Kretch, K. S., Franchak, J. M., & Adolph, K. E. (in press). Crawling and walking infants see theworld differently. Child Development. Preprint formatted by authors.

Crawling and walking infants see the world differently

Kari S. Kretch, John M. Franchak, & Karen E. AdolphNew York University

AbstractHow does visual experience change over development? To investigate changes in visual input overthe developmental transition from crawling to walking, 30 13-month-olds crawled or walked down astraight path wearing a head-mounted eye-tracker that recorded gaze direction and head-centeredfield of view. Thirteen additional infants wore a motion-tracker that recorded head orientation.Compared with walkers, crawlers’ field of view contained less walls and more floor. Walkers directedgaze straight ahead at caregivers, whereas crawlers looked down at the floor. Crawlers obtainedvisual information about targets at higher elevations—caregivers and toys—by craning their headsupward and sitting up to bring the room into view. Findings indicate that visual experiences areintimately tied to infants’ posture.

Much of what infants learn depends on what theysee: Natural vision provides opportunities forlearning about the properties and affordances ofplaces, surfaces, objects, and people (Franchak,Kretch, Soska, & Adolph, 2011), and for estab-lishing words and concepts that provide cogni-tive links with the visible denizens of the envi-ronment. For example, infants’ understanding ofcausal and self-propelled motion is related to thefrequency with which they observe these types ofmotion in their everyday environment (Cicchino,Aslin, & Rakison, 2011). Similarly, toddlers aremore likely to learn the name of an object if theobject is large and prominent in their field ofview at the moment it is named—visual inputthat occurs naturally when infants hold objectsup for visual inspection (Yu & Smith, 2012).

How do opportunities for learning from visualinput change from moment to moment and overdevelopment? Shifts in body posture may con-tribute to real-time changes in visual input. Thisis particularly relevant to development becausethe amount of time infants spend in differentactivities such as lying prone, sitting, crawling,cruising, and walking changes with developmen-tal improvements in motor skill. In particular,

crawling and walking have unique effects on in-fants’ experiences and cognitive outcomes (e.g.,Adolph et al., 2012; Campos et al., 2000; Walle& Campos, in press). Many researchers havespeculated that such effects stem from differ-ences in visual input (Adolph, 1997; E. J. Gib-son & Pick, 2000; Iverson, 2010; Karasik, Tamis-LeMonda, & Adolph, 2011; Newcombe & Lear-month, 1999), but the presence of such differ-ences has never been confirmed empirically.

Different postures change infants’ vantage point,but these differences could be negated or exag-gerated by infants’ own head and eye movementswithin a given posture. For example, crawlinginfants might crane their heads upwards to com-pensate for being low to the ground. Whetherposture has real, functional effects on what in-fants see and where they choose to look hasnot been studied. Advances in head-mountedeye-tracking technology have made it possible todescribe infants’ visual experiences while theymove around the world, and have challengedmany of our long-held assumptions about whereinfants look during everyday activities (Franchaket al., 2011). Here, we take advantage of this newtechnology to ask whether and how infants’ vi-

2

sual experiences differ while crawling, walking,and sitting.

Locomotor Development Affects Op-portunities for Learning

The onset of crawling is a major milestone, andcrawling experience facilitates psychological ad-vances. Crawling allows self-initiated access tothe larger world (Campos et al., 2000; E. J. Gib-son, 1988; Piaget, 1954), which is linked withimprovements in cognitive skills such as spatialsearch (Horobin & Acredolo, 1986; Kermoian& Campos, 1988), position constancy (Bai &Bertenthal, 1992; Bertenthal, Campos, & Bar-rett, 1984), optic flow perception (Anderson etal., 2001; Higgins, Campos, & Kermoian, 1996;Uchiyama et al., 2008), and memory retrieval(Herbert, Gross, & Hayne, 2007). The changesbrought about by independent mobility also haveimplications for social development: Crawlersdisplay more attachment behaviors (Campos,Kermoian, & Zumbahlen, 1992) and are moreadept at following a gaze/pointing gesture (Cam-pos, Kermoian, Witherington, Chen, & Dong,1997) than pre-crawlers.

Opportunities for learning change again withthe transition from crawling to walking. Novicewalkers take more steps, travel greater distances,and visit more places than experienced crawlers(Adolph et al., 2012; Clearfield, 2011). Walk-ers are more likely to cross the room to en-gage with distal objects, to carry objects acrossthe room, and to cross the room to share ob-jects with caregivers (Karasik, Adolph, Tamis-LeMonda, & Zuckerman, 2012; Karasik et al.,2011). As a consequence, walkers receive moreverbal feedback from their mothers (Karasik,Tamis-LeMonda, & Adolph, in press). Walk-ers also engage in more bids for social interac-tion, produce more caregiver-directed vocaliza-tions and gestures, spend more time interactingwith caregivers, and experience more frequentemotional interactions with caregivers (Biringen,Emde, Campos, & Applebaum, 1995; Clearfield,2011; Clearfield, Osborne, & Mullen, 2008). And

recent studies suggest that language develop-ment is accelerated when infants begin to walk(Ellis-Davies, Sakkalou, Fowler, Hilbrink, & Gat-tis, 2012; Walle & Campos, in press).

Vision Is a Whole Body Process

To describe how infants acquire visual infor-mation, researchers have spent decades study-ing infants’ eye movements (Aslin & McMurray,2004; Haith, 1969; Johnson, Amso, & Slemmer,2003; Salapatek & Kessen, 1966; von Hofsten &Rosander, 1997). But vision involves more thaneye movements. It is a whole body process in-volving movements of the eyes, head, and body(J. J. Gibson, 1966, 1979; Land, 2004). To seea target—an upcoming flight of stairs, a toy ona shelf, or a parent’s face—infants must orienttheir bodies toward the target, rotate their headsto bring the target into the field of view, andpoint their eyes to fixate the target. The geome-try of the body can facilitate or hinder this pro-cess by constraining where the eyes are in spaceand where they point. In particular, the projec-tion of the visual field depends on both the heightof the head relative to the ground (imagine rais-ing or lowering a flashlight without changing theangle) and the up-and-down (pitch) rotation ofthe head relative to parallel (akin to tilting theflashlight toward the ceiling or floor). Althoughthe eyes rotate within the head, they are biasedto return to the more comfortable center position(Fuller, 1996; Pare & Munoz, 2001). Therefore,developmental changes in eye height and rangeof head motion may influence what infants see.

Height differences alone may be sufficient to alterthe visual world. When facing straight ahead,a 174-cm tall adult loses sight of an obstacleon the floor at a distance of 75 cm, whereas a128-cm tall child does not lose sight of an obsta-cle until it is 55 cm ahead. Accordingly, adultsare less likely to fixate upcoming obstacles thanchildren and adults fixate objects from fartheraway (Franchak & Adolph, 2010). Toddlers areeven more likely to fixate obstacles and some-times sustain fixation through the moment of

3

foot contact. In a crawling posture, infants arestill more likely to fixate upcoming obstacles andto keep them in their field of view (Franchak etal., 2011). While crawling, infants’ eyes are evenlower to the ground. But how much lower, andpossible consequences for visual input, are notknown.

Differences in the head’s range of motion whilecrawling versus walking may also contribute todifferences in access to visual information. Walk-ers’ erect spines are perpendicular to the ground,whereas crawlers’ spines are closer to paral-lel. Thus, with the neck in a neutral position,walkers’ faces are pointed straight ahead, butcrawlers’ faces are pointed toward the ground.The total range of up-down motion of the neck isapproximately 150 degrees, comprising about 70degrees of flexion to rotate the head downwardand 80 degrees of extension to rotate the headupward (Klinich & Reed, 2013; Lewandowski& Szulc, 2003; Lynch-Caris, Majeske, Brelin-Fornari, & Nashi, 2008; Ohman & Beckung,2008; Youdas et al., 1992). To point their facesstraight ahead (that is, to bring the absolutepitch angle of the head to parallel), crawlerswould need to extend their necks to the outerreaches of the range of motion. This fight againstgravity is not trivial, because infants’ heads arelarge and heavy relative to their bodies (Oun-sted & Moar, 1986; Snyder, Spencer, Owings, &Schneider, 1975). How much crawlers activelyextend their necks while they crawl is still anopen question.

Possibly, visual input is not different betweencrawling and walking. The height gained with anupright posture may not be enough of a changein vantage point to cause functional differences invisual input. Moreover, while crawling, infantsmay be able to compensate for the difference inheight by craning their necks upward and point-ing their faces toward the ceiling. Conversely,new walkers might fail to exploit their increasein height by pointing their heads down to helpmaintain balance or avoid obstacles (Franchak etal., 2011).

Current Studies

The current studies investigated whether thereare differences in visual input between crawl-ing and walking, and which physical factors con-tribute to these potential differences. Two exist-ing studies indicate that visual input may differbetween the two postures during natural, spon-taneous activity: Infants are more likely to havefaces in their field of view while walking thanwhile crawling (Frank, Simmons, Yurovsky, &Pusiol, in press), and are more likely to fixateobstacles on the floor while crawling than whilewalking (Franchak et al., 2011). But to deter-mine whether differences are due directly to real-time, physical differences in locomotor posture,it is necessary to standardize the testing situa-tion for both crawlers and walkers, holding ev-erything constant (age, task, social information)except locomotor posture.

In a naturalistic setting, informative visual stim-uli might be at any height relative to infants’bodies: on the floor, on a shelf or table, or atopadult shoulders. Differences in visual access topeople and objects while crawling and walkingmay depend on the location of people and ob-jects in the environment. Thus, it was impor-tant to examine how crawling and walking in-fants are affected by different visual target lo-cations. We manipulated the height of visualstimuli (caregivers’ faces and toys) to examinewhether crawlers and walkers gain view of visualtargets at different locations and whether theyadapt their eye, head, and body movements todo so.

Experiment 1: What Infants See

In Experiment 1, we used head-mounted eyetracking to describe the visual information ac-cessed by infants while crawling and walking.Infants crawled or walked down a uniform, con-trolled path so that we could use calibratedmarkings on the floor and walls to measure howmuch of the environment was in view. We also

4

examined functional consequences of differentfields of view—how frequently people and objectswere visible for crawlers and walkers—and wherein the room crawlers and walkers directed theirgaze.

Method

Participants Thirty infants participatedwithin a week of their 13-month birthday.Infants were recruited from local hospitalsand families received small souvenirs for theirparticipation. Fifteen infants crawled on handsand knees as their primary form of locomotion(8 boys, 7 girls; M = 4.6 months of crawlingexperience dating from the first day infantscrawled 10 feet continuously, as reported byparents) and 15 had begun to walk (7 boys, 8girls; M = 1.9 months of walking experiencedating from the first day infants walked 10feet continuously). Data from an additional17 infants were excluded: 6 infants refused tocrawl or walk on the raised walkway duringthe warm-up period, 2 refused to wear thehead-mounted eye tracker, 6 became fussy whilewearing the eye tracker, and 3 infants’ data werelost due to equipment failure.

Head-mounted eye tracker We used a Pos-itive Science head-mounted eye tracker to recordinfants’ eye gaze and head-centered field of view(Franchak et al., 2011; Franchak, Kretch, Soska,Babcock, & Adolph, 2010). The eye tracker con-sisted of two miniature cameras mounted on arounded band that attached securely to a fittedspandex hat with Velcro (Figure 1a). The scenecamera was mounted on the band slightly abovethe infant’s right eye, and pointed outward torecord the scene in front of the infant (with a54.4◦ horizontal × 42.2◦ vertical field of view).The infrared eye camera was mounted on a flexi-ble wire and pointed inward to record the infant’sright eye, which was illuminated by an infraredemitting diode (IRED).

Infants also wore a fitted vest with a small (8 ×4 × 2 cm) breakout box attached to the back.

Video feeds from the eye tracker were routedthrough the breakout box into a laptop com-puter several feet away. The wires were longenough to allow unconstrained movement on thewalkway, and an assistant followed behind in-fants to hold the wires out of the way. Videosfrom both camera feeds—the scene video and theeye video—were captured on the laptop for laterprocessing.

After the session, Yarbus software (Positive Sci-ence) was used to calculate infants’ point of gazewithin the scene video using estimations of thelocation of the center of the pupil and the cornealreflection. The software constructed a gaze video(30 frames/s) with a crosshair overlaid on eachframe indicating the point of gaze with a spatialaccuracy of 2◦ (Figure 1b).

Walkway and procedure Infants were testedon a raised walkway (65 cm high × 98 cm wide× 490 cm long; Figure 1c). To facilitate precisemeasurement of locations, the walkway was cov-ered with high-density foam with stripes of al-ternating colors spaced 15 cm apart (32 stripesin total). Colors repeated every five stripes,and numbered labels (1-7) were placed along thelength of the walkway at every fifth stripe so thatstripes could be uniquely identified by the com-bination of number and color. An 82-cm widecurtain hung from the ceiling 63 cm from thefar end of the walkway and was covered with14 stripes, identified by the same five colors andthe numbers 1-3. The curtain stretched from theceiling to the top of the walkway so that the low-est stripe was 65 cm high (level with the walkwaysurface) and the highest stripe was 260 cm high(195 cm above the walkway surface).

Caregivers sat on a chair placed between thewalkway and the curtain. To familiarize infantswith the experimental setup, caregivers encour-aged infants to crawl or walk the length of thewalkway several times without the eye track-ing equipment, while an experimenter followedalongside to ensure infants’ safety. Then, anassistant distracted infants with toys while theexperimenter placed the eye-tracking equipment

5

highest

point visible

closest

point visible

look to

caregiver

scene

camera

eye

camera

A. B.

C.

Figure 1: (A) Infant wearing head-mounted eye-tracker. (B) Example gaze video frame exported from Yarbussoftware. Caregiver is pictured in the middle toy height condition. Red crosshairs indicate infant’s point ofgaze; white dotted line indicates gaze locations that were scored as looks to the caregiver. (C) Striped walkwayand curtain apparatus. Caregiver is pictured in the low toy height condition. The experimenter followedbehind infants to ensure their safety on the walkway and an assistant followed to keep the wires out of theway.

6

on the infant and adjusted the eye camera andIRED until the image of the eye was centered andbrightly illuminated. Once the equipment was inplace and infants were comfortable, calibrationdata were collected: The assistant called infants’attention to various locations by squeaking toysor shaking rattles in windows cut out of a largeposter board. Infants sat approximately four feetaway from the calibration board, and the experi-menter calibrated the scene camera by adjustingits position so that the entire board was visiblein the scene video. Although we could not guar-antee that the angle of the scene camera was ex-actly the same for every infant, the small range ofmotion of the scene camera and the constraintson the placement of the attached eye camera forgaze calculations ensured that the inter-infantdifferences were minimal—most likely within 5-10 degrees. These variations likely added noiseto the data but would not be expected to differbetween crawlers and walkers.

After calibration, infants were again encouragedto crawl or walk several times toward their care-givers. Caregivers enticed infants by calling tothem from the end of the walkway and hold-ing attractive toys and snacks. To present in-fants with different visual target locations, wevaried the height of caregivers and toys on alter-nating blocked trials. In the low toy condition,caregivers sat in the chair and held toys on thesurface of the walkway. In the middle toy con-dition, caregivers sat and held toys up at theirown eye height. And in the high toy condition,caregivers stood up and held toys at their ele-vated eye height. Each infant received 1-4 blocksof three trials (M = 8.20 total trials) and trialorder was counterbalanced across infants. Num-ber of trials did not differ significantly betweencrawlers (M = 7.27) and walkers (M = 9.13),t(28) = 1.59, p = .12. An assistant wheeled avideo camera on a dolly alongside the walkwayto record infants’ locomotion; these videos werelater synchronized with gaze videos to facilitatedata coding.

Data coding With a head-mounted eyetracker, infants are free to locomote and move

their heads, so that the field of view shifts contin-uously throughout the session. Thus, automaticdata processing methods developed for remoteeye tracking systems are not feasible, and datamust be scored by hand from video. We usedthe open source video coding software Datavyu(www.datavyu.org) to score infants’ locomotorand visual behaviors. A primary coder scored100% of the data, and a second coder scored33% of each infant’s data to ensure inter-raterreliability.

Locomotion and posture. Coders first identi-fied each video frame when infants took a for-ward step: the first frame when a hand (forcrawlers) or a foot (for walkers) contacted thewalkway surface. For each step, coders scoredinfants’ location on the walkway based on thestripe region in which infants placed their handor foot. Coders also scored each time crawlersshifted into a sitting posture (walkers never didthis), and identified the stripe region where thefarthest forward body part fell (knees or feet).Coders agreed within one stripe on 99.2% ofsteps, and on the exact stripe on 96.5% of steps(κ = .96).

Field of view. For each video frame identifiedas a step, the coders scored various aspects ofthe scene video as an approximation of infants’field of view. Preliminary data indicated that thecontents of the scene video showed little changewithin the duration of a step, so more detailedframe-by-frame coding was unnecessary. Addi-tionally, to obtain samples of field of view datawhen infants were sitting, coders selected thefirst and last frames when infants were sittingand facing forward, and scored all outcome mea-sures for those video frames. In total, codersscored field of view measures for 6,313 videoframes.

To measure the orientation of infants’ field ofview, coders used the colored stripes to identifywhere the top of the scene camera field of viewintersected the vertical plane (the curtain), andwhere the bottom of the scene camera field ofview intersected the horizontal plane (the floor).

7

A

A

B B

Figure 2: Scale drawings of the boundaries of the scene camera field of view for crawlers and walkers. (A)Highest point visible. (B) Closest point visible. Dimensions were calculated using group means for the middletoy height condition.

If the curtain was in view, the coders identifiedthe highest point visible: the uppermost stripe atthe top of the scene video (Figure 1b-c). Codersagreed within one stripe on 97.7% of steps, andon the exact stripe on 89.7% of steps (κ = .89).If the floor was in view, the coders identified thelowermost stripe at the bottom of the scene video(Figure 1b-c); the closest point visible was thedistance between that stripe and the stripe atinfants’ hands or feet. Coders agreed within onestripe on 98.1% of steps, and on the exact stripeon 92.8% of steps (κ = .93). Frames where in-fants had their heads turned to the side such thatno stripes were visible were eliminated from anal-ysis. Finally, coders scored whether the care-giver’s face and the toy held by the caregiverwere visible in the scene video—coders agreedon 98.3% of samples, κ = .96.

Gaze location. Frame-by-frame coding of the lo-cation of the crosshair is necessary to determinewhere infants direct their gaze within the field ofview. However, because such coding is extremelylaborious, coders scored only a subset of trials forgaze location. Four infants did not produce use-able gaze data because of poor calibration. Forthe remaining 26 infants, 57/190 trials were notuseable because of poor eye tracking data qual-ity (due to suboptimal camera placement, inad-equate illumination of the eye, or infants cryingor squinting). Coders scored eye gaze data forthe first useable trial in each condition, and forcrawlers, coders also scored each trial that con-tained a sit (88 trials in total). Overall, coders

scored gaze location in 28,962 video frames.

For each frame of those trials, coders scoredwhether the gaze crosshair was on the caregiver,the floor, or the wall ahead. We allowed a mar-gin of error for looks to the caregiver by includingany frames where the crosshair was on the cur-tain, within 2 stripes from the top of the care-giver’s head (1b); these criteria are similar toautomated areas of interest commonly used inremote eye-tracking (e.g., Johnson, Slemmer, &Amso, 2004). Coders agreed on 98% of frames.Frames where infants were looking off the sideof the walkway or outside the boundaries of thescene video were excluded from analysis.

Results and Discussion

Our primary questions were whether visual in-put differs between crawlers and walkers and howinfants are affected by changes in target loca-tion. We tested for effects of posture and targetlocation using generalized estimating equations(GEEs), a type of linear model that accountsfor covariance between repeated measures andallows for testing of non-normal distributions.All categorical dependent variables (field of viewcontents and eye gaze targets) were analyzed us-ing a binomial probit model. Entering infants’location on the walkway as a covariate in themodels did not change any of the findings for pos-ture and target location, so we did not include lo-

8

Table 1: Wald χ2 values for all GEE model effects.

Variable Posture Toy Height Condition Posture x Condition

Highest point visible 52.84** 40.19** 0.45

Closest point visible 112.35** 26.34** 7.56*

Face in view 46.17** 12.62** 1.75

Toy in view 77.95** 12.30** 5.99*

Looking toward floor 5.81* 16.09** 8.18*

Looking toward caregiver 3.31# 10.02** 5.55#

Note: #p < .7, ∗p < .05, ∗∗p < .01.

cation in the analyses reported here. Data fromthe first 75 cm and the last 75 cm of the walkwaywere eliminated because few infants contributeddata at those locations. We used Sidak-correctedpairwise comparisons to follow up on significanteffects. Main effects and interactions from theGEEs are presented in Table 1.

Crawling vs. walking: Scene camera fieldof view Portions of the environment in viewdiffered for crawlers and walkers. Figure 2 isdrawn to scale based on our data to illustratethe average visual fields of crawlers and walkersin the current study.

Highest point visible. For us to measure thepoint of intersection with the curtain (Point Ain Figure 2), infants had to hold their headshigh enough that the curtain was visible in thescene camera. For walkers, this was nearly al-ways the case, but crawlers repeatedly lost sightof the curtain: On 25.7% of steps, crawlers’ en-tire scene video contained only the floor, indicat-ing that while crawling, infants may miss visualinput from distal parts of the room around them.

When crawlers did see across the room, they sawmuch less of it than walkers: The highest pointvisible in the scene camera was about twice ashigh for walkers (M = 135.17cm) as for crawlers

(M = 65.94cm; Figure 3a; Table 1, row 1). Bothcrawlers and walkers responded to changes intarget location, as indicated by a main effect oftoy height condition and the absence of a pos-ture × condition interaction: As the location ofpeople and objects moved upward, both crawlersand walkers tilted their heads up to compen-sate. Pairwise comparisons indicated that thehigh toy condition (with the caregiver standingup) was significantly different from both otherconditions (ps < .01), and the difference betweenthe other two conditions was marginally signifi-cant (p = .08).

The findings represent a conservative estimateof the differences between crawlers and walk-ers. For measurements of the highest point vis-ible, the range was restricted: The largest pos-sible value was the top of the curtain. However,the highest point visible, particularly for walk-ers, was often higher than the top of the cur-tain (418 video frames for walkers and 36 videoframes for crawlers were scored as the maximum195 cm), so that the data suffered from a lit-eral ceiling effect. If we had tested infants ina room with an infinitely high ceiling, the dis-parity between crawlers and walkers would havebeen even larger. Although on average, walkers’view was higher than crawlers’ view, infants inboth groups showed substantial within-subject

9

Toy Height ConditionLow Middle High

Cm fr

om H

ands

/Fee

t

0

20

40

60

80

100

Toy Height ConditionLow Middle High

Ppn

of F

rame

s

0.00

0.25

0.50

0.75

1.00

Toy Height ConditionLow Middle High

Cm fr

om F

loor

0

50

100

150

200

Toy Height ConditionLow Middle High

Ppn

of F

rame

s

0.00

0.25

0.50

0.75

1.00

A. Highest Point Visible

C. Caregiver Face in View D. Toy in View

B. Closest Point Visible

WalkersCrawlers

Figure 3: Outcome measures by posture and toy height condition. (A) Highest point visible in the scenecamera (in cm from the floor). (B) Closest point visible from the scene camera (in cm from the infant). (C)Proportion of steps where the caregiver’s face was visible in the scene camera. (D) Proportion of steps wherethe toy was visible in the scene camera. Error bars denote standard errors.

variability (Figure 4). The large ranges of valuesobtained for each infant indicate that constraintson crawlers’ and walkers’ visual fields were notabsolute. Crawlers could tilt their heads upand occasionally did, giving them intermittentglimpses of parts of the environment that weretypically out of view.

Closest point visible. Whereas walkers hadgreater visual access to distal and elevated lo-cations, crawlers had a better view of the floorin front of their hands. The closest point visibleat the bottom of the scene camera was closer tocrawlers’ hands (M = 20.89cm) than to walk-ers’ feet (M = 83.00cm; Figure 3b; Table 1, row2). The GEE revealed a main effect of condi-tion and a posture × condition interaction: Aswalkers tilted their heads up in the high toy con-dition, the lower boundary of the field of viewwas pushed farther forward than in the low con-

dition (p < .01); the conditions did not differ forcrawlers (ps > .10).

Crawling vs. walking: Caregivers and toysin the field of view Differences in the fieldof view for crawlers and walkers had importantfunctional consequences. Walkers had caregiversand toys—the goal at the end of the walkway—intheir field of view, but crawlers frequently didnot.

Caregiver. Walkers had their caregiver’s face inview twice as often as crawlers (M = 89.3% vs.43.7%; Figure 3c; Table 1, row 3). The analysisrevealed only a main effect of condition, but noposture × condition interaction: Both crawlersand walkers were less likely to have their care-givers in view in the high toy condition, whenthey stood up, than in the low toy condition,when they were sitting and closer to infants’ eyelevel, p < .01.

10

High

est P

oint V

isible

(cm)

0

50

100

150

200

Crawlers Walkers

Figure 4: Individual ranges for the highest point visible. Each line represents one infant. Open circles denoteminimum and maximum values, and filled circles denote mean values.

Toy. Crawlers had toys in view less frequentlythan walkers (M = 92.2% vs. 52.5%), andonly crawlers were affected by toy location: Forcrawlers, all three conditions differed from eachother (ps < .02), but none differed for walkers(Figure 3d; Table 1, row 4). This finding sug-gests that crawlers have adequate visual accessto objects on the floor, but objects placed on fur-niture or affixed to walls are unlikely to be seenwhile crawling.

Crawling vs. walking: Gaze location Theavailability of information in the field of viewinfluenced where crawlers and walkers directedtheir gaze (Figure 5; Table 1, rows 5-6). Crawlersspent more time than walkers—in fact, a major-ity of video frames—looking down at the floor(M = 54.4% of frames overall vs. M = 28.5%of frames), and less time than walkers lookingahead at the caregiver (M = 36.6% of framesvs. M = 54.1% of frames; the remaining 9%of frames for crawlers and 17.4% of frames forwalkers were spent looking at the wall behindthe caregiver). This finding is particularly strik-ing because the experimental procedure biasedinfants to look at their caregivers, who called totheir infants throughout the trial and waved toysfor them to retrieve. The GEE confirmed a maineffect of toy height condition: Infants looked lessfrequently at the floor and more frequently at thecaregiver in the high toy condition than the other

two conditions (ps < .05), when their attentionwas pulled upwards. Significant (looking towardfloor) and marginally significant (looking towardcaregiver) posture × condition interactions andpairwise comparisons suggest that the conditioneffects were only reliable for walkers, ps < .01.The field of view data indicate shifts in the avail-ability of visual information; the crawler-walkerdifferences in gaze behavior are notable becausethey indicate shifts in infants’ active visual at-tention.

Sitting Nine of the 15 crawlers switched fromall fours to a sitting posture mid-trial. Eight in-fants sat once or twice, one sat four times, and

Low Mid High Low Mid High

Ppn

of F

rame

s

0.00

0.25

0.50

0.75

1.00

WallCaregiverFloor

Crawlers Walkers

Figure 5: Proportion of video frames where eye gazewas directed toward the floor (dark gray), the care-giver (white), and the wall behind the caregiver (lightgray), by posture and toy height condition.

11

one sat seven times. What prompted infants tosit up? Possibly, it was the lack of visual infor-mation obtained in a crawling posture: Out ofthe 22 sits in the data set, eight occurred in thehigh toy condition and 12 in the middle toy con-dition, but only two occurred in the low toy con-dition. A GEE with a Poisson link function con-firmed a significant effect of condition on the rateof infants’ spontaneous sits: Infants were signifi-cantly more likely to sit in the middle conditionthan the low condition, OR = 6.33, p = .01, andmarginally more likely to sit in the high condi-tion than the low condition, OR = 4.50, p = .07.Although infants may have intentionally sat upto see the toys (sit-to-see), it is also possible thatinfants sat up for other reasons (e.g., discomfortof crawling) and then took advantage of the moreupright posture to obtain a different view (sit-and-see). Regardless of their intentions, the datafrom the head-mounted eye tracker illustrate animportant consequence of this phenomenon.

When infants sat up, their visual world changeddramatically. For two infants (and three trialsfrom two other infants), field of view data werenot available because infants faced away from thecurtain when they sat up. The seven infants forwhom data are available had more of the curtainin view while sitting (M = 160.22cm) than whilecrawling (M = 72.21cm), t(6) = 4.86, p < .01.While sitting, infants were also more likely tohave their caregiver’s face (M = 90.8% of frames)and the toy in view (M = 92.8%) than whilecrawling (Ms = 51.1% and 58.2%, for face andtoy, respectively), ts(6) > 3.24, ps ≤ .02. Thesix infants for whom eye tracking data were avail-able were less likely to look at the floor whilesitting (M = 12.0% of frames) than while crawl-ing (M = 48.4%), t(5) = 2.72, p = .04. Theywere also more likely to look at the caregiverwhile sitting (M = 62.6%) than while crawling(M = 40.0%), but this effect did not reach sig-nificance in the small subsample, t(5) = 1.65,p = .16.

Summary Even in an identical physical andsocial context, the visual experiences of 13-month-old crawlers and walkers were very dif-

ferent. Crawlers’ visual world was dominated bythe floor, whereas walkers had continuous visualaccess to distal and elevated people and objects.Crawlers got a better view of the room whenthey interrupted locomotion to sit up and take alook around.

Although the primary comparisons in this exper-iment were between subjects, the effects shouldbe attributed to the real-time physical con-straints of different postures rather than groupdifferences between crawlers and walkers. Thetransition to sitting immediately transformedthe visual world of crawlers; while sitting up,those same crawlers had a similar view of theroom, people, and objects as did walkers. Infact, crawlers spontaneously stop crawling tosit up after only a few seconds of crawling inboth standard crawling tasks and free play situ-ations (Soska, Robinson, & Adolph, 2013), in-dicating that crawlers’ floor-centered view ofthe world is intermittent. Moreover, in every-day life, crawlers have additional postures avail-able—standing or cruising upright—and theycan view the world from a higher vantage pointwhile being carried. Similarly, we were able toentice three of the crawlers and three of thewalkers to perform a trial or two in the oppo-site posture (one crawler was just barely able tostring several independent steps together acrossthe walkway; the others happily walked holdingthe experimenter’s hands). These infants’ dataresembled the crawler group while crawling andthe walker group while walking. Thus, the visualeffects of the transition from crawling to walkingappear to be instantaneous.

Experiment 2: BodyConstraints on Visual

Experience

Differences in visual experience between crawlersand walkers could, in theory, stem from two sep-arate body constraints: eye height and headangle. However, the relative contributions of

12

these two factors have not been examined. Eyeheight is clearly highest in walking, lowest incrawling, and somewhere in between in sitting,but how different are the three postures and towhat extent are differences in visual input drivenby height differences alone? Is crawlers’ heightdisadvantage compounded by incomplete neckextension so that their faces point toward theground, or do they strain their necks to compen-sate? Where do infants point their heads in theupright postures of walking and sitting?

To examine the contributions of height and headangle on the different visual experiences illus-trated in Experiment 1, we ran a small sampleof infants in the same crawling/walking protocoland measured the position and orientation of thehead throughout each trial.

Method

Participants Thirteen 13-month-old infantswere recruited and compensated in the samemanner as Experiment 1. Nine infants werecrawlers (M = 3.38 months crawling experience)and 4 infants were walkers (M = 1.81 monthswalking experience). Data from 4 additional in-fants were excluded for refusal to crawl on theraised walkway (n = 3) or equipment failure(n = 1).

Motion capture system and procedureThe walkway and task were identical to Ex-periment 1: Infants crawled or walked to theircaregivers several times over the striped plat-form. We used an electromagnetic motion track-ing system (Ascension TrakSTAR) to track theposition and orientation of infants’ heads whilecrawling and walking. The system collects six-degree-of-freedom measurements—three dimen-sions of position (x, y, z) and three axes of ori-entation (pitch, yaw, and roll)—of small sensorswith reference to a fixed magnetic field trans-mitter. Measurements were collected at 240 Hzusing custom software.

Because sensor accuracy is a function of proxim-

ity to the transmitter, the transmitter was placednext to the raised platform and elevated 43 cmfrom the ground on a plastic stand. The mostaccurate measurements were obtained within a150-cm portion of the walkway (approximately378 cm from the curtain to 228 cm from the cur-tain) and only data within this portion were an-alyzed.

After infants crawled or walked a few times onthe platform to become comfortable with thetask, the experimenter placed the equipment. In-fants wore the same hat and vest used with thehead-mounted eye tracker (see Figure 1A), anda sensor was secured to the right side of the hatwith Velcro. The sensor was oriented parallel toan imaginary line running from the junction ofthe ear and the head to the corner of the eye.This is approximately parallel to the standardanatomical Frankfort plane, thus ensuring thatwhen the sensor was parallel to the ground thehead was in the anatomically neutral position.The sensor was connected to the computer via along cable that was secured with Velcro to the in-fants’ backs to allow enough slack for head move-ments; the cable was held out of the way by anassistant who followed behind infants during thetrials.

Measurements from only one sensor provide dataonly about the orientation of the sensor relativeto the ground surface. Because it was also impor-tant for us to know how much infants extendedtheir necks, that is, the orientation of crawlers’heads relative to their trunks, four crawlers werealso outfitted with a second sensor attached tothe back of the Velcro vest.

Data reduction As in Experiment 1, codersscored the timing and location of infants’ stepsfrom video. Readings from the motion capturesensors were averaged over the duration of eachstep for analysis, so that the data resolution wasthe same as in Experiment 1. Coders also identi-fied periods of time when crawling infants spon-taneously sat up (there were 14 spontaneous sitsproduced by 5 crawlers); these data were ana-lyzed separately.

13

Results and Discussion

Height To our knowledge, this is the firststudy to quantify infants’ functional height dur-ing crawling and walking. We found that walkerswere twice as tall while walking (M sensor height= 68.78cm) than crawlers were while crawling(M = 33.94cm). When they sat up, crawlers’height was larger than while crawling but stillrelatively low, ranging from 35.77-55.26 cm de-pending on whether infants knelt with their bot-toms resting on their heels or sat with their bot-toms on the platform surface.

Head pitch angle Figure 6a-b shows the dis-tributions of head pitch angles for crawlers andwalkers, with 0 degrees representing a head ori-entation parallel to the ground. Head pitch an-gles for walkers were centered around 0 degrees,as would be expected from a neutral head posi-tion in an upright posture. Crawlers’ head pitchangles were lower than walkers on average, butthere was a great deal of overlap and the differ-ence did not reach statistical significance in thissample (p = .12)—surprisingly, most crawlersmanaged to bring their heads to parallel at somepoint. Crawlers did not point their heads towardthe ceiling to compensate for being half as tall ina crawling posture. However, they nearly alwaysdid so while sitting (Figure 6c).

Crawlers displayed more variability in head posi-tion than did the walkers (Figure 6a). Crawlershad larger within-trial standard deviations inhead pitch angle Ms = 6.90 for crawlers and 4.33for walkers, Wald χ2 = 7.57, p < .01. Crawlersalso displayed larger deviations in head pitch an-gle between consecutive steps, Ms = 4.87 de-grees for crawlers and 3.47 degrees for crawlers,Wald χ2 = 18.34, p < .01. This indicates thatcrawlers flexed and extended their necks repeat-edly, but walkers kept their heads more stable.

Trunk orientation and neck extensionCrawlers’ shoulders were propped up slightlyhigher than their bottoms, resulting in averagetrunk angles ranging from 15.56 to 18.63 degreesfrom parallel for the four infants tested with the

Head Pitch Angle (degrees)-60 -40 -20 0 20 40 60

Numb

er o

f Ste

ps

0

30

60

90

120

Head Pitch Angle (degrees)-60 -40 -20 0 20 40 60

Numb

er o

f Ste

ps

0

30

60

90

Head Pitch Angle (degrees)-60 -40 -20 0 20 40 60

Numb

er o

f Sits

0

1

2

3

4

5

A. Crawlers

B. Walkers

C. Sits

Figure 6: Distribution of head pitch angle measure-ments pooled over infants and trials for (A) crawlers,(B) walkers, and (C) crawlers while sitting.

extra marker on their backs. To obtain a roughmeasure of neck extension, we compared infants’trunk angle to the complement of their headpitch angle. While crawling, infants craned theirnecks at about the maximum extension (60-90degrees) reported in the literature for childrenand adults in a stationary standing or sittingposition (Klinich & Reed, 2013; Lewandowski &Szulc, 2003; Lynch-Caris et al., 2008; Ohman& Beckung, 2008; Youdas et al., 1992): aver-age extension for the four infants ranged from

14

49.19 to 81.65 degrees. This indicates that in-fants failed to point their heads up while crawlingbecause they had reached their physical limits.Note that the infants in this study were mature,experienced crawlers. Earlier in development, asyounger, weaker, less coordinated crawlers, lift-ing the head may be even more difficult; it maybe especially difficult for infants who crawl ontheir bellies.

Calculated head rotation for Experiment1 infants Once we had a measure of averageheight (hhead) for crawlers and walkers, we wereable to use the field of view data to estimate headpitch angles (θ) for infants in Experiment 1. Foreach step, we used measurements of the highestpoint visible (hvisible) and the distance from thecurtain (d) at each step to create a triangle thatcould be solved using the following equation:

θ = tan−1 hvisible − hheadd

− 21

The first part of this equation yields the pitchangle of the upper boundary of the scene camerafield of view; to estimate the pitch angle of thecenter of the scene camera, the constant of 21degrees (half the vertical field of view of the scenecamera) was subtracted.

Walkers (M = −9.13 degrees) in Experiment 1pointed their heads significantly higher than didcrawlers (M = −21.76 degrees), t(28) = 5.29,p < .01. Although reliable, the difference be-tween crawlers and walkers was far smaller thanwould have been expected if crawlers had kepttheir heads in an effortless, neutral posture.

General Discussion

Using head-mounted eye tracking and motiontracking, we documented differences in visual in-put during crawling and walking. While crawl-ing, infants mostly see the ground in front oftheir hands; while walking, they see the wholeroom and its inhabitants. These differences invisual input result directly from the different

constraints of infants’ bodies while crawling andwalking.

Human bodies are not well built for quadrupedallocomotion. The evolution of upright walkingproduced modifications in the anatomy of thehuman spine (Tobias, 1992). The spine at-taches to the base of the skull in humans. Incontrast, the spine attaches to the back of theskull in horses, cats, nonhuman primates, andother mammals. Our anatomy is optimal for thebipedal walking and other upright postures, butrestricts visual access to the environment whilecrawling. Crawlers struggle to exploit the fullrange of motion of their necks—straining againstgravity to do so—but still see the world differ-ently than walkers do; that is, unless they sit upand tilt their heads toward the ceiling.

Visual input in everyday situations

Should we expect the results from this controlledlaboratory experiment to generalize to infants’actual everyday experiences? We tested infantsin a highly controlled setting for two reasons:to isolate postural contributions to infants’ vi-sual experience, and to facilitate measurementof infants’ visual fields. Spontaneous locomotiondiffers dramatically between 12- to 13-month-old crawlers and walkers: Walkers spend moretime in motion, travel three times the distance,explore more areas of the environment, and in-teract differently with their caregivers (Adolphet al., 2012; Clearfield, 2011; Karasik et al., inpress). So in the current study, we held num-ber of trips, path, and the caregiver’s interactionconstant. We also covered the walkway with col-ored stripes to make it easier for coders to scoreinfants’ field of view from video. But could ourparticular experimental setup have biased the re-sults?

One possibility is that the colored stripes on thewalkway encouraged crawlers to look down morethan they otherwise would have. However, walk-ers traveled down the same walkway with thesame colored stripes and looked at the floor sig-

15

nificantly less frequently than crawlers. If thestripes did indeed draw crawlers’ attention down,this would suggest that salient objects on thefloor are more likely to attract the attention ofcrawlers than walkers, another potential conse-quence of their different views of the world.

Another artificial aspect of the design was thatthe caregiver sat or stood on a lower surface thanthe infants. This means that our experimentalsetup actually made it easier for crawlers to seethe caregiver’s face than if they were on the samesurface. In the low and middle conditions, care-givers’ faces were closer to infants’ eye level thanif they sat on the floor with their infants. De-spite contriving the situation so that caregiver’sfaces were maximally available to crawling in-fants, crawlers only had their caregivers’ face inview 43.7% of the time and only fixated the care-giver 36.6% of the time. In a natural environ-ment, the height of caregivers relative to infantschanges as the caregivers themselves change pos-ture; we manipulated caregiver height to sim-ulate these changes. An intriguing possibilitycurrently being investigated in our laboratory isthat caregivers are sensitive to infants’ point ofview and adapt their posture to impose them-selves in infants’ visual fields. In this way, thetransition from crawling to walking might reor-ganize parents’ behavior.

We tested infants moving down an uncluttered,uniform path. Would the differences we docu-mented between crawlers and walkers generalizeto a more complex natural environment? For-tunately, our findings converge with recent datafrom naturalistic studies. During spontaneousplay in a laboratory playroom, toddlers are morelikely to look at the floor while crawling thanwhile walking (Franchak et al., 2011), and walk-ers are more likely to have faces in their field ofview than are crawlers (Frank et al., in press).Our kinematic data suggests that the differencesbetween crawlers and walkers are a product ofthe shape of their bodies in motion, rather thanthe specific experimental context.

Advantages of walkers’ viewpoint

Our findings indicate that while walking, infantssee distal and elevated objects and people. Vi-sual access to these particular parts of the en-vironment during locomotion may contribute topreviously documented psychological advancesthat accompany the onset of walking.

Visual access to distal parts of the environmentmight increase engagement with distal objectsand people. Previous research revealed thatwalkers travel across the room to retrieve ob-jects and share them with caregivers, whereascrawlers interact with proximal objects and care-givers (Karasik et al., 2011). As a result, walk-ers receive more verbal feedback from caregivers(Karasik et al., in press). This may in turn be re-lated to recent findings that the onset of walkingaccelerates language development (Ellis-Davieset al., 2012; Iverson, 2010; Walle & Campos, inpress).

Increased visual access to distal objects may alsofacilitate the development of spatial cognition.Place learning (the coding of locations relative todistal landmarks) is not evident until about 21months of age (Newcombe, Huttenlocher, Drum-mey, & Wiley, 1998). Possibly, the ability to seemore of the environment while walking increasesinfants’ attention to the relations between distalobjects and landmarks necessary for place learn-ing (Newcombe & Learmonth, 1999).

Differences in eye level have previously been im-plicated in toddlers’ surprisingly low rates ofvisual attention to their mothers’ faces (Fran-chak et al., 2011). The current study and pre-vious work (Frank et al., in press) suggests thatcrawlers experience even less frequent visual ac-cess to faces than walkers. The transition fromcrawling to walking may therefore increase op-portunities for some types of social learning suchas joint attention or social referencing.

Although crawlers can access some of the samevisual information as walkers, doing so incurs agreater cost and is organized differently in time:

16

Crawlers must repeatedly crane their necks upand down or stop moving and sit up to samplevisual information that can be accessed contin-uously and largely for free while walking. Thisdistinction is important, because some learningexperiences may depend on viewing objects orevents at specific times or viewing a continu-ously changing display. For example, learningthe mapping between an object and its namepresumably depends on seeing the object withina certain time frame of the naming event (Yu& Smith, 2012). Walking infants may be morelikely to have objects in view at just the rightmoment, which may make these mappings eas-ier. Additionally, processes that rely on continu-ous visual tracking of objects, such as spatial re-orientation following body movement (Acredolo,Adams, & Goodwyn, 1984), may be improved inwalkers compared to crawlers.

Our findings also help to explain an enduringquestion in motor development: Why do expertcrawlers forsake a skill they have mastered for anew, initially difficult mode of locomotion? Pre-walking infants have plenty of access to the vi-sual world while sitting, cruising, and being car-ried. However, the promise of enhanced visualaccess the wider environment during locomotionmay motivate infants to stand up and begin towalk. The ability to see more of the room whilemoving may contribute to the fact that novicewalkers take more steps, travel father distances,and spend twice as much time in motion thanexperienced crawlers (Adolph et al., 2012).

Advantages of crawlers’ viewpoint

Crawlers’ view of the world has its own bene-fits. We found that while crawling, infants hadeffortless visual access to the floor in front oftheir hands; while walking, they lost sight ofthe floor close to their feet. This representsa significant advantage for crawlers for visualguidance of locomotion. Upcoming obstacles orchanges in the ground surface can be easily de-tected if they remain in the field of view through-out the approach, and appropriate locomotor re-

sponses may be more easily planned if visual con-tact is maintained during the action (Adolph,1997). Accordingly, infants display different vi-sual strategies for guiding locomotion over obsta-cles depending on their posture: They are morelikely to fixate obstacles in advance while crawl-ing than while walking (Franchak et al., 2011).

Different views of the ground ahead may con-tribute to posture-specific learning in visu-ally guided locomotion. Whereas experiencedcrawlers accurately perceive affordances for lo-comotion over slopes and cliffs, novice walkersmake large errors in their new upright posture(Adolph, 1997; Kretch & Adolph, 2013). Thecurrent study suggests that an important chal-lenge of the transition from crawling to walkingis learning to interpret substantially different vi-sual information for guiding locomotion. Differ-ent viewing angles create different correlationsamong visual, proprioceptive, and vestibular in-formation; different viewing angles also generatedifferent patterns of optic flow to specify the lay-out of objects and surfaces and infants’ move-ment through the environment (Bertenthal &Campos, 1990; J. J. Gibson, 1950, 1979). In ad-dition, timing of locomotor planning might needto be adjusted because obstacles or edges exit thebottom of the field of view earlier while walkingthan while crawling, similar to the earlier exitof obstacles from the visual field of adults com-pared with shorter children (Franchak & Adolph,2010). Indeed, the timing of obstacle fixationsis different between adults and children to takethe flow of available information into account.Newly walking infants may need weeks of prac-tice to successfully use visual information aboutthe surface layout from their new upright pos-ture.

In some ways, a limited view of the wider envi-ronment while crawling may also be an advan-tage for learning about the world. Fewer objectsand surfaces in view at a time also means fewerdistractions, and may serve as a kind of spot-light on the immediate surrounds and the cur-rent task. Moreover, not being able to see ob-jects during locomotion may encourage infants

17

to hold them in memory; repeated practice maycontribute to improvements in memory followingcrawling onset (Herbert et al., 2007).

Conclusion

Many researchers have argued that “travelbroadens the mind” (Bertenthal et al., 1984;Campos et al., 2000; E. J. Gibson, 1988).Our findings demonstrate that different forms oftravel are not on equal footing. Different viewsof the world may lead infants to have divergentexperiences and different opportunities for learn-ing while crawling or walking.

Infants experience a variety of postures at everypoint in development: lying supine or prone, be-ing carried, sitting, crawling, cruising, and walk-ing upright. Motor development changes thefrequency with which infants experience differ-ent postures; in particular, the onset of inde-pendent walking increases the amount of timeinfants spend upright. This leads to different ex-periences that facilitate a variety of developmen-tal outcomes. Our data suggest that differencesin visual experience may be a part of the suiteof changes that accompany the transition fromcrawling to walking.

Author Note

The project was supported by Award# R37HD033486 to from the Eunice KennedyShriver National Institute of Child Health &Human Development to Karen Adolph andby Graduate Research Fellowship # 0813964from the National Science Foundation to KariKretch. The content is solely the responsibilityof the authors and does not necessarily repre-sent the official views of the Eunice KennedyShriver National Institute of Child Health &Human Development, the National Institutesof Health, or the National Science Foundation.Portions of this work were presented at the 2012International Conference on Infant Studies, the2012 meeting of the Vision Sciences Society, and

the 2013 meeting of the Society for Research inChild Development. We gratefully acknowledgeJulia Brothers and members of the NYU InfantAction Lab for helping to collect and codedata. We thank Gladys Chan for her beautifulillustration of the experimental apparatus.

References

Acredolo, L., Adams, A., & Goodwyn, S. W.(1984). The role of self-produced movementand visual tracking in infant spatial orien-tation. Journal of Experimental Child Psy-chology, 38, 312-327.

Adolph, K. E. (1997). Learning in the develop-ment of infant locomotion. Monographs ofthe Society for Research in Child Develop-ment, 62 (3, Serial No. 251).

Adolph, K. E., Cole, W. G., Komati, M., Garci-aguirre, J. S., Badaly, D., Lingeman, J. M.,Chan, G. L. Y., & Sotsky, R. B. (2012).How do you learn to walk? Thousands ofsteps and hundreds of falls per day. Psycho-logical Science, 23, 1387-1394.

Anderson, D. I., Campos, J. J., Anderson, D.E., Thomas, T. D., Witherington, D. C.,Uchiyama, I., & Barbu-Roth, M. A. (2001).The flip side of perception-action coupling:Locomotor experience and the ontogenyof visual-postural coupling. Human Move-ment Science, 20, 461-487.

Aslin, R. N., & McMurray, B. (2004). Auto-mated corneal-reflection eye tracking in in-fancy: Methodological developments andapplications to cognition. Infancy, 6, 155-163.

Bai, D. L., & Bertenthal, B. I. (1992). Locomo-tor status and the development of spatialsearch skills. Child Development, 63, 215-226.

Bertenthal, B. I., & Campos, J. J. (1990).A systems approach to the organizing ef-

18

fects of self-produced locomotion during in-fancy. In C. K. Rovee-Collier & L. P. Lipsitt(Eds.), Advances in infancy research (Vol.6, pp. 1-60). Norwood, NJ: Ablex.

Bertenthal, B. I., Campos, J. J., & Barrett, K.C. (1984). Self-produced locomotion: Anorganizer of emotional, cognitive, and so-cial development in infancy. In R. N. Emde& R. J. Harmon (Eds.), Continuities anddiscontinuities in development (pp. 175-210). New York: Plenum Press.

Biringen, Z., Emde, R. N., Campos, J. J., &Applebaum, M. I. (1995). Affective reorga-nization in the infant, the mother, and thedyad: The role of upright locomotion andits timing. Child Development, 66, 499-514.

Campos, J. J., Anderson, D. I., Barbu-Roth,M. A., Hubbard, E. M., Hertenstein, M.J., & Witherington, D. C. (2000). Travelbroadens the mind. Infancy, 1, 149-219.

Campos, J. J., Kermoian, R., Witherington,D., Chen, H., & Dong, Q. (1997). Activ-ity, attention, and developmental transi-tions in infancy. In P. J. Lang & R. F. Si-mons (Eds.), Attention and orienting: Sen-sory and motivational processes (pp. 393-415). Mahwah, NJ: Lawrence Erlbaum As-sociates.

Campos, J. J., Kermoian, R., & Zumbahlen,M. R. (1992). Socioemotional transforma-tions in the family system following infantcrawling onset Emotion and its regulationin early development. New directions forchild development (pp. 25-40).

Cicchino, J. B., Aslin, R. N., & Rakison, D.H. (2011). Correspondences between whatinfants see and know about causal and self-propelled motion. Cognition, 118, 171-192.

Clearfield, M. W. (2011). Learning to walkchanges infants’ social interactions. InfantBehavior & Development, 34, 15-25.

Clearfield, M. W., Osborne, C. N., & Mullen,

M. (2008). Learning by looking: Infants’ so-cial looking behavior across the transitionfrom crawling to walking. Journal of Exper-imental Child Psychology, 100, 297-307.

Ellis-Davies, K., Sakkalou, E., Fowler, N.,Hilbrink, E., & Gattis, M. (2012, June). Ican walk, I can talk: Emerging motor devel-opment predicts language development. Pa-per presented at the International Confer-ence on Infant Studies, Minneapolis, MN.

Franchak, J. M., & Adolph, K. E. (2010). Visu-ally guided navigation: Head-mounted eye-tracking of natural locomotion in childrenand adults. Vision Research, 50, 2766-2774.

Franchak, J. M., Kretch, K. S., Soska, K. C., &Adolph, K. E. (2011). Head-mounted eyetracking: A new method to describe infantlooking. Child Development, 82, 1738-1750.

Franchak, J. M., Kretch, K. S., Soska, K. C.,Babcock, J. S., & Adolph, K. E. (2010).Head-mounted eye-tracking in infants’ nat-ural interactions: A new method. Paperpresented at the Proceedings of the 2010Symposium on Eye Tracking Research andApplications, Austin, TX.

Frank, M. C., Simmons, K., Yurovsky, D., &Pusiol, G. (2013). Developmental and pos-tural changes in children’s visual access tofaces. In M. Knauff, M. Pauen, N. Sebanz,& I. Wachsmuth (Eds.), Proceedings of the35th Annual Meeting of the Cognitive Sci-ence Society (pp. 454-459). Austin, TX:Cognitive Science Society.

Fuller, J. H. (1996). Eye position and target am-plitude effects on human visual saccadic la-tencies. Experimental Brain Research, 109,457-466.

Gibson, E. J. (1988). Exploratory behavior inthe development of perceiving, acting, andthe acquiring of knowledge. Annual Reviewof Psychology, 39, 1-41.

Gibson, E. J., & Pick, A. D. (2000). An eco-

19

logical approach to perceptual learning anddevelopment. New York: Oxford UniversityPress.

Gibson, J. J. (1950). The perception of the vi-sual world. Oxford: Houghton Mifflin.

Gibson, J. J. (1966). The senses consideredas perceptual systems. Boston: Houghton-Mifflin.

Gibson, J. J. (1979). The ecological approach tovisual perception. Boston: Houghton Mif-flin Company.

Haith, M. M. (1969). Infrared television record-ing and measurement of ocular behavior inthe human infant. American Psychologist,24, 279-283.

Herbert, J., Gross, J., & Hayne, H. (2007).Crawling is associated with more flexiblememory retrieval by 9-month-old infants.Developmental Science, 10, 183-189.

Higgins, C. I., Campos, J. J., & Kermoian, R.(1996). Effect of self-produced locomotionon infant postural compensation to opticflow. Developmental Psychology, 32, 836-841.

Horobin, K., & Acredolo, L. (1986). The role ofattentiveness, mobility history, and separa-tion of hiding sites on Stage IV search be-havior. Journal of Experimental Child Psy-chology, 41, 114-127.

Iverson, J. M. (2010). Developing language in adeveloping body: The relationship betweenmotor development and language. Journalof Child Language, 37, 229-261.

Johnson, S. P., Amso, D., & Slemmer, J. A.(2003). Development of object concepts ininfancy: Evidence for early learning in aneye tracking paradigm. Proceedings of theNational Academy of Sciences, 100, 10568-10573.

Johnson, S. P., Slemmer, J. A., & Amso, D.

(2004). Where infants look determines howthey see: Eye movements and object per-ception performance in 3-month-olds. In-fancy, 6, 185-201.

Karasik, L. B., Adolph, K. E., Tamis-LeMonda,C. S., & Zuckerman, A. (2012). Carry on:Spontaneous object carrying in 13-month-old crawling and walking infants. Devel-opental Psychology, 48, 389-397.

Karasik, L. B., Tamis-LeMonda, C. S., &Adolph, K. E. (2011). Transition fromcrawling to walking and infants’ actionswith objects and people. Child Develop-ment, 82, 1199-1209.

Karasik, L. B., Tamis-LeMonda, C. S., &Adolph, K. E. (in press). Crawling andwalking infants elicit different verbal re-sponses from mothers. Developmental Sci-ence.

Kermoian, R., & Campos, J. J. (1988). Loco-motor experience: A facilitator of spatialcognitive development. Child Development,59, 908-917.

Klinich, K. D., & Reed, M. P. (2013). Pe-diatric anthropometry. In J. R. Crandall,B. S. Myers, D. F. Meaney & S. ZellersSchmidtke (Eds.), Pediatric injury biome-chanics (pp. 1-32). New York: Springer Sci-ence+Business Media.

Kretch, K. S., & Adolph, K. E. (2013). Cliff orstep? Posture-specific learning at the edgeof a drop-off. Child Development, 84, 226-240.

Land, M. F. (2004). The coordination of rota-tions of the eyes, head and trunk in saccadicturns produced in natural situations. Ex-perimental Brain Research, 159, 151-160.

Lewandowski, J., & Szulc, P. (2003). The rangeof motion of the cervical spine in childrenaged from 3 to 7 years—an electrogonio-metric study. Folia Morphologica, 62, 459-461.

20

Lynch-Caris, T., Majeske, K. D., Brelin-Fornari, J., & Nashi, S. (2008). Establish-ing reference values for cervical spine rangeof motion in pre-pubescent children. Jour-nal of Biomechanics, 41, 2714-2719.

Newcombe, N. S., Huttenlocher, J., Drummey,A. B., & Wiley, J. G. (1998). The devel-opment of spatial location coding: Placelearning and dead reckoning in the secondand third years. Cognitive Development,13, 185-200.

Newcombe, N. S., & Learmonth, A. (1999).Change and continuity in early spatial de-velopment: Claiming the ”radical middle”.Infant Behavior & Development, 22, 457-474.

Ohman, A. M., & Beckung, E. R. E. (2008).Reference values for range of motion andmuscle function of the neck in infants. Pe-diatric Physical Therapy, 20, 53-58.

Ounsted, M., & Moar, V. A. (1986). Propor-tionality changes in the first year of life:The influence of weight for gestational ageat birth. Acta Paediatrics Scandinavia, 75,811-818.

Pare, M., & Munoz, D. P. (2001). Expressionof a re-centering bias in saccade regulationby superior colliculus neurons. Experimen-tal Brain Research, 137, 354-368.

Piaget, J. (1954). The construction of reality inthe child. New York: Basic Books.

Salapatek, P., & Kessen, W. (1966). Visualscanning of triangles by the human new-born. Journal of Experimental Child Psy-chology, 3, 155-167.

Snyder, R. G., Spencer, M. L., Owings, C. L.,& Schneider, L. W. (1975). Anthropome-try of U.S. infants and children. (Paper No.750423). Warrendale, PA: Society of Auto-motive Engineers.

Soska, K. C., Robinson, S. R., & Adolph, K. E.

(2013). A new twist on old ideas: How sit-ting reorients crawlers. Manuscript underreview.

Tobias, P. V. (1992). The upright head in ho-minid evolution. In A. Berthoz, W. Graf &P. Vidal (Eds.), The head-neck sensory mo-tor system (pp. 5-16). New York: OxfordUniversity Press.

Uchiyama, I., Anderson, D. I., Campos, J. J.,Witherington, D., Frankel, C. B., Lejeune,L., & Barbu-Roth, M. (2008). Locomotorexperience affects self and emotion. Devel-opmental Psychology, 44, 1225-1231.

von Hofsten, C., & Rosander, K. (1997). De-velopment of smooth pursuit tracking inyoung infants. Vision Research, 37, 1799-1810.

Walle, E. A., & Campos, J. J. (in press). In-fant language development is related to theacquisition of walking. Developmental Psy-chology.

Youdas, J. W., Garrett, T. R., Suman, V. J.,Bogard, C. L., Hallman, H. O., & Carey,J. R. (1992). Normal range of motion ofthe cervical spine: An initial goniometricstudy. Physical Therapy, 72, 770-780.

Yu, C., & Smith, L. B. (2012). Embodied atten-tion and word learning by toddlers. Cogni-tion, 125, 244-262.