Vision Research 50 (2010) 2163–2175

Contents lists available at ScienceDirect

Vision Research

journal homepage: www.elsevier .com/locate /v isres

Functional anatomy of predictive vergence and saccade eye movements in humans:A functional MRI investigation

Tara L. Alvarez a,⇑, Yelda Alkan a, Suril Gohel b, B. Douglas Ward c, Bharat B. Biswal b,⇑⇑a Department of Biomedical Engineering, New Jersey Institute of Technology, University Heights, Newark, NJ 07102, USAb Department of Radiology, University of Medicine and Dentistry of New Jersey, University Heights, Newark, NJ 07102, USAc Department of Biophysics, Medical College of Wisconsin, Milwaukee, WI 53226, USA

a r t i c l e i n f o

Article history:Received 12 January 2010Received in revised form 11 August 2010

Keywords:VergenceSaccadesPredictionFrontal eye fieldsSupplementary eye fieldDorsolateral prefrontal cortexCingulateCerebellum

0042-6989/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.visres.2010.08.018

⇑ Corresponding author. Fax: +1 973 596 5222.⇑⇑ Corresponding author.

E-mail addresses: tara.l.alvarez@njit.edu (T.L. AlvarBiswal).

a b s t r a c t

Purpose: The purpose of this study is to investigate the functional neural anatomy that generates ver-gence eye movement responses from predictive versus random symmetrical vergence step stimuli inhumans and compare it to a similar saccadic task via the blood oxygenation level dependent signal fromfunctional MRI.Methods: Eight healthy subjects participated in fMRI scans obtained from a 3 T Siemens Allegra scanner.Subjects tracked random and predictable vergent steps and then tracked random and predictable saccad-ic steps each within a block design. A general linear model (GLM) was used to determine significantly(p < 0.001) active regions of interest through a combination of correlation threshold and cluster extent.A paired t-test of the GLM beta weight coefficients was computed to determine significant spatial differ-ences between the saccade and vergence data sets.Results: Predictive saccadic and vergent eye movements induced many common sites of significant func-tional cortical activity including: the dorsolateral prefrontal cortex (DLPFC), parietal eye field (PEF),cuneus, precuneus, anterior and posterior cingulate, and the cerebellum. However, differentiation in spa-tial location was observed within the frontal lobe for the functional activity of the saccadic and vergentnetwork induced while studying prediction. A paired t-test of the beta weights from the individual sub-jects showed that peak activity induced by predictive versus random vergent eye movements was signif-icantly (t > 2.7, p < 0.03) more anterior within the frontal eye field (FEF) and the supplementary eye field(SEF) when compared to the functional activity from predictive saccadic eye movements.Conclusion: This research furthers our knowledge of which cortical sites facilitate a subject’s ability topredict within the vergence and saccade networks. Using a predictive versus random visual task, saccadicand vergent eye movements induced activation in many shared cortical sites and also stimulated differ-entiation in the FEF and SEF.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

There are five major types of eye movements originally de-scribed by Dodge in 1903. Three adjust the position of the eye tokeep the object of interest on the fovea and two stabilize the eyeduring head movement (Dodge, 1903; Goldberg, Eggers, & Gouras,2000). Saccades are fast, tandem, conjugate movements which rap-idly shift the fovea to a new target. Smooth pursuit movementskeep the image of a moving target on the fovea. Vergence is the in-ward (convergence) and outward (divergence) turning of the eyesto track targets at different depths. Numerous studies have beenconducted to study saccade and vergence anatomy.

ll rights reserved.

ez), bbiswal@gmail.com (B.B.

Prediction in the visual system dates to the research of Dodge in1931 and is a strategy that the brain utilizes in oculomotor controlto reduce the response latency and generate a movement withgreater peak velocity (Dodge, 1931). Predictive behaviors havebeen reported in saccade, smooth pursuit and vergence eye move-ments (Barnes & Asselman, 1991; Kowler & Steinman, 1979; Ku-mar, Han, Garbutt, & Leigh 2002; Rashbass & Westheimer, 1961;Ron, Schmid, & Orpaz, 1989; Stark, Vossius, & Young, 1962). Stud-ies of saccadic eye movements have reported that when predictionis utilized responses show reduced latencies, even as small as zeromsec and some responses show anticipatory movements beforestimulus onset (Kowler & Steinman, 1979). Rashbass and Westhei-mer first analyzed the use of predictable sinusoids varying in depthin 1961 and reported that predictive vergence sinusoidal responsesshowed a decrease in latency compared to step or pulse responseswhen the subject did not know when the target would changepositions (Rashbass & Westheimer, 1961). A step input is an abrupt

2164 T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175

change in vergence disparity such as when a person fixates on a faraway target then fixates to a target located at close range. Theysuggested that this behavior is due to the anticipation of future dis-parity changes. Another study showed that the latency in conver-gent and divergent repetitive vergence step stimuli decreased,especially when the frequency was less than 1 Hz, providing evi-dence of a prediction operator that was most effective at 0.5 Hz(Krishnan, Farazian, & Stark, 1973). Our group has also reported adecrease in latency, an increase in peak velocity, and anticipatorymovements when comparing vergence responses from a predict-able symmetrical step disparity vergence stimulus where subjectsknew the timing and magnitude information of the stimulus com-pared to a random vergence step stimulus (Alvarez, Semmlow,Yuan, & Munoz, 2002). Furthermore, Kumar et al. (2002) showedthat the anticipatory movements observed in vergence eye move-ments to predictable step stimuli were influenced by the previousvisual stimuli suggesting that working memory is involved inanticipatory drifts (Kumar et al., 2002). These results suggest dif-ferent cortical resources may be recruited when prediction is uti-lized resulting in reduced latency, increased peak velocity andanticipatory movements.

Cortically, investigators report that a predictive controller re-sides in the dorsolateral prefrontal cortex (DLPFC) (Pierrot-Deseil-ligny, Müri, Nyffeler, & Milea, 2005). Pierrot-Deseilligny andcolleagues studied patients with a lesion limited to the DLPFCand report a significant decrease in anticipatory saccades com-pared to control subjects when studying predictive saccadic move-ments. They report that the DLPFC is involved, specifically in thetiming control of predictive saccades; however, vergent eye move-ments were not investigated in their study.

Single cell primate studies of vergence have reported cellularactivity evoked by using vergence stimuli in the primary visual cor-tex (Poggio, 1995), the posterior parietal area (Genovesio & Ferra-ina, 2004; Gnadt & Mays, 1995), the bilateral frontal eye fields(FEF) (Akao, Mustari, Fukushima, Kurkin, & Fukushima, 2005; Gam-lin & Yoon, 2000), the cerebellum, (Gamlin & Clarke, 1995; Miles,Fuller, Braitman, & Dow, 1980; Nitta, Akao, Kurkin, & Fukushima,2008; Zhang & Gamlin, 1998), and the midbrain (Judge & Cum-ming, 1986; Mays & Porter, 1984; Mays, Porter, Gamlin, & Tello,1986). Several behavioral vergence eye movement studies of pre-diction have been conducted; yet they do not provide functionalcortical insight. Prediction can easily be studied in humans viafMRI. An fMRI study using predictive versus random vergenceeye movements has not been conducted previously and hence willbe the focus of this study.

Numerous behavioral, animal, fMRI and clinical investigationshave been reported for the saccadic system. Several review paperssummarize the functional anatomy using fMRI to study cognitivecontrol of saccades (Pierrot-Deseilligny, Milea, & Müri, 2004), its rolein spatial attention (Luna & Sweeney 1999), and its role in spatialworking memory (Curtis, 2006). Other reviews describe the corticalcontrol of saccades through a detailed investigation of single cellstudies, lesion or fMRI experiments in primates, as well as transcra-nial magnetic stimulation and case reports from humans (Gaymard,Ploner, Rivaud, Vermersch, & Pierrot-Deseilligny, 1998; Leigh & Zee,2006; Pierrot-Deseilligny et al., 2004). Hence, the influence of pre-diction upon the saccadic system will be investigated to confirmour findings with those published by others. It will also be comparedto how prediction influences the vergence system which has notbeen previously studied via fMRI investigations.

The aim of this current study is to investigate prediction in thevergence and saccade neural network in humans. Since vergentand saccadic eye movements both exhibit anticipatory movementsand reduced latencies when stimuli are predictive, we hypothesizethat the cortical resources for prediction will be similar for bothsystems. In this study, a predictive versus random symmetrical

vergence step stimulus is used to obtain vergence neural activityand is compared to the saccade neural activity induced by predic-tive versus random saccade stimulus. Hence, this is the first paperto systematically perform a whole brain study on the anatomicalnetwork responsible for vergence predictive behavior in humansusing fMRI. We hypothesize (1) functional activity will be inducedin the DLPFC from prediction in both the saccade and vergenceneural networks, (2) spatial differentiation between the two net-works will be observed within the bilateral frontal eye fields,which has been previously reported in single cell experiments fromprimates and (3) similar activation sites in the sensory area, parie-tal lobe and cerebellum will be observed.

2. Materials and methods

2.1. Subjects

Eight subjects who did not know the hypotheses of the experi-ment participated in this study (5F, 3M, mean age 26 ± 4 years). Allsubjects had normal binocular vision assessed by the Randot Stere-opsis test with a fixation disparity better than 70 s of arc and a nearpoint of convergence less than 10 cm. Six of the subjects wereemmetropes and two were corrected for normal refraction wherethe average prescription among the myopes was �1D. All subjectswere right handed. None of the subjects had a history of brain in-jury or other neurological disorder. Subjects participated in an eyemovement experiment prior to functional scanning. Each subject’seye movements were recorded to ensure the subject understoodthe task. All subjects were able to perform the task required. Sub-jects gave informed consent approved by the University of Medi-cine and Dentistry of New Jersey and the New Jersey Institute ofTechnology Institution Review Boards.

2.2. Materials and apparatus

Images were acquired using a 3.0 T Siemens Allegra MRI scan-ner with a standard head coil (Erlangen, Germany). Visual stimuliwere a set of non-ferrous LED targets that formed a line 5 cm inheight by 2 mm in width located at three positions. Eye movementrecordings confirmed that the subject could perform both the sacc-adic and vergent oculomotor tasks.

Eye movements were recorded using an infrared (k = 950 nm)limbus tracking system manufactured by Skalar Iris (model 6500,Delft, Netherlands). All of the eye movements were within thelinear range of the system (±25�). The left-eye and right-eye re-sponses were calibrated, recorded and saved separately for offlineanalysis. Digitization of the eye movements was performed with a12-bit digital acquisition hardware card using a range of ±5 volts(National Instruments 6024 E series, Austin, TX, USA). A customMatlab™ 7.0 (Waltham, MA, USA) program was used for offlineeye movement data analysis and eye movement data were plottedusing Axum (Cambridge, MA, USA).

2.3. Imaging instrumentation and procedure

The subject was positioned supine on the gantry of the scannerwith his/her head along the midline of the coil. All participantswere instructed to limit head motion. Foam padding was used torestrict additional movement and motion correction software de-scribed below was utilized to ensure head motion did not influencethe results. Ear plugs which still enabled the participant to hearinstructions from the operators were used to ensure communica-tion during the scan while reducing scanner noise by up to30 dB. In all experiments, the radio frequency power deposition

T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175 2165

and field-switching rate were kept below levels specified by the USFDA.

A quick scan was obtained and used to localize high resolutionanatomical and functional scans within the magnet. Since the cer-ebellum was an area of interest in this study, all subjects werepositioned so that images could be attained of the whole brain.All functional scans used a T2* weighted echo planar imaging(EPI) sequence. The imaging parameters were FOV = 220 mm,64 � 64 matrix, TR = 2000 ms, TE = 27 ms and flip angle = 90�. Thewhole brain was imaged in an axial configuration where 32 sliceswere collected and each slice was 5 mm thick. The resolutionwas 3.4 � 3.4 � 5 mm. There were 140 volumes acquired per scanequating to a duration of 4 min and 40 s. Between scans, the sub-jects were asked if they were comfortable and could perform thetask. Subjects confirmed they could perform each task with ease.After all functional tasks, a high resolution MPRAGE (magnetiza-tion-prepared rapid gradient-echo) data set was collected. TheMPRAGE imaging parameters were: 80 slices, FOV = 220 mm, slicethickness = 2 mm, TR = 2000 ms, TE = 4.38 ms, T1 = 900 ms, flip an-gle = 8�, and matrix = 256 � 256 which resulted in a spatial resolu-tion of 0.9 � 0.9 � 2 mm.

2.4. Functional experimental design

The saccadic visual stimulus is shown in Fig. 1A. Each visualstimulus would be present for a random duration of time between0.5 and 3.0 s where approximately 20 visual step stimuli would bepresented within each random phase. The subject could not antic-ipate the timing of the visual stimulus. A saccadic magnitude of 10�from midline was chosen because saccades less than 15� from mid-line do not evoke head motion (Ciuffreda & Tannen, 1995). Func-tional scans were obtained with a standard block design usingeither a predictable or random eye movement stimulus. The exper-iment began with a random saccadic stimulus, shown in Fig. 1B.Subjects would track targets that would randomly appear in threelocations: (1) 0� (midline); (2) 10� into the left visual field (denotedas negative) or (3) 10� into the right visual field (denoted as posi-tive). Since we had only three targets, the subject could potentiallyanticipate which target may be illuminated; however the subjectcould not predict the timing sequence or when the next targetwould be illuminated. Anticipatory movements are commonlyobserved when the visual stimulus is predictable but are not com-monly observed when the timing of the stimulus is not predictable(Kumar et al., 2002).

Subjects would track the illuminated LED which produced ran-dom saccadic step stimuli for 40 s followed by predictable saccadiceye movements for 40 s. The predictable saccadic stimulus wouldbe illuminated in the left visual field (10� from midline), along mid-line and then into the right visual field (10� from midline) and re-main in each location for 2 s, shown in Fig. 1B. This pattern wouldrepeat six times. A similar periodicity has been used in otherbehavioral studies to investigate the influence of prediction in ver-gence eye movements (Alvarez et al., 2002; Krishnan et al., 1973).Random and predictable phases were repeated for 3.5 cycles for atotal duration of 280 s or 4 min and 40 s, as shown in Fig. 1C. Thesubjects were instructed to look at the visual target and blink whenneeded without moving their heads. The operator gave an audiblecue when the predictable phase began. The subjects were in-structed to anticipate the next target when they were in the pre-dictable phase.

For vergence stimulation, subjects viewed the same LED appa-ratus used during the saccadic experiment. The custom fMRI com-patible LEDs were adjusted and centered, as shown in Fig. 1A.There were three vergence fixation points, 2�, 3� and 4� centeredalong the subject’s midline to produce symmetrical vergence stepstimuli. The vergence step stimulus was a 2� disparity change

which was chosen due to the physical constraints of the imagingcenter and to decrease the occurrence of saccades within the sym-metrical vergence response (Coubard & Kapoula, 2008; Semmlow,Chen, Granger-Donnetti, & Alvarez, 2009; Semmlow, Chen, Pedron-o, & Alvarez, 2008). Subjects confirmed they were able to comfort-ably view the visual stimuli during the imaging session. For allexperiments, only one location was illuminated at a time. For therandom phase, the subjects could not predict which of the threelocations would be shown. The time when the next target was dis-played was also randomized between 0.5 and 3 s in duration. Forthe predictable sequence, the targets began with the 4� vergencefixation followed by the 3� vergence fixation and then the 2� targetusing the non-ferrous LEDs. This sequence was repeated six timesover the duration of 40 s, see Fig. 1B. As with the saccadic experi-ment, this experiment also used a standard block design of the ran-dom and predictable eye movements, see Fig. 1C.

Half of the subjects began with the vergence experimental trialsthen performed the saccade trials while the other half of the sub-jects began with the saccade trials and then performed the ver-gence trials. A total of three saccade and three vergenceexperimental trials were collected in case head motion was a prob-lem which was not the case within this dataset.

2.5. Data analysis

Data were analyzed with AFNI (Cox, 1996). All the scans werefirst registered and motion corrected. As with any fMRI data set,the data collected were susceptible to motion artifact from headmotion or other non-neuronal influences (D’Esposito et al., 1999).Hence, all data used in this present study were analyzed for thepresence of motion-induced artifact. Subjects were instructed tolimit head motion and foam padding was used to further reducemotion. Many algorithms exist for the detection and correctionof mis-registered images. For this study, a minimum least-squareimage registration method available in AFNI was utilized to detectand correct for the presence of any motion- induced changes onthe 3D image space. Six parameters were monitored to determineif head motion was a problem within our data set. Three parame-ters indicated the movement within each plane (anterior to poster-ior, right to left, and inferior to superior, calculated in millimeter)and three parameters indicated the amount of rotation about thethree orthogonal axes (yaw, pitch and roll, calculated in degrees).Data were sync interpolated in time in order to account forphase-shifts related to slice-wise data acquisition. A recent com-parison of several software packages found that the AFNI imageregistration algorithm was both reliable and fast in comparisonwith other software (Oakes et al., 2005). The least-square imageregistration method employed in this study used the fourth imagein each data set as a reference and the motion parameters wereestimated for the time-series set.

After motion correction, the data were detrended to eliminatelinear drifts in the data set. Three trials were recorded in case headmotion was problematic in our data set. However, since head mo-tion was minimal; the three saccade data trials were concatenated.Thus, our saccade dataset was 14 min in total. The same procedurewas used with the vergence data.

The functional MRI blood oxygenation level dependent (BOLD)signal analyzed using a general linear model (GLM) method hasbeen reported to be correlated to direct neuronal measurements(Attwell & Iadecola, 2002; Logothetis & Wandell, 2004). Hence,the fMRI time series data within this study were analyzed with aGLM where each voxel of the entire brain was correlated with ahemodynamic model of our data set. Due to the variations of thehemodynamic response function, a data driven independent com-ponent analysis was used to obtain a reference vector correspond-ing to the experimental stimulus (Berns, Song, & Mao 1999;

0 40 80 120 160 200 240 280Time (sec)

0

1

Task

Random Random Random Random

0 20 40

-10

0

10

Sacc

ade

(º)

0 20 40Time (sec)

0

2

4

Verg

ence

Ang

le (º

)

Experimental Design

Time (sec)

Schematic of Visual Target Locations for fMRI Eye Movements

Predictable Predictable Predictable

Random Saccadic Task Random Vergence Task

Predictable Vergence Task

(C) Block Design of Random versus Predictable Stimuli

40 00

2

4

Verg

ence

Ang

le (º

)

2040 0

-10

0

10

Predictable Saccadic Task

20

Far Target

Middle TargetNear Target

MidlineTarget

10 DegTarget

10 DegTarget

Time (sec)Time (sec)

Sacc

ade

(º)

(B)

(A)

Fig. 1. Experimental set-up and design. The schematic of the custom fMRI compatible LEDs for the saccade (plot A left) and vergence (plot A right) experiments. The LEDswould illuminate in a random or a predictable pattern (plot B). A block design protocol is used where the ‘‘on” stimuli are the predictable and ‘‘off’ stimuli are the random eyemovements (plot C).

2166 T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175

Calhoun, Kiehl, & Pearlson, 2008; Gavrilescu et al., 2008). Probabi-listic independent component analysis available through the ME-LODIC software from FSL was used to calculate the independentsignal sources (Beckmann & Smith, 2004). The signal source thatcorresponded to our block design was the reference vector usedto correlate each voxel within our data set. Using the GLM analysis,only data that attained a minimum threshold of functional activitycorresponding to a z-score of 2.8 (two tail p = 0.005) were furtheranalyzed.

Individual anatomical and functional brain maps were trans-formed into the standardized Talairach Tournoux coordinate space(Talairach & Tournoux, 1988). Functional data (multiple regression

coefficients) were spatially low-pass filtered using a Gaussian ker-nel (6 mm full width half maximum) and then merged by combin-ing coefficient values for each interpolated voxel across allparticipants. The combination of individual voxel probabilitythreshold and the cluster size threshold (11 voxels rounded to avolume of 750 mm3 for our data set) yielded the equivalent of awhole-brain corrected for multiple comparison significance levelof a < 0.001. The cluster size was determined using the AFNI Alpha-Sim program (Ward, 2000). This program estimates the overall sig-nificance level by determining the probability of false detectionthrough Monte Carlo simulation. Through individual voxel proba-bility thresholding and minimum cluster size thresholding, the

T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175 2167

probability of false detection is determined from the frequencycount of cluster sizes. The program does assume the underlyingpopulation of voxel intensity has a normal distribution. Oursimulation used 10,000 Monte Carlo iterations, assumed a clusterconnection of the nearest neighbor, voxel dimension of 3.4 �3.4 � 5 mm and sought a significance level of 0.001. Hence, a clus-ter size of 750 mm3 or greater corresponded to p < 0.001 correctedfor multiple comparisons. Individual maps of t-statistics weresmoothed with a Gaussian kernel of 6 mm full width, half maxi-mum to account for inter-individual anatomical variation (Binder,Liebenthal, Possing, Medler, & Ward, 2004; Lewis, Brefczynski,Phinney, Janik, & DeYoe, 2005; Schmid, Rees, Frith, & Barnes,2001). The functional data are displayed as a z-score shown inthe figure scale bar. The skull was removed since it is not relevantto our experiment.

The functional activity for the saccadic network is well estab-lished and is reviewed by Pierrot-Deseilligny et al. (2004). Hence,we hypothesized that our experiment would provoke activationin the frontal eye field (FEF), the supplementary eye field (SEF),the dorsolateral prefrontal cortex (DLPFC), the parietal eye field(PEF), and the anterior and posterior cingulate cortex during thesaccadic experiment. Functional MRI studies have shown that thesaccade related area of FEF is localized in the upper portion ofthe anterior wall of the precentral sulcus (Rosano et al., 2002)and in a review paper is described as being in the vicinity of theprecentral sulcus and/or in the depth of the caudalmost part ofthe superior frontal sulcus (Paus, 1996). The human SEF is locatedon the medial surface of the superior frontal gyrus, in the upperpart of the paracentral sulcus (Grosbras, Lobel, Van de Moortele,LeBihan, & Berthoz, 1999). The dorsolateral prefrontal cortex is lo-cated within Brodmann Area (BA) 46/9 (Pierrot-Deseilligny et al.,2005). The parietal eye field is located in the lateral intraparietalarea (Pierrot-Deseilligny et al., 2004). The anterior and posteriorcingulate cortexes are located in Brodmann Areas (BA) 24 and 23respectively (Pierrot-Deseilligny et al., 2004). These regions wereinitially investigated as well as other areas within the brain. Anindividual subject analysis was performed and only regions thatshowed significant activation in all eight subjects are reported.

2.6. Statistical analysis

All data were converted to Talairach Tournoux normalizedspace which inherently smoothes the data, hence separate spatialsmoothing was not conducted prior to Tailarach transform. Wehypothesize that the vergence and saccade circuits would showspatial differentiation because single cell recordings in primatesshow differentiation within the FEF (Gamlin & Yoon, 2000). Thenull hypothesis is that no difference in the signal amplitude willbe observed between the vergence and saccade data sets. Hence,to determine if significant spatial differences existed between thesaccade and vergence data sets, the beta weights from the generallinear model were compared with a paired t-test in a voxel-wisebasis to create a statistical significance spatial map. In this situa-tion, both the paired t-test (computed with AFNI 3dttest) and amixed-effects ANOVA with two factors (experimental condition,vergence versus saccade experiments, as the fixed factor; and sub-ject as the random factor), (computed with AFNI 3dANOVA2) arethe same. Data were thresholded for an absolute t-value greaterthan 1.9 (two-tailed p-value = 0.10).

If an area had similar levels of activation (which can result froman area that is significantly activated or not activated within boththe saccade and vergence data sets) then significant differences be-tween the functional data sets would not be observed. However, ifone data set has activation and the other does not, then the statis-tical significant spatial maps would quantify these areas. For exam-ple, we hypothesize that the functional activity within FEF will be

more anterior during vergence movements compared to saccadicmovements which has been reported studying single cell record-ings from primates (Gamlin & Yoon, 2000). If the data support spa-tial differentiation within FEF, then activity more anterior withinFEF will be active during vergent but not saccadic movements.Similarly, if the data support this hypothesis then more activity lo-cated posterior within FEF will be observed with saccadic but notwith vergent movements. The statistical difference spatial mapsare displayed using the scaled t-value as the color overlay uponstandardized anatomical images to show the spatial location of sig-nificantly different areas of activation.

3. Results

3.1. Eye movement results

Typical eye movement responses recorded prior to the imagingsessions are shown in Fig. 2. Saccadic responses, defined as the sumof the calibrated left- and right-eye movements divided by two, areshown in Fig. 2 plot A, where rightward movements are positive.Vergence is defined as the difference between the calibrated left-and right-eye movements where convergent movements are plot-ted as positive (Fig. 2 plot B). Both the position (deg) and velocity(deg/s) traces as a function of time for eye movements from stepstimuli with a random onset delay (dashed lines) and from the stepstimuli with a predictable timing and magnitude sequence (solidlines) are plotted. Responses to predictable stimuli have a reducedlatency compared to responses from random stimuli. Anticipatorydrifts are observed in the responses to predictable stimuli but notin the responses to stimuli with the random onset delay. The la-tency was quantified for saccadic (plot 2C) and vergent (plot 2D)responses from predictable stimuli (light gray) and from stimuliwith the random onset delay (dark gray) from one subject. The his-togram plots support that this protocol does stimulate predictiveand non-predictive behaviors as observed by the differences inthe latencies between the two types of movements.

3.2. Functional MRI results

Six motion related parameters were computed and correctedfor each subject during each of the scans. The largest average de-gree of rotation was 0.14� ± 0.13� and 0.17� ± 0.14� in the pitchdirection for the saccade and vergence datasets respectively. Thelargest average amount of movement within a plane was0.27 ± 0.19 mm and 0.29 ± 0.27 mm in the inferior to superiorplane for the saccade and vergence datasets respectively. We donot feel head motion was problematic. Thus, all data were utilizedfor this analysis.

The averaged functional activity from the eight subjects per-forming the predictable versus random saccadic oculomotor taskis shown in Fig. 3, left portion of the figure. Fig. 3, shows three axialslices and one sagittal slice displaying the anatomy of functionalactivity located within the interior cortical and subcortical loca-tions. Table 1 lists the peak activation in Talairach Tournoux coor-dinates for a given anatomical location of the averaged data setwith the corresponding z-score and Brodmann Area (BA) fromthe saccadic task. Data are also analyzed individually to determinehow many of the eight subjects showed activation for a given ana-tomical location. Only areas that showed significant activation forall subjects are included in the tables. For the saccadic functionalactivation induced from the random versus predictable visual ocu-lomotor visual tasks, activity is observed in the vicinity of the supe-rior frontal sulcus (denoted with a blue arrow) and precentralsulcus (denoted with a green arrow), also defined as the frontaleye fields (Paus, 1996). Functional activity is also observed in the

Responses from Predictable Stimuli Responses from Random Stimuli

-0.25 0.00 0.25 0.50 0.75Time (sec)

0

1

2

Posi

tion

(deg

)

0

5

10

15Velocity (deg/sec)

AnticipatoryMovement

AnticipatoryMovement

Target Onset

-0.25 0.00 0.25 0.50 0.750

5

100

100

200

300

400

Posi

tion

(deg

)

Velocity (deg/sec)

Target Onset

AnticipatoryMovement

AnticipatoryMovement

Time (sec)

-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5Latency (sec)

0

5

10

15

20

25

Num

ber o

f Res

pons

es

-0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5-0.30

5

10

15

20

25

Latency (sec)

Num

ber o

f Res

pons

es

(A) Saccade Responses (B) Vergence Responses

(C) Saccade Latency (D) Vergence Latency

Responses from Predictable Stimuli Responses from Random Stimuli

Fig. 2. Eye movement recordings from random (dashed line) and predictable step stimuli (solid lines) for saccade (plot A) and vergence (plot B) responses. Position (deg) andvelocity (deg/s) traces are plotted. Anticipatory movements are observed with the predictable responses denoted by an arrow. Latency histograms of one subject for saccadic(plot C) and vergent (plot D) responses from predictable (light gray) and random (dark gray) stimuli are shown.

2168 T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175

medial frontal gyrus; the supplementary eye field; the dorsolateralprefrontal cortex; the ventral lateral prefrontal cortex; the intrapa-rietal sulcus, referred to as the parietal eye field (Pierrot-Deseil-ligny et al., 2004); the cuneus; the precuneus; the anterior andposterior cingulate; and the cerebellar vermis.

The vergence functional activity induced from the predictableversus random symmetrical step stimuli is shown in Fig. 3 on theright portion of the figure. To facilitate comparison between thedata sets the same axial and sagittal sections are plotted and bothfunctional and anatomical data are normalized to the stereotacticTalairach and Tournoux space. The peak activation from the groupanalysis of each region of interest, the Brodmann Area, and the cor-responding z-score is reported in Table 2 for vergence. Only regionsof interest in which activation was observed in all eight individualsubject data sets are tabulated. Similar areas are activated using apredictive versus random vergence task compared to the predic-tive saccadic data set. Those areas are the dorsolateral prefrontalcortex, the cuneus, the precuneus/lingual gyrus, the superior pari-etal lobe, as well as the anterior and posterior cingulate. Althoughsimilar cortical areas are activated in the vergence and saccadicpredictive tasks, other areas show differentiation in spatial locationfor the peak activation.

Within the frontal lobe, the two main areas that showed differ-entiation between the saccade and vergence predictive tasks werethe FEF and the SEF. The differentiation between FEF and SEF wasfirst observed within the group average data (Table 1 versus Table2) quantified via Talairach Tournoux coordinates. An individualanalysis was conducted for the saccade and vergence data setsshown in Table 3 for the FEF and Table 4 for the SEF which showsthe z-score and Talairach Tournoux coordinates of the peak activa-tion per subject. Fig. 4A shows the functional activation using theGLM analysis for vergence and saccades predictive experiment.Differences between the data sets were analyzed with the pairedt-test computed using the beta weights from the GLM analysison a voxel-wise basis plotted in Fig. 4B. The red arrows labeledwith FEFv and SEFv show the vergence activity is significantlymore anterior than the saccadic activity denoted with yellow ar-rows labeled with FEFs and SEFs. The axial slice 53 mm abovethe bicommissural plane is plotted in Fig. 4 to show significant dif-ferences between the FEF and SEF areas of vergence and saccadeactivity. The paired t-test T values to analyze spatially significantdifferences and corresponding z-score values from the functionalactivity studied using the GLM analysis for saccade and vergencedata set are shown in Table 5 for the positive paired t-test values

Vergence

SEF

PEF PEF

BA7

FEFFEF

SEF

PEF PEF

BA7

FEFFEF

DLPFC DLPFC

BA40

BA32

Cuneus

BA31

BA40

BA32

DLPFC DLPFC

BA40

Cuneus

BA40

BA31

VLPFCVLPFC VLPFC VLPFC

Vermis IV / V

Z = 35 S

Z = 3 S

Medial Frontal Gyrus

Vergence

PosteriorCingulate

PosteriorCingulate

Medial Frontal Gyrus

SaccadeL R L R

Z = 51 S

Left to Right = 4L

Vermis IV / VPAPA

Precentral Sulcus

Superior Frontal Sulcus

Superior Frontal Sulcus

Precentral Sulcus

Zmin = 2.88 Zmax = 10

Fig. 3. Functional activation for the group analysis of predictable versus random eye movements for the saccade data set (left side) and the vergence data set (right side).Nomenclature is frontal eye fields (FEF), supplementary eye fields (SEF), dorsolateral prefrontal cortex (DLPFC), parietal eye field (PEF), ventral lateral prefrontal cortex(VLPFC) and Brodmann Area (BA). The superior frontal sulcus is denoted with blue arrows and the precentral sulcus is denoted with green arrows. The number of mm abovethe bicommissural plane is indicated for each axial section. The number of millimeter into the left hemisphere from the midsagittal plane is denoted for the sagittalsection. The functional activation denoted as a z-score from a minimum of 2.88 to a maximum value of 10 is overlaid onto a standardized Talairach and Tournoux normalizedimage.

T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175 2169

Table 1Average peak activation of the eight subjects for the predictable versus random saccadic oculomotor task in Talairach Tournouxcoordinates with the level of significance denoted as a z-score. For the X axis: positive is right (R) and negative is left (L); for the Y axis:positive is anterior (A) and negative is posterior (P); for the Z axis: positive is superior (S) and negative is inferior (I).

Region BrodmannArea

X(mm)

Y(mm)

Z(mm)

z-score

Frontal eye field, superior middle frontal gyrus, precentral gyrus 8/6 30R �5P 50S 5.758/6 �24L �2P 50S 5.40

Medial frontal gyrus 8 1R 39A 48S 7.75

Supplementary eye field, medial frontal gyrus 6 �2L 4A 55S 7.75

Dorsolateral prefrontal cortex, medial frontal gyrus 9 42R 32A 35S 7.68�47L 12A 37S 9.89

Superior ventral lateral prefrontal cortex 44/45 53R 20A 20S 8.64�52L 16A 28S 9.13

Inferior ventral lateral prefrontal cortex, inferior frontal gyrus, precentralgyrus

45/47 49R 15A 3S 8.64�50L 15A 6S 8.89

Parietal eye field, inferior parietal area 40 38R �48P 51S 7.68�54L �47P 36S 8.89

Cuneus, lingual gyrus 18 �6L �72P 11S 7.92

Cuneus 17 �6L �79P 6S 7.92

Precuneus, inferior parietal area, angular gyrus 39 40R �64P 36S 7.92�43L �61P 39S 7.68

Precuneus 7 1R �74P 40S 9.13

Superior parietal area 7 30R �65P 48S 9.38�35L �61P 50S 8.64

Posterior cingulate 31 2R �56P 27S 7.2230 5R �56P 11S 8.89

�8L �56P 10S 8.6429 7R �46P 10S 9.63

�5L �45P 8S 9.38

Anterior cingulate, medial frontal gyrus 32 4R 55A 5S 8.64�7L 46A 1S 7.68

Cerebellar vermis IV/V 4R �57P �2I 9.13

Superior colliculus 4R �42P 7S 5.20

2170 T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175

and Table 6 for the negative paired t-test values. These TalairachTournoux coordinates highlighted by the red (vergence) and yel-low (saccade) arrows differentiating FEF and SEF are reported inTables 5 and 6. These results support the hypothesis that func-tional activity for FEF and SEF are significantly more anterior forvergence movements compared to saccade movements.

4. Discussion

Functional activity in the vergence and saccade networksevoked using a predictive versus a random oculomotor taskshowed both shared and spatially distinct cortical resources.Although prediction is the primary variable studied, it is also pos-sible that the differences observed may in part be due to differ-ences in frequency, amplitude and/or the speed of the movements.

4.1. Activation due to predictive eye movements

The dorsolateral prefrontal cortex (DLPFC) showed similar areasof peak activation induced via a predictive versus random oculo-motor task comparing the saccade and vergence data sets. Afore-mentioned, numerous studies suggest that the DLPFC is involvedin memory and prediction (Pierrot-Deseilligny et al. (2004) for re-view). Short-term memory studies show DLPFC activity associatedwith saccades (Baumann, Frank, Rutschmann, & Greenlee, 2007;Müri & Nyffeler, 2008; Ozyurt, Rutschmann, & Greenlee, 2006)and n-back memory tasks (McMillan, Laird, Witt, & Meyerand2007; Owen, McMillan, Laird, & Bullmore, 2005; Pierrot-Deseil-

ligny et al., 2005; Tsuchida & Fellows, 2009). Patients with lesions,specifically within DLPFC, showed a decrease in the percentage ofanticipatory saccadic movements when viewing a predictive sacc-adic sequence compared to control subjects (Pierrot-Deseillignyet al., 2004). Our investigation used predictive oculomotor taskswhere the timing sequence is potentially stored in short-termmemory. Our results (Fig. 3) support the theory that the DLPFC isa shared cortical area between the vergence and saccadic networksto facilitate prediction and potentially short-term memory.

The anterior cingulate has been suggested to be involved in pre-diction during fMRI studies using saccades (Simó, Krisky, & Swee-ney, 2005) and smooth pursuit (Schmid et al., 2001). Studiessupport that the anterior cingulate is activated in memory guidedsaccades versus visually-guided saccades using fMRI (Heide et al.,2003; Petit, Courtney, Ungerleider, & Haxby, 1998) and PET (Inoue,Mikami, Ando, & Tsukada, 2004). Our study uses prediction and vi-sual memory to remember where the visual targets are located.Hence, our data support the hypothesis that the anterior cingulateis involved in predictive and visual memory processes. Less isknown about the posterior cingulate’s role in saccades (Pierrot-Deseilligny et al. (2004) for review).

A review of the ventral lateral prefrontal cortex (VLPFC) sug-gests it is involved in working memory and task switching (Badre& Wagner, 2007). Using fMRI in humans during a visual object cat-egorization and recognition task, authors conclude this area iswhere vision and memory meet (Schendan & Stern, 2008). Our re-sults show activation in the VLPFC during the predictive versusrandom vergent and saccadic tasks. Our task did utilize visualworking memory needed for the subjects to predict the oculomotor

Table 2Average peak activation of the predictable versus random vergence oculomotor task in Talairach Tournoux coordinates with the level ofsignificance denoted as a z-score. For the X axis: positive is right (R) and negative is left (L); for the Y axis: negative is posterior (P) andpositive is anterior (A); for Z axis: positive is superior (S) and negative is inferior (I).

Region BrodmannArea

X(mm)

Y(mm)

Z(mm)

z-Score

Frontal eye field, superior middle frontal gyrus, precentral gyrus 8/6 30R 5A 50S 4.878/6 �24L 4A 50S 5.29

Medial frontal gyrus 8 1R 48A 44S 7.68

Supplementary eye field, medial frontal gyrus 6 0 14A 51S 7.45

Dorsolateral prefrontal cortex, medial frontal gyrus 9 37R 36A 31S 6.54�50L 11A 33S 6.54

Superior ventral lateral prefrontal cortex, inferior frontal gyrus 44/45 51R 12A 19S 5.87�54L 13A 19S 6.76

Inferior ventral lateral prefrontal cortex, inferior frontal gyrus, precentralgyrus

45/47 �48L 24A 5S 7.22

53R 18A 7S 5.68

Parietal eye field, inferior parietal area 40 47R �46P 47S 6.3140 �50L �46P 44S 6.76

Cuneus, lingual gyrus 18 �2L �85P 1S 6.99

Cuneus 17 �5L �88P �2I 7.68

Precuneus, inferior parietal area, angular gyrus 39 43R �61P 39S 6.31�44L �64P 38S 6.31

Precuneus 7 6R �75P 44S 7.45

Superior parietal area 7 34R �63P 50S 6.99�34L �63P 47S 6.76

Posterior cingulate 31 �2L �67P 30S 6.0930 11R �59P 15S 6.09

�10L �54P 14S 6.5429 7R �47P 4S 7.45

�8L �45P 11S 6.54

Anterior cingulate, medial frontal gyrus 32 4R 52A 3S 6.09�4L 49A 2S 5.65

Cerebellar vermis IV/V 1R �49P 1S 6.76

Table 3Individual subject analysis for the peak activation given in Talairach Tournoux spatial coordinates for FEF for the predictive versus random saccade and vergence oculomotortasks. X is the left (L) and right (R) direction, Y is the anterior (A) and posterior (P) direction and Z is the superior (S) and inferior (I) direction. The statistical significance istabulated as the z-score. The average and standard deviation is reported.

Subject Saccade task Vergence task

X (mm) Y (mm) Z (mm) z-Score X (mm) Y (mm) Z (mm) z-Score

1 36R �4P 51S 4.6 29R 10A 51S 3.9�36L �7P 51S 2.8 �26L 3A 51S 3.2

2 31R �2P 54S 6.7 35R 15A 49S 6.2�27L �6P 54S 4.9 �28L 15A 48S 5.0

3 24R �7P 54S 2.6 21R 4A 49S 2.6�31L �3P 49S 3.3 �21L 0 49S 3.8

4 21R �2P 50S 4.2 28R 5A 50S 6.1�24L �2P 50S 5.6 �31L 9A 45S 6.6

5 24R 2A 55S 5.8 31R 6A 52S 5.4�24L �1P 52S 3.4 �21L 2A 52S 4.6

6 31R �2P 48S 4.7 31R 5A 53S 6.1�21L �6P 47S 5.9 �21L 9A 53S 7.6

7 24R �5P 49S 7.2 31R 6A 53S 6.6�21L �5P 49S 4.8 �21L 9A 53S 3.9

8 22R �6P 51S 6.6 21R 5A 53S 3.7�24L 1A 49S 8.7 �27L 8A 45S 4.2

Average ± Standard Deviation 27R ± 5.3 �3.3P ± 2.9 52S ± 2.6 5.3 ± 1.5 28.4R ± 5.0 7.0A ± 3.7 51.3 ± 1.8 5.1 ± 1.5�26L ± 5.2 �3.6P ± 2.8 50S ± 2.2 4.9 ± 1.9 �24.5L ± 4.0 6.9A ± 4.9 49.5 ± 3.3 4.9 ± 1.5

T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175 2171

task; hence our data support that the VLPFC is in part involved inspatial working memory used to predict the next visual stimulus.

A recent review discusses the role of the cerebellum in saccadicmovements for motor learning when training such as prediction is

utilized (Schubert & Zee, 2010). Functional MRI studies have dem-onstrated that the cerebellum is active during tasks requiring thesubject to make a movement with predictable versus non-predict-able target timing (Sakai et al., 2000). Clinical studies comparing

Table 4Individual subject analysis for the peak activation in Talairach Tournoux spatial coordinates for the SEF for the predictive versus random saccade and vergence oculomotor tasks. X(mm) is the left (L) and right (R) direction, Y (mm) is the anterior (A) and posterior (P) direction and Z (mm) is the superior (S) and inferior (I) direction above the bicommissuralplane. The statistical significance is tabulated as the z-score.

Subject Saccade task Vergence task

X (mm) Y (mm) Z (mm) z-Score X (mm) Y (mm) Z (mm) z-Score

1 �7L �7P 51S 3.7 2R 20A 46S 3.72 �3L �2P 48S 4.6 0 19A 48S 9.43 0 4A 49S 3.7 4R 14A 50S 1.84 �7L 2A 50S 8.6 3R 12A 47S 4.55 �4L 2A 47S 7.6 0 13A 52S 6.46 �7L 5A 48S 3.5 0 8A 53S 4.17 �3L 6A 54S 9.2 3R 19A 49S 6.78 �7L 8A 54S 11.1 0 18A 53S 5.5

Average ± standard deviation �4.8L ± 2.7 2.3A ± 4.8 50S ± 2.7 6.5 ± 3.0 1.5R ± 1.7 15A ± 4.3 50S ± 2.7 5.3 ± 2.3

Saccadic Activity Vergence Activity

Saccade minus VergenceVoxel Wise Positive

Paired t-Test Analysis

L Z = 53 R

FEFV FEFVFEFS FEFS

(A) (B)General Linear Model Analysis

Tmin = 1.9 Tmax = 11

L Z = 53 R

Zmin = 2.88 Zmax = 10

FEFSFEFS

SEFS

L Z = 53 R

FEFVFEFV

Superior Frontal Sulcus

Precentral Sulcus

Superior Frontal Sulcus

SEFV

Vergence minus SaccadeVoxel Wise NegativePaired t-Test Analysis

FEFSFEFS

FEFV FEFV

L Z = 53 R

Tmin = -1.9 Tmax = -11

SEFVSEFS

Fig. 4. Axial images showing differentiation between FEF and SEF. Functional activity using the GLM analysis is shown in portion A. The voxel wise positive and negativepaired t-test shows significant differentiation between FEF and SEF for vergence and saccades (portion B). The GLM analysis reports activity using a z-score from 2.88 to 10.The paired t-test using the beta weights from the GLM analysis reports significant differences from T = ±1.9 to ±11 (two-tailed p-value = 0.10 to p < 0.0001). Functional activityand paired t-test significant differences are overlaid onto Talairach Tournoux normalized axial structural images. The number of mm above the bicommissural plane isindicated for each axial section. L: left; R: right. The superior frontal sulcus is denoted with blue arrows and the precentral sulcus is denoted with green arrows in portion A.The significant differences are denoted with red arrows for vergence (FEFv and SEFv) and yellow arrows for saccades (FEFs and SEFs).

Table 5Saccade minus vergence dataset/positive paired t-test statistics showing differentiation between FEF and SEF in comparing predictive versus random saccade and vergence tasks.

Region Talairach Tournoux stereotactic coordinates z-Score in saccadedata set

z-Score in vergencedata set

Positive pairedt-test T value

X (mm) Y (mm) Z (mm)

FEF (activity in saccades) 34R �5P 53S 3.9 <1 3.4�28L �3P 53S 3.7 <1 5.3

FEF (activity in vergence) 31R 6A 53S <1 4.2 3.4�34L 2A 53S <1 4.7 6.1

SEF (activity in saccades) �7R �2P 53S 2.4 <1 3.7

SEF (activity in vergence) 5R 9A 53S <1 2.0 3.7

2172 T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175

predictable to non-predictable saccadic responses from patientswith cerebellar lesions conclude the cerebellum is crucial forsynchronizing saccades with learned or planned temporal events(Sailer, Eggert, & Straube, 2005). Our data support that the cerebel-lar vermis IV/V is involved in temporal prediction for both saccadicand vergent movements.

One differentiation observed is activation in the superiorcolliculus for the saccadic data set but not the vergence data set.The superior colliculus has been studied through single cell record-ings to modulate its activity based upon saccadic motor preparation

(Sparks (1999) for review). Hence, we attribute the functional activa-tion in the saccadic data set to the motor preparation evoked throughthe predictive nature of our visual stimulation.

4.2. Differentiation between vergence and saccade prediction taskswithin the frontal eye field

Functional MRI studies using saccades have shown for humansthe bilateral frontal eye fields (FEF) are located at the vicinity of theprecentral sulcus and the superior frontal sulcus for intentional or

Table 6Vergence minus saccade datasets/negative paired t-test statistics showing differentiation between FEF and SEF in comparing predictive versus random saccade and vergencetasks.

Region Talairach Tournoux stereotactic coordinates z-Score in saccadedata set

z-Score in vergencedata set

Negative pairedt-test T value

X (mm) Y (mm) Z (mm)

FEF (activity in saccades) 34R �8P 53S 2.2 <1 �5.9�41L �9P 53S 2.0 <1 �3.2

FEF (activity in vergence) 31R 5A 53S <1 2.8 �4.8

�29L 2A 53S <1 2.7 �2.7SEF (activity in saccades) 7R �2P 53S 2.4 <1 �3.7

SEF (activity in vergence) 5R 9A 53S <1 2.6 �3.1

T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175 2173

voluntary saccades (Amiez, Kostopoulos, Champod, & Petrides,2006; Berman et al., 1999; Lynch & Tian, 2006; Paus, 1996; Rosanoet al., 2002; Schraa-Tam et al., 2009), predictive saccades (Con-nolly, Goodale, Goltz, & Munoz, 2005; Milea et al., 2007) or mem-ory guided saccades (Bodis-Wollner et al., 1997; Brown et al.,2004; Curtis & D’Esposito, 2006; Kastner et al., 2007; Ozyurtet al., 2006; Srimal & Curtis, 2008). Previous investigations reportdistinct anatomical locations within the FEF for smooth pursuitand saccadic eye movements when studying humans within fMRIstudies (Berman et al., 1999; Petit, Clark, Ingeholm, & Haxby,1997; Petit & Haxby, 1999; Rosano et al., 2002), studying humansusing PET (O’Driscoll et al., 2000), and single cell studies from pri-mates (Gottlieb, MacAvoy, & Bruce, 1994; Shi, Friedman, & Bruce,1998; Stanton, Friedman, Dias, & Bruce, 2005; Tian & Lynch,1996). In rhesus primates, Gamlin and Yoon report that a distinctarea located anterior to cells responsible for saccadic commandsencode for vergence and ocular accommodation; they recommendthe classic definition of the FEF be expanded to include this area(Gamlin & Yoon, 2000). Our findings using human subjects supporta distinct location within the FEF for vergence which is more ante-rior compared to the FEF for saccadic tasks when vergence and sac-cades are evoked by predictable versus random step stimuli,reported in Tables 3, 5 and 6 and Fig. 4.

4.3. Differentiation between vergence and saccade predictive taskswithin the supplementary eye field

The supplementary eye field (SEF) is located within the medialsurface of the frontal gyrus and is thought to prepare saccadic mo-tor programs (Pierrot-Deseilligny et al., 2004). Studies suggest thatthe SEF is a higher order cognitive area responsible for prediction(Nyffeler, Rivaud-Pechoux, Wattiez, & Gaymard, 2008; Uchida,Lu, Ohmae, Takahashi, & Kitazawa, 2007) or the cortical area thatcontrols the production of an error signal (Schall & Boucher,2007). Others speculate the SEF modulates attention (Konen, Kleis-er, Bremmer, & Seitz, 2007). Similar to FEF, studies have reporteddifferentiation between saccades and smooth pursuit movementswithin SEF (Lynch & Tian, 2006; Petit & Haxby, 1999). Our datashow distinct spatial locations within the SEF for the functionalactivity of saccade and vergence eye movements induced from apredictive versus random step stimulus. Vergence functional activ-ity is more anterior than the functional activity for saccades, re-ported in Tables 4–6 and Fig. 4.

4.4. Differentiation of functional activity between saccade andvergence data sets

The statistically significant differences between the positive andnegative paired t-tests showed that the datasets were not equal.There were more significant differences in the positive paired t-testor when comparing the saccade minus the vergence data sets thanin the negative paired t-test or when comparing the vergence

minus the saccade datasets (Fig. 4B). There are two potential expla-nations for this observation. First, the magnitude of the saccadicstimulus was ±10� compared to a 2� disparity change for vergence.Second, a difference in the physiology may exist. Functional imag-ing studies of smooth pursuit and saccade movements reveal a sig-nificantly smaller region of activation in both SEF and FEF (Petit &Haxby, 1999) in smooth pursuit activity compared to saccadicactivity which is also observed in primate studies (Tian & Lynch,1995; Tian & Lynch, 1996). Hence, we speculate this relationshipmay also exist between the saccadic and vergence circuits.

4.5. Saccade and vergence interaction

Numerous behavioral studies discuss the interaction betweenthe saccadic and vergent systems which began with the work ofZee, Fitzgibbon, and Optican (1992). Our laboratory and other inves-tigators have published that even when symmetrical vergence stim-uli are presented to a subject, many of the responses containsaccades (Coubard & Kapoula, 2008; Semmlow et al., 2008; Semm-low et al., 2009; van Leeuwen, Collewijn, & Erkelens, 1998). How-ever, these saccades are small in magnitude compared to thevergence stimulus where most saccades are less than 1� for a 4� ver-gence change (Semmlow et al., 2008). Similarly, studies have shownthat with saccadic movement, a transient divergent and then con-vergent movement is observed (Collewijn, Erkelens, & Steinman,1997; Vernet & Kapoula, 2009). Therefore, it can be expected thatthe vergence and saccade oculomotor systems would share somepredictive centers which is true in many of the regions of interest re-ported in the current study. However, despite these interactions be-tween the vergence and saccade circuits, our results also reportdifferentiation between the networks in the FEF and the SEF.

5. Conclusion

This study of humans using the whole brain quantifies signifi-cant spatial differentiation between vergent and saccadic areas ofpeak activation in the FEF and SEF when predictable versus randomstimuli were presented. Several cortical sites were shared suggest-ing the oculomotor systems have commonalities that are not spe-cific to one network when studying predictive versus random eyemovement responses. Furthermore, our work supports that the fol-lowing cortical regions: the dorsolateral prefrontal cortex, anteriorand posterior cingulate, and ventral lateral prefrontal cortex are in-volved in predictive and short term working memory tasks duringoculomotor predictive movements. This research furthers ourunderstanding of which cortical areas are active when predictionis utilized during vergent and saccadic eye movements.

Acknowledgments

This research was supported in part by NSF CAREER BES-044713 to TLA and NIH 5R01NS049176 to BBB. The authors appre-

2174 T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175

ciate the comments from John Semmlow, PhD and the reviewerson the initial manuscript and Anna Barrett, MD on the activitywithin the FEF.

References

Akao, T., Mustari, M. J., Fukushima, J., Kurkin, S., & Fukushima, K. (2005). Dischargecharacteristics of pursuit neurons in MST during vergence eye movements.Journal of Neurophysiology, 93, 2415–2434.

Alvarez, T. L., Semmlow, J. L., Yuan, W., & Munoz, P. (2002). Comparison of disparityvergence system responses to predictable and non-predictable stimulations.Current Psychology of Cognition, 21, 243–261.

Amiez, C., Kostopoulos, P., Champod, A. S., & Petrides, M. (2006). Local morphologypredicts functional organization of the dorsal premotor region in the humanbrain. Journal of Neuroscience, 26, 2724–2731.

Attwell, D., & Iadecola, C. (2002). The neural basis of functional brain imagingsignals. Trends in Neurosciences, 25, 621–625.

Badre, D., & Wagner, A. (2007). Left ventrolateral prefrontal cortex and the cognitivecontrol of memory. Neuropsychologia, 45, 2883–2901.

Barnes, G. R., & Asselman, P. T. (1991). The mechanism of prediction in humansmooth pursuit eye movements. Journal of Physiology, 439, 439–461.

Baumann, O., Frank, G., Rutschmann, R. M., & Greenlee, M. W. (2007). Corticalactivation during sequences of memory-guided saccades: A functional MRIstudy. NeuroReport, 18, 451–455.

Beckmann, C. F., & Smith, S. M. (2004). Probabilistic independent componentanalysis for functional magnetic resonance imaging. IEEE Transactions onMedical Imaging, 23, 137–152.

Berman, R. A., Colby, C. L., Genovese, C. R., Voyvodic, J. T., Luna, B., Thulborn, K. R.,et al. (1999). Cortical networks subserving pursuit and saccadic eye movementsin humans: An FMRI study. Human Brain Mapping, 8, 209–225.

Berns, G. S., Song, A. W., & Mao, H. (1999). Continuous functional magneticresonance imaging reveals dynamic nonlinearities of ‘‘dose-response” curvesfor finger opposition. Journal of Neuroscience, 19, 1–6. RC17.

Binder, J. R., Liebenthal, E., Possing, E. T., Medler, D. A., & Ward, B. D. (2004). Neuralcorrelates of sensory and decision processes in auditory object identification.Nature Neuroscience, 7, 295–301.

Bodis-Wollner, I., Bucher, S. F., Seelos, K. C., Paulus, W., Reiser, M., & Oertel, W. H.(1997). Functional MRI mapping of occipital and frontal cortical activity duringvoluntary and imagined saccades. Neurology, 49, 416–420.

Brown, M. R., DeSouza, J. F., Goltz, H. C., Ford, K., Menon, R. S., Goodale, M. A., et al.(2004). Comparison of memory- and visually guided saccades using event-related fMRI. Journal of Neurophysiology, 91, 873–889.

Calhoun, V. D., Kiehl, K. A., & Pearlson, G. D. (2008). Modulation of temporallycoherent brain networks estimated using ICA at rest and during cognitive tasks.Human Brain Mapping, 29, 828–838.

Ciuffreda, K. J., & Tannen, B. (1995). Saccadic eye movements. In K. J. Ciuffreda & B.Tannen (Eds.), Eye movement basics for the clinician (pp. 36–45). New York:Mosby.

Collewijn, H., Erkelens, C. J., & Steinman, R. M. (1997). Trajectories of the humanbinocular fixation point during conjugate and non-conjugate gaze-shifts. VisionResearch, 37, 1049–1069.

Connolly, J. D., Goodale, M. A., Goltz, H. C., & Munoz, D. P. (2005). FMRI activation inthe human frontal eye field is correlated with saccadic reaction time. Journal ofNeurophysiology, 94, 605–611.

Coubard, O. A., & Kapoula, Z. (2008). Saccades during symmetrical vergence. GraefesArchive for Clinical and Experimental Ophthalmology, 246, 521–536.

Cox, R. W. (1996). AFNI: Software for analysis and visualization of functionalmagnetic resonance neuroimages. Computers and Biomedical Research, 29,162–173.

Curtis, C. E. (2006). Prefrontal and parietal contributions to spatial workingmemory. Neuroscience, 139, 173–180.

Curtis, C. E., & D’Esposito, M. (2006). Selection and maintenance of saccade goals inthe human frontal eye fields. Journal of Neurophysiology, 95, 3923–3927.

D’Esposito, M., Zarahn, E., Aguirre, G. K., & Rypma, B. (1999). The effect of normalaging on the coupling of neural activity to the bold hemodynamic response.Neuroimage, 10, 6–14.

Dodge, R. (1903). Five types of eye movement in the horizontal meridian plane ofthe field of regard. American Journal of Physiology, 8, 307–329.

Dodge, R. (1931). Conditions and consequences of human variability. In R. Dodge(Ed.), Conditions and consequences of human variability (pp. 15–35). New Haven:Yale University Press.

Gamlin, P. D., & Clarke, R. J. (1995). Single-unit activity in the primate nucleusreticularis tegmenti pontis related to vergence and ocular accommodation.Journal of Neurophysiology, 73, 2115–2119.

Gamlin, P. D., & Yoon, K. (2000). An area for vergence eye movement in primatefrontal cortex. Nature, 407, 1003–1007.

Gavrilescu, M., Stuart, G. W., Rossell, S., Henshall, K., McKay, C., Sergejew, A. A., et al.(2008). Functional connectivity estimation in fMRI data: Influence ofpreprocessing and time course selection. Human Brain Mapping, 29, 1040–1052.

Gaymard, B., Ploner, C. J., Rivaud, S., Vermersch, A. I., & Pierrot-Deseilligny, C. (1998).Cortical control of saccades. Experimental Brain Research, 123, 159–163.

Genovesio, A., & Ferraina, S. (2004). Integration of retinal disparity and fixation-distance related signals toward an egocentric coding of distance in the posteriorparietal cortex of primates. Journal of Neurophysiology, 91, 2670–2684.

Gnadt, J. W., & Mays, L. E. (1995). Neurons in monkey parietal area LIP are tuned foreye-movement parameters in three-dimensional space. Journal ofNeurophysiology, 73, 280–297.

Goldberg, M., Eggers, H., & Gouras, P. (2000). The ocular motor system. In E. R.Kandel, J. H. Schwartz, & T. M. Jessell (Eds.), Principles of neural science (3rd ed.,pp. 676). New York: McGraw-Hill, Health Professions Divisions.

Gottlieb, J. P., MacAvoy, M. G., & Bruce, C. J. (1994). Neural responses related tosmooth-pursuit eye movements and their correspondence with electricallyelicited smooth eye movements in the primate frontal eye field. Journal ofNeurophysiology, 72, 1634–1653.

Grosbras, M. H., Lobel, E., Van de Moortele, P. F., LeBihan, D., & Berthoz, A. (1999). Ananatomical landmark for the supplementary eye fields in human revealed withfunctional magnetic resonance imaging. Cerebral Cortex, 9, 705–711.

Heide, W., Binkofski, F., Seitz, R. J., Posse, S., Nitschke, M. F., Freund, H. J., et al.(2003). Activation of frontoparietal cortices during memorized triple-stepsequences of saccadic eye movements: An fMRI study. NeuroReport, 14,283–287.

Inoue, M., Mikami, A., Ando, I., & Tsukada, H. (2004). Functional brain mapping ofthe macaque related to spatial working memory as revealed by PET. Journal ofNeurophysiology, 92, 3056–3068.

Judge, S. J., & Cumming, B. G. (1986). Neurons in the monkey midbrain with activityrelated to vergence eye movement and accommodation. Journal ofNeurophysiology, 55, 915–930.

Kastner, S., DeSimone, K., Konen, C. S., Szczepanski, S. M., Weiner, K. S., & Schneider,K. A. (2007). Topographic maps in human frontal cortex revealed in memory-guided saccade and spatial working-memory tasks. Journal of Neurophysiology,97, 3494–3507.

Konen, C. S., Kleiser, R., Bremmer, F., & Seitz, R. J. (2007). Different corticalactivations during visuospatial attention and the intention to perform asaccade. Experimental Brain Research, 182, 333–341.

Kowler, E., & Steinman, R. M. (1979). The effect of expectations on slow oculomotorcontrol – I: Periodic target steps. Vision Research, 19, 633–646.

Krishnan, V. V., Farazian, F., & Stark, L. (1973). An analysis of latencies andprediction in the fusional vergence system. American Journal of OptometryArchives of the American Academy of Optometry, 50, 933–939.

Kumar, A. N., Han, Y., Garbutt, S., & Leigh, R. J. (2002). Properties of anticipatoryvergence responses. Investigative Ophthalmology & Visual Science, 43,2626–2632.

Leigh, R. J., & Zee, D. S. (2006). The saccadic system. In R. J. Leigh & D. S. Zee (Eds.),Neurology of eye movements (pp. 108–187). New York: Oxford University Press.

Lewis, J. W., Brefczynski, J. A., Phinney, R. E., Janik, J. J., & DeYoe, E. A. (2005). Distinctcortical pathways for processing tool versus animal sounds. Journal ofNeuroscience, 25, 5148–5158.

Logothetis, N. K., & Wandell, B. A. (2004). Interpreting the BOLD signal. AnnualReview of Physiology, 66, 735–769.

Luna, B., & Sweeney, J. A. (1999). Cognitive functional magnetic resonance imagingat very-high-field: Eye movement control. Topics in Magnetic Resonance Imaging,10, 3–15.

Lynch, J. C., & Tian, J. R. (2006). Cortico-cortical networks and cortico–subcorticalloops for the higher control of eye movements. Progress in Brain Research, 151,461–501.

Mays, L. E., & Porter, J. D. (1984). Neural control of vergence eye movements:Convergence and divergence neurons in midbrain. Journal of Neurophysiology,51, 1091–1108.

Mays, L. E., Porter, J. D., Gamlin, P. D., & Tello, C. A. (1986). Neural control of vergenceeye movements: Neurons encoding vergence velocity. Journal ofNeurophysiology, 56, 1007–1021.

McMillan, K. M., Laird, A. R., Witt, S. T., & Meyerand, M. E. (2007). Self-pacedworking memory: Validation of verbal variations of the n-back paradigm. BrainResearch, 1139, 133–142.

Milea, D., Lobel, E., Lehéricy, S., Leboucher, P., Pochon, J. B., Pierrot-Deseilligny, C.,et al. (2007). Prefrontal cortex is involved in internal decision of forthcomingsaccades. Neuroreport, 18, 1221–1224.

Miles, F. A., Fuller, J. H., Braitman, D. J., & Dow, B. M. (1980). Long-termadaptive changes in primate vestibuloocular reflex. III. Electrophysiologicalobservations in flocculus of normal monkeys. The Journal ofNeurophysiology, 43, 1437–1476.

Müri, R. M., & Nyffeler, T. (2008). Using transcranial magnetic stimulation to probedecision-making and memory. Progress in Brain Research, 171, 413–418.

Nitta, T., Akao, T., Kurkin, S., & Fukushima, K. (2008). Involvement of the cerebellardorsal vermis in vergence eye movements in monkeys. Cerebral Cortex, 18,1042–1057.

Nyffeler, T., Rivaud-Pechoux, S., Wattiez, N., & Gaymard, B. (2008). Involvement ofthe supplementary eye field in oculomotor predictive behavior. Journal ofCognitive Neuroscience, 20, 1583–1594.

Oakes, T. R., Johnstone, T., Ores Walsh, K. S., Greischar, L. L., Alexander, A. L., Fox, A.S., et al. (2005). Comparison of fMRI motion correction software tools.Neuroimage, 28, 529–543.

O’Driscoll, G. A., Wolff, A. L., Benkelfat, C., Florencio, P. S., Lal, S., & Evans, A. C.(2000). Functional neuroanatomy of smooth pursuit and predictive saccades.NeuroReport, 11, 1335–1340.

Owen, A. M., McMillan, K. M., Laird, A. R., & Bullmore, E. (2005). N-back workingmemory paradigm: A meta-analysis of normative functional neuroimagingstudies. Human Brain Mapping, 1, 46–59.

Ozyurt, J., Rutschmann, R. M., & Greenlee, M. W. (2006). Cortical activation duringmemory-guided saccades. NeuroReport, 17, 1005–1009.

T.L. Alvarez et al. / Vision Research 50 (2010) 2163–2175 2175

Paus, T. (1996). Location and function of the human frontal eye-field: A selectivereview. Neuropsychologia, 34, 475–483.

Petit, L., Clark, V. P., Ingeholm, J., & Haxby, J. V. (1997). Dissociation of saccade-related and pursuit-related activation in human frontal eye fields as revealed byfMRI. Journal of Neurophysiology, 77, 3386–3390.

Petit, L., Courtney, S. M., Ungerleider, L. G., & Haxby, J. V. (1998). Sustained activityin the medial wall during working memory delays. Human Brain Mapping, 8,209–225.

Petit, L., & Haxby, J. V. (1999). Functional anatomy of pursuit eye movements inhumans as revealed by fMRI. Journal of Neurophysiology, 82, 463–471.

Pierrot-Deseilligny, C., Milea, D., & Müri, R. M. (2004). Eye movement control by thecerebral cortex. Current Opinion in Neurology, 17, 17–25.

Pierrot-Deseilligny, C., Müri, R. M., Nyffeler, T., & Milea, D. (2005). The role of thehuman dorsolateral prefrontal cortex in ocular motor behavior. Annals of theNew York Academy of Sciences, 1039, 239–251.

Poggio, G. E. (1995). Mechanisms of stereopsis in monkey visual cortex. CerebralCortex, 5, 193–204.

Rashbass, C., & Westheimer, G. (1961). Disjunctive eye movements. Journal ofPhysiology (London), 159, 339–360.

Ron, S., Schmid, R., & Orpaz, D. (1989). Applying a model of saccadic prediction topatients’ saccadic eye movements. Brain, Behavior and Evolution, 33, 179–182.

Rosano, C., Krisky, C. M., Welling, J. S., Eddy, W. F., Luna, B., Thulborn, K. R., et al.(2002). Pursuit and saccadic eye movement subregions in human frontal eyefield: A high-resolution fMRI investigation. Cerebral Cortex, 12, 107–115.

Sailer, U., Eggert, T., & Straube, A. (2005). Impaired temporal prediction and eye-hand coordination in patients with cerebellar lesions. Behavioural BrainResearch, 160, 72–87.

Sakai, K., Hikosaka, O., Takino, R., Miyauchi, S., Nielsen, M., & Tamada, T. (2000).What and when: Parallel and convergent processing in motor control. Journal ofNeuroscience, 20, 2691–2700.

Schall, J. D., & Boucher, L. (2007). Executive control of gaze by the frontal lobes.Cognitive, Affective, & Behavioral Neuroscience, 7, 396–412.

Schendan, H. E., & Stern, C. E. (2008). Where vision meets memory: Prefrontal–posterior networks for visual object constancy during categorization andrecognition. Cerebral Cortex, 18, 1695–1711.

Schmid, A., Rees, G., Frith, C., & Barnes, G. (2001). An fMRI study of anticipation andlearning of smooth pursuit eye movements in humans. NeuroReport, 12,1409–1414.

Schraa-Tam, C. K., van Broekhoven, P., van der Geest, J. N., Frens, M. A., Smits, M., &van der Lugt, A. (2009). Cortical and cerebellar activation induced by reflexiveand voluntary saccades. Experimental Brain Research, 192, 175–187.

Schubert, M. C., & Zee, D. S. (2010). Saccade and vestibular ocular motor adaptation.Restorative Neurology and Neuroscience, 28, 9–18.

Semmlow, J. L., Chen, Y. F., Granger-Donnetti, B., & Alvarez, T. L. (2009). Correction ofsaccade-induced midline errors in responses to pure disparity vergence stimuli.Journal of Eye Movement, 2, 1–13.

Semmlow, J. L., Chen, Y. F., Pedrono, C., & Alvarez, T. (2008). Saccadic behaviorduring the response to pure disparity vergence stimuli. I: General properties.Journal of Eye Movement Research, 1, 1–11.

Shi, D., Friedman, H. R., & Bruce, C. J. (1998). Deficits in smooth-pursuit eyemovements after muscimol inactivation within the primate’s frontal eye field.Journal of Neurophysiology, 80, 458–464.

Simó, L. S., Krisky, C. M., & Sweeney, J. A. (2005). Functional neuroanatomy ofanticipatory behavior: Dissociation between sensory-driven and memory-driven systems. Cerebral Cortex, 15, 1982–1991.

Sparks, D. L. (1999). Conceptual issues related to the role of the superior colliculusin the control of gaze. Current Opinion in Neurobiology, 9, 698–707.

Srimal, R., & Curtis, C. E. (2008). Persistent neural activity during the maintenance ofspatial position in working memory. Neuroimage, 39, 455–468.

Stanton, G. B., Friedman, H. R., Dias, E. C., & Bruce, C. J. (2005). Cortical afferents tothe smooth-pursuit region of the macaque monkey’s frontal eye field.Experimental Brain Research, 165, 179–192.

Stark, L., Vossius, G., & Young, L. R. (1962). Predictive control of eye trackingmovement. IRE Transactions on Human Factors in Electronics, 3, 52–57.

Talairach, J., & Tournoux, P. (1988). In J. Talairach & P. Tournoux (Eds.), Co-planarstereotaxic atlas of the human brain (pp. 1–75). New York: Thieme.

Tian, J. R., & Lynch, J. C. (1995). Slow and saccadic eye movements evoked bymicrostimulation in the supplementary eye field of the cebus monkey. Journal ofNeurophysiology, 74, 2204–2210.

Tian, J. R., & Lynch, J. C. (1996). Functionally defined smooth and saccadic eyemovement subregions in the frontal eye field of Cebus monkeys. Journal ofNeurophysiology, 76, 2740–2753.

Tsuchida, A., & Fellows, L. K. (2009). Lesion evidence that two distinct regionswithin prefrontal cortex are critical for n-back performance in humans. Journalof Cognitive Neuroscience, 21, 2263–2275.

Uchida, Y., Lu, X., Ohmae, S., Takahashi, T., & Kitazawa, S. (2007). Neuronal activityrelated to reward size and rewarded target position in primate supplementaryeye field. Journal of Neuroscience, 27, 13750–13755.

van Leeuwen, A. F., Collewijn, H., & Erkelens, C. J. (1998). Dynamics ofhorizontal vergence movements: Interaction with horizontal and verticalsaccades and relation with monocular preferences. Vision Research, 38,3943–3954.

Vernet, M., & Kapoula, Z. (2009). Binocular motor coordination during saccades andfixations while reading: A magnitude and time analysis. Journal of Vision, 9,1–13.

Ward, B. D. (2000). Simultaneous inference for FMRI data. <http://web.mit.edu/seven/doc/AFNI98/AlphaSim.ps> 02/2010.

Zee, D. S., Fitzgibbon, E. J., & Optican, L. M. (1992). Saccade–vergence interactions inhuman. Journal of Neurophysiology, 68, 1624–1641.

Zhang, H., & Gamlin, P. D. (1998). Neurons in the posterior interposed nucleus of thecerebellum related to vergence and accommodation. I. Steady-statecharacteristics. Journal of Neurophysiology, 179, 1255–1269.