development of social roles in elicited and spontaneous …ddl/articlescopy/watson... · 2015. 1....

Developmental Psychology1980, Vol. 16, No. 5, 483-'494

Development of Social Roles in Elicited andSpontaneous Behavior During the Preschool Years

Malcolm W. WatsonBrandeis University

Kurt W. FischerUniversity of Denver

In two experiments, children between ll/2 and 7l/2 years of age were tested ona predicted, eight-step sequence of the development of social roles. Performanceon this sequence was related to two measures of more spontaneous behavior.Nearly all children fit the predicted sequence perfectly. By 2 years of age, mostchildren could make a doll act as an independent agent. The majority of 3-year-olds could make a doll carry out several behaviors fitting the role of doctor. Atage 4 or 5, most children developed the capacity to show a social role, makinga doctor doll interact with a patient doll. The intersection of social roles for twoagents appeared at about age 6: A man doll could be both doctor and father toa patient who was also his daughter. In their spontaneous behavior, early pre-schoolers almost always showed the highest step that they were capable of, butbeginning with the step for social roles, late preschoolers seldom showed theirhighest step.

For children to act and feel socially com-petent, they need to understand roles suchas mother, father, doctor, and patient.Despite the importance of these social rolesin the child's life, the development of socialroles in the preschool years is, at best, onlyvaguely understood. The several previousstudies of social-role development haveconcluded that children do not really under-stand social roles until 7 or 8 years of age(Signer, 1974; Chambers & Tavuchis, 1976;Elkind, 1962; Emmerich, 1959,1961; Green-field & Childs, 1977; Sigel, Saltz, & Ros-

Experiment 2 was part of the first author's doctoraldissertation at the University of Denver. The researchwas supported by a predoctoral research fellowshipto the first author from the National Institute of MentalHealth, 1F31 MH05321-02; by a student researchgrant to the first author from the Grant Foundation;and by a research grant to the second author fromthe Spencer Foundation.

The authors wish to thank Daniel Bullock, JosephCampos, Paul Corbitt, Susan Harter, Sandra Pipp,and Phillip Shaver for invaluable assistance and sug-gestions in completing these studies. Special thanksto Ralph J. Roberts, Jr., for collecting the data inExperiment 1 and helping with many other partsof the research.

Requests for reprints should be sent to Malcolm W.Watson, Department of Psychology, Brandeis Univer-sity, Waltham, Massachusetts 02154.

kind, 1967). Thus they have uniformly sup-ported Piaget's (1928) early conclusion thatthe understanding of social roles is a rela-tively late development.

There is reason to believe, however, thatthis research underestimates the preschoolchild's understanding of social roles. Severalinvestigators have described behaviors orcapacities in preschool children that implysome understanding of social roles (El'Konin,1971; Piaget, 1951). Recent research on otherskills has shown that many abilities thatwere thought to first emerge at age 7 or 8 arein fact present in some form several yearsearlier (Gelman, 1978). And at least onetheory of cognitive development, called skilltheory (Fischer, in press), predicts that evenin the early preschool years children shouldbe able to understand social roles in con-crete situations. To demonstrate such skillsin young preschool children, it is necessaryto use a task that is simple and concreteand does not overload the preschool childwith complicated instructions or a need toexplain the roles verbally.

In addition, the sequence of role-conceptdevelopment needs to be determined. Withspecification of a sequence, many con-troversies about age of acquisition can be

Copyright 1980 by the American Psychological Association, Inc. 0012-1649/80/1605-0483$00.75

483

484 MALCOLM W. WATSON AND KURT W. FISCHER

resolved (Rubin & Pepler, in press). Skilltheory provides mechanisms for predictingsuch a sequence.

According to this theory, children's skillsdevelop through a set of hierarchically or-ganized levels, each of which first emergesat a certain average age within a givenpopulation. Each level is described in termsof a skill structure, an organization of ac-tions or representations that a child cancontrol. In addition, several transformationrules specify how skills at a given level canbe made more complex. In the presentstudies, the rule used most often is com-pounding, in which an additional componentis added to a skill at a given level; for example,a skill comprised of two actions is expandedto subsume three actions.

In the present research, we used skilltheory to predict a sequence for the develop-ment of social roles in pretend play for theages from approximately 1 to 7 years (seeTable 1). The sequence contains eight steps,but there is nothing special about this num-ber. The number and nature of the steps thatchildren show will vary with the contexts towhich the children are exposed.

Before determining the actual steps thatwe would assess, we observed severaldozen preschool children in pretend-playsituations. The specific steps that we testedinvolved pretend stories similar to ones thatappeared in those situations.

According to skill theory, children be-tween 1 and 7 years of age develop throughfour levels. At about 1 year of age, the levelof sensory-motor systems emerges, inwhich children can coordinate severalaspects of two or more actions into a singleskill. In the first step in the sequence to betested, self as agent, children coordinateactions such as looking, holding, and touch-ing with the lips so that they can, for example,pretend to drink from a cup.

The next two steps involve the develop-ment of skills for making agents, such asdolls, act independently of the child. Both ofthese steps require the second of the fourcognitive levels—single representations—that typically emerges at about age 2. In asimple type of representation, childrencoordinate two sensory-motor systems intoa single skill. Thus, at Step 2, active other

agent, a system for manipulating a doll iscoordinated with a system for some otherbehavior, such as drinking or walking, withthe result that the child can make the dolldrink or walk (Fischer & Corrigan, in press;Watson & Fischer, 1977). The third step,active substitute agent, is at the same cogni-tive level and is predicted by the transforma-tion rule of substitution, by which a block canbe substituted for the doll in Step 2.

Such skills for making dolls and blocks actas independent agents would seem to beprecursors of skills for social roles, sincea role involves multiple behaviors of anindependent agent. The first two steps arealso suggested by research on children'sspontaneous pretend play (Dasen, Inhelder,Lavalle'e, & Retschitzki, 1978; Watson& Fischer, 1977).

Step 4, behavioral role, is predicted by thetransformation rule of compounding. Thesimple representation in Steps 2 and 3 can beexpanded to include additional behaviors,so that the child can make the doll carryout two or more behaviors coordinatedaround a specific role, such as mother ordoctor.

The fifth step represents the emergenceof the first skills that fit the standard defini-tion of a role. In social psychology andsociology, the cluster of behaviors defining asocial role is determined by the relationshipof that role to another complementary role(Mead, 1934). For example, the role ofmedical doctor is related to the comple-mentary role of patient, with the person ineach role expecting certain behaviors of theperson in the other role. Step 4 involves arole by itself. A role related to its comple-ment first develops at Step 5, social role.This relation of a role to its complement re-quires the next cognitive level, representa-tional mappings, in which two or morerepresentations are related in a single skill.This level first emerges at about age 4, ac-cording to previous research in other domains(Fischer, in press).

In the several years after children firstdemonstrate specific social roles, they willdevelop the ability to perform the kinds ofrole tasks investigated in most previousstudies (e.g., Sigel et al., 1967). By com-pounding more than two roles into a com-

DEVELOPMENT OF SOCIAL ROLES 485

plex. representational mapping, they canrelate a social role to several complements,as in Step 6, social role with three agents.

At about 6 years of age, they will become

capable of the next cognitive level, repre-sentational systems, in which several aspectsof two or more representations are co-ordinated. They will therefore be able to

Table 1Predicted Developmental Sequence of Social Role Playing

Cognitive level/step Role-playing skill Modeled behaviors

Sensory-motor systems1

Single representations2

Representational mappings5

Representational systems7

Self as agent: A person pretends tocarry out one or more behaviors,not necessarily fitting a social role.

Active other agent: An agent performsone or more behaviors, notnecessarily fitting a social role.

Active substitute agent: An objectsubstituting for an agent performsone or more behaviors, notnecessarily fitting a social role.

Behavioral role: An agent performsseveral behaviors fitting a socialrole (in this case, doctor).

Social role: One agent behavingaccording to one social role (doctor)relates to a second agent behavingaccording to a complementarysocial role (patient).

Social role with three agents: Twocomplementary roles (patient andnurse), instead of just one, aresimultaneously related to the firstrole (doctor).

Intersection of social roles for twoagents: Two separate, agent-complement role relations arecoordinated so that one agent is intwo roles simultaneously and relatesto another agent in two comple-mentary roles (doctor-father withpatient-daughter).

Intersection of social roles for threeagents: Three separate agent-complement role relations arecoordinated so that one agent in twoor three roles simultaneouslyrelates to two other agents each inthe relevant complementary roles(doctor-father-husband withpatient-daughter and mother ofpatient-wife).

Experimenter pretends to drink froman empty cup.

Experimenter pretends that a doll istalking, walking, eating, andwashing, as if it were actuallycarrying out the actions itself.

Experimenter pretends that a block iswalking, eating, talking, and goingto sleep, as if it were a person ora doll.

Experimenter pretends that a doctor-doll uses a thermometer and anotolargyngoscope on another doll.

Experimenter pretends that a patientdoll is sick and a doctor dollexamines her and gives her somemedicine. The patient doll makesappropriate responses during theexamination.

Experimenter pretends that a doctordoll examines a sick patient doll andis aided by a nurse doll. Bothpatient and nurse respondappropriately.

Experimenter pretends that a doctordoll examines a sick patient dolland also acts as a father to thepatient, who is his daughter. Thepatient doll responds appropriatelyas both patient and daughter.

Experimenter pretends that one dollis a doctor, father, and husbandrelating to two other dolls. Thesecond doll is a sick patient andthe first doll's daughter. The thirddoll is the patient's mother and thefirst doll's wife.

Note. The stories and exact scoring criteria are available from the authors on request.


build skills for relating several intersectingsocial roles for the same agent, as in Steps7 and 8. Such role intersection seems to bea genuine concern of many preschool chil-dren, as illustrated by a question one 6-year-old asked of his father: "Daddy, when youfinish school and become a psychologist,will you still be my daddy?" Step 8 occursafter Step 7 because it requires a com-pounded representational system.

To assess this predicted sequence, wechose to use a modeling-and-imitation tech-nique employed in some previous research.Other studies have concluded that in pretend-ing, children do not blindly mimic a modelbut demonstrate only those aspects thatthey can understand or assimilate (Gottlieb,1973; Harnick, 1978; McCall, Parke &Kavanaugh, 1977; Scollon, 1976; Slobin &Welsh, 1973). In fact, the modeled actionsseem to act as general facilitators of pre-tending (Watson & Fischer, 1977) as long asthe modeling and pretending occur in acontext in which the behaviors are ap-propriate for the child (see Bloom, Rocis-sano, & Hood, 1976). Also, a similar model-ing procedure is commonly used in playtherapy with preschool children: The ther-apist acts out or explains a story or theme ofsome kind, and the child frequently picksup the story or theme and incorporates itinto his or her play (e.g., Harter, 1977).

We introduced several modifications ofthe typical modeling-and-imitation procedure,however. First, each developmental step wasindependently assessed by a separate storythat was devised expressly to test that step:The children watched the experimentermodel each story and then tried to do it them-selves. With this strong form of scalogramanalysis, we could directly test the develop-mental sequence with a cross-sectionaldesign.

Second, we tried to directly assess therelation between behavior under this im-mediate-elicited-imitation procedure andbehavior under a more spontaneous condi-tion like that used in Watson and Fischer(1977). Most studies of development in playassess only the child's spontaneous pref-erences and must assume that they are ob-serving the child's highest capabilities. Wewere able to compare capabilities under

two conditions. After the elicited-imitationassessment, each child was left to play alonewith the toys. The child's behavior underthis free-play condition was undoubtedlyaffected by the prior occurrence of theelicited condition, but the free-play condi-tion did give some indication of what thechildren preferred to do when their behaviorwas not constrained to immediate imitation.In two previous studies using only a pro-cedure similar to the free-play condition(Harnick, 1978; Watson & Fischer, 1977),the pattern of steps that 1- and 2-year-oldsspontaneously performed showed not onlythe acquisition of the steps in a develop-mental sequence but also the deletion oromission of steps in the same order as thesteps had been acquired (i.e., first in, firstout). A reasonable extrapolation from thosestudies is that in the elicited-assessmentcondition children will correctly perform allsteps up to the highest that they are capableof, but in the free-play condition they willshow a systematic combination of acquisi-tion and omission similar to that in theprevious study.

The hypotheses tested, then, were: (a) Inthe elicited-assessment condition, pre-school children will demonstrate the de-velopmental sequence for social roles de-scribed in Table 1. (b) The first steps willbe achieved by early preschool children—self as agent by about \Vi years of age, activeother agent by about 2 years. The firstdemonstration of a social role (Step 5) willemerge at about 4 years of age, and thefirst intersection of roles (Step 7) will appearat about age 6. (c) In the free-play condi-tion, children will show the same develop-mental sequence as in the elicited-assess-ment condition, but they will not show allthe earlier steps that they are capable of.The omission of earlier steps may follow thesame sequence as the acquisition of steps.

Two experiments were completed to testthese hypotheses.

Experiment 1

Method

SubjectsSubjects were 40 normal, white, middle-class chil-

dren whose parents had volunteered to have them


participate in the study. Four age groups were tested,with respective mean ages of 1 year 6 months (allthe same age), 2 years (range = 2 years to 2 years1 month), 3 years (all were 3), and 4 years 3 months(range = 3 years 11 months to 4 years 3 months).There were 10 children in each group, 5 of each sex.All children brought to the laboratory completed thestudy except 1 2-year-old and 1 4-year-old. These 2children would not cooperate in performing the elicited-assessment tasks.

Facilities and ApparatusTesting took place in a carpeted playroom, furnished

with curtains, pictures on the walls, a chair for themother, and a child's chair and round table. A micro-phone hung from the center of the ceiling.

Toys consisted of a set of two Milton-Bradleyrigid cardboard dolls in plastic stands that representedrealistic stereotypes of a male medical doctor and afemale child (patient). The doctor was 20 cm high andthe patient 12 cm high. Also used were a set of woodendoll furniture and a toy doctor's kit consisting of acontainer, syringe, thermometer, medicine bottle,otolaryngoscope, and sponge.

Each child's behavior and vocalizations throughoutthe session were recorded on videotape from anadjoining observation room. The camera was situatedbehind a filming window that was covered with blackcloth. The camera was virtually invisible to the child.

ProcedureAll children were tested in two conditions, elicited

assessment and free play. The two conditions werealways administered in the same order because theoccurrence of a high frequency of pretending isfacilitated by adult modeling (Watson & Fischer, 1977).Also, we wanted to maximize the chances that thechildren would show the same steps in the twoconditions.

After a mother and child entered the playroom andsat down, the experimenter explained to the childthat he wanted to see how children played and actedout stories and that together they would play a game,but first he had to get something ready in anotherroom, and that the child could go ahead and play withthe toys while the experimenter was gone.

Familiarization. The experimenter left the room,and the child was allowed to play for 3-4 min to be-come familar with the setting and objects.

Elicited assessment. The experimenter reenteredthe room and told the child that they would use thedolls to act out some stories and that he wanted tosee how good an actor the child was. The experi-menter asked the child to watch how he acted outeach story and then take a turn acting out his orher own story, using the same dolls and ideas that theexperimenter used. The experimenter told the childthat if he or she did a good job, he would providea surprise toy afterward; however, he also told thechild that the story need not be exactly like the experi-menter's. The presentation of the modeled stories,each followed by an imitative story from the child,

lasted approximately 15 minutes. A warm-up taskwas presented before the experimental tasks: the motherdoll was made to eat and sleep as an active agent.

Then, to test each of the steps, similar but distinctstories were modeled, one representing each step inthe sequence. In this experiment, Steps 1-5 weretested (see Table 1). In the story for the first step,the experimenter himself pretended to drink from acup. In the other stories, he made one or two dollsact and talk. If the child did nothing after the experi-menter demonstrated a story, then the experimentershowed it again. Each story took from 30 to 50 sec tomodel, and the ordering of the stories by length wasnot the same as the predicted developmental sequenceor the order of testing. The duration of the children'sstories after each modeled story ranged from 20 to180 sec.

The stories were presented in a scrambled orderso that neither practice nor fatigue effects couldartifactually produce an apparent scalable sequence;the order of the tasks, designated by step number,was 4, 2, 1, 5, 3.

Free-play condition. After the modeling and elicitedimitation were completed, the experimenter saidthat he had to leave for a short time but that thechild could continue to play. The experimenter sug-gested that the child act out some more stories on hisor her own. The child was then left to play alone for6 minutes. This behavior was later scored in terms ofthe steps in the developmental sequence. Thus itprovided a measure of what might be called "imi-tative pretend play"—pretend behavior that reflectsthe stories modeled by the experimenter but is notrigidly constrained by them. This procedure is similarto that used in the previous study by Watson andFischer (1977). The experimenter then returned to theplayroom and debriefed the mother and child.

Scoring and Interobserver ReliabilityThe experimenter scored each child's performance

from the videotapes by using established criteria to rateeach of the child's stories for the role-playing stepdemonstrated. In the elicited assessment, exact imita-tion of the experimenter's story was not requiredand was seldom shown. To pass the task for a givenstep, the child had only to meet minimum criteriaof agent use and role relations for showing that par-ticular step, as outlined in the descriptions of theskills for each step in Table 1; thus the child couldpass a step by acting out a story in which the actionswere not at all like those of the experimenter so longas they fit the general criteria for the step.

A second observer, blind to the hypothesized se-quence, the exact age of the children, and the otherobserver's ratings, scored the performances of onerandomly selected child of each sex from each agegroup. For these eight children, the interrater reli-ability for highest step passed in the elicited-assess-ment condition was perfect (r = 1.00). Across all stepsin the elicited assessment, the mean percentage ofperfect agreement was 97.5.

In the free-play condition, all children spent much oftheir time playing with the dolls, acting out a medianand mode of seven stories per child. Each episode


that they acted out was scored for the step that itshowed. Scoring criteria were the same as those for theelicited-assessment condition. The reliability of thetwo highest step passed was nearly perfect (r = .95);for all steps shown, the two observers agreed com-pletely 95% of the time.

Results

The hypothesized sequence was stronglysupported, but there were important dif-ferences between the two conditions.

Sequence

To test whether the obtained task pro-files for individual children indicated ascalable developmental sequence, we usedscalogram analysis, in particular Green's(1956) tests for reproducibility and con-sistency.

The sequence was tested separately ineach condition. For the elicited-assessmentcondition, the sequence was strongly sup-ported. Green's (1956) index of consistency(/) was .87, indicating a highly scalablesequence. As shown in Table 2, 38 of 40children fit the predicted profiles perfectly,and the other 2 each skipped only one step.

One of the five steps did not scale strongly,however. Step 3, active substitute agent,proved to show almost the same profilesas Step 2, active other agent. Only twochildren passed Step 2 without passingStep 3; one child passed Step 3 withoutpassing Step 2.

For the free-play condition, however, thesequence did not prove to be scalable (/ =

.18). Even when Step 3 was combined withStep 2, the sequence was still not scalable(/ = .29). Also, when children omittedsteps, their omissions did not follow theorder of acquisition.

Highest Step

Although the sequence did not scale in thefree-play condition, there was evidencethat the children's behavior in free playdid reflect their ability as assessed in theelicited-assessment condition. In the free-play condition, most children (35 of 40)showed the highest step that they werecapable of according to the elicited assess-ment, and none showed a higher step. WithSteps 2 and 3 combined, the correlationbetween highest step in the two conditionswas nearly perfect, r(38) = .94, p < .001.Likewise, the correlation of each measurewith age was high: for elicited assessment,r(38) = .86, p < .001, and for free play,r(38) = .85, p < .001. When age was par-tialed out, the correlation between the twoconditions was still substantial, r(23) =.71, p < .001.

In an analysis of variance with age andsex as between-subjects factors and condi-tion (assessment and free play) as a within-subjects factor, there was a significant ageeffect, F(3, 32) = 50.7, p < .001. On a 4-point scale, the mean highest steps were1.00, 1.95, 2.85, and 3.40 in order of increas-ing age. By the Tukey posthoc test, thefirst three age groups were all significantlydifferent from each other, but the third

Table 2Distribution of Children Showing Task Profiles in Experiment I

Developmentalstep of profiles

012345Nonscalable profiles

Total

Role-playing task

1 2 3 4 5 n

1+ — — — — 7+ + — — — 2+ + + — — 12+ + + + — 10+ + + + + 6+ — + — — 1+ + + — + 1

40

M age

1 year 6 months1 year 7 months1 year 10 months2 years 4 months3 years 5 months3 years 1 1 months2 years4 years 3 months

Note. Occurrence of the designated type of role playing is indicated by +.


and fourth groups did not differ significantly.There were no other main effects or inter-actions.

Despite the absence of an Age x Condi-tion interaction, more fine-grained analysisdid suggest an age change. Of the five chil-dren who did not play at their highest stepin free play, four had Step 5 as their higheststep in the elicited assessment; three otherchildren with Step 5 as their highest step didshow it in their play.

Experiment 2

The second experiment assessed the restof the predicted sequence and retested sev-eral of the steps assessed in Experiment 1. Inaddition to assessing the sequence, it pro-vided a test of the possible Age x Condi-tion interaction in Experiment 1: If thefindings for Step 5 in the free-play condi-tion indicated a reliable change that was afunction of age or cognitive level, then asimilar pattern of results should occur notonly for Step 5 but for all later steps aswell. Also, to test the generality of thefindings, we introduced another condition:After the free-play session, children wereasked to act out the best story they could.

Method

SubjectsSubjects were 68 normal, white, middle-class chil-

dren whose parents had volunteered to have themparticipate in the study. There were five age groups,with respective mean ages of 1 year 6 months (all thesame age), 3 years (range = 2 years 11 months to 3years 3 months), 4 years 6 months (range = 4 years 5months to 4 years 11 months), 6 years (range =5 years 10 months to 6 years 3 months), and 7 years 6months (range = 7 years 6 months to 7 years 7 months).In each of the middle three age groups, there were 20children (10 of each sex) for a total of 60 children.Only 4 children (2 of each sex) were tested in eachof the groups aged 1 year 6 months and 7 years 6months because these children were added to obtaindata for the extreme steps in the scalogram analysisonly.

All children brought to the laboratory completed thestudy except three 3-year-old boys and two 3-year-oldgirls. These five children were shy and did not carryout the elicited-assessment stories.

ProcedureThe procedure was the same as in Experiment 1

in most respects. The experimenter was a different

adult male experienced with preschool children (thefirst author). In addition to the doctor and childdolls, two additional dolls of the same brand andstyle were added for the later steps—a woman nurseand a woman who served as the mother of the patient.Steps 2, 4, 5, 6, 7, and 8 were tested.

To assess for order defects, we presented the elicited-assessment tasks in two different orders, from Step 2to Step 8 for half of the children and in reverse order,Step 8 to Step 2, for the other half. In each Age xSex group, half the children were randomly chosenfor each order.

After the free-play period, the experimenter re-entered the playroom, greeted the child, and carriedout the best story condition: He asked the child toact out the best story that he or she could but pro-vided no additional cues or modeling. The story waslater scored for the step in the developmental se-quence that it demonstrated, in the same manner thatthe assessment and free-play stories were scored.

Scoring and Interobserver ReliabilityTwo undergraduate psychology students inde-

pendently scored each child's performance from thevideotapes in the same manner as in Experiment 1.Both observers were unaware of the exact ages of thesubjects and the rationale for the sequence. Thesecond observer scored 40 children, randomly chosenbut including every Age x Sex group. Scoring wasagain highly reliable. The interrater correlation forhighest step passed in the elicited assessment washigh (r = .97). For all steps demonstrated, the twoobservers agreed completely 98% of the time.

The second observer also scored the occurrence ofthe six steps in the free-play and best story condi-tions, but she did not separate the scores for the twoconditions. Reliability can therefore be assessed onlyfor the combination of these two conditions. Forhighest step reached, the interrater correlation washigh (r = .93). For all steps demonstrated, the ob-servers showed complete agreement 98% of the time.

Results

The results strongly supported the hy-pothesized sequence, and there was a sharpchange in the relation between the stepsshown in the several conditions, beginningat Step 5.

Sequence

In the elicited-assessment condition, 65 of68 children fit the sequence perfectly, asshown in Table 3; the other three all skippedStep 6, but for most children Step 6 did fitthe sequence. The index of consistency was.89, indicating a highly scalable sequence.

For the free-play condition, the sequenceonce again did not prove to be scalable


Table 3Distribution of Children Showing Task Profiles in Experiment 2

Developmentalstep of profiles

0245678Nonscalable profiles

Total

Role-playing task

2 4 5 6 7 8 «

2+ — — — — — 8+ + — — — — 23+ + + — — _ 11_l_ _i_ _l_ -l_ 9+ + + + + — 3+ + + + + + 9+ + + — + — 2+ + + — + + 1

68

M age

1 year 6 months2 years 7 months4 years4 years 7 months5 years 3 months6 years6 years 3 months6 years7 years 6 months

Note, Occurrence of the designated type of role playing is indicated by +.

(/ = .38). Of the 68 children, only 53 fitthe predicted Guttman-scale profiles. Theomission of steps did not consistently occurin the order of their acquisition, and thusthere was no evidence for a sequence com-bining acquisition and omission.

Highest Step

When the dependent variable was higheststep in a given condition, an analysis ofvariance showed the expected Age x Con-dition interaction. Age, sex, and order oftasks were between-subjects variables; con-dition (assessment, free play, best story)was a within-subjects variable; and the

Assessment

Free PlayBest Story

3.0 4.5 6.0

AGE IN YEARS7.5

Figure 1. Mean highest step as a function of conditionand age in Experiment 2.

steps were treated as a 6-point scale. Onlythe middle three age groups were used inanalysis of variance because the youngestand oldest groups included only four chil-dren each. As shown in Figure 1, there wasan age effect, F(2, 48) = 14.27, p < .001,a condition effect, F(2, 96) = 16.75, p <.001, and an Age x Condition interaction,F(4, 96) = 3.20, p < .05, but no othersignificant effects. With age, the differencebetween the assessment condition and theother two conditions increased. The resultsfor the ll/2- and 71/4-year-olds fit the samepattern (see Figure 1).

The Newman-Kuels multiple range testindicated that there were no significantdifferences between the conditions at age 3years and that the elicited-assessment condi-tion was generally higher than the other two.The following paired comparisons betweenconditions were significant: For the 4Vi-year-olds, the mean highest step in theassessment condition was higher than thatin the best story condition (p < .01). For the6-year-olds, the mean highest step in theassessment condition was higher than that inboth the free-play and the best story condi-tions (p < .001). No other condition dif-ferences within age groups were significant.Across age groups within each condition,all paired comparisons were significant atthe .05 level or better except for that be-tween 3- and 41/i-year-olds in the best storycondition and that between 4Vi- and 6-year-olds in the free-play condition.


The Age x Condition interaction can alsobe analyzed in terms of developmental steprather than age. If the interaction is a resultof a change in developmental level, thenaccording to skill theory the differencesbetween conditions should first becomeevident at Step 5, the first step at the levelof representational mappings. Figure 2 showsthat the difference between the assessmentand free-play conditions first emerges at Step5 in both experiments: The proportion ofchildren playing at their highest assessedstep fell precipitously at Step 5 and continuedto fall further at higher steps.

The correlations between age and higheststep for each of the three conditions wereall high (allps < .001): Age correlated .76,.63, and .52 with assessment, free play, andbest story, respectively; assessment .75 and.71 with free play and best story, respec-tively; and free play correlated .84 with beststory. Even with age partialed out, the cor-relations among the conditions remainedhigh (allps < .001): Assessment correlated.56 with both free play and best story, andfree play correlated .78 with best story.

Discussion

The results seem to have implicationsfor two separate but related issues, thesequence of social-role development and thereflections of that sequence in spontaneousbehavior.

Development of Social Roles

The present results confirm the hypothesisthat in the early preschool years childrencan understand that people are independentagents and can begin to understand socialroles. By Wi years of age, most middle-class children can pretend that they arecarrying out some action; in this way, theyshow an understanding that they themselvesare agents. By age 2, they can make a doll dosomething, as if it were acting on its own,thus demonstrating an elementary skill forrepresenting other people as independentagents. Most 3-year-olds can make a dollcarry out several activities relating to a roleand so show a behavioral role. Most 4-year-

I 2(3) 4 5 6 7 8HIGHEST STEP IN ASSESSMENT

Figure 2. Relation between highest step in assessmentand the proportion of children showing that step in play.

olds can act out a social role, relating onebehavioral role such as doctor to a com-plementary role such as patient. Develop-ment in the late preschool years involvesmore complicated combinations of roles,culminating in the ability of most 6-year-olds to have one agent carry out severalroles simultaneously.

In the elicited assessment, the childrenclearly demonstrated the predicted eight-step developmental sequence, described inTable 1. Of 108 children tested, 103 fit thesequence perfectly, and the other 5 skippedonly one step. These results provide espe-cially strong support for the sequencebecause, contrary to previous studies,we assessed each step with an indepen-dent task.

Previous research generally supports thebroad outline of the sequence of social-roledevelopment and thus lends some generalityto it. Similar sequences occur across manyroles and tasks (e.g., Chambers & Tavuchis,1976; Emmerich, 1959, 1961; Greenfield &Childs, 1977; Piaget, 1928; Sigel et al.,1967). However, some of the specific stepsas operationally defined in the sequenceand the ages at which steps occur will varyas a function of many task and contentdifferences.

This developmental sequence is consistentwith recent research showing that prespec-


live-taking abilities begin to emerge longbefore 6 or 7 years of age. Indeed, the social-role sequence can be viewed as a measureof a type of perspective taking. Rubin andPepler (in press) have suggested that oneway of making sense of the many apparentlycontradictory studies on perspective takingis to attempt to describe the sequence inwhich perspective taking develops from anearly age. The social-role sequence seemsto provide one starting point for such adescription. Early on, children take the per-spective of a single agent carrying outindividual actions, then they take the per-spective of an agent carrying out a behavioralrole, then the perspective of an agent fillinga social role, and finally the perspectiveof a more individualized agent filling severalsocial roles at once.

Whether the sequence is conceptualizedin terms of perspective taking or only interms of agents and social roles, it describeshow the preschool child gradually constructsskills for one type of social interaction.Step by step, children build more complexskills until age 6 or 7 they have the abilityto deal with some of the complexities ofinteractions involving social roles.

These increases in complexity need to beanalyzed in terms of the child's skills, notjust in terms of simple changes in the stimuluscomplexity of the stories or simple, linear,cumulative increases in ability. For example,the number of dolls increased generally acrossthe stories, but this increase cannot explainthe sequence, since for many adjacent stepsthere was no increase, and in one casethere was actually a reversal: The number ofdolls used for the steps was 0, 1, 1,2, 2, 3,2,and 3, respectively. Similarly, the length ofthe stories did not progress consistentlyfrom shortest to longest; the length of timerequired to model a story varied from only30 to 50 sec, and the length did not increasesystematically from step to step. The verbalabilities required to act out a story likewiseincreased across the sequence, but thestructures of the behaviors and verbaliza-tions, not simply the amount, determinedwhich steps a child passed. Thus, althoughall these factors undoubtedly influencedoverall performance, they could not account

in any simple way for the invariant se-quentiality of the steps.

Relation of Elicited Assessment toSpontaneous Behavior

The free-play and best story conditionswere included to provide measures of therelation between the highly structured,elicited-imitation assessment and morespontaneous behaviors. The data indicatethat a striking developmental shift occurredin this relation. Beginning with Step 5,the first step at the cognitive level of repre-sentational mappings, half or more of thechildren did not play at their highest step(see Figure 2) or show that step in the beststory condition. The steps shown in thefree-play and best story conditions did con-tinue to increase with age, but at a lowerrate than the steps shown in the assessmentcondition (see Figure 1).

The use of modeling prior to the free-playcondition allowed us to test a hypothesisderived from previous studies that employedthe same type of procedure—that childrenwould show an acquisition and omissionsequence (Harnick, 1978; Watson & Fischer,1977). This hypothesis received no supportin either experiment. In addition, behaviorunder the free-play condition did not showGuttman scales.

Further research will be required to deter-mine the meaning of the developmentalshift in highest step shown in the spontaneousconditions, but methodological problemswill not easily account for it. Step 5, wherethe shift began, involves a type of story thatis often seen in children's play—a doctorexamining a patient. Also, Steps 5 and/or6 occurred in the play and best stories ofmost children whose highest step in theassessment was 6, 7, or 8, and Steps 7 and 8occurred in the play and best stories ofsome children.

To check the possibility that the scoringcriteria for the highest steps were somehowresponsible for the shift, the first authorrescored the videotapes of the nine childrenwho showed the largest differences—twoor more steps—between assessment andfree play. This rescoring was an open-ended


search for any activity that could possiblybe interpreted as a role intersection or otherform of representational system. None of thechildren acted out a story that could beidentified as involving a representationalsystem of any kind.

The pattern of the data also seems topreclude an explanation of the results interms of statistical artifact. Suppose that thechildren had some upper limit on the numberof steps that they would show in play; inthat case, the proportion of children demon-strating a step in play would decrease asmore steps were available to the child, sothat the proportion of children playing attheir highest step would fall gradually acrossthe entire sequence, not abruptly at Step 5.

If the developmental shift cannot beexplained away, then what could it mean?One possibility is that children at this pointundergo some sort of motivational shift suchthat they are no longer motivated to per-form at the highest step that they are capableof. Another related possibility is that theshift reflects an emerging separation ofimitative and spontaneous abilities.

Whatever the proper explanation of thedevelopmental shift, the finding does suggesta different orientation to the question of thevalidity of imitation as a tool for assessingcognitive development. Instead of askingwhether imitation is a good general assess-ment tool, perhaps it would be more produc-tive to ask how measures based on directimitation relate to other measures of develop-ment, especially those involving more spon-taneous behavior. For example, the de-velopmental shift would seem to imply thatimitation provides a good measure of achild's capacities in the early preschoolyears. Beginning at about age 4, however,imitation may systematically overestimatea child's spontaneous preferences. Or statedthe opposite way, spontaneous behaviormay underestimate capacities of imitation.

Indeed, this orientation can be appliedbeyond the realm of imitation. Rigorousmethods like the elicited-assessment pro-cedure provide not only strong tests ofdevelopmental sequences but also the pos-sibility of relating those sequences to spon-taneous behavior. The developmental

sequence of social roles that we have testedshould apply to most concrete roles thatchildren experience commonly in their dailylives, roles like mother and father, boy andgirl, teacher and pupil. This sequence,measured in various role domains, willprovide measures that can be used withsome precision to investigate the relationbetween role concepts and competence inspontaneous social interactions involvingthe same roles. In this way, perhaps re-searchers can begin to understand more fullyhow cognitive-developmental change relatesto the child's normal behavior in every-day life.

References

Bigner, J. J. Second boms' discrimination of siblingrole concepts. Developmental Psychology, 1974,10, 564-673.

Bloom, L., Rocissano, L., & Hood, L. Adult-childdiscourse: Developmental interaction betweeninformation processing and linguistic knowledge.Cognitive Psychology, 1976, 8, 521-522.

Chambers, J. C., & Tavuchis, N. Kids and kin:Children's understanding of American kin terms.Journal of Child Language, 1976, J, 63-80.

Dasen, P., Inhelder, B., Lavallee, M., & Retschitzki,J. Naissance de I'intelligence chez I'enfant baoulede Cote d'lvoire. Berne, Switzerland: Huber, 1978.

Elkind, D. Children's conception of brother andsister: Piaget Replication Study V. Journal of GeneticPsychology, 1962, 100, 129-136.

EI'Konin, D. Symbolics and its functions in the playof children. In R. E. Herron & B. Sutton-Smith(Eds.), Child's play. New York: Wiley, 1971.

Emmerich, W. Young children's discriminations ofparent and child roles. Child Development, 1959,30, 403-419.

Emmerich, W. Family role concepts of children agessix to ten. Child Development, 1961, 32, 609-624.

Fischer, K. W. A theory of cognitive development:The control and construction of hierarchies of skills.Psychological Review, in press.

Fischer, K. W., & Corrigan, R. A skill approachto language development. In R. Stark (Ed.), Languagebehavior in infancy and early childhood. The Nether-lands: Amsterdam, Elsevier North Holland, in press.

Gelman, R. Cognitive development. Annual Reviewof Psychology, 1978,29, 297-332.

Gottlieb, S. Modeling effects upon fantasy. In J. L.Singer, The child's world of make-believe: Experi-mental studies of imaginative play. New York:Academic Press, 1973.

Green, B. F. A method of scalogram analysis usingsummary statistics. Psychometrika, 1956, / , 79-88.


Greenfield, P. M., & Childs, C. P. Understandingsibling concepts: A developmental study of kinterms in Zinacantan. In P. Dasen (Ed.), Cross-cultural Piagetian psychology. New York: GardnerPress, 1977.

Harnick, F. S. The relationship between ability leveland task difficulty in producing imitation in infants.Child Development, 1978,49, 209-212.

Harter, S. A cognitive-developmental approach tochildren's expression of conflicting feelings and atechnique to facilitate such expression in playtherapy. Journal of Consulting and Clinical Psy-chology, 1977,45, 417-432.

McCall, R. B., Parke, R. D., & Kavanaugh, R. D.Imitation of live and televised models in childrenone to three years of age. Monographs of theSociety for Research in Child Development, 1977,42(5, Serial No. 173).

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

Piaget, J. Judgment and reasoning in the child (M.Warden, trans.). London: Routledge & Kegan Paul,1928.

Piaget, J. Play, dreams and imitation in childhood

(C. Gattegno & F. M. Hodgson, trans.). New York:Norton, 1951.

Rubin, K. H., & Pepler, D. J. The relationship ofchild's play to social-cognitive growth and develop-ment. In H. Foot, T. Chapman, & J. Smith (Eds.),Friendship and childhood relationships. London:Wiley, in press.

Scollon, R. Conversations with a one year old. Honolulu:University Press of Hawaii, 1976.

Sigel, I. E., Saltz, E., & Roskind, W. Variablesdetermining concept conservation in children.Journal of Experimental Psychology, 1967, 74,471-475.

Slobin, D. I., & Welsh, C. A. Elicited imitationas a research tool in developmental psycholinguistics.In C. A. Ferguson & D. I. Slobin (Eds.), Studiesof child language development. New York: Holt,Rinehart & Winston, 1973.

Watson, M. W., & Fischer, K. W. A developmentalsequence of agent use in late infancy. Child De-velopment, 1977, 48, 828-836.

Received February 2, 1980

development of social roles in elicited and spontaneous …ddl/articlescopy/watson... · 2015. 1....

Documents