gradient spin-echo echo-planar sequence sensi-tive to BOLD contrast (T2) (TR = 3 sec TE = 40 msecflip angle = 908 375 375 mm in-plane resolution)During each functional run 88 volumes of axial images(25 slices 45 mm slice thickness 1-mm skip betweenslices) were acquired

Diffusion tensor images (DTI) were acquired usinga diffusion weighted single-shot spin-echo EPI se-quence using the following parameters TR = 10 secTE 88 msec flip angle 908 slice thickness = 25 mm

Baird et al 691

FOV = 240 mm matrix size = 128 128 8 excita-tions acquisition time (9 min 20 sec) Diffusion weight-ing was performed along 6 independent directions withb value 1000 secmm2 A reference image (b = 0 secmm2) was also acquired

Data Analysis

Behavioral Data

For each trial a time to name the presented object wasrecorded These naming times were grouped by objectorientation (usual or unusual) Within each group onlynaming times within 3 SD from the mean were included(see Table 2) The time required to name usual objectswas used in the final analysis

Diffusion Tensor Imaging Data

Using software written by Dr Inati the averaged dif-fusion weighted images and 6 apparent diffusion co-efficients were calculated from which 6 independentelements of the diffusion tensor were determinedfor each voxel Eigenvalues and eigenvectors of thetensor were calculated using a Jacobi transformationFA was calculated from the eigenvalues as describedby Basser and Pierpaoli (1998) and used for all subse-quent analyses

Functional Magnetic Resonance Imaging Data

fMRI data were analyzed using Statistical ParametricMapping software (SPM99 Wellcome Department ofCognitive Neurology London UK) (Friston et al 1995)For each functional run data were preprocessed toremove sources of noise and artifact Functional datawere corrected for differences in acquisition time be-tween slices for each whole-brain volume realignedwithin and across runs to correct for head movementand coregistered with each participantrsquos anatomical dataFunctional data were then transformed into a standardanatomical space (3-mm isotropic voxels) based on theICBM 152 brain template (Montreal Neurological Insti-tute) which approximates the atlas space of Talairachand Tournoux (1988) Normalized data were then spa-tially smoothed (6 mm full width half maximum) usinga Gaussian kernel Analyses took place at 2 levels Firstwithin-subject activity was examined using a fixed-effects model second a random effects model (Holmesamp Friston 1998) was used to explore the data acrosssubjects

Statistical analyses were performed on individual sub-jects using a general linear model incorporating taskeffects modeled with a canonical hemodynamic re-sponse function and its temporal derivative (Fristonet al 1998) as well as mean linear and quadratic trends

for each of the 2 runs This model was used to computeparameter estimates (b) and t contrast images for eachcomparison (usual and unusual views) at each voxel

To identify regions for which the level of activationacross participants was related to the time required toname objects shown at unusual perspectives a simpleregression analysis was performed on the average im-ages for the unusual view condition Individual contrastimages were submitted to a second-level random-effectsregression analysis to create mean t images (thresholdp lt 005 t = 291 uncorrected for multiple compari-sons) Based on previous reports of increased BOLDresponse during sustained attention and cognitive effort(Coull Frackowiak amp Frith 1998 Barch et al 1997)regions that were increasingly active with longer timeon task were selected An automated peak-search algo-rithm identified the location of peak activations basedon voxelwise t values An extent threshold of 5 contig-uous voxels was applied to activated clusters meetingthe voxel-level threshold

For each participant hemodynamic response func-tions (10 frames long) for each trial type were thenestimated across each ROI using a finite impulse re-sponse formulation of the general linear model (Ollingeret al 2001 Burock amp Dale 2000) The parameterestimates for this model (calculated using the least-squares solution to the general linear model) are esti-mates for the temporally evolving response magnitudeat each of the 10 points in peristimulus time selectivelyaveraged across all occurrences of that peristimulustime interval This approach has recently been imple-mented by Poldrack et al as an add-on toolbox to theSPM analysis software (SPM ROI Toolbox sourceforgenetprojectsspm-toolbox ) This method resulted in asingle value for each region per individual These valueswere later used in a regression analysis to predictindividual differences in diffusion anisotropy

Regression Analysis

Single BOLD values in 4 ROIs (left and right frontal leftand right parietal) per subject were entered as indepen-dent variables in a regression analysis using SPM the FAimages of each subject were used as the dependentmeasure This analysis yielded ROIs that were predictedby BOLD response FA values were sampled per subjectfrom these regions

Acknowledgments

This research was supported in part by the Dartmouth BrainImaging Center the National Institutes of Health and theMcDonnell-Pew Foundation MKC is supported by a graduateresearch fellowship from the National Science Foundation Wethank W Kelley G Wolford T Laroche J Woodward andS Ketay for their assistance and the 2002 Summer Institutein Cognitive Neuroscience fellows who participated in thisproject

692 Journal of Cognitive Neuroscience Volume 17 Number 4

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693

tion in each region for each subject These values werethen submitted to a second analysis to examine func-tional connectivity

To isolate white matter regions that might underliethe functional connectivity associated with this task weperformed a voxelwise multiple regression analysis inSPM using the changes in BOLD signal measured in the4 cortical ROIs as the independent variables and the FAas the dependent measure This method was used to ex-plore the idea that individual variation in RT and con-comitant cortical BOLD response during the task wouldpredict the extent of white matter organization withinthe pathways that enable communication between theseregions Our results from this analysis revealed a largeand highly significant region (threshold set to p lt 005uncorrected) within the splenium area of the corpuscallosum a region known to connect the right and leftparietal cortices This region was significantly associatedwith relatively faster RTs and relatively less BOLD re-sponse When the directionality of the regression wasreversed such that significant cortical regions were as-sociated with longer RTs and increased BOLD responsean area in the genu portion of the callosum was observed(threshold set to p lt 005 uncorrected) This region ofthe corpus callosum has been consistently implicatedin information transfer between the frontal lobes (seeFigure 2AB and Table 2)

The time required to recognize objects at unusualviews was also used in a simple regression analysis to in-vestigate regions of white matter integrity that might bepredicted by individual differences in RTs (see Table 2)When the statistical threshold was set at p lt 005 nosignificant regions were observed however when thethreshold was lowered to p lt 1 it became evidentthat RT was related to differences in regions of the cor-pus callosum very similar to those detected by our mul-tiple regression analysis

DISCUSSION

The present study set out to examine the feasibility ofcombining fMRI and DTI data sets Individual differencesin RT on a task requiring subjects to name objects

presented from an unusual perspective were used todetermine cortical activity Longer RTs were found to beassociated with increases in both frontal and parietalcortices Statistical analysis yielded 4 cortical regions ofactivity (in the frontal and parietal cortices) from whichindividual activation values were extracted using an ROIanalysis These values were then used as independentvariables in a voxelwise multiple regression analysiswhere each subjectrsquos FA values were the dependentmeasure This analysis yielded a region in the spleniumof the corpus callosum that was significantly associatedwith decreased BOLD response in the 4 ROIs Addition-ally a region in the genu of the corpus callosum wasfound to significantly relate to increased BOLD responsein the 4 ROIs

Information transfer across the hemispheres was ex-amined by means of a task known to require both rightparietal and left inferior frontal participation Previousresearch has established that the successful namingof objects viewed from an unusual perspective relieson the functional integrity of both the right parietal(Warrington amp Taylor 1973) and left inferior frontalcortex (Humphreys et al 1999) as well as the presenceof the splenial portion of the corpus callosum (Rudge ampWarrington 1991) Although our results affirm thesefindings they significantly extend the previous litera-ture by demonstrating the functionality of the above-described network in the intact brain Furthermorethere appears to be a behavioral continuum for perfor-mance on this task that is neurologically discernableThe current results suggest that an individualrsquos ability toname objects viewed from unusual perspectives can bepredicted by the efficiency of this network measured byregional cortical activity as indexed by BOLD responseand white matter integrity as indexed by FA measure-ments in the splenium of the corpus callosum

Although previous neuropsychological research hassuggested that an intact splenium is necessary fornaming objects viewed in unusual orientations addi-tional research on the object naming abilities of pa-tients with posterior callosal lesions has providedmixed results Some patients with selective spleniallesions can name objects (Intriligator Henaff amp Michel

Table 1 Regions of Interest

Coordinates

Region of Activation Brodmannrsquos Area x y z Z score Volume (mm 3)

Left parietal 7 39 78 39 260 216

Right parietal 7 57 54 33 275 513

Left frontal 9 45 30 27 286 216

Right frontal 9 45 12 36 247 162

Clusters of 5 or more contiguous voxels whose global maxima meet a Z threshold of 291 p lt 005 uncorrected are reported Coordinates are fromthe atlas of Talairach and Tournoux (1988)

Baird et al 689

2000) whereas others demonstrate complete left hemi-anomia which is not limited to objects shown in un-usual perspectives (for a review see Suzuki et al1998) The fact that increased FA in the anterior callosalregion was significantly associated with increasinglypoor performance on the task (see Figure 2) may re-

f lect the existence of multiple callosal channels be-tween the perceptual identification systems of theright hemisphere and the semantic association systemsof the left hemispheres This would be consistent withGeschwindrsquos (1965) proposal that objectsrsquo rich som-esthetic associations enable object naming via anterior

Figure 2 Z scale statistical

maps illustrating regions of

significance based on FA

Positive Z scores are depictedin warm color scale negative in

cool color scale Results from a

multiple regression analysisisolated the splenium of the

corpus callosum as being

associated with relatively faster

RTs and decreased BOLDresponse In contrast the genu

of the corpus callosum was

significantly associated with

longer RTs and increase BOLDresponse in cortical ROIs on a

task that required individuals to

name objects presented fromunusual perspectives Results

are presented in both the (A)

sagittal and (B) axial planes

Table 2 Individual Data

SubjectNumber

Genu FA(Dimensionless Units) SD

Splenium FA(Dimensionless Units) SD

Unusual RT(msec) SD

Usual RT(msec) SD

1 06065 01148 07554 00936 124959 33360 85076 19040

2 07027 00738 06681 00531 110910 39400 81859 19910

3 07301 01206 07950 00506 92717 26420 68787 15060

4 07742 00905 06974 00504 128273 29470 98132 14590

5 07721 00786 07087 00626 92132 22860 69074 8880

6 07960 01335 08176 00693 123676 30120 89166 41260

7 08073 01032 05993 00730 106009 28450 81092 15850

8 06241 00818 04758 01807 133725 33480 99500 30200

9 03223 00805 04500 01646 120221 9580 91694 11960

10 04804 00743 04066 01426 112544 30810 90148 33880

11 07172 00930 08088 00694 71453 15630 59244 6610

12 07988 00935 06820 00704 107052 26630 77610 12230

13 07912 01040 07392 00750 105319 32760 85654 24960

14 03465 01144 07162 00492 147916 26420 139732 30020

15 06558 00709 07282 00512 107933 30190 81989 22050

Total Mean 06617 00952 06699 00837 112323 27705 86584 20433

Total SD 01606 00193 01302 00432 18767 7345 18320 9910

ROI data were collected per individual from both the splenium and genu of the corpus callosum The anterior region was 2160 mm3 and the posteriorregion was 1701 mm3 both regions contained voxels whose global maxima meet a z threshold of 291 p lt 005

690 Journal of Cognitive Neuroscience Volume 17 Number 4

callosal channels when the splenium is damaged Thisidea is further supported by the work of Sidtis VolpeHoltzman Wilson and Gazzaniga (1981) who demon-strated that following partial posterior callosal commis-surotomy transfer of sensory information between thehemispheres was nonexistent however the transfer ofmore cognitive information (ie semantic and episodicinformation) remained intact The authors concludedthat this preservation of function was the result ofsparing of anterior portions of the callosum during thesectioning The present results lend strong support tothis clearly demonstrating that in the intact brain thereexist multiple routes by which information can travelfrom one hemisphere to the other and that these routesvary in their efficiency In the case where performanceon the naming of objects presented at unusual angleswas slower it is conceivable that individuals were re-cruiting greater cognitive resources (as evidenced by therelatively increased BOLD response) and relying princi-pally on the interaction of the frontal cortices to performthe task Conversely it is likely that individuals whodemonstrate relatively faster performance are able to doso because they are relying more heavily on the transferof information between the parietal cortices and thusmaking relatively greater use of relatively faster percep-tual systems to accomplish the task more efficiently

Conclusion

By using behavior to predict cortical activity which wasthen used to predict white matter integrity these datarepresent an important advance in the available methodsfor combining magnetic resonance data sets to examinefunctional connectivity Traditional methods have usedbehavior (eg RT) to predict BOLD response andor FAacross subjects The present study underscores the utilityof a stepwise process whereby more of the varianceacross individuals was accounted for by using an indexof cortical activity (derived from RT) to index FA Theseresults suggest that behavioral performance on a task isrelated to cortical activity in multiple regions as well asthe integrity of the fibers connecting them Although theinterpretation of these results must remain preliminarythe present findings demonstrate the feasibility of usingunique and convergent neuroimaging techniques toexamine specific behavioral phenomena

METHODS

Participants

Fifteen normal right-handed adults (9 women meanage = 259 years) participated in the experiment Allparticipants gave informed consent Handedness wasassessed using the Edinburgh Handedness Inventory(Oldfield 1971) Participants either received monetarycompensation or volunteered as part of an fMRI work-

shop Data from 3 additional participants were excludedbecause of technical difficulties

Experimental Task

Trial onsets were time-locked to the beginning ofvolume acquisition so that each trial was the length of1 TR (3 sec) During each trial a digital photographof a common object was presented in either a prototypi-cal (usual) or an unusual perspective for 1500 msecStimuli were created based on those used by Warringtonand Taylor (1973) The present study replicated thesestimuli in digital photograph form (instead of linedrawings) On each trial participants immediatelypressed the response button upon determining thename of the object and then said the name aloud Tominimize priming effects the unusual orientation alwayspreceded the usual orientation of the same objectBecause our a priori hypotheses were based on responseto the unusual views analyses were based exclusivelyon data from the unusual view condition

During each of the 2 functional runs 25 usual 25unusual and 24 fixation trials were pseudorandomlyintermixed The fixation trials were included to in-troduce jitter into the time series so that unique esti-mates of the hemodynamic responses for the trial typesof interest could be computed (Ollinger Shulman ampCorbetta 2001)

Behavioral Apparatus and Imaging Parameters

Visual stimuli were presented using an Apple G3 laptopcomputer running PsyScope 125 software (CohenMacWhinney Flatt amp Provost 1993) Stimuli were pro-jected to participants with an Epson (model ELP-7000)LCD projector onto a screen positioned at the head endof the bore and viewed through a mirror mounted onthe head coil Participantsrsquo naming times were recordedusing hand-held fiber-optic buttons interfaced with aPsyScope button box (New Micros Dallas TX)

All images were acquired using a 15-T scanner(General Electric Medical Systems Signa CVNvi LX84Waukesha WI) with a standard head coil Anatomicalimages were acquired using a high-resolution 3-D spoiledgradient recovery sequence (128 sagittal slices TR =77 sec TE = 3 msec flip angle = 158 voxel size = 1 1 12 mm) Functional images were collected in runsusing a gradient spin-echo echo-planar sequence sensi-tive to BOLD contrast (T2) (TR = 3 sec TE = 40 msecflip angle = 908 375 375 mm in-plane resolution)During each functional run 88 volumes of axial images(25 slices 45 mm slice thickness 1-mm skip betweenslices) were acquired

Diffusion tensor images (DTI) were acquired usinga diffusion weighted single-shot spin-echo EPI se-quence using the following parameters TR = 10 secTE 88 msec flip angle 908 slice thickness = 25 mm

Baird et al 691

FOV = 240 mm matrix size = 128 128 8 excita-tions acquisition time (9 min 20 sec) Diffusion weight-ing was performed along 6 independent directions withb value 1000 secmm2 A reference image (b = 0 secmm2) was also acquired

Data Analysis

Behavioral Data

For each trial a time to name the presented object wasrecorded These naming times were grouped by objectorientation (usual or unusual) Within each group onlynaming times within 3 SD from the mean were included(see Table 2) The time required to name usual objectswas used in the final analysis

Diffusion Tensor Imaging Data

Using software written by Dr Inati the averaged dif-fusion weighted images and 6 apparent diffusion co-efficients were calculated from which 6 independentelements of the diffusion tensor were determinedfor each voxel Eigenvalues and eigenvectors of thetensor were calculated using a Jacobi transformationFA was calculated from the eigenvalues as describedby Basser and Pierpaoli (1998) and used for all subse-quent analyses

Functional Magnetic Resonance Imaging Data

fMRI data were analyzed using Statistical ParametricMapping software (SPM99 Wellcome Department ofCognitive Neurology London UK) (Friston et al 1995)For each functional run data were preprocessed toremove sources of noise and artifact Functional datawere corrected for differences in acquisition time be-tween slices for each whole-brain volume realignedwithin and across runs to correct for head movementand coregistered with each participantrsquos anatomical dataFunctional data were then transformed into a standardanatomical space (3-mm isotropic voxels) based on theICBM 152 brain template (Montreal Neurological Insti-tute) which approximates the atlas space of Talairachand Tournoux (1988) Normalized data were then spa-tially smoothed (6 mm full width half maximum) usinga Gaussian kernel Analyses took place at 2 levels Firstwithin-subject activity was examined using a fixed-effects model second a random effects model (Holmesamp Friston 1998) was used to explore the data acrosssubjects

Statistical analyses were performed on individual sub-jects using a general linear model incorporating taskeffects modeled with a canonical hemodynamic re-sponse function and its temporal derivative (Fristonet al 1998) as well as mean linear and quadratic trends

for each of the 2 runs This model was used to computeparameter estimates (b) and t contrast images for eachcomparison (usual and unusual views) at each voxel

To identify regions for which the level of activationacross participants was related to the time required toname objects shown at unusual perspectives a simpleregression analysis was performed on the average im-ages for the unusual view condition Individual contrastimages were submitted to a second-level random-effectsregression analysis to create mean t images (thresholdp lt 005 t = 291 uncorrected for multiple compari-sons) Based on previous reports of increased BOLDresponse during sustained attention and cognitive effort(Coull Frackowiak amp Frith 1998 Barch et al 1997)regions that were increasingly active with longer timeon task were selected An automated peak-search algo-rithm identified the location of peak activations basedon voxelwise t values An extent threshold of 5 contig-uous voxels was applied to activated clusters meetingthe voxel-level threshold

For each participant hemodynamic response func-tions (10 frames long) for each trial type were thenestimated across each ROI using a finite impulse re-sponse formulation of the general linear model (Ollingeret al 2001 Burock amp Dale 2000) The parameterestimates for this model (calculated using the least-squares solution to the general linear model) are esti-mates for the temporally evolving response magnitudeat each of the 10 points in peristimulus time selectivelyaveraged across all occurrences of that peristimulustime interval This approach has recently been imple-mented by Poldrack et al as an add-on toolbox to theSPM analysis software (SPM ROI Toolbox sourceforgenetprojectsspm-toolbox ) This method resulted in asingle value for each region per individual These valueswere later used in a regression analysis to predictindividual differences in diffusion anisotropy

Regression Analysis

Single BOLD values in 4 ROIs (left and right frontal leftand right parietal) per subject were entered as indepen-dent variables in a regression analysis using SPM the FAimages of each subject were used as the dependentmeasure This analysis yielded ROIs that were predictedby BOLD response FA values were sampled per subjectfrom these regions

Acknowledgments

This research was supported in part by the Dartmouth BrainImaging Center the National Institutes of Health and theMcDonnell-Pew Foundation MKC is supported by a graduateresearch fellowship from the National Science Foundation Wethank W Kelley G Wolford T Laroche J Woodward andS Ketay for their assistance and the 2002 Summer Institutein Cognitive Neuroscience fellows who participated in thisproject

692 Journal of Cognitive Neuroscience Volume 17 Number 4

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693

2000) whereas others demonstrate complete left hemi-anomia which is not limited to objects shown in un-usual perspectives (for a review see Suzuki et al1998) The fact that increased FA in the anterior callosalregion was significantly associated with increasinglypoor performance on the task (see Figure 2) may re-

f lect the existence of multiple callosal channels be-tween the perceptual identification systems of theright hemisphere and the semantic association systemsof the left hemispheres This would be consistent withGeschwindrsquos (1965) proposal that objectsrsquo rich som-esthetic associations enable object naming via anterior

Figure 2 Z scale statistical

maps illustrating regions of

significance based on FA

Positive Z scores are depictedin warm color scale negative in

cool color scale Results from a

multiple regression analysisisolated the splenium of the

corpus callosum as being

associated with relatively faster

RTs and decreased BOLDresponse In contrast the genu

of the corpus callosum was

significantly associated with

longer RTs and increase BOLDresponse in cortical ROIs on a

task that required individuals to

name objects presented fromunusual perspectives Results

are presented in both the (A)

sagittal and (B) axial planes

Table 2 Individual Data

SubjectNumber

Genu FA(Dimensionless Units) SD

Splenium FA(Dimensionless Units) SD

Unusual RT(msec) SD

Usual RT(msec) SD

1 06065 01148 07554 00936 124959 33360 85076 19040

2 07027 00738 06681 00531 110910 39400 81859 19910

3 07301 01206 07950 00506 92717 26420 68787 15060

4 07742 00905 06974 00504 128273 29470 98132 14590

5 07721 00786 07087 00626 92132 22860 69074 8880

6 07960 01335 08176 00693 123676 30120 89166 41260

7 08073 01032 05993 00730 106009 28450 81092 15850

8 06241 00818 04758 01807 133725 33480 99500 30200

9 03223 00805 04500 01646 120221 9580 91694 11960

10 04804 00743 04066 01426 112544 30810 90148 33880

11 07172 00930 08088 00694 71453 15630 59244 6610

12 07988 00935 06820 00704 107052 26630 77610 12230

13 07912 01040 07392 00750 105319 32760 85654 24960

14 03465 01144 07162 00492 147916 26420 139732 30020

15 06558 00709 07282 00512 107933 30190 81989 22050

Total Mean 06617 00952 06699 00837 112323 27705 86584 20433

Total SD 01606 00193 01302 00432 18767 7345 18320 9910

ROI data were collected per individual from both the splenium and genu of the corpus callosum The anterior region was 2160 mm3 and the posteriorregion was 1701 mm3 both regions contained voxels whose global maxima meet a z threshold of 291 p lt 005

690 Journal of Cognitive Neuroscience Volume 17 Number 4

callosal channels when the splenium is damaged Thisidea is further supported by the work of Sidtis VolpeHoltzman Wilson and Gazzaniga (1981) who demon-strated that following partial posterior callosal commis-surotomy transfer of sensory information between thehemispheres was nonexistent however the transfer ofmore cognitive information (ie semantic and episodicinformation) remained intact The authors concludedthat this preservation of function was the result ofsparing of anterior portions of the callosum during thesectioning The present results lend strong support tothis clearly demonstrating that in the intact brain thereexist multiple routes by which information can travelfrom one hemisphere to the other and that these routesvary in their efficiency In the case where performanceon the naming of objects presented at unusual angleswas slower it is conceivable that individuals were re-cruiting greater cognitive resources (as evidenced by therelatively increased BOLD response) and relying princi-pally on the interaction of the frontal cortices to performthe task Conversely it is likely that individuals whodemonstrate relatively faster performance are able to doso because they are relying more heavily on the transferof information between the parietal cortices and thusmaking relatively greater use of relatively faster percep-tual systems to accomplish the task more efficiently

Conclusion

By using behavior to predict cortical activity which wasthen used to predict white matter integrity these datarepresent an important advance in the available methodsfor combining magnetic resonance data sets to examinefunctional connectivity Traditional methods have usedbehavior (eg RT) to predict BOLD response andor FAacross subjects The present study underscores the utilityof a stepwise process whereby more of the varianceacross individuals was accounted for by using an indexof cortical activity (derived from RT) to index FA Theseresults suggest that behavioral performance on a task isrelated to cortical activity in multiple regions as well asthe integrity of the fibers connecting them Although theinterpretation of these results must remain preliminarythe present findings demonstrate the feasibility of usingunique and convergent neuroimaging techniques toexamine specific behavioral phenomena

METHODS

Participants

Fifteen normal right-handed adults (9 women meanage = 259 years) participated in the experiment Allparticipants gave informed consent Handedness wasassessed using the Edinburgh Handedness Inventory(Oldfield 1971) Participants either received monetarycompensation or volunteered as part of an fMRI work-

shop Data from 3 additional participants were excludedbecause of technical difficulties

Experimental Task

Trial onsets were time-locked to the beginning ofvolume acquisition so that each trial was the length of1 TR (3 sec) During each trial a digital photographof a common object was presented in either a prototypi-cal (usual) or an unusual perspective for 1500 msecStimuli were created based on those used by Warringtonand Taylor (1973) The present study replicated thesestimuli in digital photograph form (instead of linedrawings) On each trial participants immediatelypressed the response button upon determining thename of the object and then said the name aloud Tominimize priming effects the unusual orientation alwayspreceded the usual orientation of the same objectBecause our a priori hypotheses were based on responseto the unusual views analyses were based exclusivelyon data from the unusual view condition

During each of the 2 functional runs 25 usual 25unusual and 24 fixation trials were pseudorandomlyintermixed The fixation trials were included to in-troduce jitter into the time series so that unique esti-mates of the hemodynamic responses for the trial typesof interest could be computed (Ollinger Shulman ampCorbetta 2001)

Behavioral Apparatus and Imaging Parameters

Visual stimuli were presented using an Apple G3 laptopcomputer running PsyScope 125 software (CohenMacWhinney Flatt amp Provost 1993) Stimuli were pro-jected to participants with an Epson (model ELP-7000)LCD projector onto a screen positioned at the head endof the bore and viewed through a mirror mounted onthe head coil Participantsrsquo naming times were recordedusing hand-held fiber-optic buttons interfaced with aPsyScope button box (New Micros Dallas TX)

All images were acquired using a 15-T scanner(General Electric Medical Systems Signa CVNvi LX84Waukesha WI) with a standard head coil Anatomicalimages were acquired using a high-resolution 3-D spoiledgradient recovery sequence (128 sagittal slices TR =77 sec TE = 3 msec flip angle = 158 voxel size = 1 1 12 mm) Functional images were collected in runsusing a gradient spin-echo echo-planar sequence sensi-tive to BOLD contrast (T2) (TR = 3 sec TE = 40 msecflip angle = 908 375 375 mm in-plane resolution)During each functional run 88 volumes of axial images(25 slices 45 mm slice thickness 1-mm skip betweenslices) were acquired

Diffusion tensor images (DTI) were acquired usinga diffusion weighted single-shot spin-echo EPI se-quence using the following parameters TR = 10 secTE 88 msec flip angle 908 slice thickness = 25 mm

Baird et al 691

FOV = 240 mm matrix size = 128 128 8 excita-tions acquisition time (9 min 20 sec) Diffusion weight-ing was performed along 6 independent directions withb value 1000 secmm2 A reference image (b = 0 secmm2) was also acquired

Data Analysis

Behavioral Data

For each trial a time to name the presented object wasrecorded These naming times were grouped by objectorientation (usual or unusual) Within each group onlynaming times within 3 SD from the mean were included(see Table 2) The time required to name usual objectswas used in the final analysis

Diffusion Tensor Imaging Data

Using software written by Dr Inati the averaged dif-fusion weighted images and 6 apparent diffusion co-efficients were calculated from which 6 independentelements of the diffusion tensor were determinedfor each voxel Eigenvalues and eigenvectors of thetensor were calculated using a Jacobi transformationFA was calculated from the eigenvalues as describedby Basser and Pierpaoli (1998) and used for all subse-quent analyses

Functional Magnetic Resonance Imaging Data

fMRI data were analyzed using Statistical ParametricMapping software (SPM99 Wellcome Department ofCognitive Neurology London UK) (Friston et al 1995)For each functional run data were preprocessed toremove sources of noise and artifact Functional datawere corrected for differences in acquisition time be-tween slices for each whole-brain volume realignedwithin and across runs to correct for head movementand coregistered with each participantrsquos anatomical dataFunctional data were then transformed into a standardanatomical space (3-mm isotropic voxels) based on theICBM 152 brain template (Montreal Neurological Insti-tute) which approximates the atlas space of Talairachand Tournoux (1988) Normalized data were then spa-tially smoothed (6 mm full width half maximum) usinga Gaussian kernel Analyses took place at 2 levels Firstwithin-subject activity was examined using a fixed-effects model second a random effects model (Holmesamp Friston 1998) was used to explore the data acrosssubjects

Statistical analyses were performed on individual sub-jects using a general linear model incorporating taskeffects modeled with a canonical hemodynamic re-sponse function and its temporal derivative (Fristonet al 1998) as well as mean linear and quadratic trends

for each of the 2 runs This model was used to computeparameter estimates (b) and t contrast images for eachcomparison (usual and unusual views) at each voxel

To identify regions for which the level of activationacross participants was related to the time required toname objects shown at unusual perspectives a simpleregression analysis was performed on the average im-ages for the unusual view condition Individual contrastimages were submitted to a second-level random-effectsregression analysis to create mean t images (thresholdp lt 005 t = 291 uncorrected for multiple compari-sons) Based on previous reports of increased BOLDresponse during sustained attention and cognitive effort(Coull Frackowiak amp Frith 1998 Barch et al 1997)regions that were increasingly active with longer timeon task were selected An automated peak-search algo-rithm identified the location of peak activations basedon voxelwise t values An extent threshold of 5 contig-uous voxels was applied to activated clusters meetingthe voxel-level threshold

For each participant hemodynamic response func-tions (10 frames long) for each trial type were thenestimated across each ROI using a finite impulse re-sponse formulation of the general linear model (Ollingeret al 2001 Burock amp Dale 2000) The parameterestimates for this model (calculated using the least-squares solution to the general linear model) are esti-mates for the temporally evolving response magnitudeat each of the 10 points in peristimulus time selectivelyaveraged across all occurrences of that peristimulustime interval This approach has recently been imple-mented by Poldrack et al as an add-on toolbox to theSPM analysis software (SPM ROI Toolbox sourceforgenetprojectsspm-toolbox ) This method resulted in asingle value for each region per individual These valueswere later used in a regression analysis to predictindividual differences in diffusion anisotropy

Regression Analysis

Single BOLD values in 4 ROIs (left and right frontal leftand right parietal) per subject were entered as indepen-dent variables in a regression analysis using SPM the FAimages of each subject were used as the dependentmeasure This analysis yielded ROIs that were predictedby BOLD response FA values were sampled per subjectfrom these regions

Acknowledgments

This research was supported in part by the Dartmouth BrainImaging Center the National Institutes of Health and theMcDonnell-Pew Foundation MKC is supported by a graduateresearch fellowship from the National Science Foundation Wethank W Kelley G Wolford T Laroche J Woodward andS Ketay for their assistance and the 2002 Summer Institutein Cognitive Neuroscience fellows who participated in thisproject

692 Journal of Cognitive Neuroscience Volume 17 Number 4

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693

callosal channels when the splenium is damaged Thisidea is further supported by the work of Sidtis VolpeHoltzman Wilson and Gazzaniga (1981) who demon-strated that following partial posterior callosal commis-surotomy transfer of sensory information between thehemispheres was nonexistent however the transfer ofmore cognitive information (ie semantic and episodicinformation) remained intact The authors concludedthat this preservation of function was the result ofsparing of anterior portions of the callosum during thesectioning The present results lend strong support tothis clearly demonstrating that in the intact brain thereexist multiple routes by which information can travelfrom one hemisphere to the other and that these routesvary in their efficiency In the case where performanceon the naming of objects presented at unusual angleswas slower it is conceivable that individuals were re-cruiting greater cognitive resources (as evidenced by therelatively increased BOLD response) and relying princi-pally on the interaction of the frontal cortices to performthe task Conversely it is likely that individuals whodemonstrate relatively faster performance are able to doso because they are relying more heavily on the transferof information between the parietal cortices and thusmaking relatively greater use of relatively faster percep-tual systems to accomplish the task more efficiently

Conclusion

By using behavior to predict cortical activity which wasthen used to predict white matter integrity these datarepresent an important advance in the available methodsfor combining magnetic resonance data sets to examinefunctional connectivity Traditional methods have usedbehavior (eg RT) to predict BOLD response andor FAacross subjects The present study underscores the utilityof a stepwise process whereby more of the varianceacross individuals was accounted for by using an indexof cortical activity (derived from RT) to index FA Theseresults suggest that behavioral performance on a task isrelated to cortical activity in multiple regions as well asthe integrity of the fibers connecting them Although theinterpretation of these results must remain preliminarythe present findings demonstrate the feasibility of usingunique and convergent neuroimaging techniques toexamine specific behavioral phenomena

METHODS

Participants

Fifteen normal right-handed adults (9 women meanage = 259 years) participated in the experiment Allparticipants gave informed consent Handedness wasassessed using the Edinburgh Handedness Inventory(Oldfield 1971) Participants either received monetarycompensation or volunteered as part of an fMRI work-

shop Data from 3 additional participants were excludedbecause of technical difficulties

Experimental Task

Trial onsets were time-locked to the beginning ofvolume acquisition so that each trial was the length of1 TR (3 sec) During each trial a digital photographof a common object was presented in either a prototypi-cal (usual) or an unusual perspective for 1500 msecStimuli were created based on those used by Warringtonand Taylor (1973) The present study replicated thesestimuli in digital photograph form (instead of linedrawings) On each trial participants immediatelypressed the response button upon determining thename of the object and then said the name aloud Tominimize priming effects the unusual orientation alwayspreceded the usual orientation of the same objectBecause our a priori hypotheses were based on responseto the unusual views analyses were based exclusivelyon data from the unusual view condition

During each of the 2 functional runs 25 usual 25unusual and 24 fixation trials were pseudorandomlyintermixed The fixation trials were included to in-troduce jitter into the time series so that unique esti-mates of the hemodynamic responses for the trial typesof interest could be computed (Ollinger Shulman ampCorbetta 2001)

Behavioral Apparatus and Imaging Parameters

Visual stimuli were presented using an Apple G3 laptopcomputer running PsyScope 125 software (CohenMacWhinney Flatt amp Provost 1993) Stimuli were pro-jected to participants with an Epson (model ELP-7000)LCD projector onto a screen positioned at the head endof the bore and viewed through a mirror mounted onthe head coil Participantsrsquo naming times were recordedusing hand-held fiber-optic buttons interfaced with aPsyScope button box (New Micros Dallas TX)

All images were acquired using a 15-T scanner(General Electric Medical Systems Signa CVNvi LX84Waukesha WI) with a standard head coil Anatomicalimages were acquired using a high-resolution 3-D spoiledgradient recovery sequence (128 sagittal slices TR =77 sec TE = 3 msec flip angle = 158 voxel size = 1 1 12 mm) Functional images were collected in runsusing a gradient spin-echo echo-planar sequence sensi-tive to BOLD contrast (T2) (TR = 3 sec TE = 40 msecflip angle = 908 375 375 mm in-plane resolution)During each functional run 88 volumes of axial images(25 slices 45 mm slice thickness 1-mm skip betweenslices) were acquired

Diffusion tensor images (DTI) were acquired usinga diffusion weighted single-shot spin-echo EPI se-quence using the following parameters TR = 10 secTE 88 msec flip angle 908 slice thickness = 25 mm

Baird et al 691

FOV = 240 mm matrix size = 128 128 8 excita-tions acquisition time (9 min 20 sec) Diffusion weight-ing was performed along 6 independent directions withb value 1000 secmm2 A reference image (b = 0 secmm2) was also acquired

Data Analysis

Behavioral Data

For each trial a time to name the presented object wasrecorded These naming times were grouped by objectorientation (usual or unusual) Within each group onlynaming times within 3 SD from the mean were included(see Table 2) The time required to name usual objectswas used in the final analysis

Diffusion Tensor Imaging Data

Using software written by Dr Inati the averaged dif-fusion weighted images and 6 apparent diffusion co-efficients were calculated from which 6 independentelements of the diffusion tensor were determinedfor each voxel Eigenvalues and eigenvectors of thetensor were calculated using a Jacobi transformationFA was calculated from the eigenvalues as describedby Basser and Pierpaoli (1998) and used for all subse-quent analyses

Functional Magnetic Resonance Imaging Data

fMRI data were analyzed using Statistical ParametricMapping software (SPM99 Wellcome Department ofCognitive Neurology London UK) (Friston et al 1995)For each functional run data were preprocessed toremove sources of noise and artifact Functional datawere corrected for differences in acquisition time be-tween slices for each whole-brain volume realignedwithin and across runs to correct for head movementand coregistered with each participantrsquos anatomical dataFunctional data were then transformed into a standardanatomical space (3-mm isotropic voxels) based on theICBM 152 brain template (Montreal Neurological Insti-tute) which approximates the atlas space of Talairachand Tournoux (1988) Normalized data were then spa-tially smoothed (6 mm full width half maximum) usinga Gaussian kernel Analyses took place at 2 levels Firstwithin-subject activity was examined using a fixed-effects model second a random effects model (Holmesamp Friston 1998) was used to explore the data acrosssubjects

Statistical analyses were performed on individual sub-jects using a general linear model incorporating taskeffects modeled with a canonical hemodynamic re-sponse function and its temporal derivative (Fristonet al 1998) as well as mean linear and quadratic trends

for each of the 2 runs This model was used to computeparameter estimates (b) and t contrast images for eachcomparison (usual and unusual views) at each voxel

To identify regions for which the level of activationacross participants was related to the time required toname objects shown at unusual perspectives a simpleregression analysis was performed on the average im-ages for the unusual view condition Individual contrastimages were submitted to a second-level random-effectsregression analysis to create mean t images (thresholdp lt 005 t = 291 uncorrected for multiple compari-sons) Based on previous reports of increased BOLDresponse during sustained attention and cognitive effort(Coull Frackowiak amp Frith 1998 Barch et al 1997)regions that were increasingly active with longer timeon task were selected An automated peak-search algo-rithm identified the location of peak activations basedon voxelwise t values An extent threshold of 5 contig-uous voxels was applied to activated clusters meetingthe voxel-level threshold

For each participant hemodynamic response func-tions (10 frames long) for each trial type were thenestimated across each ROI using a finite impulse re-sponse formulation of the general linear model (Ollingeret al 2001 Burock amp Dale 2000) The parameterestimates for this model (calculated using the least-squares solution to the general linear model) are esti-mates for the temporally evolving response magnitudeat each of the 10 points in peristimulus time selectivelyaveraged across all occurrences of that peristimulustime interval This approach has recently been imple-mented by Poldrack et al as an add-on toolbox to theSPM analysis software (SPM ROI Toolbox sourceforgenetprojectsspm-toolbox ) This method resulted in asingle value for each region per individual These valueswere later used in a regression analysis to predictindividual differences in diffusion anisotropy

Regression Analysis

Single BOLD values in 4 ROIs (left and right frontal leftand right parietal) per subject were entered as indepen-dent variables in a regression analysis using SPM the FAimages of each subject were used as the dependentmeasure This analysis yielded ROIs that were predictedby BOLD response FA values were sampled per subjectfrom these regions

Acknowledgments

This research was supported in part by the Dartmouth BrainImaging Center the National Institutes of Health and theMcDonnell-Pew Foundation MKC is supported by a graduateresearch fellowship from the National Science Foundation Wethank W Kelley G Wolford T Laroche J Woodward andS Ketay for their assistance and the 2002 Summer Institutein Cognitive Neuroscience fellows who participated in thisproject

692 Journal of Cognitive Neuroscience Volume 17 Number 4

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693

FOV = 240 mm matrix size = 128 128 8 excita-tions acquisition time (9 min 20 sec) Diffusion weight-ing was performed along 6 independent directions withb value 1000 secmm2 A reference image (b = 0 secmm2) was also acquired

Data Analysis

Behavioral Data

For each trial a time to name the presented object wasrecorded These naming times were grouped by objectorientation (usual or unusual) Within each group onlynaming times within 3 SD from the mean were included(see Table 2) The time required to name usual objectswas used in the final analysis

Diffusion Tensor Imaging Data

Using software written by Dr Inati the averaged dif-fusion weighted images and 6 apparent diffusion co-efficients were calculated from which 6 independentelements of the diffusion tensor were determinedfor each voxel Eigenvalues and eigenvectors of thetensor were calculated using a Jacobi transformationFA was calculated from the eigenvalues as describedby Basser and Pierpaoli (1998) and used for all subse-quent analyses

Functional Magnetic Resonance Imaging Data

fMRI data were analyzed using Statistical ParametricMapping software (SPM99 Wellcome Department ofCognitive Neurology London UK) (Friston et al 1995)For each functional run data were preprocessed toremove sources of noise and artifact Functional datawere corrected for differences in acquisition time be-tween slices for each whole-brain volume realignedwithin and across runs to correct for head movementand coregistered with each participantrsquos anatomical dataFunctional data were then transformed into a standardanatomical space (3-mm isotropic voxels) based on theICBM 152 brain template (Montreal Neurological Insti-tute) which approximates the atlas space of Talairachand Tournoux (1988) Normalized data were then spa-tially smoothed (6 mm full width half maximum) usinga Gaussian kernel Analyses took place at 2 levels Firstwithin-subject activity was examined using a fixed-effects model second a random effects model (Holmesamp Friston 1998) was used to explore the data acrosssubjects

Statistical analyses were performed on individual sub-jects using a general linear model incorporating taskeffects modeled with a canonical hemodynamic re-sponse function and its temporal derivative (Fristonet al 1998) as well as mean linear and quadratic trends

for each of the 2 runs This model was used to computeparameter estimates (b) and t contrast images for eachcomparison (usual and unusual views) at each voxel

To identify regions for which the level of activationacross participants was related to the time required toname objects shown at unusual perspectives a simpleregression analysis was performed on the average im-ages for the unusual view condition Individual contrastimages were submitted to a second-level random-effectsregression analysis to create mean t images (thresholdp lt 005 t = 291 uncorrected for multiple compari-sons) Based on previous reports of increased BOLDresponse during sustained attention and cognitive effort(Coull Frackowiak amp Frith 1998 Barch et al 1997)regions that were increasingly active with longer timeon task were selected An automated peak-search algo-rithm identified the location of peak activations basedon voxelwise t values An extent threshold of 5 contig-uous voxels was applied to activated clusters meetingthe voxel-level threshold

For each participant hemodynamic response func-tions (10 frames long) for each trial type were thenestimated across each ROI using a finite impulse re-sponse formulation of the general linear model (Ollingeret al 2001 Burock amp Dale 2000) The parameterestimates for this model (calculated using the least-squares solution to the general linear model) are esti-mates for the temporally evolving response magnitudeat each of the 10 points in peristimulus time selectivelyaveraged across all occurrences of that peristimulustime interval This approach has recently been imple-mented by Poldrack et al as an add-on toolbox to theSPM analysis software (SPM ROI Toolbox sourceforgenetprojectsspm-toolbox ) This method resulted in asingle value for each region per individual These valueswere later used in a regression analysis to predictindividual differences in diffusion anisotropy

Regression Analysis

Single BOLD values in 4 ROIs (left and right frontal leftand right parietal) per subject were entered as indepen-dent variables in a regression analysis using SPM the FAimages of each subject were used as the dependentmeasure This analysis yielded ROIs that were predictedby BOLD response FA values were sampled per subjectfrom these regions

Acknowledgments

This research was supported in part by the Dartmouth BrainImaging Center the National Institutes of Health and theMcDonnell-Pew Foundation MKC is supported by a graduateresearch fellowship from the National Science Foundation Wethank W Kelley G Wolford T Laroche J Woodward andS Ketay for their assistance and the 2002 Summer Institutein Cognitive Neuroscience fellows who participated in thisproject

692 Journal of Cognitive Neuroscience Volume 17 Number 4

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693

Reprint requests should be sent to Abigail A Baird Depart-ment of Psychological and Brain Sciences Dartmouth CollegeMoore Hall Hanover NH 03755 or via e-mail abigailabairddartmouthedu

The data reported in this experiment have been depositedwith The fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-118E6

REFERENCES

Barch D M Braver T S Nystrom L E Forman S DNoll D C amp Cohen J D (1997) Dissociating workingmemory from task difficulty in human prefrontal cortexNeuropsychologia 35 1373ndash1380

Basser P J amp Pierpaoli C (1998) A simplified method tomeasure the diffusion tensor from seven MR imagesMagnetic Resonance in Medicine 39 928ndash934

Burock M A amp Dale A M (2000) Estimation and detectionof event-related fMRI signals with temporally correlatednoise A statistically efficient and unbiased approachHuman Brain Mapping 11 249ndash260

Cohen J D MacWhinney B Flatt M amp Provost J (1993)PsyScope A new graphic interactive environment fordesigning psychology experiments Behavioral ResearchMethods Instruments and Computers 25 257ndash271

Coull J T Frackowiak R S amp Frith C D (1998) Monitoringfor target objects Activation of right frontal and parietalcortices with increasing time on task Neuropsychologia 361325ndash1334

Davidoff J amp Warrington E K (1999) The bare bones ofobject recognition Implications from a case of objectrecognition impairment Neuropsychologia 37 279ndash292

Friston K J Fletcher P Josephs O Holmes A Rugg M Damp Turner R (1998) Event-related fMRI Characterizingdifferential responses Neuroimage 7 30ndash40

Friston K J Holmes A P Worsley K J Poline J-PFrith C D amp Frackowiak R S J (1995) Statisticalparametric maps in functional imaging A general linearapproach Human Brain Mapping 2 189ndash210

Geschwind N (1965) Disconnexion syndromes in animalsand man Part II Brain 88 585ndash644

Guye M Parker G J Symms M Boulby PWheeler-Kingshott C A Salek-Haddadi A Barker G Jamp Duncan J S (2003) Combined functional MRI andtractography to demonstrate the connectivity of thehuman primary motor cortex in vivo Neuroimage 191349ndash1360

Holmes A P amp Friston K J (1998) Generalisability randomeffects and population inference Neuroimage 7 S754

Humphreys G W Price C J amp Riddoch M J (1999)From objects to names A cognitive neuroscience approachPsychological Research 62 118ndash130

Intriligator J Henaff M-A amp Michel F (2000) Able to nameunable to compare The visual abilities of a posteriorsplit-brain patient NeuroReport 11 2639ndash2642

Klingberg T Vaidya C J Gabrieli J D Moseley M E ampHedehus M (1999) Myelination and organization of thefrontal white matter in children A diffusion tensor MRIstudy NeuroReport 10 2817ndash2821

Kosslyn S M Alpert N M Thompson W L Chabris C FRauch S L amp Anderson A K (1994) Identifying objectsseen from different viewpoints A PET investigation Brain117 1055ndash1071

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D NHoppel B E Cohen M S Turner R et al (1992)Dynamic magnetic resonance imaging of human brainactivity during primary sensory stimulation Proceedingsof the National Academy of Sciences USA 895675ndash5679

LeBihan D (2003) Looking into the functional architectureof the brain with diffusion MRI Neuroscience 4 469ndash480

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsicsignal changes accompanying sensory stimulationFunctional brain mapping with magnetic resonanceimaging Proceedings of the National Academy of SciencesUSA 89 5951ndash5955

Oldfield R (1971) The assessment and analysis ofhandedness The Edinburg Inventory Neuropsychologia 997ndash113

Oleson P J Nagy Z Westerberg H amp Klingberg T (2003)Combined analysis of DTI and fMRI data reveals a jointmaturation of white and grey matter in a fronto-parietalnetwork Cognitive Brain Research 18 48ndash57

Ollinger J M Shulman G L amp Corbetta M (2001)Separating processes within a trial in event-related functionalMRI Neuroimage 13 210ndash217

Rudge P amp Warrington E K (1991) Selective impairment ofmemory and visual perception in splenial tumours Brain114 349ndash360

Sereno M I (1998) Brain mapping in animals and humansCurrent Opinions in Neurobiology 8 188ndash194

Sidtis J J Volpe B T Holtzman J D Wilson D H ampGazzaniga M S (1981) Cognitive interaction after stagedcallosal section Evidence for transfer of semantic activationScience 212 344ndash346

Sugio T Inui T Matsuo K Matsuzawa M Glover G Hamp Nakai T (1999) The role of the posterior parietalcortex in human object recognition A functionalmagnetic resonance imaging study NeuroscienceLetters 276 45ndash48

Suzuki K Yamadori A Endo K Fujii T Ezura M ampTakahashi A (1998) Dissociation of letter and picturenaming resulting from callosal disconnection Neurology 511390ndash1394

Talairach J amp Tournoux P (1988) Coplanar stereotaxic atlasof the human brain Thieme Stuttgart

Toosy A T Ciccarelli O Parker G J M Wheeler-KingshottC A M Miller D H amp Thompson A J (2004)Characterizing functionndashstructure relationships in thehuman visual system with functional MRI and diffusiontensor imaging Neuroimage 21 1452ndash1463

Warrington E K amp James M (1988) Visual apperceptiveagnosia A clinico-anatomical study of three cases Cortex24 13ndash32

Warrington E K amp Taylor A M (1973) The contributionof the right parietal lobe to object recognition Cortex 9152ndash164

Warrington E K amp Taylor A M (1978) Twocategorical stages of object recognition Perception 7695ndash705

Werring D J Clark C A Parker G J M Miller D HThompson A J amp Barker G J (1999) A directdemonstration of both structure and function in the visualsystem Combining diffusion tensor imaging withfunctional magnetic resonance imaging Neuroimage 9352ndash361

Baird et al 693