Abstract The development of visual and vestibular con-trol of smooth gaze adjustments was studied longitudi-nally in 3- to 18-week-old infants. Eye and head move-ments were measured with electro-oculography (EOG)and an optoelectronic system, respectively. The infantwas placed in a chair providing full support to the trunkbut allowing relatively free head movements. The chairwas positioned at the center of a striped-patterned drum.The chair and the drum were oscillated sinusoidally, ei-ther individually or in synchrony at 0.25 Hz. When thedrum oscillated around the infant (the optokinetic re-sponse condition, OKR), the gain of both smooth eyeand head tracking components was low up to 6 weeks ofage, after which the eye gain increased dramatically andthe lag decreased. The most substantial increase in headgain was observed at 13–18 weeks of age. When onlythe chair was oscillated (visual VOR, VVOR), the com-pensatory eye gain was high at 3 weeks and the headcontributed significantly to the compensation (vestibulo-collic reflex, VCR). The head gain increased significant-ly at 13–18 weeks of age as in the OKR case. When thedrum and the chair were oscillated synchronously (inhi-bition of VOR, VORINHIB), the compensatory eye gainwas significantly lower than in the VVOR, indicatingsuppression of VOR. This effect was considerable at3 weeks. However, VCR was not suppressed but compa-rable to the VVOR condition at all ages studied. In sum-mary, we found that the vestibular control of smoothgaze adjustment functions earlier than the visual control.At 2 months, the visual control improves dramaticallyand at 3–4 months head participation increases consider-ably. The eye gain in the VORINHIB condition could bewell predicted by vector addition of the eye position sig-nals in the OKR and VVOR conditions.

Key words Infant vision · Vestibulo-ocular response ·Optokinetic response · Vestbulocollic reflex ·Head control · Eye-head coordination

Introduction

Both visual and vestibular mechanisms operate to stabi-lize gaze during head movements. The visual one aims atstabilizing gaze on the optic array by minimizing retinalslip (optokinetic response, OKR), while the vestibularone aims at stabilizing gaze in space (vestibulo-ocularresponse, VOR). The OKR is designed to work at slowoptical changes and its performance begins to deteriorateat frequencies above 0.6 Hz (Benson and Barnes 1978;Hydèn 1983). The VOR functions most optimally above1 Hz, where the gain approaches unity and the phase dif-ference approaches zero (Barnes 1993). However, it isprimarily below 1 Hz where the visual and vestibularsystems jointly contribute to gaze stabilization duringsubject motion.

Problems of coordination between the VOR and OKRarise in this region of the workspace when both the subjectand the fixated visual structure move. Then the OKR reactsto the difference between the subject and object motion,which is appropriate, while the VOR continues to strive atmaintaining a fixed gaze direction in space, which is inap-propriate. Maintaining a stable gaze on the object of interestunder such conditions requires the VOR to be suppressed.In adults, the completeness of this suppression differs withrespect to whether the subjects move their heads actively orpassively. During active head movements, as in the casewhen the subject tracks an object with both head and eyes,the suppression is complete. During passive movements thesuppression is incomplete (McKinley and Peterson 1985;Hydén 1983). It is important to note that suppression of theVOR is only adequate at the frequencies where it interactswith the OKR. Head movements at frequencies above 1 Hzshould still be compensated. Such high-frequency headmovements arise, for instance, in postural stabilization ofthe head (Gresty and Ell 1982; Skavenski et al. 1979).

K. Rosander (✉ ) · C. von HofstenUppsala University, Department of Psychology,Box 1225, 751 42 Uppsala, Swedene-mail: kerstin.rosander@psyk.uu.seFax: +46 184712123

Exp Brain Res (2000) 133:321–333Digital Object Identifier (DOI) 10.1007/s002210000413

R E S E A R C H A RT I C L E

Kerstin Rosander · Claes von Hofsten

Visual-vestibular interaction in early infancy

Received: 13 April 1999 / Accepted: 7 March 2000 / Published online: 7 June 2000© Springer-Verlag 2000

Passive movements of the head and body may not onlygive rise to compensatory eye movements but also com-pensatory head movements, the vestibulo-collic reflex(VCR) (Peterson and Richmond 1988; Keshner and Peter-son 1995). When the subject is rotated relative to a fixedtarget, the VOR and the VCR must be complementary tomaintain fixation on the target. If the object and subjectrotate in synchrony, both the VCR and the VOR should besuppressed in order to maintain fixation on the target.

It is not known how these control systems are estab-lished and how they interact during the early develop-ment of the human child. Do the VOR and the OKR de-velop independently to start with and solve their interac-tion problems later or are they continuously integratedwith each other? How is the VCR expressed during pas-sive body movements with a movable head?

The slow-frequency VOR seems to function ratherwell at birth but with some intermittence. The gain ap-proaches 1.0 for short intervals (Finocchio et al. 1991).Little is known about the early human development ofthe VCR. Jenkyn et al. (1975) described a head-stabiliz-ing response in the newborn, the nuchocephalic reflex,which consisted of a counterrotation of the head whenthe shoulders were turned. The reflex was still observedin 2-year-old infants, but disappeared at 4 years of age.As discussed by Fukushima (1997), Leigh and Zee(1991) suggested that this reflex is similar to an expres-sion of the VCR.

The OKR functions to some degree at birth (Daytonand Jones 1964; Finocchio et al. 1990), but improvessubstantially in terms of stability and reliability over thefirst few months of life (Aslin 1981; Shea and Aslin1990; von Hofsten and Rosander 1996). However, vonHofsten and Rosander found that, even at 3 months ofage, the gain of the OKR when tracking a verticallystriped pattern covering most of the visual field was low-er than when tracking a distinct target in smooth pursuit.Von Hofsten and Rosander (1997) and Phillips et al.(1997) found that smooth pursuit in human infants devel-ops rapidly between 2 and 3 months of age.

There are almost no data published regarding infants'ability to suppress the VOR when they rotate or oscillatepassively in synchrony with a wide-angle visual struc-ture in front of them. Goodkin (1980) reported that 3-but not 2-month-old infants were able to suppress VORin this conflict situation. However, a closer examinationof Goodkin's data raises a number of questions. First, inthe Goodkin study head movements were not measured.Thus the recorded eye movements could, to some de-gree, have been responses to head movements. Secondly,Goodkin presented no quantitative data analysis of thefindings. It was only reported in the form of a figure. Al-though not reported as such, it is clear from this figurethat some systematic suppression occurred in the 2-month-olds as well. Thus, the questions of when andhow VOR suppression develops in the human infant arestill to a large degree unanswered.

There are several possible scenarios. First, some adultstudies suggest that when tracking an object with both

eyes and active head movements, the VOR is suppressedby the smooth pursuit (SP) system (Barnes and Lawson1989; Lanman et al. 1978). If these principles also applyto passive head movements, then the VOR suppressionshould not appear until smooth pursuit has been firmlyestablished at 3 months of age, which would also supportGoodkin's suggestion. Another possibility has been pre-sented by Huebner et al. (1992). The results they ob-tained when oscillating an adult passively in synchronywith a visual structure in front of them could be well ac-counted for by an additive model in which the SP andVOR signals were simply superimposed. In principle, anOKR and a VOR signal could also be superimposed. Ifthis is how VOR suppression works in young infants,there is no reason to expect the smooth pursuit system toplay a privileged role in its development. As long as theVOR compensates for head rotation and the infant tracksvisual structures, these signals can superimpose.

In the present longitudinal study, infants were fol-lowed from 3 to 18 weeks of age. At each visit they wereplaced inside a striped-patterned cylinder with an at-tached, centered target. Their head and eye movementswere measured during three experimental conditions; thecylinder oscillated while the infant was stationary(OKR), the infant oscillated and the cylinder was station-ary (“visual VOR,” VVOR), and finally both the cylin-der and the infant oscillated in phase (“inhibition ofVOR,” VORINHIB).

Materials and methods

Subjects

Six full-term and healthy infants were followed over their first4 months of life. Five of the infants were males and one was fe-male. The first time they were seen, one infant was 1.5 weeks old,four were 3 weeks old and one was 3.5 weeks old. They were thenseen when 5.5±1, 7.5±1, 10±1, 13.5±1.5, and 17.5±1.5 weeks old.All subjects increased normally in body weight during the periodof study. It happened twice that insufficient data were collected ata specific visit because the infant was drowsy. The parent(s) werethen asked to come back later during the same age period. In boththose cases, there was at least one trial in each condition fromwhich data could be extracted.

Apparatus

The experiment was performed with the same apparatus as de-scribed earlier (von Hofsten and Rosander 1996). The infant wasplaced in an infant chair, especially designed to give full supportof the trunk, while allowing free movements of the limbs. It wasplaced at the center of a drum, 100 cm in diameter and 100 cmhigh (Fig. 1). The rotational axis of the drum corresponded ap-proximately to the dorsal column of the infant. The head of thesubject was lightly supported with pads so that it could rotatewithout falling aside. During the experiments, the chair was in-clined at an angle of 40°. The axis on which the infant chair waspositioned was coupled to a motor that could either rotate thedrum, the chair, or both synchronously.

The cylinder covered most of the infant's peripheral visual fieldand was patterned with red and white stripes oriented in its axialdirection. At the position of the infant, the spatial frequency of thestripes was approximately 0.14 c/°. In the middle of the drum sur-

322

face, right in front of the infant, a circular yellow happy face sur-rounded with a black contour was placed measuring 7.2° of visualangle. Below it a black stripe of the same width as the happy faceextended downwards 20°. It was placed there to provide an addi-tional landmark for the subject. The happy face together with theblack stripe could be moved manually from outside the drum, aprocedure performed to calibrate the EOG equipment.

In the middle of the schematic face, a mini video camera(Panasonic WV-KS152) was placed. Its black front had a diameterof 15 mm or 2.2° of visual angle. During the experiments, this vid-eo camera recorded the infant's looking behavior. As it movedwith the target and was always directed at the face of the infant, itwas possible to determine whether the infant fixated the target.Two red light diodes constituted the eyes of the happy face andthey could be made to blink in order to attract the gaze of the sub-ject. When calibrating the EOG, the target was moved manually tovarious positions along the slit. The video signal was fed into avideo tape recorder and a monitor placed to the side of the appara-tus so that the parents and the experimenter could observe the in-fant during the experiment.

Stimuli

Each infant was presented with three different stimuli. The firstwas an oscillation of the surrounding cylinder (OKR), the secondan oscillation of the chair (VVOR), and the third an oscillation ofthe cylinder and chair in synchrony (VORINHIB). In all threecases the oscillations were sinusoidal, the frequency was 0.25 Hzand the amplitude 15° of visual angle. The maximum velocity dur-ing each cycle was thus 22°/s. During the experiment, the lightlevel in the laboratory was kept constant.

Measurements

Head, cylinder, and chair motions

An optoelectronic device, Selspot (Selcom AB, Partille, Sweden),was used to monitor the rotational movements of the head, the cyl-

inder, and the infant chair. The signal-emitting part of the systemconsisted of infrared light-emitting diodes (LEDs), 4 mm in diam-eter. For further details, see von Hofsten and Rosander (1996).Two LEDs, placed midsagittally on the skull 7–9 cm apart, wereused to measure head rotation. Two LEDs were placed on the topof the infant chair to measure the chair/body rotation and, finally,one LED was positioned on the happy face to measure the move-ment of the cylinder.

EOG measurements and calibration

The EOG system was designed in collaboration with G. Westling(Department of Physiology, Umeå University, Umeå, Sweden) andhas been described previously (von Hofsten and Rosander 1996).The electrodes were of the miniature type (Beckman) and hadbeen soaked in 0.9% NaCl solution for at least 30 min before useand then filled with conductive electrode cream. They were ap-plied to the outer canthi of each eye, which had previously beenrubbed gently with a 20% alcohol solution. The ground, a standardEEG child electrode, was attached to the ear lobe.

The EOG was calibrated as described earlier by moving thehappy face together with the attached black vertical stripe below itto various positions while recording the EOG and LED positions(see von Hofsten and Rosander 1997). Data were collected atstraight ahead and at 16° and 24° to the left and to the right. Ateach position the light diodes on the happy face were flashed andthe target was shaken a little and then left in the same position for1–2 s. If there was any doubt that the infant's fixation was on thetarget, it was shaken a little further and the light diodes flashedagain.

Experimental procedure

At each visit, the same routine was applied. Before starting the ex-periment it was ensured that the infant was recently fed and in analert state. The parent signed a consent form of being informedabout the procedure to be applied. One parent and two experi-menters were always present in the laboratory during the experi-

323

Fig. 1a,b The experimentalsetup. a Dorsal and b lateralview

ment. It started with a calibration procedure (von Hofsten andRosander 1997) followed by the experimental conditions (OKR,VOR, VORINHIB) presented in randomized order. The durationof each presentation was 30 s. The base level of the EOG signalwas adjusted when the infant looked straight ahead between trials.If the infant fuzzed or fell asleep during a trial, the experiment wasinterrupted temporarily, after which the last trial was repeated andthe experiment continued. Such interruptions were uncommon forthe ages above 6 weeks. The experimental session had a durationof around 15 min. Data were sampled at 200 Hz and fed into theSelspot-designed software system and stored.

Accuracy of experimental data

After filtering, the positional accuracy of each Selspot measure-ment was about ±0.2° of the visual angle and the velocity data ap-proximately ±0.8°/s. Corresponding estimates for the EOG wereabout ±0.4° and ±1.5°/s.

The EOG calibration procedure was as described in vonHofsten and Rosander (1997), who also discussed the accuracy oftheir procedure. They argued that the method used gave a reason-ably unbiased estimate.

Data analysis

The videotapes and the collected data were inspected in order toidentify intervals where the subjects were drowsy, inattentive, orwhere they closed their eyes. These intervals were disregarded inthe analysis. If the infant was attentive for less than two cycles(8 s) of continuous recording, the whole trial was disregarded.This happened in one case. Data analysis was performed using aprogram especially designed for time series (Lars Bäckström, De-partment of Physiology, Umeå University). SPSS was used for thestatistical analysis.

All calculations were performed in the rotational plane (xz-plane, see Fig. 1), which had its origin on the rotation axis (y-axis). The signals from the LEDs on the head, the chair, and thetarget were therefore converted to angular coordinates. For detailson the method of calculating head angle and visual angle, see vonHofsten and Rosander (1996). The infants often moved their headsduring the rotation of the chair, and these head movements wereextracted as the difference between the head and the chair direc-tions in space, respectively.

Gain and phase

Gaze direction was estimated as the sum of the eye and head di-rections. The gaze was defined relative to the chair (i.e., the body)by adding the eye position and the position of the head relative tothe chair. Angular velocities were estimated as the difference be-tween consecutive coordinates. Each velocity was the mean differ-ence between four time samples corresponding to a 50-Hz low-pass filter. The rotational velocities of the eye, head, cylinder, andchair motions were routinely calculated in this way. In order to ex-amine the smooth component of the eye tracking, saccades definedas periods of velocity greater than 50°/s were eliminated from thecomposite raw eye movement record. The periods cut out werethen replaced with interpolated data and integrated to obtain thedesaccaded tracking position. The number of saccades over condi-tions and age levels were also analyzed. Only saccades exceeding100°/s were then counted. The gain of eye, head and gaze angularpositions relative to the object (or the chair) was estimated as theratio of the amplitudes of the signals at 0.25 Hz. The amplitudewas determined from the Fourier spectrum of the signal. The max-imal gain is then 1, which corresponds to a gaze that is on the tar-get in the OKR and the VVOR conditions. However, in theVORINHIB condition a gaze gain of 0 will indicate a high VORinhibition (gaze calculated relative to the body).

Cross-correlation analysis was used to calculate phase shift(timing).

Results

The OKR condition

Examples of eye and head movements for a 1.5-week-old and an 18-week-old infant are shown in Fig. 2. It canbe seen that, at the younger age, the tracking has a lowgain and is intermittent. The head movements are insig-nificant. At the older age, the tracking is smoother, has ahigh gain and head movements constitute a prominentpart of the tracking. Figure 3 shows the gains of the de-saccaded gaze and eye position as a function of age. Thedifference between the eye and gaze records representsthe contribution of the head movements. A considerableimprovement of the gaze gain can be seen between 5.5and 7.5 weeks of age (F(5,25)=8.622, P<0.001). It is pri-marily due to an increase in eye position gain(F(5,25)=5.195, P<0.01). The contribution of head move-ments also improved over age (F(5,25)=4.134, P<0.01),which was especially noticeable at 17 weeks (Fig. 4).The mean lag of the smooth eye tracking component de-creased significantly with age (F(5,25)=3.713, P<0.02). Itwas 26° or 290 ms at 3 weeks, 17° or 190 ms at5.5 weeks, and 8° or 90 ms at 10 weeks of age. Thereaf-ter it leveled out (see Fig. 5). The head lag was not pos-sible to calculate for the two youngest ages because oftoo small gains and too low cross-correlations (below0.2). For the four older ages no significant changes wereseen. The lag was 23° or 260 ms at 7.5 weeks and 38° or420 ms at 17 weeks. The saccades improved the gazegain at all ages (F(1,5)=30.52, P<0.01), but its average ef-fect did not change significantly over the age periodstudied. Up to 17 weeks, the gaze gain was approximate-ly constant at around 0.75 (without saccades) or 0.85(with saccades).

The VVOR condition

Examples of eye and head movements of a 1.5-week-oldand a 5-week-old infant are shown in Fig. 6. It is ob-served that the compensatory gain is high and that headmovements contribute significantly to the compensatorygaze stabilization. Figure 7 depicts the average gain ofthe desaccaded gaze and eye positions as a function ofage. It is observed that the gain of the gaze is high at allages, indicating that the VVOR mechanism is rather wellestablished already at the youngest age level. The gazegain increases over age (F(5,25)=2.765, P<0.05) and thateffect can be accounted for by an increasing head gain(F(5,25=2.849, P<0.05). Head movements (Fig. 4) con-tribute to the gaze gain at all ages (F(5,25)=68.59,P<0.001).

The desaccaded smooth eye adjustments were, on av-erage, close to counter phase (180°) at all ages as shownin Fig. 5. In fact at five of the six ages the eye laggedless than 180°, or, on average, 174°, corresponding to acompensational lead of 70 ms. The head adjustments

324

were also close to 180° compensations at all ages exceptat the two youngest ones (3 and 5.5 weeks).

Saccades were mostly of the returning kind and theyhad a deteriorating effect on the gaze gain for the olderbut not for the younger ages (F(5,25)=4.686, P<0.01). Thefrequency of saccades >100°/s was analyzed separatelyas a function of age and condition. It was found that thefrequency increased with age (F(5,25)=4.832, P<0.01) andthat it varied with condition (F(2,10)=4.142, P<0.05).Overall, there were more saccades in the VVOR condi-tion than in the other ones. There were no interactionsbetween age and condition (F<1.0). This effect is shownin Fig. 8.

The VORINHIB condition

Figure 9 shows examples of eye and head movements ofa 4-week-old and a 14-week-old infant. It is clear fromthese two examples that the VOR is almost totally sup-

325

Fig. 2 The OKR condition: position data for the object, head, eyeand gaze for a 1.5- (a) and an 18- (b) week-old infant

Fig. 3 The OKR condition: calculated position gain without sac-cades (with SEM) for gaze and eye respectively as a function ofage

pressed during parts of the trials but the VCR is not. Onthe contrary, the head movements seem unaffected by themoving visual target. Figure 10 shows the average gainsof the desaccaded gaze and eye positions as a function ofage. The figure indicates that the various measured gainsdo not change over age and this was confirmed by thestatistical analyses (all Fs <1.0).

The phase of smooth eye tracking component was, onaverage, less than counter phase (Fig. 5). At the twoyoungest ages it averaged about 140°, and at the threeoldest ages around 160°. The linear trend with age wasmarginally significant (F(5,25)=2.417, P<0.07). However,more importantly, there was a clear non-linearity in thedevelopmental curve at 7.5 weeks. At that age the phaseshift of the smooth eye tracking component was 104°,which means that it had almost no compensational effectat all. Figure 5 also shows that the phase shifts weremore variable at 7.5 and 10 weeks than at the other ages.

The head movements contributed significantly to thegaze gain (F(1,5)=8.958, P<0.05), and this contributionincreased with age (F(5,25)=2.828, P<0.05) (Fig. 4). Thephase of the head adjustments at the younger ages wasvery variable (SEMs at 5.5, 7.5, and 10 weeks were40–50°), but at the oldest age the head adjustments wererather consistently counter phase with an average phase

326

Fig. 4 Head position (relativeto the chair) gain for OKR,VVOR, and VORINHIB withSEM, as a function of age

Fig. 5 Average phase (with SEM) for eye position without sac-cades relative to the chair

lag of 167° and an SEM of 10°. The saccades did not al-ter the gaze gain significantly.

Relation between OKR, VVOR, and VORINHIB

In order to illustrate the gaze gain for the conditionsVVOR and VORINHIB, the gaze relative to the body (asseen in the earlier figures) and the gaze gain relative tospace for all ages and subjects are shown in Fig. 11. Thedata are clustered relatively close to the diagonal refer-ence line, and in the VVOR condition the gaze adjust-ments relative to the body are clustered close to 1, whilethey are clustered closer to 0 in the VORINHIB condi-tion. The figure confirms that the two conditions indeedreflect different gaze adjustments.

To what degree can the effects in the OKR and VVORconditions predict the effects in the VORINHIB condi-tion? A simple additive model was tested. It assumed thatthe effects in the VORINHIB condition were accounted

327

Fig. 6 The VVOR condition: position data for head, eye and gazefor a 1.5- (a) and a 5- (b) week-old infant

Fig. 7 The VVOR condition: calculated position gain without sac-cades for gaze (relative to the body) and eye respectively as afunction of age

328

Fig. 8 The frequency of saccades >100°/s in OKR, VVOR, andVORINHIB

Fig. 9 The VORINHIB condition: position data without saccades for head, eye and gaze respectively for a 4- (a) and a 14- (b) week-old infant

Fig. 10 The VORINHIB condition: calculated position gain with-out saccades for gaze (relative to the body) and eye respectively asa function of age

for by the vector sum of the effects in the VVOR and theOKR conditions. The eye vector sum in the OKR andVVOR conditions was compared with the obtained (ex-perimental) eye vector in the VORINHIB. In Fig. 12, thethus calculated and normalized eye vector amplitudes, theeye position gains, for each subject are shown togetherwith the corresponding experimental ones. As can beseen in Fig. 12, insignificant differences appear betweenthe calculated and experimental VORINHIB gains. It isnoted, however, that the subjects develop individually.The mean values for each age indicate no difference be-tween calculated and experimental eye gains.

Discussion

The present study replicates several earlier findings re-garding the development of smooth visual tracking ofwide-angle stimuli (von Hofsten and Rosander 1996).The gains at the older age levels are comparable in thetwo studies and so is the decrease in lag of the trackingat 2 months. The significant improvement at 2 monthsparallels earlier findings regarding the smooth pursuittracking for a distinct target against a homogeneousbackground (von Hofsten and Rosander 1997).

The compensational smooth eye tracking componentsin the VVOR condition had high gain and adequate tim-ing already at the youngest age level. The eyes actuallyhad a small compensatory phase lead that is in agree-ment with earlier research on neonates being subject toslow body oscillations (Weissman et al. 1989). The re-sults thus confirm earlier research suggesting that thevestibular control of eye movements at slow oscillation

329

Fig. 11 VVOR and VORINHIB conditions: desaccated gaze gainsrelated to space [G(space)], and to the body [G(body)], plotted forall ages and subjects (S1–S6). The more the gaze adjustments arealigned with the chair motion, the closer will G(space) approachthe indicated lines

is rather well developed in normal neonates (Eviatar etal. 1974; Finocchio et al. 1991).

Saccades

In the OKR condition, saccades were generally of thecatching-up kind and improved fixation. However, espe-cially at the younger age levels, these saccades were in-sufficient to maintain gaze on the central part (the happyface) of the moving wide-angle pattern. There was nosystematic tendency towards OKN at any age. Obvious-ly, the central target (the happy face) was attractiveenough to inhibit such tendencies. In the VVOR condi-tion the saccades were mostly of the returning kind,which resulted in lower compensatory gaze gain for ageshigher than 10 weeks. In spite of the different effects ofsaccades on gain in the different conditions, the increasein frequency of high-amplitude saccades was similar inall conditions. It was as if the specific experimental con-ditions did not determine whether saccades were goingto be elicited or not, but only the direction and magni-tude of the response.

Inhibition of VOR

In the present study, the compensatory eye movements inthe VORINHIB condition were substantially smallerthan in the VVOR condition for all ages. The resultsdemonstrate that vision has a significant moderating ef-fect on the VOR. The compensatory smooth eye trackingcomponents were, however, never totally eliminated. Infact, the suppression of the VOR did not change signifi-cantly over age. Eventually, with development the sup-pression will become more complete. Children at 9 yearsof age show VOR suppression in an experimental situa-tion as described in the present paper (Herman et al.1982). Larsby et al. (1984) found that the eye movementgain was 0.07 for a group of adult subjects being oscil-lated in synchrony with a fixation target in front of them.

The present result does not lend support to the studyby Goodkin (1980) that 3- but not 2-month-old infantssuppress VOR in a conflict situation. Neither do the re-sults support the suggestion that the onset of VOR sup-pression is linked to the onset of smooth pursuit. VonHofsten and Rosander (1997) found that smooth pursuitemerges between 2 and 3 months of age. It is importantto point out, however, that the visual stimulus in thepresent study covers the whole visual field. It is not un-reasonable to expect that VOR suppression for smallertargets does not function until smooth pursuit for suchtargets has been established. This idea receives supportfrom the fact that the vector sum of the OKR and theVVOR is similar to the obtained gain of smooth eyetracking components in the VORINHIB condition. At1 month of age, small distinct targets are mainly trackedwith saccades (see, e.g., Aslin 1981). A vector additionas the one above for such a stimulus would leave theVVOR rather intact.

330

The fact that no age effects were observed in the de-gree of VOR suppression does not imply that the sup-pression is unaffected by the dramatic improvement invisual tracking at 2 months of age. On the contrary, theeffect is substantial but it is mostly reflected in the phaseof the gaze adjustments rather than the gain (see Fig. 5).The mean phase difference between the eyes and thechair movement at that age was 104°, which implies that,on average, the gaze adjustments were almost not com-pensatory at all. In addition, the variance in the data atthat age more than tripled compared to the previous age,suggesting a transition in the functioning of the VORsuppression. The variance continued to be high at10 weeks, after which it decreased and settled into amuch more compensatory mode of functioning. Our in-terpretation is that the significant increase in smoothtracking at this age induces a temporary turbulence in thetracking system with a subsequent decrease in the preci-sion of the gaze-orienting system.

Is it possible that the transition in the functioning ofthe VOR suppression at 2 months is affected by otherchanges in the control of gaze? For instance, it has alsobeen reported that infants at 2 months of age, whenplaced in a situation where their head is stationary whiletheir body is turned, utilize their cervico-ocular reflex(COR) significantly more than at 4 months (Reisman andAnderson 1989). In the present study, the effect of CORwould be to compensate the active participation of thehead in gaze adjustments. Thus, if COR was a more ef-fective part of the gaze adjustments at this age than be-fore or after, this would result in a lower gaze gain. Asthis was not the case, it is concluded that the COR is notof decisive importance for the transition in the visualvestibular interaction at 2 months of age.

At all ages there were some residual compensatoryeye movements in the VORINHIB condition. These re-sidual smooth eye tracking components could be well ac-counted for by a simple additive model like the one sug-gested by Huebner et al. (1992). This model makes nocommitment to the smooth pursuit system. In principle,any signals activating smooth eye movements could beadded. Thus, whether this signal is an OKR or a smoothpursuit signal makes no difference. Such a model canalso easily handle different combinations of passive headmovements and target motions. Active head movements,however, might require a somewhat different kind ofmodel. This is what the present results suggest.

Head movements

The development of gaze stabilization in young infantscannot be understood without considering the contribu-tion of active head movements. Head movements consti-tuted an active part of the gaze adjustments in all the

three conditions of the present study, demonstrating thateyes and head are coordinated from a very early age.

At the two youngest ages, the gain of the head move-ments was dependent on the mode of control. In the OKRcondition, the head-movement gain was close to zero andsignificantly lower than in the two conditions where thesubject was oscillated (VVOR and VORINHIB). This re-sult indicates that the limitations in head control duringthe first 2 months of life are modality dependent. Vestibu-lar control of head movements, the VCR, seems to befunctioning at an earlier age than visual control. Further-more, the VCR-initiated head movements in theVORINHIB condition do not seem to be influenced by theincreased efficiency of visual stimuli to regulate smoothtracking in contrast to the effects in the OKR condition.

From 2 months of age, however, the increase in thegain of head movements in all conditions seemed to be afunction of a more general improvement in head mobili-ty (see Fig. 4). It is expected that the role of head move-ments for gaze adjustments continues to grow in impor-tance beyond the age range studied. Von Hofsten andRosander (1997) and Daniel and Lee (1990) found thathead movements often dominate the smooth tracking of5- and 6-month-old infants. They found that the headgain was sometimes close to 1.0.

Although the magnitude of the head movements in-creased in a similar way over conditions, their contribu-tions to gaze stabilization differed radically. Head move-ments contributed to gaze stability on the visual patternin the VVOR condition but deteriorated gaze stabiliza-tion in the VORINHIB condition. As the vestibular stim-ulus was the same in these two conditions whereas thevisual ones were different, the results suggest that thehead movements were controlled by the vestibular input.This, in turn, suggests a more direct vestibular-colliccoupling in the young infant. As this system becomesmore established with age, its deteriorating effect onVOR suppression increases during passive rotation. It isthus concluded that compensatory head movement is anintegrated part of the vestibular reaction to body dis-placement, at least in these young infants.

There are also significant differences in the timing ofthe head movements between the different conditions. Inthe OKR condition, the head had a systematic tendencyto lag the target motion at all ages. These lags corre-sponded closely to those found by von Hofsten andRosander (1997). In the VVOR condition, however, thecompensatory head movements were either simultaneouswith or leading the inducing motion. This indicates thatthe head lag in the OKR condition cannot be explainedby the greater inertia of the head or by any other simplebiomechanical principles. Stabilizing gaze by head andeye movements is a control problem.

Neurobiological considerations

The presented results reflect the maturation of neuroana-tomic pathways and how vestibular and visual informa-

331

Fig. 12 The calculated (vector addition of eye positions in theOKR and VVOR conditions) and experimental eye position gainsin the VORINHIB condition for each subject and age

▲

tion is utilized in early development. At the onset of thisstudy, the infants, then 3 weeks old, showed a VOR withhigh gain and small phase lag coordinated with a well-developed VCR. These results suggest that the vestibulo-cerebellar pathways are well established at that age.However, vestibular control of eye movements continuesto improve as demonstrated earlier (von Hofsten andRosander 1996). They found that the high-frequencyVOR of 4-week-old infants showed no systematic lagbut insufficient gain. The gain improved up to 3 monthsof age and the timing variance decreased.

The visual control of smooth eye and head movementswas found to be rather poor at 3 weeks of age but im-proved dramatically at 2 months. Functionally this makesense. As the fovea becomes more focused on the retina(Banks and Bennett 1988) and binocular vision becomesestablished (Held 1991), there is a greater need for direct-ing gaze more precisely on the point of interest. Neuro-logically, several important changes are associated withthis development. First, it marks the onset of functioningof the MT and the MST areas. The fact that smooth eyetracking not only improves its gain but also its phase sug-gests that other areas in the CNS are involved as well. Ithas been suggested that the proactive corticopontocere-bellar pathway (Altman and Bayer 1997) targeting theflocculus and emerging from extrastriate and parietal cor-tex (Glickstein 1997; Stone and Lisberger 1990) is centralfor prospective control and could thus be involved in thepredictive eye tracking observed (Keeler 1990). Theemergence of such pathways from the cerebral cortex isin agreement with the fact that at 2 months of age thesynaptogenesis of the cortex approaches its peak (Hut-tenlocher 1994). Further support for this hypothesiscomes from PET studies that indicate activity in the pari-etal cortex at 2 months of age, and during the followingmonth activity is also observed in the cerebellar cortexand the thalamus (Chugani et al. 1994). The ability ofVOR suppression seen already at 3 weeks of age mightreflect the relative maturity of pathways such as the floc-culus-brainstem nuclei, as the Purkinje cells modulatesmooth pursuit and VOR suppression (Buttner andWaespe 1984). In the adult, several cortical areas takepart in the control of vestibular nuclei complex and it hasbeen suggested that such connections influence the VCRand the VOR (Akbarian et al. 1994). Thus, with increasedexperience, the cortical control and the cerebellar regula-tion of spatial orientation based on visual and vestibularinformation will become more refined.

From a developmental point of view, the substantialincrease in the contribution of head movements to gazestabilization, especially for the 3- to 4-month-old infantsstudied, reflects the development of a more flexiblemode of gaze stabilization in which head and eye move-ments are interchangeable. It will provide a more versa-tile platform for perception and action in space (Danieland Lee 1990). Such a mode of gaze control requires amore complex organization of the eye-head-body systemwith respect to the oculomotor output (Collewijn 1989;Leigh and Brandt 1993). The present study only taps the

very beginning of the development of this mode of con-trol. Even at 19 weeks of age, head movements are stilldominated by vestibular information that hampers theability to fixate a moving target during self-motion. Fur-thermore, head tracking of a visual target shows a sub-stantial lag at this age. As these coordination problemsare solved, new challenges are presented at the emer-gence of every new stage of mobility from crawling towalking. It is therefore expected that the development ofgaze control in the service of action will continue to de-velop over childhood.

Acknowledgements We would like to thank all the enthusiasticparents who made this study possible. This research was support-ed by the Swedish Council for Research in the Humanities and So-cial Sciences and NIH Grant 5 R01 HD 16195-17.

References

Akbarian S, Grusser OJ, Guldin WO (1994) Corticofugal connec-tions between the cerebral cortex and brainstem vestibular nu-clei in the macaque monkey. J Comp Neurol 339:421–437

Altman J, Bayer SA (1997) Development of the cerebellar systemin relation to its evolution, structure, and functions. CRCPress, Boca Raton

Aslin R (1981) Development of smooth pursuit in human infants.In: Fischer DF, Monty RA, Senders EJ (eds) Eye movements:cognition and visual development. Erlbaum, Hillsdale, NJ, pp31–51

Banks MS, Bennett PJ (1988) Optical and photoreceptor immatu-rities limit the spatial and chromatic vision of human neonates.J Opt Soc Am [A] 5:2059–2079

Barnes GR (1993) Visual-vestibular interaction in the control ofhead and eye movement: the role of visual feedback and pre-dictive mechanisms. Progr Neurobiol 41:435–472

Barnes GR, Lawson JF (1989) Head-free pursuit in the human of avisual target moving in a pseudo-random manner. J PhysiolLond 410:137–155

Benson AJ, Barnes GR (1978) Vision during angular oscillations:the dynamic interaction of visual and vestibular mechanisms.Aviat Space Environ Med 49:340–345

Buttner U, Waespe W (1984) Purkinje cell activity in the primateflocculus during optokinetic stimulation, smooth pursuit eyemovements and VOR-suppression. Exp Brain Res 55:97–104

Chugani HT, Phelps ME, Mazziotta JC (1994) Positron emissiontomography study of human brain functional development. In:Johnson M (ed) Brain development and cognition. Blackwell,Oxford, pp 124–143

Collewijn H (1989) The vestibular-ocular reflex: is it an indepen-dent subsystem? Rev Neurol 145:502–512

Daniel BM, Lee DN (1990) Development of looking with headand eyes. J Exp Child Psychol 50:200–216

Dayton GO, Jones MH (1964) Analysis of characteristics of fixa-tion reflex in infants by use of direct current electrooculogra-phy. Neurology 14:1152–1156

Eviatar L, Eviatar A, Naray I (1974) Maturation of neuro-vestibu-lar responses in infants. Dev Med Child Neurol 16:435–446

Finocchio DV, Preston KL, Fuchs AF (1990) Obtaining a quantita-tive measure of eye movement in human infants: a method ofcalibrating the electrooculogram. Vision Res 30:1119–1128

Finocchio DV, Preston KL, Fuchs A (1991) Infant eye move-ments: quantification of the vestibulo-ocular reflex and visual-vestibular interactions. Vision Res 10:1717–1730

Fukushima K (1997) Corticovestibular interactions: anatomy,electrophysiology, and functional considerations. Exp BrainRes 117:1–16

Glickstein M (1997) Mossy-fibre sensory input to the cerebellum.In: De Zeeuw CI, Strata P, Voogd J (ed) Progress in brain re-search, vol 114. Elsevier, Amsterdam, pp 251–259

332

Goodkin F (1980) The development of mature patterns of head-eye coordination in the human infant. Early Hum Dev4:373–386

Gresty MA, Ell JJ (1982) Normal modes of head and eye coordina-tion and their symptomatology of their disorders. In:Lennerstrand G, Zee DS, Keller EL (eds) Functional basis ofocular motility disorders. Pergamon Press, Oxford, pp 391–405

Held R (1991) Development of binocular vision and stereopsis. In:Regan D, Cronly-Dillon JR (eds) Vision and visual dysfunc-tion, vol 9. Macmillan, London, pp 170–178

Herman R, Maulicci R, Stuyck J (1982) Development and plastici-ty of visual and vestibular generated eye movements. ExpBrain Res 47:69–78

Huttenlocher PR (1994) Morphometric study of human cerebralcortex development. In: Johnson M (ed) Brain developmentand cognition. Blackwell, Oxford, pp 112–124

Huebner WP, Leigh RJ, Seidman SH, Thomas CW, Billian C, Disc-enna AO, Dell'Osso LF (1992) Experimental tests of a superpo-sition hypothesis to explain the relationship between the vestib-ular reflex and smooth pursuit during horizontal combined eye-head tracking in humans. J Neurophysiol 68:1775–1792

Hydén D (1983) The broad frequency-band rotatory test. Linko-ping University Medical Dissertations, No. 152. Linkoping,Sweden

Jenkyn LR, Walsh DB, Walsh BT, Culver CM, Reeves AG (1975)The nuchocephalic reflex. J Neurol Neurosurg Psychiatr38:561–566

Keeler JD (1990) A dynamical system view of cerebellar function.Physica D 42:396–410

Keshner EA, Peterson BW (1995) Mechanisms controlling humanhead stabilization. I. Head-neck dynamics during random rota-tions in the horizontal planes. J Neurophysiol 73:2293–2301

Lanman J, Bizzi E, Allum J (1978) The coordination of eye andhead movement during smooth pursuit. Brain Res 153:39–53

Larsby B, Hydén D, Ödkvist LM (1984) Gain and phase charac-teristics of compensatory eye movements in light and dark-ness. A study with a broad frequency-band rotatory test. ActaOtolaryngol (Stockh) 97:223–232

Leigh RJ, Brandt T (1993) A reevaluation of the vestibulo-ocularreflex: new ideas of its purpose, properties, neural substrate,and disorders. Neurology 43:1288–1295

Leigh RJ, Zee DS (1991) The neurology of eye movements, 2ndedn. Contemporary neurology series. Davis, Philadelphia, pp232–263

McKinley PA, Peterson BW (1985) Voluntary modulation of thevestibuloocular reflex in humans and its relation to smoothpursuit. Exp Brain Res 60:454–464

Peterson BW, Richmond FJ (1988) Control of head movements.Oxford University Press, New York

Phillips JO, Finocchio DV, Ong L, Fuchs AF (1997) Smooth pur-suit in 1- to 4-month-old human infants. Vision Res37:3009–3020

Reisman JE, Anderson JH (1989) Compensatory eye movementsduring head and body rotation in infants. Brain Res 484:119–129

Shea SL, Aslin GN (1990) Oculomotor responses to step-ramp tar-gets by young infants. Vision Res 30:1077–1092

Skavenski AA, Hansen RM, Steinman RM, Winterson BJ (1979)Quality of retinal image stabilization during small natural andartificial rotations in man. Vision Res 19:657–683

Stone LS, Lisberger SG (1990) Visual responses of Purkinje cellsin the cerebellar flocculus during smooth-pursuit eye move-ments in monkeys. I. Simple spikes. J Neurophysiol 63:1241–1261

von Hofsten C, Rosander K (1996) The development of gaze con-trol and predictive tracking in young infants. Vision Res36:81–96

von Hofsten C, Rosander K (1997) Development of smooth pur-suit tracking in young infants. Vision Res 37:1799–1810

Weissman BM, DiScenna AO, Leigh RJ (1989) Maturation of thevestibulo-ocular reflex in normal infants during the first 2months of life. Neurology 39:534–538

333