Coordinating Attention to People and Objectsin Mother-Infant and Peer-Infant Interaction

Roger Bakeman and Lauren B. AdamsonGeorgia State University

BAKEMAN, ROGER, and ADA.MSON, LAUREN B. Coordinating Attention to People and Objects inMother-Infant and Peer-Infant Interaction. CHILD DEVELOPMENT, 1984, 55, 1278-1289. In alongitudinal study, inlants 6-18 months of age were observed in their homes playing with theirmothers and with peers. Of primary concern was how they coordinated their attention to peopleand objects. Observations were coded using a state-based scheme that included a state of coordinatedjoint engagement as well as states of person engagement, object engagement, onlookiug, audpassive joint engagement. All developmental trends observed were similar regardless of partner:person engagement declined with age, while coordinated joint engagement increased. Passivejoint engagement, object engagement, and onlooking did not change with age. However, theabsolute amount of some engagement states was affected by partner: both passive and coordinatedjoint engagement were much more likely when infants played with mothers. We conclude thatmothers may indeed support or "scaffold" their infants' early attempts to embed objects in socialinteraction, but that as attentional capabilities develop even quite unskilled peers may beappropriate partners for the exercise of these capacities.

The emergence of the ability to 1974; Tronick, Als, & Adamson, 1979). When,coordinate attention toward a social partner as is increasingly the case by the middle ofand an object of mutual interest is often the first year, infants turn away from face-to-regarded as an important developmental face interactions toward object exploration,milestone. In several accounts of early they still seem to attend to only a single aspectcommunication development, this pattern of of their surroundings (Kaye & Fogel, 1980;attention marks a pivotal change in the Trevarthen & Hubley, 1978); now they focusinfant's communicative competence (Werner intently on objects, providing few indications& Kaplan, 1963). It has been viewed both as that they desire to share this new interestthe "developmental heir" of the very young with their partners,infant's interactions with caregivers /-> j n . r > . . i(Trevarthen & Hubley, 1978) and as the . Gradually, infants interactions becomefoundation for subsequent symbolically *"^f"^/f object-focused attention becomes

1. . J. .. ,„ in'7r:\ embedded in social contexts. Around 6mediated conversations (Bruner, 1975). i r , , , , ,

months or age, babies begin to switch theirAccording to most accounts, the gaze back and forth between caregiver and

developmental course of this pattern of object (Newson & Newson, 1975). Sooncoordinated attention spans much of infancy. gestures become clear signals that infants arePrior to its advent, infants engage solely in actively attempting to share attention todyadic interactions. When they are engaged something external to the social interaction,with a person, their attention seems confined establishing that object or event as the "topic"to the process of interaction itself. Thus, while of joint concern (Leung & Rheingold, 1981;young infants can participate in finely tuned Murphy & Messer, 1977). Finally, by aboutexchanges of affect, the system of 13 months of age, this ability to coordinatecommunication is essentially expressive attention becomes consolidated, allowing(Brazelton, Koslowski, & Main, 1974; Stern, infants to enter readily into nonverbal

This research was supported by a grant from the National Science Foundation (BNS-8012068).Many people have contributed to this work, but we would like especially to thank our projectadministrator, Pamela Rutherford, whose many contributions were often above and beyond thecall of duty; Kathy Buck and Kathy Brown, who aided us when developing the coding scheme;Kathy Brown, Carol Calvert, Vickie Reade, and Emily Simerly, who faithfully coded engagementstate; Darlene Meador and Salvador Macias III, who operated our video cameras; and of coursethe mothers, who allowed us to come into their homes to observe. A preliminary version of thispaper was presented at the biennial meeting of the Society for Research in Child Development,Detroit, April 1983. Requests for reprints may be sent to either author. Department of Psychology,Georgia State University, Atlant;i, Georgia 30303.

[Child Development. 1984, 55, 1278-1289. © 1984 by the Society for Research in Child Development, Inc.All rights reserved. 0(X)9-3920/84/5504-0024$01.00]

Bakeman and Adamson 1279

referential communication with a personabout a present object (Bates, 1979; Harding& Golinkoff, 1979).

The foregoing account of thedevelopmental course of coordinatedattention is drawn from several sources. Thepurpose of the research reported here is todocument—within the confines of one studyand one sample followed longitudinally—thesequence proposed. In particular, we areconcerned with two main issues: (a) thesequence and timing of normative changesin the patterning of infant attention inpotentially social situations, and (b) theuniqueness of the role adults may play infostering rudimentary forms of coordinatedattention. To address these issues, weperformed an observational study aimed atdescribing how 6—18-month-old infantsdistributed their attention to people andobjects as they played in their homes withmothers and peers.

An important aspect of the first issueconcerns when coordinated attention appears.In particular, it seems likely that aconsiderable gap exists between when infantsare first able to display coordinated attentionand when they do so routinely during freeplay. Knowing when infants typically beginto integrate social and object realms wouldbe helpful in two ways. First, such norms forthe performance of coordinated attentionwould be useful during clinical assessmentsof infants experiencing developmentaldifficulties (Seibert, Note 1). Second, suchinformation could help clarify the relationshipbetween major changes in the attentionalstructure of social interaction and thedevelopment of specific communicativeskills. In particular, it may help explain why,even after the capacity to coordinate attentionappears to be consolidated, it is often severalmonths before the tiansition to a linguisticallybased communication system occurs (Nelson,1979).

The second issue addressed hereconcerns how partners may affect the wayinfants deploy their attention. Someinvestigators of early communicationdevelopment emphasize the essential natureof partner support when infants first begin tomaintain engagement with an object to whichthe partner is also attending. For example,Bruner (1982) suggests that caregiversprovide a necessary "scaffold" or supportivestructure for infants as they begin to employreferential communicative actions duringshared activities such as picture-book reading(Ninio & Bruner, 1978) and object hide-and-

seek (Ratner & Bruner, 1978). If this view iscorrect, we would expect considerably morecoordinated attention when infants areobserved with mothers than with same-agedpeers, assuming, as seems reasonable, thatmothers are both more "competent" thanpeers in assisting shared activities and moremotivated to complement their infants'attention.

Indeed, mothers may be such eager andable supporters of joint attention that they ineffect free their infants of the need, at leastinitially, to shift attention back and forthbetween the mother and the object of mutualconcern. To assess this possibility, wedistinguish between what we call "passivejoint engagement" and "coordinated jointengagement." In both cases, infants attendto the same object as their partners do. Forexample, during an episode of passive jointengagement, a baby might look intently at astring of beads, attempting to grasp it as themother dangles it invitingly. Although theinfant might be interacting with the beads ina way that would not be possible if the motherwere not also involved, the baby's attentionappears to be primarily on the beads whilethe mother complements this engagement.During an episode of coordinated jointengagement, however, the baby might notonly attempt to manipulate tlie beads but alsoto glance briefly at the partner, perhapssmiling at her and pointing to the beads asshe moves them about. Because passive jointengagement places fewer attentionaldemands on infants, we expect it would occurmore often at the younger ages thancoordinated joint engagement. We also expectthat, like coordinated joint engagement,passive joint engagement would occur moreoften when infants are with their presumablymore supportive mothers than when withpeers.

The present study segments infants'attention during play with mothers and peersinto sequences of "engagement states." Theseinclude not only the two kinds of jointengagement just described but also personengagement, object engagement, onlooking,and unengaged (cf. Parten, 1932). Pilot workwith our videotapes suggested that passivejoint engagement, although qualitativelydifferent from coordinated joint engagement,was more similar to it than solitary objectengagement (which it clearly resembles). Totest this notion, we examined howengagement states were sequenced,reasoning that if the two kinds of jointengagement were similar—in the sense of

1280 ChUd Development

"functioning" the same way within the streamof behavior observed—then they should bothoccupy similar "positions" in that stieam, thatis, be preceded and followed by the sameother states. Finally, differences insequencing between mothers and peers areexamined. If mothers deliberately foster bothpassive and coordinated joint engagement farmore than peers do, then we might expectthat antecedents of these states of jointattention would be considerably moresystematic when playing with mothers thanwith peers.

Methods

SubjectsTwenty-eight infants participated in this

study. They were divided into two cohorts of14 infants each, based on their age at the firstof four recording sessions. Infimts in cohort1 were observed when they were within 2weeks of 6, 9, 12, and 15 months of age (ex-cept for one 9-month and one 12-month ses-sion that occurred 8 and 10 days, respec-tively, after the 2-week time interval). Allinfants in cohort 2 were observed when theywere within 2 weeks of 9, 12, 15, and 18months of age. An equal number of boys andgirls were in each cohort; cohort 2 comprisedan equal number of first- and later-born in-fants, while cohort 1 contained five first- andnine later-born infants.

The infants' mothers were volunteerswho had been contacted initially throughseveral community organizations. All sub-jects were living in intact, middle or profes-sional class, nuclear families. The mothers'mean age was 29.7 years (range = 23-38years), and 75% had completed college. Al-though only two of the mothers were em-ployed full-time, 64% worked outside thehome or attended school on at least a part-time basis.

Each mother recruited a peer who hadhad some previous contact with her infant.This was done to avoid the "stranger" or"novelty" effects an unfamiliar peer mightevoke. Mothers reported that in the monthprior to 77% of the sessions, the infant andpeer had played together two or more times;an additional 20% of the sessions had beenpreceded by one contact in the past month,while no contact had occurred in the priormonth for 3% of the sessions. In 23 of the 28cases, the same peer was retained for all foursessions. The mean age difference betweenthe peers and infants was 3.3 weeks (rangeof age difference = 1—55 days).

Recording SessionsProcedure.—All recording sessions took

place in the infants' homes. They werescheduled at the mother's convenience andlasted between I1/2 and 2 hours. The mothersof both the infants and peers were informedthat we were interested in observing healthyinfants as they developed ways of commu-nicating with adults and peers. They weretold that we needed videotapes of their ba-bies' typical performances, not records of their"best" ones. Yet, ifthey thought an infant wasbecoming overly tired or upset, we urgedthem to suggest we stop our videotapingeither briefly or for the day. In 12% of thesessions, we did terminate our videotapingbefore the full session was completed. Tenof these interrupted sessions were com-pleted within the next week, two of the re-maining three were finished within the next3 weeks, and only one (that of a 15-month-old girl from cohort 1) could not be resched-uled, resulting in missing data for one peer-infant observation.

Conditions.—Each session consisted ofthree conditions, (a) In the mother condition,the infant and mother were observed whilethey played on the floor with a set of toys weprovided. Mothers were asked to play withtheir infants as they might if they had a fewminutes to devote to a spontaneous play pe-riod, (b) In the peer condition, the two in-fants were placed in the center of the roomnear an array of toys. Their mothers wereasked to sit to the side of the room and tocarry on a conversation wth minimal atten-tion to their infants, (c) In the alone condi-tion, the infant was placed on the floor in theroom's center near a group of toys and themother sat ofFto one side on a chair or couch.Data from this condition are not used in thisreport.

The mother and alone conditions werecounterbalanced over sessions and subjects.The peer condition alternated between beingthe first and the last condition of a session;it was not scheduled as the middle conditionin order to minimize the disruption that mightbe caused by peer's arrival and departure.

Toys.—Three similar but nonidenticalsets of toys were used. Each set contained atoy telephone, a picture book, a four-piecewooden puzzle, a set of colorful nesting cups,a doll, a rattle with movable parts, and a softplastic toy with wheels. The order of pre-sentation of the three sets was counterbal-anced over conditions by sessions and bysubjects.

Bakeman and Adamson 1281

Video recording.—Two black-and-whitevideo cameras (Sanyo VCM-20N50 withNewvicon tube), capable of recording withno more light than is available in most homes,were set up facing each other in the largestroom in which the infant typically played. Inthis way, two different views of the sameevents were obtained; later, coders couldeither view both tapes or select the one thatgave the best view ofthe particular behaviorbeing coded. Every effort was made to be asunobtrusive as possible. The images from thecamera were recorded on two %-inch por-table video cassette recorders (JVC CR-4400U). A time-code generator (For.A TCG-3200) was connected to both recorders so thatthe same moment in time could be locatedon playback on both tapes. The time codewas recorded on one of each cassette's twoaudio tracks so that it was "internal" to thetape, not part of the picture. The standardtime code defined by the Society of MotionPicture and Television Engineers (SMPTE)was used; it consists of eight digits, two forhours, two for minutes, two for seconds, andtwo for frame number. The other audio trackof each cassette was used to record soundfrom one of two microphones (Shure 571Lavalier) that were placed on the floor at twoopposing points ofthe room's perimeter.

Camera operators were instructed to ob-tain approximately 10 min of usable record-ing in each condition, stopping when theyestimated this time period had elapsed, re-gardless ofthe subjects' activities at that mo-ment. They were also to stop recording ifevents moved substantially beyond theboundaries of the conditions as we definedthem or if the mother requested an interrup-tion. In the mother condition, breaks in tap-ing were relatively rare; most ofthe sessionsconsisted of a single segment of uninter-rupted recording such that the mean numberof interruptions per mother condition was 0.2.Interruptions were more frequent in the peerand the alone conditions, when the meannumbers of interruptions per condition were2.8 and 1.6, respectively.

Once the times that the baby was off-camera or times an adult intervened wereexcluded (see definitions below), the meantime per condition was 10.1 min (SD = 1.8).Reasons for variable times include: somechildren became unduly fussy before 10 minhad expired, when interruptions occurredcamera operators did not keep track of elapsedtime accurately, ojierators could not knowwhat later would be coded an adult inter-ruption instead of adult intervention (see be-

low), and operators were constrained to fittwo ofthe conditions on one 20-min cassetteeven though part of a condition might laterbe coded as "off-condition."

CodingThe coding scheme used here segments

the infant's activity exhaustively into distinctand mutually exclusive periods that are char-acterized in terms ofthe infant's engagementwith objects and/or people in the environ-ment. We conceptualize these periods as"states" that have some duration, and so verybrief fluctuations in attention, those lastingless than 3 sec, are not regarded as indicatingan engagement state change.

Categories.—Six categories of engage-ment were defined. The categories are as fol-lows, (a) Unengaged: The infant appears un-involved with any specific person, object, oractivity, although he or she might be scan-ning the environment as though looking forsomething to do. (b) Onlooking: The infantis observing another's activity, often quite in-tently, but is not taking part in that activity.(c) Persons: The infant is engaged just withthe other person. Typically such engagementinvolves face-to-face or person play. For ex-ample, a baby giggles and coos as his motherplaces her face close to his and tickles him,or the baby looks at the peer's face, vocalizes,and then reaches toward the peer, (d) Ob-jects: The infant is involved in playing withobjects alone, attending just to the toys athand, (e) Passive joint: The infant and theother person (the mother in the mother con-dition, the peer in the peer condition) areactively involved in the same object, but thebaby evidences little awareness ofthe other'sinvolvement or even presence. Mothers oftenattempt to induce this state in their babies.They will manipulate objects (e.g., shake rat-tles, ring toy telephones, etc.) in ways thatseem designed to capture their infants' at-tention by making the objects "come alive"for them. (/) Coordinated joint: The infant isactively involved with and coordinates his orher attention to both another person and theobject that person is involved with. For ex-ample, the baby pushes the truck the motherhas been pushing and then looks back andforth between the mother's face and the ti uck;or the baby bangs his hand onto the same toythe peer is manipulating and then looks atthe peer, bangs- the toy, and tlien looks at thepeer once again. This is similar to Sugarman-Bell's (1978) "co-ordinated person-objectorientation" and it would include what Eck-erman has called "coordinate play" (Ecker-man, Whatley, & Kutz, 1975).

1282 Child Development

Three additional codes were used to in-dicate times when we did not regard the in-fant as in any ofthe six states described above,(a) Off-camera: In spite ofthe best efforts ofour camera operators, neither cassette con-tained an adequate view of the infant; thiswas a rare event, (b) Adult intervention: Theinfant (or peer) became sufficiently upset thatadult intei"vention was required to soothe andcomfort him or her. This usually occurredonly during the alone or peer conditions, andtypically the cameras were turned ofT untilthe baby's equilibrium was restored, (c) Adultinterruption: The adult became concernedby something the infant was doing and triedto redirect his or her attention. By definition,this could occur only in the alone or peerconditions. Periods coded as off-camera oradult intervention were not included in sub-sequent analyses. Periods coded as adult in-terruption were retained because they didnot represent a major break in the condition.

Coding procedures.—Two teams of twocoders each coded engagement state. Tapeswere viewed using an editing VCR (JVC CR-8200U), a SMPTE time-code reader (For.ATCR-3100), and a 19-inch monitor (Pana-sonic TR-195 MB). With this equipment,coders can play tape at variable speeds, fromone-fifth real time to five times real time,forward and backward, while both maintain-ing a picture and displaying the current timecode; playback can also be stopped or "fro-zen," again with the current time code dis-played. In practice, coders would play a tapein real time until they decided that the cur-rent engagement state had changed. Theywould then reverse the tape, positioning itbefore the suspected "break point" in en-gagement states, and then play forward againin real time. This fast-backward, real-time-forward procedure might be repeated severaltimes until the team felt certain about thetiming of the break point. They would thenstop the tape at the break point and recordthe time code, accurate to the second. In sum,the coders' task was to segment the streamof behavior into sequences of the engage-ment states we defined, identifying "breakpoints" in the stream of behavior in the sensethat Newtson (1973) and others use that term.

Interobserver reliability.—Somewhatmore than 15% of the sessions were codedindependently by both teams. These reli-ability checks were not announced to thecoding teams and were interspersed ran-domly throughout coding. Each team's codingwas compared with a "standard" version pre-pared by the two principal investigators. The

amount of agreement was gauged with Co-hen's kappa (Cohen, 1960), a statistic thatcorrects for chance agreement and so pro-duces lower values than the percentageagreement statistics often given (Hollen-beck, 1978). For the 20 mother sessionschecked, kappas ranged from .62 (77% agree-ment in this case) to .91 (93% agreement),with a mean of .77 (85% agreement); for the18 peer sessions checked, kappas ranged from.62 (75% agreement) to .92 (94% agreement),with a mean of .79 (88% agreement); and forthe 17 alone sessions checked, kappas rangedfrom .58 (82% agreement) to 1.00 (100%agreement), with a mean of .86 (96% agree-ment). An agreement was tallied only if agiven second was coded in the same state asthe standard. The unit used—a second—seems reasonable, if arbitrary, but changingit to a half second, for example, would notchange values for kappa.

Results

Not all babies were observed in eachengagement state in both mother and peerconditions at all ages. At each age, most ba-bies were observed at least once unengaged,onlooking, and engaged in objects, while onlyabout two-thirds were observed once or moreengaged with persons. All babies were ob-served at least once in the passive joint statewhen with their mothers, but only about two-thirds were when with a peer. The percentof babies engaging at least once in coordi-nated joint activity increased quite sharplywith age and was usually greater when withthe mother than when with a peer.

The average percents of time infantsspent in the various engagement states aregiven in Table 1. These percents were ana-lyzed with sex X partner (mother/peer) xage (6, 9, 12, 15, or else 9, 12, 15, 18) re-peated-measures analyses of variance, runseparately for each cohort. A second set ofotherwise identical analyses substituted par-ity for sex so that the effects of parity couldbe examined. To examine order effects, a thirdset substituted order—peer first or peer last—for sex. When age effects were significant,they were decomposed into linear, quadratic,and cubic trends; we then report which trendswere significant because this provides infor-mation as to the exact form of the age effect.Effects of all kinds are mentioned only if theywere significant in both cohoits, thus pro-viding within-sample replication. For econ-omy of presentation, only p values and notF values are given here, although completeANOVA statistics are available from the au-

Bakeman and Adamson 1283

TABLE 1

PERCENT OF TIME IN ENGAGEMENT STATE

AGE (in Months)

STATE AND CONDITION 6 9 12 15 18

Unengaged:Mother 20.9 13.3 13.1 6.1 2.9Peer 27.3 25.1 31.5 16.4 9.7

Onlooking;Mother H.9 13.5 12.9 14.0 7.5Peer 24.7 10.3 12.5 15.5 9.8

Persons:Mother 11.7 12.2 7.6 4.6 4.6Peer 11.8 8.2 4.6 5.3 3.3

Objects:Mother 36.7 42.1 43.4 40.7 36.7Peer 32.1 50.0 44.9 51.5 58.7

Passive joint:Mother 16.6 16.9 19.3 23.1 21.5Peer 3.4 3.4 3.2 1.9 4.3

Coordinated joint:Mother 2.3 2.0 3.6 11.2 26.6Peer 3 1.7 1.8 4.2 7.2

NOTE.—Numbers are mean percents, based on 14 infants at 6and 18 months. 27 infants at 15 months in the peer condition, and28 infants otherwise. Because adult interruption occupied some time,especially for older infants in the peer condition, not all percentsadd to 100.

thors. (Data for the one missing condition—a 15-month-old infant in the peer condition—were replaced with means.)

Age effects.—As the babies became older,they spent more time in coordinated jointplay (linear trend p < .05 and .01 for cohorts1 and 2, respectively), less time engaged withthe other person (i.e., the mother in themother condition and the peer in the peercondition; linear trend p < .05 and .01), andless time unengaged (linear trend p < .01and .001). In general, the amount of time spentengaged in passive joint activity, in onlook-ing, and with objects did not change signif-icantly as the infants became older.

Partner effects.—When with their motherinstead of a peer, the infants spent more timein both coordinated joint (p < .05 and .01)and passive joint activity (p < .001 for both),and less time unengaged (p < .01 and .001).There were no significant differences be-tween mother and peer conditions in theamount of time spent onlooking, engagedwith objects, or engaged with the otherperson.

Other effects.—Finally, there was noconsistent pattern of either sex, parity, or or-der effects with respect to percent of timespent in the various engagement states.

Durations.—Mean durations—that is, theaverage number of seconds each kind of en-gagement state lasted, computed for each in-fant—were also analyzed. These resultslargely parallel the ones reported above:mean durations for coordinated joint play be-came longer with age (linear trend p < .01for both cohorts), and when with their motherinstead of a peer, episodes of both cooixli-nated joint (p < .01 and .05) and passive jointplay (p < .001 for both) lasted longer. Meandurations for episodes of coordinated jointengagement were 7.1, 5.7, 10.6, 19.5, and 33.8sec at 6, 9, 12, 15, and 18 months when withmothers and 1.6, 5.3, 4.7, 5.3, and 8.8 sec forthose ages when with peers. Mean durationsfor episodes of passive joint engagement were15.1 and 5.6 sec when with mothers and peers,respectively (averaged over all ages). Therewere no consistent effects for the other states;mean durations, averaged over all ages andconditions, for episodes of unengaged, on-looking, person, and object engagement were10.9, 8.8, 9.3, and 19.6 sec, respectively.

At most ages, more infants were ob-served at least once in passive and in coor-dinated joint play when with the mother in-stead of a peer. This raises the possibility thatthe differences claimed above for both per-cents and mean durations between the mother

1284 Child Development

and peer conditions might not hold if infantswho were never observed in the passive orcoordinated joint states at a particular agewere excluded from the analysis. In fact,when such infants are excluded the differ-ences reported remain.

Sequences.—The analyses of variancejust reported indicate that, as infants in ourstudy hecame older, they became more likelyto coordinate their attention to people andobjects. They also spent more time in bothpassive and coordinated joint engagementwhen with mothers than when with peers.Thus these data tell us something about theamount of time spent in different engage-ment states, but they do not tell us whetherinfants' attention to people and/or objects wassequenced in any systematic way within anongoing period of play. In particular, they donot reveal whether there were characteristicways of attending to people and objects justbefore periods of joint engagement were be-gun or just after such periods ceased.

To answer such questions, we next ex-amined how episodes of passive and coor-

dinated joint engagement were positionedrelative to other engagement states withinthe observation sessions. The primary statis-tic for these analyses is a z score comparingobserved values for particular engagementstate transitions to their expected ones. Anadvantage of this z score is that it takes baserate into account; if the z score is positive,then that particular transition is more likelyto occur than would be expected just bychance, given the base rate ofthe consequentstate. In addition, z scores can be tested forsignificance (Allison & Liker, 1982; Gottman6f Bakeman, 1979; and Sackett, 1979).

For the first set of analyses reportedhere—those concerned with the antecedentsand consequents of passive and coordinatedjoint engagement—subjects' data were pooledwithin each age and condition; even so, therewere insufficient numbers of some se-quences to report results (see Table 2). Theseanalyses indicate that both passive and co-ordinated joint engagement were likely to bepreceded by object play; when with bothmothers and peers, the sequences object en-

TABLE 2

TRANSITIONS TO PASSIVE AND COORDINATED JOINT ENGAGEMENT

ACE (in Months)

12 15 18ANTEGEDENT-CONSEQUENT STATE CONDITION 6 9

Unengaged—passive joint:Mother - 1.3 - 2.3*Peer - 2.4*

Onlooking—passive joint:Mother 4.6*** 7.3***Peer 2.8**

Person—passive joint:Mother - 2 . 1 * -1.6Peer X X

Object—passive joint:Mother 2.8** 3.6***Peer 1.4 1.5

Unengaged—coordinated:Mother X XPeer X -1.6

Onlooking—coordinated:Mother X XPeer X

Object—coordinated:Mother X 1.4Peer X 2.0*

NOTE.—Scores are z scores based on episodes pooled across subjects. In ellect, these 2 scores compare agiven transition to its base-rate expected value. Only z scores significant at the .20 level or better arc shown.Z scores not computed because too few episodes were coded are marked with "X." To save spaee, scores lorPerson-Coordinated and Passive Joint—Coordinated are not given in the table; at most ages there were too lewepisodes to compute scores, and when there were, none was significant at the .20 level.

* ;) < .05.•• p < .01.••• p < .001.

- 2 . 1 *-3.3***

7.6***2.2*

X

2.9**3.0**

- 2 . 1 *-2.3*

3.7***3.6***

-4.4***-1.6

6.0***

5.1***2.5*

-1.3

-2 .3*

4.5***2.4*

XX

1.8

XX

3.83.4

X-2.2

4.42.7

Bakeman and Adamson 1285

TABLE 3

TRANSITIONS FROM PASSIVE AND COORDINATED JOINT ENGAGEMENT

ACE (in Months)

ANTECEDENT-CONSEQUENT STATE CONDITION 6 9 12 15 18

Passive joint—unengaged:Mother . . . . . . -1.8 XPeer -1.9 -1 .3 -2.3** X

Passive joint—onlooking:Mother -1.6 -3.5***Peer . . . . . . - 2.0*

Passive joint—person:Mother - 2.5* -1.9 . . . . . . XPeer X X X . . . X

Passive joint—object:Mother 7.5*** 9.4*** 12.3*** 9.9*** 4.6***Peer 1.9 3.5*** 3.8*** 3.2** 4.0***

Coordinated—unengaged:Mother X X . . . . . . XPeer X . . . -1.4

Coordinated—onlooking:Mother X X - 2 . 1 * -1.2 -1.7Peer X -1.5 -1.9 -2 .3*

Coordinated—object:Mother X 3.0** . . . 2.5* 4.6***Peer X 3.4*** 3.8*** 2.9** 4.4***

NOTE.—Scores are z scores based on episodes pooled across svibjects. In effect, these z scores compare agiven transition to its base-rate e.xpected value. Only z scores significant at the .20 level or better arc shown. Zscores not computed because too few episodes were coded are marked with "X." To save space, scores ior Person-Coordinated and Passive Joint—Coordinated are not given in the table; at most ages there were too iew episodesto compute seores, and when there were, none was significant at the .20 level.

* p < .05.** p < .01.*•* p < .001.

gagement to passive joint and object en-gagement to coordinated joint occurred atgreater than chance rates for most ages. Inaddition, a period of onlooking occurring justhefore a period of passive joint engagementwas also a likely sequence when 6—15-month-olds played with their mothers and when 9-and 12-month-olds played with their peers.Certain other sequences, however, did notcharacterize the flow of attention into jointstates. In particular, person engagement wasnot a systematic precursor of either passiveor coordinated joint engagement; when thissequence occurred sufficiently often to eval-uate, rates were close to those expected bychance. Finally, the sequences unengaged topassive and unengaged to coordinated jointappeared significantly less often than onewould expect by chance.

Object play was not only a likely ante-cedent to both passive and coordinated jointengagement, it was a likely consequent aswell. When with both mothers and peers,both the passive joint to object and the co-ordinated joint to object sequences charac-

terized the flow of attention at most ages. Noother engageinent states seemed to system-atically follow passive or coordinated joint(see Table 3).

The analyses just reported suggest whichsequences were more likely to occur thantheir base rates would suggest, but they donot tell us if the tendency to differ from baserate changed with age or differed signifi-cantly when infants were observed withmothers instead of peers. To answer thesequestions, repeated-measures analyses ofvariance were run (sex x partner x age, asbefore, on the two cohorts separately); ana-lyzed were individual z scores associated withthose transitions that the previous analysisidentified as likely sequential patterns.

Neither infant age nor partner system-atically affected the extent to which objectengagement tended to precede passive or co-ordinated joint engagement or the extent towhich unengaged tended not to precede pas-sive joint engagement. In other words, thesesequences tended to characterize all obser-

1286 Child Development

vational sessions. The only qualificationarises when we consider the two cohorts sep-arately. Then, both partner and age effectsare significant (p < ,05) for the sequence ob-ject engagement to coordinated joint en-gagement, but only in cohort 2, primarily be-cause of the 18-nionth-old infants there.

The partner did affect the likelihood ofthe sequence onlooking to passive joint; thissequence was significantly more likely whenwith mothers than with peers {p < .01 and,05 for cohorts 1 and 2). Moreover, the partneralso affected the organization of attention aftera period of passive (but not coordinated) jointengagement; the sequence passive joint toobject play was significantly more likely whenwith mothers {p < ,001 for both cohorts).

Finally, age did not seem to affect thestrength of any of the patterns found in theearlier sequential analysis, with the follow-ing qualification. There is some suggestionthat the sequence coordinated joint to objectplay differed more strongly from base rate atlater ages (linear trend p < ,001 and ,10 forcohorts 1 and 2),

DiscussionOur interpretation of these results is

guided by the general notion that new com-municative forms appear first embedded inand supported by a social context. This ideais central to the theories of Mead (1934) andVygotsky (1978) and has been espoused re-cently by a number of investigators con-cerned with early communication develop-ment, including Bruner (1982), Newson andNewson (1975), and Trevarthen (1977),Within this social interactionist perspective,the achievement of joint engagement hasparticular significance for two reasons. First,it is thought to provide the social setting forthe emergence of new forms of referentialcommunication (Adamson & Bakeman, 1982),Second, the capacity to coordinate attentionto objects and people is viewed as the cul-mination of a developmental sequence thatbegins with face-to-face interaction (Tronick,1982) and continues as infants turn their at-tention increasingly toward objects (Trevar-then & Hubley, 1978).

The fact that we found a developmentaldecrease in person play coupled with an in-crease in coordinating attention to a partnerand to the object the two share—and not anincrease in solitary object play—providessupport for this view of how early commu-nication development is sequenced. Coor-dinated joint engagement may well be the

developmental heir to face-to-face play, asTrevarthen suggests, but we would place thistransition later than he and others do. Bates(1979) and Sugarman-Bell (1978), for exam-ple, report that by the end ofthe first year oflife, infants can alternate glances back andforth between a person and a shared object.This may be so, but given our data we con-clude that the performance of coordinatedjoint engagement does not become routineuntil several months later. After all, the av-erage amount of time infants spent with theirmothers in coordinated joint engagement didnot exceed 10% until 15 months of age, andnot until 18 months of age were all infantsobserved in this state at least once. Indeed,we probably did not follow the subjects inour study long enough to observe the amountof coordinated joint activity during free playstabilize, (With both partners, the prevalenceof this state increased dramatically from 15to 18 months, doubling both in terms oftheproportion of time it occupied and the meanlength of individual episodes,)

This relatively late appearance of coor-dinated joint engagement during free play isinteresting for at least two reasons. First, suchnormative data are valuable for their ownsake, if for no other reason than that they caninform clinical expectations. Second, theymay have important implications for under-standing communication development. Forexample. Nelson (1979) has asked why, sinceinfants by the end ofthe first year of life seemequipped with cognitive and social skills re-quired for language, they do not really beginto talk until several months later. She sug-gests that one reason for this apparent de-velopmental plateau in communication de-velopment is that infants require this time tocoordinate social and object realms beforelanguage development proper can proceed.In our study, we found that there were infant-based constraints on the fiow of their atten-tion between object and social domains dur-ing the 6-18-month-old age range. Not onlywere certain routes into and out of joint en-gagement more likely to be traveled than oth-ers, but also infants did not become jointlyengaged unless they were first focusing onsome aspect ofthe environment. Even moreto the point, we found differences in atten-tional organization in the many months thatseparate the first performance of nonverbalreferential actions and entrance into the lin-guistic realm: the coordination of people andobjects, regardless of partner, was not mas-tered until near the end of infancy.

In addition to the normative develop-ment of coordinated joint engagement, we

Bakeman and Adamson 1287

are also interested in the role ofthe partner.Our suspicion is that mothers—or, in prin-ciple, any competent and willing adult orolder child—may act in ways that support orscaffold performances by their infants evenbefore the infants are capable of such per-formances unaided. And, indeed, we ob-served higher rates when with mothers in-stead of peers for those engagement statesthat offer some potential for referential com-munication—passive and coordinated jointengagement—and no partner effects for statesthat do not—person play, onlooking, and ob-ject play.

Part of the mothers' positive influenceon joint engagement states can be viewedsimply as a product of her greater skill infacilitating coordinated joint engagementonce her infant is able to coordinate atten-tion. Thus, although the absolute amount oftime spent in coordinated joint engagementand the average length of an episode weremarkedly greater in the mother condition,coordinated joint engagement did increasewith age for both peer and mother condi-tions. Moreover, sequential analyses of theongoing flow of attention did not suggest thatthe mother was providing qualitatively dif-ferent support that could draw the infant intoa state of coordinated joint engagement.

The mothers' most striking contributionis evident when the mother/peer differencesfor passive joint engagement are examined.This state occurred during every mother-in-fant observation, and it typically character-ized the infant's engagement for about 20%of the period, regardless of the infant's age.With a peer, however, about a third of the10-min-long observations did not containeven a single episode of passive joint en-gagement; moreover, episodes that did occurtended to be very brief and to account forvery little ofthe overall observation period.Mothers also seemed to work deliberately toinduce passive joint engagement. Onlook-ing, for example, often led to a period of pas-sive joint engagement with inothers, less sowith peers. This suggests that mothers weremore likely to use their capacity to capturetheir infants' attention to an interesting spec-tacle as a prelude to a period of mutual ex-ploration of an object.

The passive joint engagement state thusseems to be closely tied to nnothers' actions.It also appears to be an attention state that,unlike coordinated joint engagement, couldbe sustained by even the youngest infants inour study. These two characteristics suggestthat it is this form of joint engagement that

adults are most likely to foster as they scaf-fold their infants' activities during the periodof development when new communicativeskills are just forming (Bruner, 1975; Kaye,1982), In a sense, during passive joint en-gagement, adults may provide an implicit so-cial context for referential communication, asocial context that will only become explicitonce the infant can coordinate attention acrossthe social and object domains—and whichwill then result in increasing amounts of co-ordinated joint engagement, such as weobserved.

Why are adults and not peers better ableto support passive joint engagement? Partly,no doubt, because adults are simply morewilling and able to complement their infants'attention. In addition, they may provide pre-dictably repeating patterns of actions as theyinteract with infants—patterns that may havedevelopmental roots in the person-focusedgames of earliest infancy (Tronick et al,, 1979),Further, adults may be able to foster passivejoint engagement because they maintain ushared inemory system with the infant (Kaye,1982), which, like a repeating pattern, mayalso free the infant from a need to attendsimultaneously to both partner and object, Apeer, we expect, provides neither a predict-able pattern of action nor access to a sharedmemory system. Indeed, even a familiar peeris a remarkably unpredictable partner, proneto sudden movements of self and toys. In asense, then, infants can depend upon theirmother's actions, and so they can join withher to examine objects beyond their inter-personal relationship, Incontrast, rarely cansuch "trust" be extended to peers until theinfant is able to also "keep an eye" on thisrelatively unsupportive partner, which mayaccount for the fact that when with peers,passive joint engagement remained at uni-formly low levels, while coordinated jointengagement steadily increased with age.

In some respects, however, infants' ex-perience with mothers and peers was similar.With both, episodes of passive joint and ofcoordinated joint engagement were unlikelyto follow unengaged episodes and were likelyto be bracketed by periods of object engage-ment. Thus, while many more episodes ofpassive joint and coordinated joint engage-ment occurred with mothers, to be sure, whenthey did occur with peers, the moment-by-moment sequential pattems were similar,suggesting that at least some differences be-tween mothers and peers may be more quan-titative than qualitative.

1288 ChUd Development

In summary, communicating with othersabout objects demands attention to both so-cial and object aspects of one's surroundings.Integrating attention is a process that is mas-tered only slowly during the latter portion ofinfancy. Moreover, it is a process that maywell depend on the primacy of social rela-tionships, which builds on skills nurturedduring the earlier period of face-to-face playand which expands during the emergence ofa "triadic" (infant-object-other) interactivesystem. Skilled adults clearly tailor their ac-tions to suit their infant's changing attentioncapabilities. Thus they can "socialize" objectattention, embedding it within the interper-sonal sphere well before infants can struc-ture this integration by themselves. But, asattentional capabilities develop, even quiteunskilled peers may be appropriate partnersfor the exercise of these capacities. The in-fant interacting with a peer need not reca-pitulate a developmental course first traveledwith an adult guide; rather, albeit less fre-quently and perhaps more clumsily, by themiddle of the second year infants can alsojoin with peers in a shared exploration oftheenvironment.

Reference Note1, Seibert, J, The development of pre-linguistic

communication skills: A neo-Piagetian anal-ysis based on levels of cognitive organization.Paper presented at the tenth annual Interdis-ciplinary Conference on Piagetian Theory andthe Helping Professions, Los Angeles, 1980,

ReferencesAdamson, L,, & Bakeman, R, Affectivity and ref-

erence: Concepts, methods, and techniques inthe study of communication development of6 to 18 month old infants. In T, Field & A,Fogel (Eds,), Emotion and early interactions.Hillsdale, N.J.: Erlbaum, 1982,

Allison, P,, & Liker, J, Analyzing sequential cate-gorical data on dyadic interaction: A commenton Gottman, Psychological Bulletin, 1982, 9 1 ,393-403,

Bates, E, The emergence of symbols: Cognitionand communication in infancy. New York: Ac-ademic Press, 1979,

Brazelton, T, B,, Koslowski, B,, & Main, M, Theorigins of reciprocity. In M, Lewis & R, L,Rosenblum (Eds,), The effect ofthe infant onits caregiver. New York: Wiley, 1974,

Bruner, J, The ontogenesis of speech acts. Journalof Child Language, 1975, 2, 1-19,

Bruner, J. The organization of action and the na-ture ofthe adult-infant transaction. In E, Tron-ick (Ed,), Social interchange in infancy: Af-

fect, cognition, and communication.Baltimore: University Park Press, 1982,

Cohen, J, A coefTicient of agreement for nominalscales. Educational and Psychological Mea-surement, 1960, 20, 37-46,

Eckerman, C , Whatley, J,, & Kutz, S, Crowth ofsocial play with peers during the second yearof life. Developmental Psychology, 1975, 11 ,42-49,

Cottman, J, M,, & Bakeman, R, The sequentialanalysis of observational data. In M, E, Lamb,S, J, Suomi, & C, R, Stephenson (Eds,), Socialinteraction analysis: Methodological issues.Madison: University of Wisconsin Press, 1979, .

Harding, C , & Colinkoff, R, The origins of inten-tional vocalizations in prelinguistic infants.Child Development, 1979, 50, 33^0 ,

Hollenbeck, A, R, Problems of reliability in ob-servational research. In C. P. Sackett (Ed,),Observing behavior (Vol, 2), Baltimore: Uni-versity Park Press, 1978,

Kaye, K, The mental and social life of babies. Chi-cago: University of Chicago Press, 1982,

Kaye, K,, & Fogel, A, The temporal structure offace-to-face communication between mother,sand infants. Developmental Psychology, 1980,16, 454-464,

Leung, E., & Rheingold, H, Development of point-ing as a social gesture. Developmental Psy-chology, 1981, 17, 215-220,

Mead, G. H, Mind, self, and society. Chicago: Uni-versity of Ghicago Press, 1934,

Murphy, C , & Messer, D, Mothers, infants, andpointing: A study of gesture. In H, R, SchafTer(Ed,), Studies in mother-infant interaction.London: Academic Press, 1977,

Nelson, K, The role of language in infant devel-opment. In M, Bornstein & W, Kessen (Eds,),Psychological development from infancy: Im-age to intention. Hillsdale, N,J,: Erlbaum,1979,

Newson, J,, & Newson, E, Intersubjectivity andthe transmission of culture. Bulletin of theBritish Psychological Society, 1975, 28, 437-445,

Newtson, D, Attribution and the unit of ongoingbehavior. Journal of Personality and SocialPsychology, 1973, 28, 28-38.

Ninio, A,, & Bruner, J, The achievement and an-tecedents of labeling. Journal of Child Lan-guage, 1978, 5, 1-15,

Parten, M, B, Social participation among preschoolchildren,yourna/ of Abnormal and Social Psy-chology, 1932, 27, 243-269,

Ratner, N,, & Bruner, J, Games, social exchange,and the acquisition of language. Journal ofChild Language, 1978, 5, 391^01,

Sackett, G, The lag sequential analysis of contin-gency and cyclicity in behavioral interactionresearch. In J, Osofsky (Ed,), Handbook of in-fant development. New York: Wiley, 1979,

Bakeman and Adamson 1289

Stem, D. The goal and structure of mother-infant play. Journal of The American Acad-emy of' Child Psychiatry, 1974, 13, 402-421,

Sugarman-Bell, S, Some organizational aspects ofpreverbal communication. In I, Markova (Ed,),The social context of language. Chichester:Wiley, 1978,

Trevarthen, G, Descriptive analyses of infant com-municative behavior. In H, R, SchafTer (Ed,),Studies in mother-infant interaction. Lon-don: Academic Press, 1977,

Trevarthen, C , & Hubley, P, Secondary intersub-jectivity: Confidence, confiding and acts ofmeaning in the first year. In A, Lock (Ed,),

Action, gesture, and symbol. London: Aca-demic Press, 1978,

Tronick, E, Z, Aflectivity and sharing. In E, Tron-ick (Ed,), Social interchange in infancy: Af-fect, cognition, and cotJiniunication. Balti-more: University Park Press, 1982,

Tronick, E,, Als, H,, & Adamson, L, Structure ofearly face to face commimicative interactions.In M, Bullowa (Ed,), Before speech: The be-ginning of interpersonal communication.Cambridge: Gambridge University Press, 1979,

Vygotsky, L. Mind in society. Cambridge, Mass,:Harvard University Press, 1978,

Werner, H,, & Kaplan, B, Symbol formation. NewYork: Wiley, 1963,